Reports on Progress in Physics Rep. Prog. Phys. 86 (2023) 026901 (207pp) https://doi.org/10.1088/1361-6633/ac9cef Review Teleparallel gravity: from theory to cosmology Sebastian Bahamonde1,2, Konstantinos F Dialektopoulos3,4,5, Celia Escamilla-Rivera6, Gabriel Farrugia7,8, Viktor Gakis7,8,9, Martin Hendry10, Manuel Hohmann1, Jackson Levi Said7,8,∗, Jurgen Mifsud7,8 and Eleonora Di Valentino11 1 Laboratory of Theoretical Physics, Institute of Physics, University of Tartu, W. Ostwaldi 1, 50411 Tartu, Estonia 2 Department of Physics, Tokyo Institute of Technology, 1-12-1 Ookayama, Meguro-ku, Tokyo 152-8551, Japan 3 Center for Gravitation and Cosmology, College of Physical Science and Technology, Yangzhou University, Yangzhou 225009, People’s Republic of China 4 Laboratory of Physics, Faculty of Engineering, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece 5 Department of Physics, Nazarbayev University, 53 Kabanbay Batyr Avenue, 010000 Astana, Kazakhstan 6 Instituto de Ciencias Nucleares, Universidad Nacional Autónoma de México, Circuito Exterior C.U., A.P. 70-543, México D.F. 04510, Mexico 7 Institute of Space Sciences and Astronomy, University of Malta, Msida, Malta 8 Department of Physics, University of Malta, Msida, Malta 9 Department of Physics, National Technical University of Athens, Zografou Campus GR 157 73, Athens, Greece 10 SUPA, School of Physics and Astronomy, University of Glasgow, Glasgow G12 8QQ, United Kingdom 11 Institute for Particle Physics Phenomenology, Department of Physics, Durham University, Durham DH1 3LE, United Kingdom E-mail: jackson.said@um.edu.mt Received 15 September 2021, revised 30 September 2022 Accepted for publication 24 October 2022 Published 1 February 2023 Corresponding editor: Dr Jun’ichi Yokoyama Abstract Teleparallel gravity (TG) has significantly increased in popularity in recent decades, bringing attention to Einstein’s other theory of gravity. In this Review, we give a comprehensive introduction to how teleparallel geometry is developed as a gauge theory of translations together with all the other properties of gauge field theory. This relates the geometry to the broader metric-affine approach to forming gravitational theories where we describe a systematic way of constructing consistent teleparallel theories that respect certain physical conditions such as local Lorentz invariance. We first use TG to formulate a teleparallel equivalent of general relativity (GR) which is dynamically equivalent to GR but which may have different behaviors for other scenarios, such as quantum gravity. After setting this foundation, we describe the plethora of modified teleparallel theories of gravity that have been proposed in the literature. We attempt to connect them together into general classes of covariant gravitational theories. Of particular interest, we highlight the recent proposal of a teleparallel analogue of Horndeski gravity which ∗ Author to whom any correspondence should be addressed. 1361-6633/23/026901+207$33.00 1 © 2023 IOP Publishing Ltd Printed in the UK Rep. Prog. Phys. 86 (2023) 026901 Review offers the possibility of reviving all of the regular Horndeski contributions. In the second part of the Review, we first survey works in teleparallel astrophysics literature where we focus on the open questions in this regime of physics. We then discuss the cosmological consequences for the various formulations of TG. We do this at background level by exploring works using various approaches ranging from dynamical systems to Noether symmetries, and more. Naturally, we then discuss perturbation theory, firstly by giving a concise approach in which this can be applied in TG theories and then apply it to a number of important theories in the literature. Finally, we examine works in observational and precision cosmology across the plethora of proposal theories. This is done using some of the latest observations and is used to tackle cosmological tensions which may be alleviated in teleparallel cosmology. We also introduce a number of recent works in the application of machine learning to gravity, we do this through deep learning and Gaussian processes, together with discussions about other approaches in the literature. Supplementary material for this article is available online Keywords: teleparallel gravity, observational cosmology, astrophysics (Some figures may appear in colour only in the online journal) Contents 3.6.3.3. Real tetrad for k ⩽ 0—SO(1,3) 3.6.4. Maximally symmetric spacetimes—ISO(1,3), SO(1,4), SO(2,3) 3.7. Conformal and disformal transformations 3.8. The teleparallel Gauss–Bonnet invariant 3.9. Analogue with other branches of physics 4. The teleparallel formulation of gravity and its GR equivalent 4.1. General properties of TG 4.1.1. Action and field equations 4.1.2. Local Lorentz invariance of the matter action 4.1.3. Local Lorentz invariance of the gravitational action 4.1.4. Palatini approach 4.1.5. Energy-momentum conservation 4.1.6. Bianchi identities 4.1.7. Premetric approach 4.2. The ADM formalism 4.2.1. 3 + 1 decomposition of teleparallel geometry 4.2.2. Lapse and shift variables 4.2.3. Lorentz gauge fixing 4.2.4. Canonical variables 4.2.5. Irreducible decomposition 4.3. The teleparallel equivalent of GR 4.3.1. The TEGR action and field equations 4.3.2. The gravitational energy problem 4.3.3. The DoF of TEGR 4.4. Outlook toward quantum TG and other dimensions 5. Modified TG theories 5.1. Foundations of modified teleparallel theories and how to make them fully invariant 5.2. New general relativity 1. Introduction 4 1.1. Original results and corrections to the literature 6 1.2. Review structure 7 1.3. Notation and conventions 8 2. Metric-affine approaches 10 2.1. Metric-affine geometry 10 2.1.1. Fundamental variables 10 2.1.2. Characteristic tensors 10 2.1.3. Bianchi identities 10 2.1.4. Connection decomposition 11 2.1.5. Metric-affine subclasses 11 2.2. Tetrads and the spin connection 13 2.2.1. Relation to metric and connection 13 2.2.2. Characteristic tensors 14 2.2.3. Local Lorentz transformations and Weitzenböck gauge 14 2.2.4. Differential form formulation 15 2.3. Geometric trinity of gravity 15 3. Torsional teleparallel geometries 17 3.1. TG as a gauge theory of translations 17 3.2. Gravitational coupling prescription 18 3.3. The field strength of gravity 19 3.4. Inertial and gravitational effects 19 3.5. Coupling to matter 22 3.5.1. Scalar fields 23 3.5.2. Fermion fields 23 3.5.3. Boson fields 24 3.5.4. Effects on the geodesic equation 25 3.5.5. A note on the energy-momentum tensor 25 3.6. Symmetries of teleparallel geometries 26 3.6.1. Axial symmetry—SO(2) 27 3.6.2. Spherical symmetry—SO(3) 27 3.6.3. FLRW cosmological symmetry 29 3.6.3.1. Flat case (k = 0)—ISO(3) 30 3.6.3.2. Real tetrad for k ⩾ 0—SO(4) 30 2 31 32 33 34 35 37 37 37 39 40 42 42 43 44 45 45 46 47 48 49 50 50 52 52 53 54 54 56 Rep. Prog. Phys. 86 (2023) 026901 5.2.1. Action and field equations 5.2.2. Hamiltonian analysis and DoF 5.3. f (T) gravity 5.3.1. Action and field equations 5.3.2. Conformal and disformal transformations in f (T) gravity 5.4. f(T, B) gravity 5.5. Extensions of NGR 5.6. Gauss–Bonnet theories 5.7. Higher-order derivatives 5.8. Scalar-tensor theories 5.8.1. Scalar-tensor theories with couplings with T and B 5.8.1.1. Action and field equations 5.8.1.2. Conformal transformations teleparallel dark energy 5.8.2. Teleparallel Horndeski gravity—BDLS theory 5.9. Theories non-minimally coupled with matter 5.10. Non-local theories 5.11. Other approaches 5.12. Good tetrad spin connection examples 5.12.1. Static spherically symmetric case 5.12.2. Time-dependent spherically symmetric case 5.12.3. Good tetrad-spin connection pair: axial symmetry 5.12.4. Cylindrical and other symmetries 6. Cosmology in TG 6.1. f (T) cosmology 6.1.1. Reconstruction method 6.1.2. Noether’s symmetry approach 6.1.3. Dynamical system approach 6.2. f(T, B) cosmology 6.2.1. Reconstruction method 6.2.2. Noether’s symmetry approach 6.2.3. Dynamical system approach 6.3. Teleparallel Gauss–Bonnet cosmology 6.3.1. Reconstruction method 6.3.2. Noether’s symmetry approach 6.3.3. Dynamical system approach 6.4. Teleparallel cosmology with non-minimal couplings to matter 6.4.1. Reconstruction method 6.4.2. Dynamical system approach 6.5. Teleparallel scalar-tensor cosmology 6.5.1. Reconstruction method 6.5.2. Noether’s symmetry approach 6.5.2.1. Minimally coupled scalar fields 6.5.2.2. Non-minimally coupled scalar fields 6.5.3. Dynamical system approach 6.5.3.1. Minimally coupled scalar fields 6.5.3.2. Non-minimally coupled scalar fields 6.6. Other theories Review 56 58 61 61 6.6.1. Non-local theories 106 6.6.1.1. Noether’s symmetry approach 106 6.6.1.2. Dynamical system approach 107 6.6.2. Higher-order derivative TG 107 6.6.2.1. Noether’s symmetry approach 107 6.6.2.2. Dynamical system approach 107 6.6.3. Mimetic and unimodular teleparallel cosmology 108 6.6.4. Teleparallel axions 108 6.7. Inflation 109 6.7.1. Born–Infeld inspired inflation 109 6.7.2. Power spectra in LQC f (T) gravity 111 6.7.3. Higgs inflation 112 6.8. Essential conditions for viability in cosmological models 112 6.9. Perspectives of teleparallel cosmology 113 7. Cosmological perturbations in TG 115 7.1. 3 + 1 split in cosmology 116 7.2. Irreducible scalar-vector-tensor decomposition 117 7.3. Gauge transformations 118 7.4. Tensor perturbations 119 7.5. Vector perturbations 121 7.6. Scalar perturbations 121 7.6.1. f(T, B) gravity 121 7.6.2. f(T, Θ) gravity 122 7.7. Growth index and the fσ8 parameter 123 7.7.1. f (T) gravity 124 7.7.2. f(T, B) gravity 124 7.7.3. f(T, Θ) gravity 125 7.7.4. Closing remarks 126 8. Polarization of GWs in TG 127 8.1. Polarization modes in metric theories of gravity 127 8.2. Polarization modes in NGR 129 8.2.1. Linearized regime and field equations 129 8.2.2. Principal symbol approach 130 8.2.3. Newman–Penrose formalism and polarizations 132 8.3. Polarization modes in f(T, B) gravity 132 8.4. Polarization modes in the teleparallel analogue of Horndeski gravity 134 9. Astrophysical systems 137 9.1. Spherical symmetry and solutions 137 9.2. Generalized Birkhoff’s theorem 140 9.3. Solar system tests 141 9.3.1. Solar system tests using particle motion observables 141 9.3.2. Parametrized post-Newtoninan formalism 145 9.4. Astrophysical compact objects 147 9.4.1. Stars 147 9.4.2. Black holes 149 9.4.3. Wormholes 150 9.5. Galaxy rotation curves 151 9.6. Overview and future work 152 10. Observational and precision cosmology in TG 153 10.1. State-of-the-art 155 10.2. Viability of teleparallel cosmology 157 63 65 65 66 67 67 67 67 69 69 71 73 74 75 75 77 79 81 82 83 84 86 86 90 90 90 91 94 95 96 96 97 98 98 98 99 100 100 100 102 102 103 106 3 Rep. Prog. Phys. 86 (2023) 026901 Review 10.3. Compendium of teleparallel cosmologies 158 10.3.1. Precision cosmology for teleparallel scenarios 159 10.3.2. Precision cosmology for f (T) 160 10.3.3. Precision cosmology for f(T, B) gravity 161 10.3.4. Cosmography for TG gravity 164 10.3.5. Probing f (T) gravity with the induced variation of fundamental constants 167 10.4. Beyond data-driven analysis for TG cosmologies 168 10.4.1. Gaussian processes and the reconstruction of f(T) gravity 170 10.4.2. Machine learning for teleparallel cosmology 174 10.5. Perspectives in observational cosmology 177 11. Conclusion 178 Data availability statement 180 Acknowledgments 180 Appendix A. Variations in TG 181 A.1. Basic teleparallel quantities 181 A.2. Calculation of the gauge current 181 A.3. Example teleparallel theories 182 A.3.1. f (T,B,ϕ,X). 182 A.3.2. f (T,B,T G ,BG ) theory. 183 A.3.3. Teleparallel analogue of Horndeski theory. 183 Appendix B. Geff calculation in f(T,B) gravity 185 Appendix C. PPN parameters for teleparallel analogue of Horndeski 186 Appendix D. EoS terms for f(T,B) gravity 186 References 188 CMB SVT VLBI SH0ES HST TRGB TDCOSMO SLACS LISA ACT SPT DR Gaia EDR3 WL DES SDSS dF BOSS JLA KiDS BBN GP CMASS MCMC KVNO LiteBIRD CORE PIXIE PRISM PICO SKA DESI BINGO Nomenclature GR GW CDM TG TEGR DoF NGR wrt FLRW EoS LHS RHS STEGR TEGB PPN SN BAO CC DE HDE PlDE QCD General Relativity Gravitational Waves Cold Dark Matter Teleparallel Gravity Teleparallel Equivalent of General Relativity Degrees of Freedom New General Relativity with respect to Friedmann–Lemaı̂tre–Robertson–Walker equation of state Left hand side Right hand side Symmetric teleparallel equivalent of general relativity Teleparallel equivalent of the Gauss–Bonnet Parameterized post-Newtonian formalism Supernova (we use abbreviations SNeIa and SNIa to refer to supernova type 1a) Baryonic Acoustic Oscillations Cosmic Chronometers Dark energy Holographic Dark Energy Pilgrim Dark Energy Quantum Chromodynamics Cosmic Microwave Background Scalar-Vector-Tensor Very Long Baseline Interferometry Supernova H0 for the equation of State Hubble Space Telescope Tip of the red-giant branch Time-delay Cosmography Sloan Lens ACS Laser Interferometer Space Antenna Atacama Cosmology Telescope South Pole Telescope Data Release Gaia Early Data Release 3 Weak Lensing Dark Energy Survey Sloan Digital Sky Survey degree field Baryon Oscillation Spectroscopic Survey Joint Light-curve Analysis Kilo-Degree Survey Big Bang Nucleosynthesis Gaussian processes Constant stellar MASS Markov chain Monte Carlo Keck+VLT+UVES+Oklo measurements Lite (Light) satellite for the studies of B-mode polarization and Inflation from cosmic background Radiation Detection Cosmic Origins Explorer Primordial Inflation Explorer Polarized Radiation Imaging and Spectroscopy Mission Probe of Inflation and Cosmic Origins Square Kilometre Array Dark Energy Spectroscopic Instrument Baryon acoustic oscillation In Neutral Gas Observations 1. Introduction General relativity (GR) is an astonishing theory both in its simplicity and its ability to retain a theoretically consistent framework while satisfying observational tests at almost all scales of physics. Together with quantum field theory, it forms one of the two pillars of modern physics. Its precise predictions and influence over physical processes in the Universe has set a standard for which all other physical theories aspire to. After framing gravity through the perspective of his strong equivalence principle, Einstein then formulated his field equations by adopting the Riemann tensor as the fundamental building block of geometric deformation, as suggested by Marcel Grossmann [1]. On the demand that the field equations be conserved, Einstein worked out their exact form through a process of reconstruction and elimination. Later on, it was noticed that they could have also been derived from the Bianchi identities, from which one can derive an Einstein tensor ◦ ◦ ◦ Gµν that is covariantly conserved ∇µ Gµν = 0 (for a covari◦ ant derivative ∇µ defined using the Levi-Civita connection ◦ Γρµν which is further explained in section 1.3) [2]. However, 4 Rep. Prog. Phys. 86 (2023) 026901 Review Einstein did not have this information at the time. Using his approach, he was able to write down the field equation as [3] ◦ Gµν = gravitational section needs further revisiting. This may take the form of extensions from GR or even modifications beyond GR that may be an alternative to its original formulation. In the literature there have been numerous proposals for new theories of gravity [22–26] which are motivated either through phenomena or some theoretical approach, as well as for other reasons such as from quantum physics. One interesting possibility that has been gaining momentum in the literature in recent decades is that of teleparallel gravity (TG) where curvature is replaced by torsion as the mechanism by which geometric deformation produces a gravitational field. It does this by dislodging the uniquely curvature-based Levi-Civita connection with a torsion-based teleparallel connection. In fact, there are now thousands of publications related to the topic in the literature. One of the many theories that torsion-based approaches to gravity has produced is the teleparallel equivalent of general relativity (TEGR) which is dynamically equivalent to GR, meaning that they cannot be distinguished through classical experiments. In fact, even Einstein himself, while not aware of the observational motivations for modifying GR, was also one of the first proposers for modifications of GR shortly after its inception. This took various forms with varying degrees of novelty. However, Einstein was captured by teleparallelism due to its reliance of the tetrad frames and its absolute teleparallelism feature [27]. Similar to the largely silent period between his papers on special and GR, Einstein devoted himself to teleparallelism and its potential for a unified field theory. He spent this time in correspondence with many of the main proponents for the new geometric framework at the time such as Roland Weitzenböck, Élie Cartan, Luther P Eisenhart, Herman Müntz, Jakob Grommer and Cornelius Lanczos to mention some. The main objective for Einstein was to use this new approach of absolute teleparallelism to unify gravity and electromagnetism since tetrad fields can support 16 independent components representing ten degrees of freedom (DoF) that are related to the metric tensor while the remaining six would correspond to a separate connection that he wanted to show could be yielded to produce the electromagnetic sector. However, after three intense years of research between the 1928 summer and 1931 spring, together with eight publications on the topic, Einstein eventually abandoned the idea. He failed to establish a relationship between this new connection and electromagnetism which was his ultimate goal of having a unified field theory approach to gravity and electromagnetism. Instead he found that these seemingly extra DoF were related to the Lorentz group and simply a manifestation of the local Lorentz freedom of gravity. The unified approach was largely abandoned after that time and only revived much later on by Møller [28] and then Pellegrini and Plebanski [29] where they formulated a Lagrangian that satisfies the concepts behind absolute parallelism. This was also followed by [30] where the first ideas of a well defined energy-momentum tensor for gravitation were conceptualized, and which led to significant advances in later decades as compared with the analogous case in GR, which we now know to be quasi-local [31]. Work on the topic remained in the foundations sector for the next couple of decades with 8πG Θµν , c4 for an energy-momentum tensor Θµν , and where G is Newton’s constant of gravity and c the speed of light [4]. These equations remain unchanged from their original form as presented by Einstein to the Prussian Academy of Sciences in November 1915 [5]. It is these equations which have been used to explore the expansion and evolution of the Universe from its infinitesimal beginnings to its currently accelerating expansion phase, as well as the nature of black holes, the propagation of gravitational waves (GW) [6] and the formation of galactic and compact structures in the Universe. There are even several attempts to extend the theory to meet the growing challenges from quantum theory and high energy particle physics. The development of GR has undergone astounding success over the decades. However, the establishment of quantum mechanics in the 1920s and then quantum field theory later on already started to expose some of the limitations of the theory early on. Observational constraints have also focused attention on some of the failings of GR which not only occur on the very small scale of the physical spectrum but also on larger scales such as for galactic systems where modifications of the standard model of particles physics [7–9] are necessary to preserve GR as the underlying theory of gravity. This has led to the realization that dark matter plays an important role in galaxies and their clusters [10–12] in the astrophysical context but also has a crucial role in the cosmological context in the dynamics of the early Universe [13]. The best macroscopic version of this is cold dark matter (CDM). GR also needs the addition of further corrections in the matter sector when considering the very early Universe where a rapid period of inflation [14, 15] is necessitated to explain the extremely flat and homogeneous Universe we observe at present. Saying that, there are other solutions to explain how inflation came about without any modification to the matter description of the standard model such as the Starobinsky’s model [16] which modified the early Universe evolution in comparison to the standard model of cosmology. It may also be the case that modifications to the gravity sector may also explain this feature of galactic rotation curves such as in the case of Starobinsky gravity [17] and conformal Weyl gravity [18]. In the same vein, the late-time Universe is undergoing a period of accelerated expansion [19, 20] which requires a cosmological constant Λ [21] to adequately describe the observed expansion using GR as the fundamental to describe gravitational phenomena. Altogether, GR can thus explain most phenomena provided adequate modifications are considered in the matter sector so that we consider the widely accepted concordance model in ΛCDM plus inflation cosmology. However, the exotic nature of these particle species remains a total mystery in terms of observations despite significant theoretical advances in physics beyond the standard model of particle physics. In the same vein, it may also be the case that the standard model of particle physics does not require a significant a restructuring for meeting these observational challenges and it is that the 5 Rep. Prog. Phys. 86 (2023) 026901 Review compelling works such as the review by Hehl et al [32]. Independently, Hayashi and Nakano [33] also started to revive a form of Einstein’s origin work. In their work, they formulated absolute teleparallelism through a gauge theory of translations on the Lorentz group. This was further extended to a fully fledged gravitational theory in [34] where a new general relativity (NGR) formalism was presented together with its weak field limit. It was at this time that the motivations for looking for theories of gravity beyond GR was starting to take off with many new toy model proposals tackling various theoretical and observational issues in the topic. In this background, absolute teleparallelism was not actively studied beyond very select groups. This was partly due to its very foundational nature as compared to other popular toy model theories at the time, which appealed to a much larger audience. This naturally leads us to the question of how would modern gravitational research look like if Einstein had continued to work on the concepts surrounding absolute teleparallelism, or if he had formulated GR in terms of torsion from the beginning (namely, TEGR)? This construction of gravity would have been written in terms of gauge theories thus eliminating the need for many future works on formulating GR as a gauge theory within the curvature-based Levi-Civita connection. This is all the more prescient due to the fact that both GR and TEGR both predict the same observational outcomes in their classical predictions (due to their dynamical equivalence). Absolute teleparallelism was finally revived in its latest incarnation in the numerous works by Aldrovandi and Pereira (among others) which culminated in two pivotal books on the topic. The first is [35] where the underlying geometry of GR is revisited and the foundational ideas behind various concepts in TG are put on a concrete mathematical basis, while the second is [36] where the physics on which TEGR is fully fleshed out together with potential future work on quantum gravity and quantum field theory. These works again resurrect Einstein’s original goal of reconciling gravity with other branches of physics [37]. TEGR continues to be studied for potential benefits as compared with GR such as in its quantum regime, and others. On the other hand, TG has since undergone a renaissance with respect to (wrt) potential modified theories of gravity. The first such modification was presented by Hayashi and Shirafuji [34] where GR was slightly modified to allow for small deviations from TEGR through a decomposition of its Lagrangian. However, the most impactful suggested modification came from Ferraro and Fiorini [38] where a form of f (T) gravity first appeared in the literature, and later by Linder [39]. The rationale here follows the same reason- Firstly, [43] where the foundations of TG and its particular application to TEGR, f (T) and some other teleparallel theories was explored. This work built on foundational work represented in the book in [36] and covered important advances on the issue of covariance in TG [44], among other topics. Another hugely important work is [45] where foundational aspects, such as the Poincaré gauge gravity origins of the theory, are explored but also many applications, both astrophysical and cosmology. These are important works in the community but the literature has since drastically evolved with even more clarity on the covariant formulation of TG and its relationship to metric-affine theories being more laid out. TG also now hosts a significantly expanded number of new theories of gravity, of particular interest we point out advances in the scalartensor section where an analogue of Horndeski gravity has also started gain popularity. In terms of observational astrophysics and cosmology, there has been serious new results such as in the regime of precision cosmology, observational constraints in new teleparallel models and novel approach to cosmology such as machine learning, as well as important progress in galactic and stellar physics. At the same time, observational physics, and particularly cosmology, is reaching the border of the region in which GR can satisfy measurements such as with the late-time cosmological tensions as well as ambiguities in early Universe physics. Given these critical advances in the literature and the resolution of many more foundational issues in TG as well as the growing pressure from observational cosmology for competitive new theories of gravity, we are motivated to reevaluate the topic and to bridge the gap between foundational theory and it observational consequences in astrophysical and cosmological systems. 1.1. Original results and corrections to the literature This Review contains some original results that have not been reported in the literature yet. In addition, throughout the process of reviewing the most important literature of TG, we have noticed some mistakes that we have corrected. We list them here for each section: • In section 4.1 we provide a general description for any TG theory in order to have a theory which certain physical properties such as having local Lorentz invariance of both the matter and gravitational action. • In section 5 we re-derived the field equations for the most popular teleparallel theories and express them with the same notation. • In section 5.12.1 we found the two most general good tetradspin connection pair in spherical symmetry for f(T, B, ϕ, X) gravity when one assumes that the teleparallel connection and the tetrad respect spherical symmetry. • Equation (5.8) corrects a sign mistake in [46]. • In section 6 we derived the flat Friedmann–Lemaı̂tre– Robertson–Walker (FLRW) equations for each model in a consistent way. We also corrected the reported FLRW equations in higher order derivative TG (see section 6.6.2). ◦ ing as in the very popular f(R) theory of gravity [40–42]. The spectrum of modified teleparallel theories has since drastically expanded to also include nonminimal couplings to matter, scalar fields as well as vector and tensor field additions, giving a plethora of teleparallel theories in which gravity can be formulated and studied. The literature on TG features a wide range of works with an increasing intensity of new additions to the growing body of publications on the topic. In this considerable body of works, we would like to point out two important reviews on the topic. 6 Rep. Prog. Phys. 86 (2023) 026901 Review • We verified the validity of the reconstructed solutions especially in cases where the modified gravity component behaves as a dark energy fluid. This appears in f (T) in section 6.1.1 and for f(T, Θ) in section 6.4.1. Other reported solutions have been checked and corrected where applicable. • In section 6.3.3, we have corrected the stability conditions for the Minkowski solution in the case of f(T, TG ) gravity. • We corrected the Noether symmetry conditions and the dynamical systems approach reported in the minimal f (T)scalar coupling in sections 6.5.2.1 and 6.5.3.1. • We corrected the dynamical analysis for the non-local f (T) gravity approach in section 6.6.1.2. • In the supplementary annexes (supplementary 1), the finding good tetrad-spin connection pairs approach for different anisotropic cosmological scenarios has been performed. Sets of good tetrad-spin connection pairs are also presented. We also corrected instances where bad tetrad-spin connection pairs have been used in literature. • In anisotropic cosmologies, some reconstructed solutions such as in the supplementary annexes (supplementary 1) have been corrected. Moreover, in another part of the supplementary annexes (supplementary 1), the FLRW limiting behavior in the Noether symmetry approach has been corrected. • A brief new discussion about the generalized Birkhoff’s theorem in f(T, B) gravity is provided in section 9.2. • In section 10.1 we provided a full compendium of TG models and their analysis perspectives in precision and observational cosmology. • In section 10.4.2 we analyzed TG theories using machine learning techniques. • In section 10.3.4 we corrected the kinetic equations for f (T) cosmography. This is manifested in equations (10.15c) and (10.15d) where we corrected the snap parameters terms, which changes the current results already reported in the literature up to third order. For future analysis, this can change the kinetic evolution of the Universe at higher redshifts. as having an associated gauge field strength. We also discuss symmetries and transformations in teleparallel geometry and discuss how it couples to matter with the minimal coupling prescription. GR is built on the geometric foundations set by Riemann and others at the around the beginning of the twentieth century. In a similar vein, TG is build on the work by Weitzenböck and many others who laid the geometric foundations for a torsional rather than curvature-based formulation of gravity. In section 4, we explore the foundational construction of teleparallel theories and its application to the formulation of the TEGR, which is dynamically equivalent to GR while being born out of an entirely distinct geometric setting. Due to the tetrad-spin connection pair basis in which TG is constructed, we revisit a number of tenets of this theory of gravity such as its ADM formalism and the possible effects it may have on quantum gravity (where it may no longer be equivalent to GR). As in curvature-based theories of gravity, extensions to TEGR are interesting to study, as are modifications and alternative to TEGR. We explore the most promising of gravitational theories in section 5 where we introduce theories through their actions and determine many of their ensuing field equations. After laying the foundations of TG and exploring some of its most popular manifestations in theory, we move to the astrophysics sector in which we first describe the impact of TG on gravitational radiation. We here focus on the polarization structure of GW in section 8 where we reexamine some of the theories promoted in section 5, in this context. In section 9, we then move on to other aspects of astrophysical systems in TG such as the issue of spherically symmetric solutions and issues related to time-dependent solutions, as well as other symmetry settings. We use this information to discuss physical stars and to discuss open problems in this regime of the literature. Finally we close this area with a discussion of some work in the regime of galactic rotation curves in a particular setting of f (T) gravity. Finally, we tackle the cosmological consequences for some of the modified teleparallel theories of gravity explained in section 5. We start with an analysis of the numerous background cosmology analyses in section 6 which forms the majority of works in the literature. This section delves into the many works on reconstruction methods, Noether symmetries, and dynamical systems. Following this lengthy exploration, we turn to the perturbative side of TG in section 7, where we explain how perturbations can be taken consistently in TG and what form they should take. We also tackle the issue of perturbations in a cosmological background and some impacts on possible inflationary theory. Finally, section 10 contains the core observational predictions from TG in the cosmological regime. We here review works that constrain some of the most popular models within TG through late-time, as well as give new perspectives on prospects of machine learning applications in teleparallel cosmology. We also discuss the state of the art in terms of the near-term prospects for observational cosmology and its possible impact in discriminating between models in gravity beyond ΛCDM. There are also some other minor new additions or corrected mistakes from the literature that we believe, are not so important to mention above. 1.2. Review structure The Review is intended to cover the main developments in TG and its impact on astrophysics, GW and cosmology. To facilitate this, we divide the Review into four main thematic areas covering the most important results coming from the last few decades while still comprehensively describing its origins. We show the Review breakdown graphically below and then go on to describe how each section is broken down. We first open with an exploration of the geometric underpinnings of metric-affine geometry in section 2 where the socalled trinity of gravity is cemented as our starting off base. In section 3, we then focus on the specific foundations surrounding torsional teleparallel geometry. We here motivate teleparallel geometry as a gauge theory of translations and describe it 7 Rep. Prog. Phys. 86 (2023) 026901 Review ds 2 = ηµν dxµ dxν = dt 2 − dx 2 − dy 2 − dz 2 , 1.3. Notation and conventions We lay out our conventions in this part of the Introduction. The literature on TG contains numerous different combinations of conventions since there exist a number of non-competing choices one can take with respect to various definitions throughout the layers on which TG is constructed. In the remainder of this work, we refer to the definitions declared here as the review convention. Firstly, we use a metric signature (+1, −1, −1, −1) which is the dominant signature in the TG literature. Also, we choose to write spacetime indices using lower case Greek alphabet letters (α, β, . . .) and capital Latin letters (A, B, . . .) for the Minkowski spacetime (Lorentz indices). Capital Latin letters starting from (I, J, . . .) will denote purely spatial indices on the Minkowski spacetime (see section 8), while lower case Latin letters from the middle of the alphabet (i, j, . . .) are reserved for purely spatial indices (see sections 2.2.1 and 4.2.1). We also define the following commonly used terms related to the foundational formulation of TG (1.1) where d denotes the differential operator and ηµν the Minkowski metric. All metric tensor line elements are produced by the tetrads eAµ and inverse tetrads EAµ (see section 3.2) gµν = ηAB eAµ eBν , gµν = η AB EAµ EBν , and (1.2) where the tetrad is a fundamental dynamical variable in the theory. The tangent spacetime is described by trivial tetrads which are denoted by hAµ with inverses HAµ . We will also define the one-forms e A = eAµ dxµ and vector tetrads EA = EAµ ∂µ (together with the trivial one-forms h A = hAµ dxµ and trivial vector tetrads HA = HAµ ∂µ ) (see section 3.1). Tensor and operator quantities that have a dependence on the connection being employed are distinguished by the following definitions for the different possible connection options (see section 2) b α e.g. R bα ; • The general connection Γ µν βµν • Teleparallel connection—the general torsional teleparallel connection that is composed of both the tetrad and spin connection defined in equation (3.30); • Weitzenböck connection—the teleparallel connection in the Weitzenböck gauge (see section 2.2.3) where the spin connection vanishes (see section 5.1); • Weitzenböck gauge—the case where the spin connection vanishes (see section 4.2). ◦ ◦ • The Levi-Civita Connection Γ αµν e.g. R αβµν ; ♢ ♢ • The symmetric teleparallel connection Γ αµν e.g. R αβµν ; • The teleparallel connection Γ αµν e.g. R αβµν , which are used throughout the Review. We also define the following derivative operators • The covariant derivative ∇µ (with over-symbols for the various connection choices—as described above) which acts only on the general spacetime; The line element for flat spacetime in Cartesian coordinates is then described by 8 Rep. Prog. Phys. 86 (2023) 026901 Review √ L = L −g, and where Sm refers to the matter action. We take special note here for objects calculated using the Levi-Civita connection (such as Rρσµν ) in that we refer to these as standard gravity objects since they are determined using the Levi-Civita connection. For minimally coupled matter system, the matter action will then be given by • The Fock–Ivanenko derivative Dµ which acts on Lorentz indices only (see section 3.3); • The exterior covariant derivative D and exterior derivative d (see section 2.2.4); • The d’Alembertian operator □ = ∇µ ∇µ = gµν ∇µ ∇ν . ˆ For any connection, we define the Riemann curvature tensor (2.26) as [2] bµ bµ bµ bµ bσ bµ Sm := ˆ d4 x L m , (1.9) bσ R ναβ := ∂α Γ νβ − ∂β Γ να + Γ σα Γ νβ − Γ σβ Γ να , (1.3) which then defines the energy-momentum tensor through the variation   ′ −2 δLm 1 δLm A Θµν := √ =e µ gνν ′ := eA µ ΘA ν gνν ′ . µν A ′ −g δg e δe ν (1.10) which defines the Levi-Civita Einstein tensor as ◦ ◦ 1◦ Gαβ := Rαβ − Rgαβ , 2 ◦ √ d4 x −gLm = (1.4) ◦ where R = gµν Rµν . On the other hand, the teleparallel connection is defined through  Γσ νµ := EAσ ∂µ eAν + ω ABµ eBν , (1.5) The trace of the energy-momentum tensor will be denoted by Θ = Θµµ . When writing the FLRW line-element, we use t to denote cosmic time (coordinate time as measured by a comoving observer in a homogeneous matter field) and τ (defined as dτ = N(t)dt/a(t)) to denote conformal time (see section 6.7.2). Thus, the FLRW metric in spherical-like spatial coordinates is written as (6.1) where ω ABµ represents the components of the teleparallel spin connection (see section 2.2). This can then be used to build the torsion (3.28), contortion (3.64) and superpotential (4.163) tensors as well as the torsion scalar (4.162) defined through  TA µν := 2ΓA [νµ] = eA ν,µ − eA µ,ν + ω A Bµ eB ν − ω A Bν e µ = −2Γ [µν] , B A ◦ Kρ µν := Γρ µν − Γρ µν = ds = N(t) dt − a(t) 2 (1.6a) 1 T := Sρ µν Tρ µν , 2 (1.6c) κ2 = 8πG, ˆ √ ◦ 1 d x −gR + Sm = 2 2κ ˆ 1 + Sm = 2 d4 x L + Sm , 2κ 4 (1.12a) (1.12b) which then gives the Hubble parameter and the conformal Hubble parameter definitions respectively as ȧ , Na a′ H(τ ) := , Na H(t) := (1.7) ˆ   dr 2 2 2 2 2 + r dϑ + sin ϑdφ , 1 − kr 2 (1.11) da , dt da a ′ (τ ) := , dτ which will be mainly used for the foundations part of the Review, while the observational sections will mainly use SI units (as will be indicated). As an example, consider the Einstein–Hilbert action (2.51) 1 SGR := 2 2κ 2 ȧ(t) := (1.6d) where square brackets represent the anti-symmetric operator, i.e. A[µν] := 21 (Aµν − Aνµ ). Also, all comma separated lower indices will refer to partial derivatives, such as f,T = ∂f/∂T. Unless otherwise stated, this Review observes the geometric units convention where c ≡ 1, 2 where N(t) is the lapse function and a(t) is the scale factor. We would like to point out that the lapse function cannot a priori be absorbed into the definition of cosmic time since it may be dynamical in some branches of certain theories. We can then define the cosmic and conformal time derivatives 1 (Tµ ρ ν + Tν ρ µ − Tρ µν ) , (1.6b) 2 Sρ µν := Kµν ρ − δρµ Tσ σν + δρν Tσ σµ = −Sρ νµ , 2 (1.13a) (1.13b) where the lapse gauge (in which N = 1) can be taken to recover the corresponding regular definitions, namely H = ȧ/a and H = a ′ /a. Throughout this Review we will refer to the energy density of a fluid as ρ, and its isotropic pressure as p. The equation of state (EoS) is defined as √ d4 x −gL (1.8) p ≡ wρ, where the GR subscript will be replaced for other theories, the Lagrangian L is related to the Lagrangian density L through (1.14) where w is the EoS parameter which often varies over the cosmic history of the Universe in modified theories of gravity. 9 Rep. Prog. Phys. 86 (2023) 026901 Review 2. Metric-affine approaches These notions are illustrated in figure 1 by their effect on the parallel transport of a vector defined by the affine connection b µ νρ . Curvature causes the parallel transport along a closed Γ curve to be non-trivial, i.e. to change the transported vector. With torsion, the parallel transport is not symmetric under exchanging the transported vector and the direction of transport. With non-metricity, the length of the vector, as measured by the metric, changes along the transport. Note that for the general affine connection Γ̂ρ µν , any of these tensorial quantities may be nonvanishing. More restricted classes of connections can be obtained by demanding that certain quantities vanish; we will discuss such cases in section 2.1.5. We start our Review with a brief overview of metric-affine geometry and its fundamental properties, which lie at the foundation of TG theories. In section 2.1, we provide a summary of general metric-affine geometry in its most commonly used formulation in terms of a metric and an affine connection. In the restricted case of TG, however, which is the main subject of this Review, a formulation in terms of a tetrad and spin connection is more common; this is Reviewed in section 2.2. Finally, in section 2.3, we discuss how different classes of geometries can be used as alternative formulations of GR, whose field equations agree with the usual, curvature-based formulation. The characterizing tensors of a general affine connection are not completely independent from each other. In particular, the curvature and torsion are related by the Bianchi identities [49] 2.1.3. Bianchi identities. 2.1. Metric-affine geometry We now give a summary here of metric-affine geometry and its characteristic properties. In section 2.1.1 we summarize the fundamental fields that constitute the metric-affine geometry. Their characterizing, tensorial properties are displayed in section 2.1.2. These tensorial quantities are related by the Bianchi identities, which we display in section 2.1.3. Further, we discuss how they can be used to decompose the connection in section 2.1.4. An overview of particular subclasses of metric-affine geometries is given in section 2.1.5. b [ω R bµ |ν|ρσ] = −R bµ ντ [ω T b τ ρσ] . ∇ (2.5) This can easily be seen by direct calculation. For the first identity (2.4), one has b [ν T b µ ρσ] − 2Γ b µ ω[ν Γ b ω ρσ] + 2Γ b ω [ρν Γ b µ |ω|σ] bµ ρσ] = −2∂[ν Γ ∇ b ω [σν Γ b µ ρ]ω , (2.6) + 2Γ In the most general metricaffine setting, which is more comprehensively reviewed in [47], the fundamental variables are a metric (i.e. a symmetric rank-two tensor field) gµν of Lorentzian signature, as defined b ρ µν in the conventions section 1.3, as well as the coefficients Γ of an affine connection, with associated covariant derivative b both defined on a four-dimensional spacetime manifold M. ∇, It thus follows that the metric has ten independent components, while the connection has 64 independent components, and there is no a priori relation between these two objects. Here and in the following we will use a hatbwhenever we refer to quantities which are related to this general affine connection. We have collected these notes on notation in section 1.3 for each of access. and b µ ω[ν Γ b ω ρσ] − 2Γ b µ [ν|ω| Γ b ω ρσ] . bµ ω[ν T b ω ρσ] = 2Γ T (2.7) Combining these two terms, and sorting the indices, one obtains the expression b [ω R b µ νσ] + 2Γ b µ ω[ρ Γ b ω νσ] bµ |ν|ρσ] + T bµ ω[ν T b ω ρσ] = 2∂[ρ Γ ∇ bµ [νρσ] , =R (2.8) which is the LHS of the Bianchi identity (2.4). Similarly, the second identity (2.5) can be proven. Expanding the covariant derivative on the LHS yields The most general metricaffine geometry defined by a metric and affine connection is characterized by a number of tensorial quantities, whose definition and conventions we use in this Review are given below. From the connection alone one obtains the curvature tensor 2.1.2. Characteristic tensors. b [ω R b µ τ [ω R b τ ν[ω R bµ |ν|ρσ] = ∂[ω R bµ |ν|ρσ] + Γ b τ |ν|ρσ] − Γ bµ |τ |ρσ] ∇ b τ [ρω R b τ [σω R bµ |ντ |σ] − Γ bµ |ν|ρ]τ . (2.9) −Γ bµ τ ρ Γ b τ νσ − Γ bµ τ σ Γ b τ νρ , b µ νσ − ∂σ Γ b µ νρ + Γ bµ νρσ = ∂ρ Γ R (2.1) Now it is easy to see that the last two terms are identical, b τ [ρω R b τ [σω R b τ [σω R bµ |ντ |σ] . (2.10) bµ |ντ |ρ] = Γ bµ |ν|ρ]τ = −Γ Γ as well as the torsion tensor (2.2) Further using the definition (2.2) of the torsion, as well as cyclically rearranging the indices within square brackets, the last two terms of the expansion (2.9) can be combined to Using also the metric, one further defines the non-metricity tensor b µ gνρ = ∂µ gνρ − Γ b σ νµ gσρ − Γ b σ ρµ gνσ . b µνρ = ∇ Q (2.4) and 2.1.1. Fundamental variables. b µ ρν − Γ b µ νρ . bµ νρ = Γ T b [ν T bµ [νρσ] = ∇ bµ ρσ] + T bµ ω[ν T b ω ρσ] , R b τ [ρω R bµ |ντ |σ] = −R bµ ντ [ω T bµ ρσ] , −2Γ (2.3) 10 (2.11) Rep. Prog. Phys. 86 (2023) 026901 Review Figure 1. Schematic geometrical representation of the curvature, torsion and non-metricity tensors by their effect on the parallel transport of vectors [48]. Reproduced with permission from [48]. © EESTI TEADUSTE AKADEEMIA. which is the RHS of the Bianchi identity (2.5). It remains to show that the first three terms on the RHS of the expansion (2.9) vanish. Expanding the curvature in the derivative term using its definition (2.1) yields An important property of metric-affine geometry is the fact that in the presence of a metb ρ µν of the general affine connection can ric, the coefficients Γ uniquely be decomposed into three parts in the form 2.1.4. Connection decomposition. b µ |τ |ρ Γ b τ |ν|σ] b µ |ν|σ] + 2∂[ω Γ bµ |ν|ρσ] = 2∂[ω ∂ρ Γ ∂[ω R b µ τ [ρ ∂ω Γ b τ |ν|σ] . + 2Γ (2.12) ◦ b ρ µν := Γρ µν + K b ρ µν + L bρ µν , Γ The first term on the RHS vanishes, since partial derivatives commute. In the remaining two terms one can substitute the derivatives on the connection coefficients by curvature tensors, which yields where we have introduced the coefficients ◦ 1 µσ µ Γ νρ := g (∂ν gσρ + ∂ρ gνσ − ∂σ gνρ ) , 2 b µ |τ |ρ Γ b τ |ν|σ] = Γ b τ ν[σ R b τ ν[σ Γ b µ |ϕ|ω Γ b ϕ |τ |ρ] , bµ |τ |ωρ] − 2Γ 2∂[ω Γ (2.13a)   b µ νρ := 1 T bν µ ρ + T bρ µ ν − T bµ νρ , K 2 After permuting the indices in square brackets, one finds that the two curvature terms found here cancel the second and third term on the RHS of the expansion (2.9). Finally, for the two terms cubic in the connection coefficients, one can exchange the dummy indices τ ↔ ϕ in the first term to find b [ω R bµ |ν|ρσ] = 0. ∇   b µ νρ − Q b ν µρ − Q b ρµν . bµ νρ := 1 Q L 2 (2.20) (2.14) ◦ In this regime, the Christoffel symbols (Γµ νρ ), the contorb µ νρ ), and the disformation tensor (L bµ νρ ) respecttion tensor (K ively quantify curvature, torsion, and non-metricity. Note that although the torsion (2.2) is independent of the metric, this does not hold for the contortion (2.19), due to the fact that indices have been raised and lowered with the metric. (2.15) This is the case in particular for the Levi-Civita connection. bµ νρσ = 0, and so the second For a flat connection one has R Bianchi identity becomes trivial, while the first one reduces to b [ν T bµ ρσ] + T bµ ω[ν T b ω ρσ] = 0. ∇ (2.19) as well as the disformation tensor due to the antisymmetry in the indices ρ and ω. This completes the proof of the second Bianchi identity (2.5). There are two notable special cases. For a symmetric conbµ νρ = 0, and so the Bianchi identities reduce to the nection, T well known relations bµ [νρσ] = 0, R (2.18) of the Levi-Civita connection, the contortion tensor b µ τ [ρ ∂ω Γ b τ |ν|σ] = Γ b µ τ [ρ R b µ τ [ρ Γ b τ |ϕ|ω Γ b ϕ |ν|σ] . b τ |ν|ωσ] − 2Γ 2Γ (2.13b) b τ |ϕ|ω Γ b ϕ |ν|σ] = 0, b µ τ [ρ Γ b τ |ϕ|ρ Γ b ϕ |ν|σ] + Γ b µ τ [ω Γ Γ (2.17) By demanding that certain of the tensorial quantities vanish, one may reduce the most general class of metric-affine geometries, which we introduced above and which is characterized by a general connection Γ̂µ νρ next to the metric gµν , and thereby obtain particular subclasses. Various of these restricted geometries are relevant and used to model theories of gravity in the literature. By 2.1.5. Metric-affine subclasses. (2.16) If both torsion and curvature vanish, both Bianchi identities become trivial. 11 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 2. Relation between different metric-affine geometries. Starting from the most general class of metric-affine geometries in the lower, left, front corner, this diagram shows how more specific geometries are obtained by imposing that one or more of the tensorial quantities defined in section 2.1.2 vanish. In the most restricted case, which is Minkowski space in the upper, right, back corner, all three quantities vanish. successive restriction, one finds the following special cases of metric-affine geometries: teleparallelism. We will denote the symmetric teleparallel connection and its associated quantities with a diamond on ♢ top, i.e. Γρ µν . b µνρ ≡ 0: The most well known class of geobµ νρ ≡ 0, Q (f) T metries, which is also most commonly used in the description of gravity theories, including GR and its extensions, is the symmetric and metric-compatible connection. In this case the connection is uniquely determined as the Levi- bµ (a) R νρσ ≡ 0: In the case of vanishing curvature, the connection is said to be flat. This case is also known as (general) teleparallel geometry, if both torsion and non-metricity are present. bµ νρ ≡ 0: If the torsion vanishes, the connection is (b) T called symmetric, since the connection coefficients satisfy b ρ [µν] ≡ 0. This case is considered, e.g. in Weyl gravity. Γ b µνρ ≡ 0: For vanishing non-metricity, the connection is (c) Q called metric-compatible, and the corresponding geometry is known as Riemann–Cartan geometry. This choice of the connection is used in Poincaré gauge theory. b µνρ ≡ 0: If only torsion is nonvanishing, bµ νρσ ≡ 0, Q (d) R one obtains the case of metric teleparallel geometry (also called torsional geometry), which is the main topic of this Review. For the metric teleparallel connection and its related quantities, we will omit the hat, and simply write Γρ µν . Moreover, from section 3 we will just just the word teleparallel to refer to this geometry. bµ νρσ ≡ 0, T bµ νρ ≡ 0: An alternative class of geometries (e) R is obtained by demanding that only non-metricity is nonvanishing. This class of geometries is known as symmetric ◦ Civita connection, which we denote by a circle, i.e. Γρ µν . b µνρ ≡ 0: Finally, if all tensorial bµ νρσ ≡ 0, T bµ νρ ≡ 0, Q (g) R quantities vanish, also the metric is constrained, and one finds the case of Minkowski space. In this case the metricaffine geometry is fixed up to diffeomorphisms and does not carry any gravitational DoF. The relations between the most general metric-affine geometry, in which none of the tensorial quantities vanish, and the special cases listed above, may be visualized in different ways. Figure 2 shows how one may obtain the aforementioned special cases by imposing restrictions on the most general metricaffine geometry. Another visualization is given by the Venn diagram in figure 3. Each of the three colored circles represents one of the tensorial quantities, which is nonvanishing for 12 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 3. Classification of metric-affine geometries. Every circle Figure 4. Graphical representation of the tetrad, reduced to a encloses those geometries in which one of the three tensorial quantities defined in section 2.1.2 is nonvanishing. On the intersections of the circles, more that one tensorial quantity is nonvanishing. Not shown is Minkowski space, where all three quantities vanish, and which is represented by the white area outside all three circles. two-dimensional model manifold. A coordinate basis (∂x , ∂y ) of the tangent space is, by definition, tangent to the coordinate lines at a given base point, but not necessary orthogonal or normalized with a given metric. In contrast, the tetrad corresponds to an orthonormal basis (E1 , E2 ) = (E1 µ ∂µ , E2 µ ∂µ ). mathematical description of the tetrad and spin connection in terms of differential forms. all geometries inside the corresponding circle. The white outer region represents Minkowski space, where curvature, torsion and non-metricity vanish. In the remainder of this Review we will mostly be working with a metric teleparallel connection Γρ µν , and omit the accent on all quantities such as the torsion tensor Tµ νρ which are derived from this connection. Any other connection, such b µ νρ , the Levi-Civita as the general metric-affine connection Γ An alternative description of the metric-affine geometry outlined in the preceding section is to use a tetrad e A µ and a spin connection ω b A Bµ , where capital Latin letters A, B = 0, . . . , 3 denote Lorentz indices, while small Greek letters µ, ν = 0, . . ., 3 denote spacetime indices (where numbers refer to spacetime coordinates on the respective spaces). From the tetrad one constructs the Lorentzian metric via the relation 2.2.1. Relation to metric and connection. ◦ connection Γµ νρ and the symmetric teleparallel connection ♢ Γµ νρ , and the quantities derived from these connections, will be distinguished by denoting them with the corresponding accent. gµν = ηAB e A µ e B ν , (2.21) where ηAB = diag(1, −1, −1, −1) denotes the Minkowski metric. For the metric gµν to be non-degenerate, one requires that the tetrad possesses an inverse EA µ , which satisfies EA µ e A ν = δνµ and EA µ e B µ = δAB , and defines the inverse metric as 2.2. Tetrads and the spin connection In the field of TG, which is the primary topic of this Review, the underlying geometry of spacetime is conventionally described in the tetrad-spin connection formulation, which we summarize here. In section 2.2.1, we display how these quantities define the metric and the affine connection, which we had introduced as independent quantities in the previous section. The characteristic tensors describing the metric-affine geometry are then expressed in terms of the tetrad and spin connection in section 2.2.2. An important property of this formulation is the local Lorentz invariance of the geometry, which we discuss in section 2.2.3, and which allows to choose a particular gauge for the spin connection in the teleparallel case. Finally, in section 2.2.4, we provide an equivalent gµν = η AB EA µ EB ν . (2.22) Note that as a consequence, the vectors (EA ) form an orthonormal basis of the tangent space, i.e. g(EA , EB ) = gµν EA µ EB ν = ηAB . (2.23) This is visualized in figure 4. While the coordinate basis (∂µ ) is defined by the coordinate lines, and is in general neither orthogonal nor normalized, the opposite is true for the tetrad basis (EA ). 13 Rep. Prog. Phys. 86 (2023) 026901 Review The frame coefficients EA µ are also required in order to calb µ νρ of the affine connection from the culate the coefficients Γ A spin connection ω b Bµ via in analogy to the decomposition (2.17) of the affine connection displayed in section 2.1.4. Here the Levi-Civita spin connec◦ tion ω A Bµ is obtained from its affine connection counterpart  b ρ µν = EA ρ ∂ν e A µ + ω Γ b A Bν e B µ , Γµ νρ from definition (2.18) by making use of the tetrad postulate (2.25), with the general connection replaced by the Levib A Bµ and Civita connection. It then follows that the contortion K b A Bµ are again related to their counterparts with disformation L spacetime indices by the tetrad and its inverse using the transformation rule (2.30) for tensor indices. ◦ (2.24) which is the unique affine connection satisfying the so-called ‘tetrad postulate’ b ρ νµ e A ρ = 0. ∂µ e A ν + ω b A Bµ e B ν − Γ (2.25) This condition necessitates that vectors remain parallel even at a distance, which is the source of the term TG. It is this condition that forces the distinction between curvature and teleparallel based connections. Finally, note that for a given metricaffine geometry, a tetrad, in general, exists only locally, and that global tetrads exist only if the spacetime manifold M is parallelizable; this fact can be understood by applying the theory of fiber bundles [50]. 2.2.3. Local Lorentz transformations and Weitzenböck gauge. It is important to note that the tetrad e A µ and spin connection ω b A Bµ are not uniquely determined by the metric gµν and affine b ρ µν . From the definition in equation (2.21), one connection Γ ′ can see that the metric is invariant, gµν = gµν , if one replaces the tetrad by e A µ 7→ e ′A µ = Λ A B e B µ , where Λ A B is a local (spacetime point dependent) Lorentz transformation, i.e. it must satisfy One advantage of the formulation in terms of a tetrad and spin connection, which makes it particularly useful for the description of teleparallel geometries, is the fact that the curvature and non-metricity defined in section 2.1.2, and hence also the answer to the question whether they vanish or not, become properties of the spin connection only, and are independent of the choice of the tetrad. In this case, one defines the curvature tensor 2.2.2. Characteristic tensors. ηAB Λ A C Λ B D = ηCD . ω b A Bµ 7→ ω b ′A Bµ = Λ A C (Λ−1 )D B ω b C Dµ + Λ A C ∂µ (Λ−1 )C B . (2.34) the torsion tensor (2.27) It thus follows that the tetrad and spin connection which model a given metric-affine geometry are uniquely determined only up to a local Lorentz transformation, and that any tetrad and spin connection related by the transformations (2.32) and (2.34) define the same metric-affine geometry. Hence, the tetrad and spin connection variables come together with a Lorentz gauge symmetry. The Lorentz gauge freedom is of particular relevance for metric teleparallelism, with a flat, metric-compatible spin connection. In this case it is always possible to choose a gauge (at least locally, on a simply connected domain) such that the spin connection vanishes identically, ω A Bµ ≡ 0. This gauge is known as the Weitzenböck gauge. It further follows that in any other gauge the spin connection takes the simple form as well as the non-metricity tensor b µAB := −ηAC ω Q b C Bµ − ηCB ω b C Aµ . (2.28) These are related to the previously defined quantities by b A Bµν = e A ρ EB σ R bρ σµν , R bµ νρσ = EA µ e B ν R b A Bρσ , (2.29a) R b A µν = e A ρ T bρ µν , T bµ νρ = EA µ T b A νρ , T (2.29b) b µAB = EA ν EB ρ Q b µνρ , Q b µνρ = e A ν e B ρ Q b µAB , Q (2.29c) so that spacetime and Lorentz indices are simply exchanged with the help of the tetrad and its inverse. The same convention may also be applied to any other tensor, by defining ω ′A Bµ = Λ A C ∂µ (Λ−1 )C B , ZA1 ···An B1 ···Bm := eA1 α1 · · · eAn αn EB1 β1 · · · EBm βm Z α1 ···αn β1 ···βm . (2.30) (2.35) and is thus a pure gauge DoF. It is also helpful to note how the tetrad and spin connection transform under infinitesimal Lorentz transformations, which are of the form This allows to decompose the spin connection in the form ◦ b A Bµ + L b A Bµ , ω b A Bµ = ω A Bµ + K (2.33) Note, however, that the connection coefficients (2.24) are not invariant under this transformation of the tetrad alone, unless the Lorentz transformation is global, ∂µ Λ A B ≡ 0. To obtain invariance of the connection also under local Lorentz transformations, one must additionally transform the spin connection by the rule b A Bµν := ∂µ ω R b A B ν − ∂ν ω b A Bµ + ω b A Cµ ω b C Bν − ω b A Cν ω b C Bµ , (2.26) b A µν := ∂µ e A ν − ∂ν e A µ + ω T b A Bµ e B ν − ω b A Bν e B µ , (2.32) Λ A B = δBA + λBA , (2.31) 14 (2.36) Rep. Prog. Phys. 86 (2023) 026901 Review where the condition (2.33) translates to   ηCD = ηAB δCA + λ A C δDB + λ B D = ηCD + 2λ(CD) + O(λ 2 ), (2.37) However, in this Review we will mostly use the tensor component notation introduced earlier, which is more common in physical applications, and resort to differential forms only when referring to relevant literature. and so λ(AB) = 0. In this case we find the transformation δλ e A µ = e ′A µ − e A µ = λBA e B µ , 2.3. Geometric trinity of gravity (2.38) The existence of a unique decomposition (2.17) of the connecb µ νρ for every metric-affine geometry leads tion coefficients Γ to an interesting consequence. It follows that the curvature bµ νρσ allows for a similar decomposition tensor R for the tetrad, and δλ ω b A Bµ = ω b ′A Bµ − ω b A Bµ = −∂µ λBA − ω A Cµ λC B + ω C Bµ λ A C = −Dµ λBA , (2.39) for the spin connection. We will make use of these relations later, when we consider the variation of the teleparallel spin connection and the notion of spacetime symmetries. Dµ is called the Fock–Ivanenko derivative and is formally defined in section 3.3. ◦ where we used the shorthand ◦ b µ νρ − Γµ νρ = K b µ νρ = Γ b µ νρ + L bµ νρ , D ◦ ◦ ◦ Rµ νρσ = Kµ τ σ K τ νρ − Kµ τ ρ K τ νσ + ∇σ Kµ νρ − ∇ρ Kµ νσ , (2.47) in terms of the contortion tensor. The second special case is ♢ b µ νρ ≡ Γµ νρ . In that of a symmetric teleparallel connection Γ this case the Riemann tensor is expressed through the disformation tensor as ♢ ◦ ♢ ♢ ◦ ♢ ◦ ♢ ♢ Rµ νρσ = Lµ τ σ L τ νρ − Lµ τ ρ L τ νσ + ∇σ Lµ νρ − ∇ρ Lµ νσ . (2.48) From these relations one further finds that the Ricci scalar can be expressed as (2.40) the torsion ◦ ◦ ◦ R = Kµ ρν Kρν µ − Kµ ρµ Kρν ν − 2∇µ Kµν ν = −T + 2∇µ Tν νµ , (2.49) 1 bA µ ν A A b A B ∧ e B , (2.41) T := T µν dx ∧ dx = De = de + ω 2 bA in the case of metric teleparallelism, and as and finally the non-metricity b AB := Q b µAB dxµ = DηAB = dηAB − ω b C A ∧ ηCB − ω b C B ∧ ηAC , Q (2.42) ◦ ◦A b AB + L b AB, b A B := ω B + K ω ♢ ♢ ◦ ♢ ♢ ◦ ♢ ◦ ♢ = −Q + ∇ν Qµ µν − ∇µ Qµν ν , ♢ ◦ ♢ (2.50) in the symmetric teleparallel case. Recall that the Ricci tensor is the central element of the gravitational part of the Einstein–Hilbert action ˆ √ ◦ 1 SEH = 2 d4 x −gR, (2.51) 2κ (2.43) b A B and disformation L b A B . Finally, the using the contortion K Bianchi identities are more commonly encountered in the form b A B = 0. DR ♢ R = Lµνρ Lρµν − Lµ µρ Lρν ν + ∇ν Lµ µν − ∇µ Lµν ν where we defined the covariant exterior derivative D. Similarly, the spin connection decomposes in the form bA = R b AB ∧ e B, DT (2.46) called the distortion tensor. Two special cases of this relation deserve particular attention, which have in common that bµ νρσ ≡ 0. The first they lead to a vanishing curvature tensor R b µ νρ ≡ Γµ νρ , case is given by a metric teleparallel connection Γ which allows to express the Riemann tensor of the Levi-Civita connection as A more compact description of the relevant objects in the tetrad and spin connection formulation, which can be found mostly in the more mathematically oriented literature on metric-affine and TG, makes use of the language of differential forms [49]. We give a brief guide to the reader on how to translate the most important quantities and relations from the differential form formulation to the component notation we use in this review. The starting point is the fact that the tetrad and the spin connection constitute the components of one-forms e A = e A µ dxµ b AB = ω and ω b A Bµ dxµ , where the presence of Lorentz indices indicates that the former takes values in the Minkowski space, while the latter takes values in the Lie algebra gl(4) of the general linear group. In this language, also the tensorial quantities we introduced before are expressed as differential forms. In particular, we have the curvature 2.2.4. Differential form formulation. b A B := 1 R b A Bµν dxµ ∧ dxν = db b AC ∧ ω b CB, R ω AB + ω 2 ◦ ◦ bµ νρσ = Rµ νρσ + ∇ρ D b µ νσ − ∇σ D b µ νρ + D b µτ ρD b τ νσ R b µτ σ D b τ νρ , −D (2.45) of GR. Together with the previous findings, one thus sees that by assuming a metric or symmetric teleparallel geometry, one (2.44) 15 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 5. Trinity of gravity. Arrows indicate the relations (2.49) and (2.50) relating the Ricci scalar of the Levi-Civita connection to the torsion and non-metricity scalars. may transform the Einstein–Hilbert action into alternative formulations. For this purpose, note that the expressions (2.49) and (2.50) for the Ricci scalar both involve a total divergence part, which turns into a boundary term when these relations are used in the Einstein–Hilbert action (2.51). Omitting these boundary terms as well as the matter contribution, one arrives at the actions ˆ √ 1 STEGR = − 2 d4 x −gT, (2.52) 2κ weak equivalence principle. This means that all matter fields retain their universal coupling to the metric and possibly its Levi-Civita connection (in the case of spinor fields), with the only possible substitution which arises from considering the metric as a derived quantity obtained from the tetrad instead of a fundamental field. It then follows that the dynamics of the matter fields in TEGR and STEGR is unaltered compared to GR, and also the energy-momentum tensor as the source of gravity is retained, which establishes the equivalence between these theories also at the matter side of the gravitational field equations. Other couplings between matter and the symmetric teleparallel connection may be conceived [53]; however, these would exceed the scope of this Review. The coupling between matter and the metric teleparallel framework will be discussed further in section 3.2. Another remark must be made regarding the omission of the boundary terms in the TEGR and STEGR actions. While this does not influence the field equations, which are the Euler– Lagrange equations derived from the action, it does have an influence when spacetime boundaries are considered. This is the case, for example, for the Casimir effect, black hole entropy or Hamiltonian formulation. Hence, attention must be paid in these cases, and possible non-equivalence may arise. Here we will not further discuss the STEGR action; TEGR, however, will be discussed in detail in section 4.3. Further, as discussed in section 2.2, in the metric teleparallel case, in which only torsion is nonvanishing and which encompasses the TEGR action (2.52), it is more common to formulate the geometry in terms of a tetrad and spin connection, so that the metric determinant is replaced by the tetrad determinant through the relation of the TEGR, and SSTEGR = − 1 2κ2 ˆ √ ♢ d4 x −gQ, (2.53) of the symmetric teleparallel equivalent of general relativity (STEGR) [51], respectively. As the nomenclature suggests, these theories are equivalent to GR in the sense that they lead to equivalent dynamics for the metric, which is the only fundamental field variable in the latter theory. This mutual equivalence, which is summarized in figure 5, has been coined the geometric trinity of gravity [52]. A few comments regarding the mentioned notion of equivalence are nevertheless necessary. First, it must be emphasized that in the transition from the Einstein–Hilbert action to the TEGR or STEGR actions the flat, teleparallel connection has been employed only in the gravitational part of the action, where it is used to substitute the Levi-Civita curvature scalar by the torsion or non-metricity scalars. The matter action, however, remains unchanged, and no direct matter coupling to the teleparallel connections is introduced, in order to preserve the 16 Rep. Prog. Phys. 86 (2023) 026901 e= Review √ −g (2.54) this section. TG introduces the tetrad fields in a much more natural way and can be interpreted as a native gauge theory of translations [33, 36]. Through Noether’s theorem [59] this results in a covariantly conserved energy-momentum tensor, which would not necessarily be the case for a gauge theory of rotation. Through their existence, tetrads connect a general manifold to its local Minkowski space coordinates where a gauge translation transformation will appear as in the action functional of TG theories. Finally, we remark that also a more general teleparallel, i.e. flat connection may be employed in the relation in ◦ equation (2.45), in order to express the Riemann tensor Rµ νρσ b µ νρ , leading to the notion in terms of the distortion tensor D of the general teleparallel equivalent of general relativity (GTEGR) [54]. In this case, imposing the flatness condition bµ νρσ ≡ 0 implies that the Ricci scalar can be written as R ◦ ◦ x̃ A → x A + ϵ A , ◦ b µρν D b ρνµ − D b µ ρµ D b ρν ν + ∇ν D b µν µ − ∇µ D b µν ν R=D ◦ b + ∇µ (2T b ν νµ − Q b µν ν ), bν νµ + Q = −G (3.1) in which ϵ A = ϵ A (xµ ) characterizes the infinitesimal local transformation. The differential operators PA = ∂A = ∂/∂xA will then be the generators of these infinitesimal transformations, which straightforwardly satisfy the commutation relations. In this way, the infinitesimal transformation can be written as (2.55) where b µνρ T b = 1T bµνρ T bµνρ + 1 T bµνρ T bµρν − T b αµα T bβµ + Q bρµν G β 4 2 b µαα T b ααµ T b µνρ Q b µνρ bβ + Q bβµ + 1 Q −Q µβ β 4 1b bµ β + 1 Q b β µ. (2.56) b νµρ − 1 Q b µαα Q b αQ − Q µνρ Q β β 2 4 2 µα δ ϵ x A = ϵ B ∂B x A . (3.2) For a generic source field Ψ(x̃ A (xµ )), infinitesimal transformation on the local Minkowski space will be δϵ Ψ(x̃ A (xµ )) = ϵA ∂A Ψ(x̃ B (xµ )), Omitting the boundary term, as for the TEGR and STEGR actions, as well as the matter contribution, one thus arrives at the GTEGR action ˆ √ 1 b SGTEGR := − 2 d4 x −gG. (2.57) 2κ (3.3) where the Minkowski point xB will appear for every Minkowski space at point xµ . However, the derivative of this field will transform as  δϵ (∂µ Ψ) = ϵ A ∂A (∂µ Ψ) + ∂µ ϵ A ∂A Ψ, (3.4) Comparing with figure 5, this theory would be joining the lower two corners of the triangle. We end this section with a comment that also more general theories may be conceived, in which other tensorial quantities mediate the gravitational interaction, and which still reproduce the Einstein equations as the metric field equations obtained from the Einstein–Hilbert action in GR, and that also for theories beyond GR various equivalent and related teleparallel theories in their different flavors can be constructed [55]. which is only covariant when the parameter ϵ A = ϵ A (xµ ) is a constant. The second term breaks the gauge invariance and is removed by the introduction of a gauge potential Bµ = BA µ ∂A , (3.5) which is a one-form taking values from the Lie algebra of the translation group. To remedy the gauge invariance violating term in equation (3.4), the gauge potential can be used to define the gauge derivative 3. Torsional teleparallel geometries eµ Ψ := ∂µ Ψ + BAµ ∂A Ψ, In this section, we discuss the foundations of TG and how it connects with a fundamental theory of gauge translations. We also describe how TG has been used in other branches of physics and its impact on the formation of gravitational scalar invariants. (3.6) where inertial effects are initially ignored for clarity, they are incorporated into this description in the explanation that follows. In this description inertial is referring to Lorentz frames and particularly to the invariance of the theory under the action of boosts and rotations in a Lorentz framework setting which is a necessity for any eventual physical theory of gravity. This derivative directly leads to the correct gauge invariance relation under an infinitesimal gauge translation 3.1. TG as a gauge theory of translations Gauge theory is a powerful framework in which to express physical theories as is well known. In GR, the first attempt in [56] which led to the forced introduction of tetrads and the Lorentz group being linked to an antisymmetric gauge current. Despite the coupling of gravity to the symmetric energymomentum tensor in GR, it is also possible to formulate GR as a gauge theory of translations [57, 58]. However, TG seems to find more benefits for this treatment as will be explored in δϵ (eµ Ψ) = ϵ A ∂A (∂µ Ψ) , (3.7) where the transformation appears as δϵ BAν = −∂ν ϵ A (xµ ). Thus, the translational coupling prescription replaces the partial derivative with eµ through ∂µ Ψ → eµ Ψ, 17 (3.8) Rep. Prog. Phys. 86 (2023) 026901 Review which observes the translational gauge invariance. For an arbitrary point xµ on the manifold, the covariant derivative eµ can be written in terms of the local Minkowski space as  eAµ ∂A Ψ = eµ Ψ = (∂A Ψ) ∂µ x A + BAµ ∂A Ψ, (3.9) which raises the partial derivative to its local Lorentz invariant, while the non-trivial tetrad also includes the gauge potential eAµ = hAµ + BAµ , where Dµ is the inertial covariant derivative, i.e. the covariant derivative wrt the Lorentz group which is associated with the spin connection. Finally, to retain tetrad invariance under  infinitesimal gauge translations (δϵ eAµ = 0) implies that where eAµ = ∂µ x A + BAµ is a non-trivial tetrad field. The close relationship between the manifold and Minkowski space indices through the tetrad field eAµ is called soldering. A trivial tetrad would mean Hµ = hAµ ∂A (BAµ = ∂µ ϵ A ), and so the translational coupling prescription becomes hAµ ∂A Ψ → eAµ ∂A Ψ, δϵ BAµ = −Dµ ϵ A , 3.2. Gravitational coupling prescription (3.11) In GR, through Einstein’s choice of the Levi-Civita connection as the expression of geometric deformation (through Riemann geometry), the coupling prescription associated with gravity becomes which is reminiscent of the Minkowski metric being raised to an arbitrary metric tensor for the general manifold ηµν → gµν . ◦ (3.12) ∂µ → ∇ µ , This is now a consequence of the prescription in equation (3.11) due to the definitions ηµν = ηAB hAµ hBν and gµν = ηAB eAµ eBν . Thus far, inertial effects have been ignored to make the argument clearer for the interpretation of TG as a gauge theory of translations. We can now include inertial effects by considering arbitrary Lorentz transformations ΛAB from the Lorentz group O(1, 3). These transformations only apply to the local Minkowski space and their index representations, which means that local coordinates will transform as x A → ΛAB (x)x B = x ′A , ◦ uµ ∇ µ uν = (3.13) (3.14) (3.15) where ω ABµ is the teleparallel spin connection, which is a purely inertial Lorentz connection. By observing the local Lorentz invariance (such as through the transformation eAµ = ΛAB e ′Bµ ), it follows that the spin connection takes on components as we discussed in section 2.2.3 ω ACµ = ΛAB (x)∂µ ΛCB (x), hA µ → eA µ . (3.21) (3.22) The relationship with the inertial and general manifold metrics now appear as results of this prescription where (3.16) ηAB hAµ hBν → ηAB eAµ eBν , wrt the Lorentz transformations, which indeed is totally inertial. The fuller tetrad can be reinterpreted as a local Lorentz invariant trivial and spin connection pair tetrad hAµ = ∂µ x A + ω ABµ x B = Dµ x A , d 2 xν ◦ ν dxα dxβ + Γ αβ = 0, dλ 2 dλ dλ from which it can be said that in GR gravitation is not expressed in a force-like manner, i.e. its acceleration ◦ vanishes (aν = 0). The coupling prescription described in equation (3.20) is called the minimal coupling prescription (namely, there the covariant derivative is coupled with the Levi-Civita connection) since it lays out a scenario where the spin connection of GR is used, in all other cases such as that in teleparallel geometries, the coupling prescription is not necessarily minimal [36]. This point is important since the coupling prescription of teleparallel geometries can make use of the Levi-Civita connection components, while its spin connection will always be different to that of GR. The situation in TG is entirely different in that gravitation emerges through a gauge theory of translations. In this setting, the inertial covariant derivative hµ is raised to eµ in the presence of gravity, leading directly to equation (3.10) where the inertial covariant derivative is raised to its non-trivial tetrad field form in the presence of gravitation. The end result is the translational gauge coupling prescription will directly lead to the generalization of equation (3.9) to eµ Ψ = eAµ ∂A Ψ = ∂µ Ψ + ω ABµ x B ∂A Ψ + BAµ ∂A Ψ, (3.20) by choice rather than a result of a gauge theory. This then results in the well known geodesic equation so that scalar and tensor fields will similarly transform appropriately. Analogously, the gauge potential will also undergo local Lorentz transformations on its inertial index, meaning that BAµ → ΛAB (x)BBµ , (3.19) which can be viewed as a generalization of the partial derivative in equation (3.7). (3.10) where BAµ 6= ∂µ ϵ A for the non-trivial tetrad field, and from which it follows that hAµ → eAµ , (3.18) (3.23) which is equivalent to saying that ηµν is raised to the general manifold gµν . In this way, we have a direct link between the metric tensor and the tetrad fields that can be used to relate the (3.17) 18 Rep. Prog. Phys. 86 (2023) 026901 Review components of the two objects. Another important feature of these relations is that the mechanics of metric tensors (such as ansatz choices) can be used to solve these relations for the tetrad field components through equation (3.23). which acts as the field strength in TG. Another important feature of this geometric formulation is that the Riemann measure of curvature vanishes for any frame in this regime, namely RABνµ ≡ ∂ν ω ABµ − ∂µ ω ABν + ω ACν ω CBµ − ω ACµ ω CBν = 0, (3.32) 3.3. The field strength of gravity which can be seen as the substitution of the teleparallel spin connection (3.16) into the metric-affine Riemann tensor in equation (2.26). In the context of the gauge structure of TG, a gauge invariant field strength can be defined as a measure of the amount of geometric deformation which is then expressed as a gravitational or Maxwell-like force in the dynamical equations. Taking the example of classical electromagnetism, the vector potential Aµ induces a gauge field strength ◦ ◦ Fµν := ∇µ Aν − ∇ν Aµ , 3.4. Inertial and gravitational effects The concept of inertia played a crucial role in Einstein’s formulation of GR through the application of Mach’s principle [61, 62]. To this end, Einstein’s attempt was to remove the privileged role that inertial frames of reference play in classical mechanics by removing the dependence of measurements on the choice of coordinates. On the other hand, GR does not fully resolve the foundational issues at the core of the definition of inertial such as the need, at times, to include absolute elements. For instance, consider rotating matter wrt its local inertial system. At its core, this points to a spacetime structure that is not entirely governed by matter fields [63, 64]. The dynamical variable in TG for gravitation is the tetrad field which builds up to the metric tensor. Given that tetrad frames relate the general manifold with their Minkowski spaces, TG offers a different perspective to this issue at the core of GR. Consider again the concept of trivial tetrad (hAµ ) and spin connection pairs which relate inertial frames with their Minkowski spaces. Tetrads can also be represented by their tetrad fields which represent a basis of vectors for the tangent space of an arbitrary point xµ on a manifold, i.e. (3.24) on the general manifold, which also obeys field conservation laws [60]. Now, taking the translational covariant derivative in equation (3.15), the gauge field strength can be similarly defined from the gauge derivatives eµ as [eµ , eν ] = TAµν ∂A , (3.25) where [eµ , eν ] := 21 (eµ eν − eν eµ ), and the translational field strength is defined as TAµν := ∂µ BAν − ∂ν BAµ + ω ABµ BBν − ω ABν BBµ , (3.26) and ∂A represents the generators of infinitesimal translations. While direct, equation (3.26) can also be recast in terms of the tetrad field eAµ by identifying the vanishing antisymmetric identity   [Dµ , Dν ] xA ≡ Dµ Dν xA − Dν Dµ xA = 0, (3.27) HA = HAµ ∂µ so that the translational field strength turns out to be expressed as   TAµν = 2 ∂[µ eAν] + ω AB[µ eBν] = ∂µ eAν − ∂ν eAµ + ω ABµ eBν − ω ABν eBµ , ∂µ = hA µ H A (3.28) and dxµ = HAµ hA , (3.34) h A (hB ) = δBA . (3.35) This orthogonality condition directly leads to tetrad frame conditions hAµ HAν = δµν (3.29) and hAµ HBµ = δBA . (3.36) These frames and their associated conditions hold for any differentiable manifold. Another way of characterizing the trivial tetrad and spin connection pairs is through the so-called structure coefficients which are defined through the commutation relation [35, 65] (3.30) In the class of frames in which the spin connection components vanish, the connection reduces to Γρνµ := EAρ ∂µ eAν which will later be introduced as the Weitzenböck gauge connection, so that the torsion tensor becomes TAµν = ∂µ eAν − ∂ν eAµ , (3.33) when requiring that where Γρµν is the teleparallel connection defined as  Γρνµ := EAρ ∂µ eAν + ω ABµ eBν ≡ EAρ Dµ eAν . h A = hAµ dxµ , which conversely result in which will also inherit the property of invariance under gauge transformations T ′Aµν = TAµν , where square brackets are used to denote the antisymmetric operator. Thus, the teleparallel field strength turns out to be exhibited as an antisymmetric expression which can be written in terms of only general manifold indices as Tρµν = EAρ TAµν = Γρνµ − Γρµν = 2Γρ[νµ] , and [HA , HB ] = f CAB hC , (3.37) which can also be interpreted as the coefficients of the anholonomy of the tetrad fields. The dual expression, where (3.31) 19 Rep. Prog. Phys. 86 (2023) 026901 Review the inverse tetrads are used, is found using the Cartan structure equation which has already been used in the general geometry framework around equation (2.35). We can recognize this as the inertial covariant derivative from equation (3.17) which lets us write hAµ ≡ Dx A . In this way, by initially considering an inertial frame, in which ω ′ABµ = 0, Lorentz transformations can be performed in a local (position-dependent) way through ΛAB which produces nonvanishing spin connection components. For each class of frames, a global (position-independent) Lorentz transformation (ΛAB = constant) can relate the frames for every global point. Noticing the property between Lorentz transformed trivial frames  1 1 dhC = − f CAB hA ∧ hB = ∂µ h Cν − ∂ν h Cµ dxµ ∧ dxν , 2 2 (3.38) where the exterior (wedge) product is used. The structure coefficients can be related to the tetrad frames straightforwardly through f CAB ≡ [HA , HB ] hC (3.39a)  = HAµ HBν ∂ν h Cµ − ∂µ h Cν , (3.39b) ΛAB = hAµ H ′ Bµ , where the last equations is a result of the orthogonality conditions equation (3.36). A special class of frames hA′ exist in which f ′CAB = 0, which is derived through the transformation in equation (3.43). The coefficients of the anholonomy can then be written as Lorentz matrices by appropriately transforming some of the tetrad frames in equation (3.39b) to give (3.40) which directly sets dhA′ = 0 meaning that hA′ can be expressed in closed differential form, and that for some x ′A , h ′A = dx ′A locally. In this case, the basis h ′A turns out to be integrable, or holonomic. However, hA refers to holonomic frames meaning that all such frames observe this vanishing property. On the other hand, non-holonomic frames do not generally observe this vanishing condition, and indeed f CAB 6= 0 in such cases. In special relativity, the inertial effects within a frame can be represented by a so-called spin connection which is a Lorentz connection coming from the Lorentz group of transformations. As is well known, these effects do not occur in every inertial frame such as in Minkowski frames. Spin connection components can be generically induced by first considering a general coordinate system xµ with the holonomic form h ′Aµ = ∂µ x ′A , f CAB = ω CBA − ω CAB , A ′B x = Λ B (x)x , TAνµ ≡ ∂ν hAµ − ∂µ hAν + ω ACν h Cµ − ω ACµ h Cν = 0, (3.48) which one would expect for such frames. This condition guarantees that gravity vanishes, however, frames do exist that exhibit nonvanishing torsion tensor components but which do not feature non-inertial frames. Considering again the spin connection, one can show that it features an antisymmetry in its first two indices (3.41) ωABµ = ηAD ω DBµ = ηAD ΛDC ∂µ ΛBC = −ηBE ω EAµ = −ωBAµ , (3.49) which leads to a number of properties. By taking three different combinations of the vanishing torsion tensor together with the antisymmetry property for the spin connection, it follows that (3.42) so that the frame gets transformed to hAµ = ΛAB (x)h ′Bµ . 0 = fBCA + fACB − f CBA = ωBA C − ωB CA + ωAB C  − ωA CB − ω CAB − ω CBA , (3.50) (3.43) By writing this transformation in terms of the transformed xA , the tetrad frame can be written in terms of the Lorentz matrices as  hAµ = ΛAB (x)h ′Bµ = ΛAB (x)∂µ ΛCB (x)xC = ∂µ x A + ω ACµ xC , (3.44) which directly gives the inverse relation to equation (3.47) as ω ABC =  1 A f + f A − f A BC . 2 B C C B (3.51) Through the spin connection, TG observes local Lorentz invariance in all relations which is a fundamental symmetry of Nature. Local Lorentz invariance should not produce dynamics in a gravitational theory, but in TG it does produce a second coupling prescription which will be related by the strong equivalence principle. This emerges as a correction term to the where the inverse transformation of equation (3.43) and identity ΛAB ΛCB = δCA was used, and where the spin connection is explicitly defined as ω ACµ := ΛAB (x)∂µ ΛCB (x), (3.47) where the general manifold index was raised with the inverse tetrad though ω CBA = hAµ ω CBµ . For inertial frames hAµ which are related by the Lorentz connection, i.e. where x ′A = x ′A (xµ ) is position dependent. Taking an arbitrary Lorentz transformation ΛAB of vectors in this basis results in A (3.46) (3.45) 20 Rep. Prog. Phys. 86 (2023) 026901 Review derivative terms for inertial frames which are raised to Lorentz covariant derivatives. To see this transition, consider a vector field ϕ ′C . Using the tetrad field where the coefficients of the anholonomy are now raised to  f CAB = EAµ EBν ∂ν e Cµ − ∂µ e Cν , (3.59) h ′ A = δAµ ∂µ , where we recall the definition in equation (3.39b). A natural consequence of the coupling prescription is that the torsion no longer vanishes organically in equation (3.48) meaning that the relation between the structure coefficients of the anholonomy and spin connection components in equation (3.47) now takes the form  f CAB = EAµ EBν T Cνµ − ω CDν eDµ + ω CDµ eDν , (3.60) (3.52) which acts on the vector field as h ′ A ϕ ′ = δAµ ∂µ ϕ ′ . C C (3.53) Transforming to an unprimed frame through ϕC = ΛDC (x)ϕ ′ gives  C h ′ A ϕ ′ = δAµ ∂µ ΛDC ϕD (3.54a) D  = ΛBA ΛDC ∂B ϕD + ΛBA ∂B ΛDC ϕD or, in a more simplified form f CAB + T CAB = ω CBA − ω CAB , (3.54b)  = ΛBA ΛDC ∂B ϕD + ω DEB ϕE (3.54c)    1 D = Λ B A Λ D C ∂B ϕD + fE B + fBDE − f D EB ϕE . 2 (3.54d) where T CAB is the torsion tensor in inertial indices. This gives a direct relationship between the structure coefficients, the torsion tensor and the spin connection. Similarly, the Riemann tensor in equation (3.32) is found to be RABCD = hCν ∂ν ω ABD − hDµ ∂µ ω ABC + ω AEC ω EBD In terms of the vector representation of the Lorentz generators A (SBC ) D = δDB δCA − ηCD η AB , − ω AED ω EBC − ω ABE f E CD .  1 C f + f C − f CAB = ω CBA + K CBA , 2 B A A B (3.56) 1 = ∂µ Ψ + ω CBµ SCB Ψ 2 1 A C = ∂µ Ψ + h µ ω BA SCB Ψ 2  1 = ∂µ Ψ + hAµ fBCA + fACB − f CBA SCB Ψ, 4 (3.63) where the so-called contortion tensor is defined as where an extra half appears due to the Lorentz generators. For a scalar field Ψ ∂µ Ψ → D µ Ψ (3.62) The same combination of indices as in inverse relation in equation (3.51) for the nonvanishing torsion tensor results in (3.55) the coupling prescription takes the form  1 D ∂B ϕ D + fE B + fBDE − f D EB SEF ϕF , 4 (3.61) K CBA := (3.57a)  1 TBCA + TACB − T CAB , 2 (3.64) which aligns with a TG version of the general contortion tensor in equation (2.19). Thus, the coupling prescription in equation (3.58) can be written fully as (3.57b) ∂µ Ψ → ∂µ Ψ + (3.57c)  1 C ω Bµ + K CBµ SCB Ψ, 2 (3.65) where full gravitational coupling prescription is laid out and where the Lorentz part of this coupling prescription is separated from the gravitational sector. In GR, the spin connection takes the form of the full expression of the coefficients of the anholonomy in equation (3.63) as [35] where SBC now represents the Lorentz generators for the Ψ field. The second term in the last relation is a compensating term that imposes the local Lorentz covariance of the derivative in the new inertial frame. This coupling prescription is hidden in the internal structure of GR and does not play an active role in many of its considerations. Thus in TG, the strong equivalence principle takes on a more active role in the construction of the theory. This concept is encapsulated in the idea behind the general covariance principle where Lorentz covariant formula in special relativity can be raised to its non-inertial form by the coupling prescription. The Lorentz covariant derivative can then be raised to its non-inertial counterpart by the coupling principle in equation (3.22) which gives ◦ ω CAB =  1 C fB A + fACB − f CAB , 2 (3.66) which is the Levi-Civita spin connection, and results in the fundamental identity ◦ ω CAB = ω CAB − K CAB . (3.67) The gravitational coupling prescription in equation (3.65) can equivalently be written as  1 ∂µ Ψ → Dµ Ψ = ∂µ Ψ + eAµ fBCA + fACB − f CBA SCB Ψ, 4 (3.58) 1◦ ∂µ Ψ → ∂µ Ψ + ω CBµ SCB Ψ, 2 21 (3.68) Rep. Prog. Phys. 86 (2023) 026901 Review which is the coupling prescription for GR. Thus, both curvature- and torsion-based approaches to gravity utilize the same coupling prescription, and are thus both consistent with the principles of covariance and strong equivalence. The potential for separating inertial effects is revealed here as a fundamental difference between torsional- and curvaturebased prescriptions for gravity. In the Levi-Civita spin connec◦ tion ω CBµ both gravitational and inertial are combined while in torsional gravity, the inertial effects of the spin connection ω CBµ is interpreted as being separated from the gravitational effects represented by the contortion tensor K CBµ . In fact, for a local frame in which the Levi-Civita spin connection vanishes, the identity implies that the inertial and gravitational effects exactly compensate for each other. The separation of gravitational and inertial effects can immediately be seen by a free particle in Minkowski spacetime in which du A dxµ ∂u A 0= = = uµ ∂µ u A , dτ dτ ∂xµ in this framework, as is laid out in this work. However, the coupling to matter must remain consistent across these possible choices when generalizing the geometric underpinnings of gravitation. As pointed out in [52] there are two important subtleties to bear in mind when considering the matter sector in a gravitational field, namely • Ambiguities in the coupling to matter may arise when changing the geometry; • The treatment of bosonic and fermionic fields may lead to important differences that may lead to inconsistencies. Considering a point particle in GR ˆ S = mc 2 (3.69) (3.70) giving   ◦ uµ ∂µ u A + ω ABµ u B = 0, (3.71) where the particle motion is dictated through the Levi-Civita spin connection rather than a separate force representing gravity. This again implies that the inertial and gravitational effects within GR are compounded together. In TG, the coupling prescription in equation (3.65) gives a different picture where  ηµν → gµν  1 C ω Bµ − K CBµ (SCB )AD uD , 2 and ◦ ∂µ → ∇ µ , (3.75) ◦  uµ ∂µ u A + (3.74) For fixed parametrization [66], this will follow the regular geodesic equation described in equation (3.71), and will depend on the Levi-Civita connection. However, in other geometric theories of gravity this point needs to be revisited as discussed in section 3.2. This is important since observational constraints on the paths of test particles have become extremely precise, making this a crucial test for the formulation of any geometric theory of gravity. The second difference to consider is the distinction between bosonic and fermionic fields since the former couples to the metric and is described by tensor representations while the latter also couples to the connection and is represented by spinors. Through the spinor representation, tetrads together with their respective gravitational spin connections are utilized to describe fermions. In GR, the minimal coupling prescription is presumed where where the four-velocity ua is subject to the Lorentz coupling prescription, and τ is the proper time. In the GR case, the coupling prescription in equation (3.68) which together with the Lorentz generators in equation (3.55) tuns out to be   1◦ uµ ∂µ u A + ω CBµ (SCB )AD uD = 0, 2 dτ. where the Levi-Civita connection, Γσµν , is taken. Moreover, the condition of vanishing torsion and non-metricity forces the m matter Lagrangian to take the regular form, Θµν = √−−2g δδgLµν (also given in equation (1.10)). If the Lagrangian had some (3.72) simplifies to ◦  uµ ∂µ u A + ω ABµ u B = KADµ uµ uD , dependence on the connection, Lm = Lm (gµν , Γσµν , ϕ), would lead to a nonvanishing hypermomentum that would necessarily have to be equal to zero once the conditions of vanishing torsion and non-metricity are imposed [52]. In TG, the change in connection associated with the covariant derivative and thus mediation of the gravitational field imply a natural choice for the coupling to matter where the covariant derivative assumes the teleparallel connection which is torsionful and satisfies vanishing curvature and zero metricity. However, this is immediately problematic, for instance consider the case of bosons through the simple example of photons. In the absence of gravitation, the field strength is described by (3.73) where the effect of gravitation on the motion of a particle is now totally different in which the gravitational force is acting through the contortion tensor term. The geodesic equations in equations (3.71) and (3.73) are equivalent by the identity in equation (3.67) but represent the fundamental difference between the effects of gravitation in GR and TG. 3.5. Coupling to matter In TG, the curvature associated with the Levi-Civita connection is replaced by torsion associated with the teleparallel connection. The gravitational sector can be readily reformulated Fµν = 2∂[µ Aν] , 22 (3.76) Rep. Prog. Phys. 86 (2023) 026901 Review where Aν is the vector potential associated with the electromagnetic field. If the teleparallel connection were taken for the coupling prescription then this would give Fµν → 2∇[µ Aν] = 2∂[µ Aν] + T σµν Aσ , Considering the simple but revealing case of a linear scalar field where the Lagrangian is given by 3.5.1. Scalar fields. Lϕ := (3.77) ∂µ ∂ µ ϕ + µ 2 ϕ = 0. Lϕ = ηµν → gµν , (3.80b) ◦ ◦ 2 □ϕ + µ ϕ = 0, (3.85) ◦ σ where □ϕ = e−1 ∂σ (e∂ is the Laplace–Beltrami operator  ϕ) √ A [36], and e = det e µ = −g. However, by noticing that √ 1 1√ ∂µ e = ∂µ −g = √ ∂µ g = −gg αβ ∂µ gαβ 2 −g 2  ◦ √ ◦ = −gΓσσµ = eΓσσµ = e Γσµσ − Kσµσ , (3.86) implies that the Laplace–Beltrami operator on the scalar field can be written as ◦ ◦ µ ◦ µ ◦ ∇µ ∇ ϕ = ∇ µ ∂ ϕ = □ϕ = e −1 [e∂µ (∂ µ ϕ) + (∂ µ ϕ) ∂µ e] (3.87a)   = e−1 e∂µ (∂ µ ϕ) + (∂ µ ϕ) e Γσµσ − Kσµσ   = ∂µ + Γσµσ − Kσµσ ∂ µ ϕ. (3.87b) (3.87c) Thus, the teleparallel Klein–Gordon equation (3.85) can be written as  (3.88) ∂µ + Γσµσ − Kσµσ ∂ µ ϕ + µ 2 ϕ = 0, from which it follows scalar fields couple to torsion through the contortion tensor. which is first introduced in equation (3.10). A consequence of this is that the contortion tensor identity presented in equation (3.67) can now be written as Kρ µν = Γρ µν − Γρ µν = (3.84) ◦ which arises so that a well-defined Dirac equation that describes the dynamics of spinor fields on a general spacetime [53]. It is this covariant derivative that is the source of the problem with regard to using the teleparallel connection for fermionic fields. In GR, the spin connection contains terms beyond the Lorentz connection and so provides the coupling required with the Levi-Civita connection. The spin connection of TG is totally inertial and can be set to zero under an appropriate gauge condition which means that the connection will no longer be coupled to the fermionic fields. This last point will mean that the energy-momentum tensor will not always be conserved which will produce an inconsistent theory. One option is to choose the coupling prescription ∂µ → ∂µ for bosons and ∂µ → Dµ , where Dµ is a fermionic covariant derivative that retains the connection coupling such that the energy-momentum tensor for matter is conserved. However, this would be quite arbitrary as a coupling prescription choice. It is for these reasons that the minimal coupling description for gravitation, as described in section 3.2, is identified for all matter coupling in TG [68, 69]. As examined below, it is this choice that most consistently provides a sound avenue for constructing teleparallel theories of gravity [70], i.e. we adopt the coupling prescription (3.80a)  ◦ 1  µν ◦ g ∇ µ ϕ∇ ν ϕ − µ 2 ϕ 2 , 2 where ∇µ ϕ = ∂µ ϕ since ϕ is a scalar. Taking the variation of this Lagrangian wrt the scalar field produces the gravitational Klein–Gordon equivalent (3.79) hAµ → eAµ , (3.83) By adopting the minimal coupling presenting in equation (3.80), the Lagrangian density is promoted to (3.78) where γ A are the Dirac matrices, m is the mass parameter, and ψ is the fermionic field. The fermionic covariant derivative is given by Dµ = ∂µ − ω ABµ [γA , γB ] , (3.82) which is defined on the tangent space, µ = mc/ℏ12 is the mass associated with the scalar field, and the ϕ = ϕ(x) position dependence is suppressed for brevity’s sake. The Lorentzian Klein–Gordon equation is produced by taking a variation wrt ϕ giving [67] in a gravitational field. This would mean that the photon is non-minimally coupled to torsion and does not observe the U(1) gauge symmetry in its standard form which would be extremely problematic. On the other hand, fermions would also pose a serious problem to a matter coupling prescription that takes on the teleparallel connection. Consider the Dirac equation [67] iγ A EAµ Dµ ψ − mcψ = 0,  1 µν η ∂µ ϕ∂ν ϕ − µ 2 ϕ 2 , 2 The half spin massive particles are described by the Dirac field which has an associated Lagrangian 3.5.2. Fermion fields. 1 (Tµ ρ ν + Tν ρ µ − Tρ µν ) , (3.81) 2 where the contortion tensor is recognized as the difference between the teleparallel and Levi-Civita connections respectively. Also, the definition in equation (3.64) was used. 12 We retain SI units for clarity here. 23 Rep. Prog. Phys. 86 (2023) 026901 Lψ := Review  icℏ  A µ ψ̄γ HA ∂µ ψ − HAµ ∂µ ψ̄γ A ψ − mc 2 ψ̄ψ, 2 (3.89) where it can be noted that the Dirac spinors couple to the axial part of the torsion tensor only, and so the other parts of the torsion tensor can vary freely without altering the fermionic sector. This is an important property to identify, it is also important to point out that ultimately, the Dirac equation will have the same solutions since TEGR is dynamically equivalent to GR, and so the differences we recognize will only pertain to the way that the TG contributions reproduce the standard equation. in the absence of gravity (where SI units are allowed so that the appearance of constants in the Dirac equation is clearer to see), and where ψ̄ denotes he conjugate transpose of ψ. This produces the flat Dirac equation iℏγ A HAµ ∂µ ψ − mcψ = 0. (3.90) Now using the spinor field covariant derivative defined in equation (3.79), the Dirac field Lagrangian that is minimally coupled to torsion will be transformed to A large class of boson fields permeate through the standard model of particle physics with the most prominent being the photon under spin 1 representation of the Lorentz group. We consider this example to show the general matter coupling for bosonic fields in teleparallel geometries [53, 71]. Taking the four-potential Aµ , the electromagnetic tensor can be written as [60] 3.5.3. Boson fields.  icℏ  A µ ψ̄γ HA ∂µ ψ − HAµ ∂µ ψ̄γ A ψ − mc 2 ψ̄ψ 2 (3.91a)  icℏ  A µ µ A 2 → ψ̄γ EA Dµ ψ − EA Dµ ψ̄γ ψ − mc ψ̄ψ, (3.91b) 2 Lψ = Fµν = ∂µ Aν − ∂ν Aµ , which conveniently separates the Levi-Civita and purely torsional terms, giving Lψ = in the absence of gravity on the tangent space. By this construction, it immediately follows that the Bianchi equation below is satisfied by this field, i.e. ◦ icℏ  A µ ◦ ψ̄γ EA ∇µ ψ − EAµ ∇µ ψ̄γ A ψ 2  −eaµ̄ gµ̄µ ebλ ecρ̄ gρ̄ρ Kλρµ {γa , [γc , γc ]}ψ − mc 2 ψ̄ψ, (3.92) ∂µ Fνσ + ∂σ Fµν + ∂ν Fσµ = 0, 1 LEM := − Fµν Fµν , 4 ◦ icℏ  A µ ◦ ψ̄γ EA ∇µ ψ − EAµ ∇µ ψ̄γ A ψ 2  −2iϵλρµν Tλρµ ψ̄γ 5 γν ψ − mc 2 ψ̄ψ. (3.99) produces the field equations (3.93) ∂µ Fµν = 0, where ϵABCD is the Levi-Civita symbol, and where the Lagrangian reduces to Lψ = (3.98) while the electromagnetic Lagrangian where {γa , γb } = 2γ(a γb) . Using several identities relating the Dirac matrices, it can be shown that [53] {γa , [γc , γc ]} = 4iϵabcd γd γ 5 , (3.97) (3.100) when varied wrt the four-potential. Together equations (3.98) and (3.99) constitute the source-less Maxwell equations [60]. Within the Lorentz gauge, where ∂µ Aµ = 0, equation (3.99) directly leads to the wave equation (3.94) ∂σ ∂ σ Aµ = 0. (3.101) The teleparallel Dirac Lagrangian can be recast in terms of the irreducible contributions to the torsion tensor. In section 5.2.1, the irreducible nature of the torsion tensor will be further probed. For the purpose of the present discussion, consider the decomposition of the torsion tensor These Lorentz equations can be raised to their gravitational analogues by applying the coupling prescription in equation (3.80) so that  1  2 λ Tλµν = t µν − tλνµ + δµλ vν − δνλ vµ + ϵλµνρ aρ , 3 3 (3.95) Given the way that the electromagnetic tensor is defined, it can be written as  ◦ ∂σ Aµ → ∇σ Aµ = ∂σ Aµ + Γµνσ − Kµνσ Aν . ◦ Lψ = ◦ Fµν = ∇µ Aν − ∇ν Aµ = ∂µ Aν − ∂ν Aµ , where aµ = 61 ϵµνσρ Tνσρ , vµ = T σ σµ and tσµν = 1 1 1 are the 2 (Tσµν + Tµσν ) + 6 (gνσ vµ + gνµ vσ ) − 3 gσµ vν axial, vector and purely tensorial parts of the torsion tensor as discussed in equation (5.6). By considering the symmetries of the torsion tensor, the Dirac Lagrangian can be further reduced to [53] (3.102) (3.103) which retains the form of the electromagnetic tensor as in Minkowski space, and where the second equation is the result of the difference of the terms in the definition being symmetric. Naturally, this means that the same Bianchi identities will continue to be satisfied, namely ∂µ Fνσ + ∂σ Fµν + ∂ν Fσµ = 0. However, the Lagrangian will be altered such that  ◦ icℏ  A µ ◦ ψ̄γ EA ∇µ ψ − EAµ ∇µ ψ̄γ A ψ − 2iψ̄γ 5 γ µ aµ ψ 2 − mc 2 ψ̄ψ, (3.96) 1 LEM = − Fµν Fµν , 4 24 (3.104) Rep. Prog. Phys. 86 (2023) 026901 Review which produce (3.105) An important nuance takes place when considering the paths of particles in gravitational theories based on nonRiemannian geometry. Given a Levi-Civita covariant derivat- when varied wrt the four-potential. This constitutes the second part of Maxwell’s equations in a teleparallel setting. ive, ∇µ , a tangent vector field, ξ σ = dxdτ , is said to be an autoparallel of Dµ provided that it remains parallel to itself along the length of the path [72, 73] ◦ ∇µ F µν = 0, ◦ ◦ In this setting, the Lorentz gauge (∇µ Aµ = 0) can only be applied when the commutator on the four-potential is considered since the covariant derivatives are not commutative, i.e. h◦ ◦ i ◦ ◦ ◦ ◦ ◦ µ µ µ µ µ ∇µ , ∇ν A = ∇µ ∇ν A − ∇ν ∇µ A = Rµν A = −Pµν A , ◦ ξ σ ∇σ ξ µ = 0. where Pµν := Pσµνσ and is defined by the Riemann tensor as calculated using both the teleparallel and Levi-Civita connections 0 ≡ Rσµνρ = Rσµνρ + Pσµνρ , ◦ (3.107) d 2 xµ ◦ µ dx α dxβ + Γ αβ = 0, dτ 2 dτ dτ In this context, the teleparallel Maxwell equation in equation (3.105) reduces to [36] d 2 xµ dx α dxβ dx α dxβ µ µ + Γ = K , αβ αβ dτ 2 dτ dτ dτ dτ ◦ ◦ ◦ (3.113) (3.114) which is the result of the transformation from the curvaturebased Levi-Civita connection to the torsion-based teleparallel connection. Thus, the geodesic equation in GR is now transformed into a Lorentz-like equation where a force term appears on the RHS featuring the contortion tensor which drives test particle paths on the LHS. Importantly, both equations represent the same paths and so feature the same solutions albeit from different perspectives. This equation highlights an important property to the GR geodesic equation in that it recovers a Lorentz force-like form in that instead of the geometry determining the path of test particles, these particles are acted upon by a force-like contribution, in the form of the contortion tension, which acts on the test particles as a force term would. For generalized teleparallel theories, the free fall paths of particles continues to be determined by equation (3.114) where the theory determines the particular geometry of the spacetime. (3.109) where a coupling with the teleparallel connection appears while continuing to satisfy the gauge invariance of the Maxwell equations. Comparing to standard gravity through the equivalency in equation (3.107), the same action is observed for the Levi-Civita connection in µ ν µν ∇µ ∇ A − R Aµ = 0, (3.112) which is the well known equation that describes the paths that test particles take. The teleparallel coupling prescription means that the geodesic equation in equation (3.113) is transformed by equation (3.81) to − Γθµν Kσθρ + Γθµρ Kσθν + Kσθρ Kθµν − Kσθν Kθµρ . (3.108) ◦ ρ results in the same equation when considering either geodesic or autoparallel equations, given by Pσµνρ := ∂ν Kσµρ − ∂ρ Kσµν + Γσθν Kθµρ − Γσθρ Kθµν ◦ ◦ ∇µ ξσ = ∂µ ξσ − Γ µσ ξρ , where the contortion tensor in equation (3.64) leads to a purely teleparallel definition of Pσµνρ through µ ν ν µ ∇µ ∇ A + P µ A = 0, (3.111) Autoparallels generalize the concept of a straight line to a general affine spacetime, while geodesics define the shortest path between two points. However, geodesics require a determination of length such as the metric tensor in Riemannian geometries. In GR, the Levi-Civita covariant derivative (3.106) ◦ σ (3.110) where the coupling also appears only in terms of the LeviCivita connection. More generally, the teleparallel coupling prescription in equation (3.80) has not been widely studied in the context of other bosonic fields. It would be interesting to further investigate how this coupling prescription may lead to further probes of TG. The coupling prescription detailed in equation (3.80) describes how matter couples to the geometry of gravity. In sections 3.5.1–3.5.3, the cases of scalar fields, fermions and bosonic fields are explored respectively through the TG analogues of the Klein–Gordon, Dirac, and Maxwell equations. In these cases, the consistency of the theory is preserved by taking the Levi-Civita covariant derivative as the teleparallel coupling to matter. 3.5.4. Effects on the geodesic equation. 3.5.5. A note on the energy-momentum tensor. matter action (1.9) ˆ ˆ √ Sm := d4 x −gLm = d4 x Lm , Consider a (3.115) which can be minimally coupled to any teleparallel theory of gravity, and where Lm and Lm are the Lagrangian density and 25 Rep. Prog. Phys. 86 (2023) 026901 Review (LXξ Γ)µ νρ = Xσξ ∂σ Γµ νρ − ∂σ Xµξ Γσ νρ + ∂ν Xσξ Γµ σρ Lagrangian respectively. The matter section is treated using the minimal coupling description (see section 3.2 [53]). Thus, the general matter Lagrangian may depend on the matter con- + ∂ρ Xσξ Γµ νσ + ∂ν ∂ρ Xµξ (3.118b) = ∇ρ ∇ν Xµξ − ∇ρ (Xσξ Tµ νσ ), (3.118c) ◦ tributions ϕ and both the tetrad eAµ and connection Γρµν . By taking variations wrt the tetrad and connections leads to the definition of the regular energy-momentum tensor −2 δLm Θµν := √ = eA µ −g δgµν  1 δLm e δeA ν ′  where we have used the flat condition Rµ νρσ ≡ 0 (see equation (2.26)). It is worth mentioning that in the standard case where the connection is the Levi-Civita one, if one assumes that the metric is invariant under a group G, (LXξ g)µν = 0, it follows that the Levi-Civita connection is ′ gνν ′ := eA µ ΘA ν gνν ′ , (3.116) ◦ also invariant under the group G, meaning that (LXξ Γ)µ νρ = 0. This property is true since the Levi-Civita connection depends directly on the metric. However, for the teleparallel connection, if we assume (LXξ g)µν = 0, it does not mean that (LXξ Γ)µ νρ = 0. Thus, these two quantities are zero if the metric and spin connection are invariant under the action of the group G. Since the TG fundamental variables are the tetrads and spin connection, one easily finds that the Lie derivatives acting on them are from which we denote the energy-momentum trace as Θ = Θµµ , and the hypermomentum as ∆ρµν := δSm ◦ δ Γρµν . (3.117) In principle, the hypermomentum contributes to a connection field equation in the same way that the regular energymomentum tensor contributes to the metrical (tetrad) variation field equation. As in the vast majority of the literature, we take a vanishing hypermomentum meaning that the matter fields are assumed to not couple to the connection. This will incur some constraints on the matter section (such as with fermions which do have nonzero hypermomentum due their coupling with the axial torsion). However, this effect is largely suppressed in real systems and is ignored in most works within and outside of TG. (LXξ e) A µ = Xνξ ∂ν e A µ + ∂µ Xνξ e A ν , (LXξ ω) A Bµ = Xνξ ∂ν ω A Bµ + ∂µ Xνξ ω A Bν . To fulfill the conditions (LXξ g)µν = (LXξ Γ)µ νρ = 0 (invariance under G), one then requires (LXξ e) A µ = −λξA B e B µ , (LXξ ω) A Bµ = Dµ λξA B , (3.120) where Dµ λξA B = ∂µ λξA B + ω A Cµ λCξ B − ω C Bµ λξA C was used and λξ is the local Lie homomorphism (see [76]) given by 3.6. Symmetries of teleparallel geometries As introduced in section 2.2, teleparallel theories have the tetrads and spin connection as their fundamental variables. Then, the notion of symmetries in TG should be taken carefully since, differently to purely metric based theories (as GR), now both quantities may or may not respect the symmetries of a certain spacetime. This study was fully described in [74] where it was assumed that both the tetrads and the teleparallel connection satisfy these symmetries. A recent quick summary of these results as long as some applications were also presented in the recent paper [75]. In the following, we explain how these symmetries can be connected to ansatz choices for the tetrad and spin connection. Let us start by defining what symmetries are in TG. A geometric object (on the manifold M) will be invariant under the group action φ : G × M → M, where G being a Lie group on M. Thus, the symmetry of TG geometries is defined as a group action φ such that both the metric and affine connection are invariant. For practical reasons, it is convenient to consider infinitesimal symmetries to find out the symmetries underlying a certain teleparallel theory. To do this, one can use the Lie derivatives wrt a vector Xξ ∈ Vect M acting on both quantities, which acting on the metric and the connection give us the following (LXξ g)µν = Xρξ ∂ρ gµν + ∂µ Xρξ gρν + ∂ν Xρξ gµρ , (3.119) λξ (x) = d . Λ (x) dt ξ̂(t) t=0 (3.121) Now, one can use local Lorentz transformations (for both the tetrads and spin connection) and choose a specific gauge ΛA B where the spin connection ω ′A Bµ (Weitzenböck gauge) vanishes to simplify the above conditions. After doing this, one gets that the conditions (3.120) become (LXξ e ′ ) A µ = −λ ′ ξ B e ′b µ , A 0 ≡ (LXξ ω ′ ) A Bµ = ∂µ λ ′ ξ B . (3.122) A One can notice that in the Weitzenböck gauge, it is much easier to solve the conditions in such a way that both the metric and the connection are invariant under the group action. We can define symmetric tetrads as the tetrads satisfying the above conditions, leading to (LXξ g)µν = (LXξ Γ)µ νρ ≡ 0, that are related by global Lorentz transformations ′ Λ̄F B A = ΛB D Λ A C ΛF′ D C . (3.123) This condition is very important since using it, one is able to find the symmetric tetrads that are compatible with the underlying symmetries. Since all computations are easier in the Weitzenböck gauge, we will omit the primes when we are (3.118a) 26 Rep. Prog. Phys. 86 (2023) 026901 Review computing the symmetries in this case and whenever we perform a Lorentz transformation, the transform quantities (tetrad and spin connection) will be denoted with the primes. It is also important to mention that the condition (LXξ Γ)µ νρ = 0 implies that 1 (LXξ Γ) σ [µν] = − (LXξ T) σ µν = 0, 2 and obtain another tetrad-spin connection pair satisfying axial symmetry e ′0 µ = C0 µ , (3.124) e0 µ = C0 µ , e ′ 3 µ = C3 µ , (3.128a) ω ′ 1 2φ = −ω ′ 2 1φ = −1. (3.128b) e1 µ = C 1 µ , e 2µ = C 2µ, e3 µ = C 3 µ , (3.129) where again C A µ are 16 arbitrary functions of t, r, ϑ, but now there is not a φ dependant in this tetrad. This branch trivially solves equation (3.122). We notice that this branch is related to the first branch via equation (3.127). However, the above tetrad has zero spin connection whereas in the first branch (3.128), the spin connection is non-zero. Thus, these two branches are different solutions satisfying axial symmetry in TG. One important remark here is that the second branch cannot recover spherical symmetry in TG in any limit. Even though the metric can be spherically symmetric in a certain limit, the connection will always be axially symmetric. This is the reason why we label this branch as solely axially symmetric branch. On the other hand, the first branch (regular one), has a spherically symmetric limit which will be the same tetrad that we will find in the next section. Let us first derive the symmetric tetrad for axial symmetry. We will work in spherical coordinates (t, r, ϑ, φ). In this case, there is only one generator given by Xz = −∂φ (only one of the generators of the case in equation (3.130)) which is related to the SO(2) group. Then, for axial symmetry, we can choose a local homomorphism in the domain of the group SO(2). One canonical choice in this domain gives us that the differential (3.121) acts on the generator X z , which in this case gives   0 0 0 0 0 0 −1 0  λ(Xz ) =  (3.125) 0 1 0 0  . 0 0 0 0 3.6.1. Axial symmetry—SO(2). In this section, we derive the most general spherically symmetric tetrads in the Weitzenböck gauge. It is convenient to start the computations using spherical coordinates as we did in the previous section. In these coordinates, one finds three generators related to the spherical symmetry 3.6.2. Spherical symmetry—SO(3). We can solve the set of equations in equation (3.122) in the Weitzenböck gauge and find the most general symmetric tetrad satisfying axial symmetry. One notices that there are two possible branches which solve equation (3.122). The first branch that we label as regular branch, reads as follows Xx = sin φ ∂ϑ + Xz = −∂φ . cos φ ∂φ , tan ϑ Xy = − cos φ ∂ϑ + sin φ ∂φ , tan ϑ (3.130) These generators are associated to the rotational group SO(3). Now, we need to choose a local homomorphism Λ in the domain of the group SO(3). A canonical choice for Λ is given by   1 0 Λ= (3.131a) 0 R(α, β, γ) e1 µ = C1 µ cos φ − C 2 µ sin φ, e 2 µ = C1 µ sin φ + C 2 µ cos φ, e ′2 µ = C 2 µ , The second branch that we label as solely axially symmetric branch which also solve the symmetry conditions (3.122) and give us the following tetrad in the Weitzenböck gauge which tells us that the torsion tensor also satisfies the symmetries. This means that symmetric tetrads will automatically impose that the torsion tensor also respect the symmetries assumed. For any TG theory of gravity constructed from a tetrad, the torsion tensor and their covariant derivatives (wrt Levi-Civita or Weitzenböck) will also respect the symmetries, then, the field equations of this theory Wµν will also satisfy the symmetry conditions. In principle, the conditions related to the connection can be relaxed and only impose the symmetries in the tetrads (or the metric). We will now use the definitions above to derive the most general tetrads in the Weitzenböck gauge satisfying axial, spherical and cosmological symmetries. e0 µ = C0 µ , e ′1 µ = C1 µ , e3 µ = C 3 µ , (3.126) with C A µ being 16 functions of the coordinates t, r, ϑ. It is possible to perform a Lorentz transformation for both the tetrad (2.32) and the spin connection (2.34) in such a way that the tetrads do not depend on φ. This can be done by choosing the following Lorentz transformation   1 0 0 0 0 cos φ sin φ 0  Λ AB =  (3.127) 0 − sin φ cos φ 0 , 0 0 0 1   1 0 0 0 1 0 0 cos α − sin α 0 0 cos β  = 0 sin α cos α 0 0 0 0 0 0 1 0 sin β   1 0 0 0 0 1 0 0   , × 0 0 cos γ − sin γ  0 0 sin γ cos γ 27  0 0 0 − sin β   1 0  0 cos β (3.131b) Rep. Prog. Phys. 86 (2023) 026901 Review where R(α, β, γ) = Rz (α)Ry (β)Rx (γ) are the rotation matrices about the Cartesian coordinates axis with angles α, β, γ. Then, the differential (3.121) of Λ acts on the generators of rotations as     0 0 0 0 0 0 0 0 0 0 0 0     , λ(Xy ) = 0 0 0 1 , λ(Xx ) =  0 0 0 −1 0 0 0 0  0 0 1 0 0 −1 0 0   0 0 0 0 0 0 −1 0  λ(Xz ) =  0 1 0 0  . 0 0 0 0 (3.132) the φ dependence. To do this, let us take the following Lorentz transformation   1 0 0 0 0 sin ϑ cos φ sin ϑ sin φ cos ϑ   Λ AB =  0 cos ϑ cos φ cos ϑ sin φ − sin ϑ , (3.135) 0 − sin φ cos φ 0 yielding the tetrad e ′0 = C1 dt + C2 dr, e ′2 = C5 dϑ − C6 sin ϑdφ, In order to find the symmetric tetrad, we need to replace the quantities in equation (3.122) and then solve the resulting equations. It is sufficient to just take the first part of that equation, giving us the following form of the tetrad in the Weitzenböck gauge e0 = C1 dt + C2 dr, e ′1 = C3 dt + C4 dr, e ′3 = C6 dϑ + C5 sin ϑdφ, (3.136) and then also having the following non-zero components of the spin connection ω ′1 2ϑ = −ω ′2 1ϑ = −1, ω ′1 3φ = −ω ′3 1φ = − sin ϑ, ω ′2 3φ = −ω ′3 2φ = − cos ϑ. (3.133a) (3.137) The pair of equations (3.136) and (3.137) is equivalent as taking the tetrad in equation (3.133) with a zero spin connection (in the Weitzenböck gauge). Let us now consider that the field equations of any arbitrary TG theory is denoted by Wµν and its antisymmetric part is W[µν] . This raises the question of which components of W[µν] are nonvanishing if we consider that the field equations respect spherical symmetry? To respond to this question, let us first take the vector X z from the rotations in equation (3.130). The invariance of W[µν] to this generator will be given by e = sin ϑ cos φ(C3 dt + C4 dr) + (C5 cos ϑ cos φ − C6 sin φ)dϑ − sin ϑ(C5 sin φ + C6 cos ϑ cos φ)dφ, (3.133b) 1 e 2 = sin ϑ sin φ(C3 dt + C4 dr) + (C5 cos ϑ sin φ + C6 cos φ)dϑ + sin ϑ(C5 cos φ − C6 cos ϑ sin φ)dφ, (3.133c) e3 = cos ϑ(C3 dt + C4 dr) − C5 sin ϑdϑ + C6 sin 2 ϑdφ, (3.133d) 0 = (LXz W)[µν] = ∂φ W[µν] , where all the six functions Ci depend only on r and t, and the ϑ dependence is fully determined. We can here notice that this tetrad is a special case of the axially symmetric regular branch (3.126) in the spherically symmetric limit. Since the symmetry condition (3.122) in spherical symmetry only has the above tetrad as a solution (in the Weitzenböck gauge), then, one can conclude that the second branch in axial symmetry (see equation (3.129)) cannot become spherically symmetric in any limit. The above symmetric tetrad is the most general tetrad satisfying spherical symmetry. The metric associated to this tetrad is (3.138) then, all the components of W[µν] do not depend on φ. Furthermore, by considering the other two rotational generators X x and X y (see (3.130)), we can compute (cos φLXx W + sin φLXy W)[µν]   0 0 = (sin ϑ)−2 W[tφ] −W[tϑ] 0 0 (sin ϑ)−2 W[rφ] −W[rϑ]  −(sin ϑ)−2 W[tφ] W[tϑ] −(sin ϑ)−2 W[rφ] W[rϑ]  ,  0 0 0 0 (3.139) ds 2 =(C12 − C32 )dt 2 − (C42 − C22 )dr 2 − (C52 − C62 ) × (dϑ 2 + sin 2 φ dφ 2 ) − 2(C3 C4 − C1 C2 )dr dt, (3.134) which gives us that W[tϑ] = W[rϑ] = W[tφ] = W[rφ] = 0. which is clearly non-diagonal. We can simplify the form of the tetrad equation (3.133) by performing a Lorentz transformation in order to eliminate all (3.140) Furthermore, from the generators X x and X y we also find that 28 Rep. Prog. Phys. 86 (2023) 026901 Review  0 ∂ϑ W[tr] (cos φLXy W − sin φLXx W)[µν] =   0 0 −∂ϑ W[tr] 0 0 0 0 0 0 − cot ϑ W[ϑφ] + ∂ϑ W[ϑφ] W[ϑφ] = W̃[ϑφ] (t, r) sin ϑ. 3.6.3. FLRW cosmological symmetry. (3.142) From here, we can conclude that the field equations Wµν of any TG theory in spherical symmetry will have only two nonvanishing antisymmetric components which are W[tr] and W[ϑφ] . This statement is valid for any TG. Let us finish this section by explicitly showing the form of the torsion tensor for the tetrad in equation (3.133) in the Weitzenböck gauge. As we pointed before, symmetric tetrads respecting spherical symmetry will impose that the torsion tensor T α µν will also respect this symmetries. The non-zero independent components for the torsion tensor T α µν for the symmetric tetrad (3.133) are Tt tr = C4 (C2,t − C1,r ) + C2 (C3,r − C4,t ) , C1 C4 − C2 C3 (3.143a) T r tr = C3 (C1,r − C2,t ) + C1 (C4,t − C3,r ) , C1 C4 − C2 C3 (3.143b) Tt ϑφ = T r ϑφ = Tϑ t ϑ = Tφ t φ = Tφ tϑ = − χ χ sin φ cos ϑ cos φ ∂ϑ − ∂φ , r r sin ϑ (3.144a) χ cos φ χ ∂φ , X2 = χ sin ϑ sin φ ∂r + cos ϑ sin φ ∂ϑ + r r sin ϑ (3.144b) χ X3 = χ cos ϑ ∂r − sin ϑ ∂ϑ , (3.144c) r √ where χ = 1 − kr 2 and k is the spatial curvature describing flat FLRW (k = 0), or open FLRW Universe (k = −1) or closed FLRW Universe (k = 1). Following the same question procedure asked before, which components of W[µν] are nonvanishing if we consider that the field equations respect cosmological symmetries? To respond to this question, we can first take the result found in the spherical symmetry case. We find that the only nonvanishing components are W[tr] and W[ϑφ] . Besides the rotational generators in spherical symmetry, in cosmological symmetries we also have the extra symmetries given by the translational generators in equation (3.144). If we apply X 3 to the remaining nonvanishing antisymmetric components of the field equations, we find X1 = χ sin ϑ cos φ ∂r + (3.143c) 2C1 C6 sin ϑ, C1 C4 − C2 C3 (3.143d) −C3 C5 + C5,t C5 + C6 C6,t , C52 + C62 (3.143e) Tϑ tφ 1 h C3 C6 − C5,t C6 + C5 C6,t i , = 2 C52 + C62 sin ϑ sin ϑ (3.143f ) Tϑ rϑ = Tφ rφ = Tφ rϑ = − 2C2 C6 sin ϑ, C2 C3 − C1 C4 −C4 C5 + C5,r C5 + C6 C6,r , C52 + C62 (3.141) Here, we derive the tetrads and spin connection satisfying cosmological symmetries in a FLRW spacetime. First, we will derive the tetrads in the Weitzenböck gauge and then as we did in previous sections, we will perform a Lorentz transformation to find an equivalent pair constructed from tetrads and a spin connection. The generators for cosmological symmetries are given by equation (3.130) plus three additional vector fields related to the translator generators, namely where we have also used equation (3.140). The above equation can be easily solved by setting W[tr] = W̃[tr] (t, r),  0  0 , cot ϑ W[ϑφ] − ∂ϑ W[ϑφ]  0 (LX3 W)[tϑ] = −χ sin ϑW̃[tr] , (3.143g) m(LX3 W)[rφ] = sin 2 ϑW̃[ϑφ] . χr 2 (3.145) The above equations then gives us that W[tr] = W[ϑφ] = 0. Thus, independent of the TG considered, the antisymmetric part of its field equations is always zero. This a consequence of the strong condition imposed by the cosmological symmetry. This statement is also valid for any rank 2-tensor. Therefore, if one is able to find symmetric tetrads satisfying cosmological symmetries, they will automatically solve the antisymmetric field equations in any TG. A recent paper [77] performed a complete classification of cosmological teleparallel geometries. Tϑ rφ 1 h C4 C6 − C5,r C6 + C5 C6,r i , = 2 C52 + C62 sin ϑ sin ϑ (3.143h) where Ci,r = ∂Ci /∂r and Ci,t = ∂Ci /∂t. Note that C1 C4 − C2 C3 6= 0 always, otherwise the inverse of the metric and the inverse of the tetrad is singular. We directly notice that the torsion tensor indeed also respects spherical symmetry. 29 Rep. Prog. Phys. 86 (2023) 026901 Review In the following subsections, we will derive the tetrads and spin connection satisfying cosmological symmetry for the three possible cases: k = 0, k > 0 and k < 0. note that the torsion tensor in the diagonal tetrad in Cartesian coordinates becomes very simple We can take the result found in spherical symmetry as a first step (3.133) in the Weitzenböck gauge and now consider the generator X 3 , giving us that the functions Ci must have the following form where i, j = 1, 2, 3 are the spatial spacetime indices and H = ȧ/(Na) is the Hubble parameter. Let us finish this section by noting that we still have the freedom to choose a different time coordinate t 7→ t̃ due to the fact that the time diffeomorphisms commute with the cosmological symmetry. By doing this, one can always choose a constant lapse function, which usually is chosen to be N(t) = 1. T i 0j = NHδ i j , 3.6.3.1. Flat case (k = 0)—ISO(3). C1 = N(t), C4 = a(t), C5 = a(t)r, C2 = C3 = C6 = 0. (3.146) We can identify N(t) as the lapse function and a(t) as the scale factor of the Universe. Thus, by replacing the above form of the functions in the tetrad (3.133), we find that the most general tetrad satisfying cosmological symmetries in the Weitzenböck gauge becomes e0 = N(t)dt, 3.6.3.2. Real tetrad for k ⩾ 0—SO(4). The first non-trivial case for a spatially curved FLRW cosmology has the following translation generators   0 0 0 0 0 0 0 0   λ± (X1 ) = ±  0 0 0 1  , 0 0 −1 0   0 0 0 0 0 0 0 −1  λ± (X2 ) = ±  0 0 0 0  , 0 1 0 0   0 0 0 0 0 0 1 0   λ± (X3 ) = ±  (3.152) 0 −1 0 0 . 0 0 0 0 (3.147a) e1 = a(t) [sin ϑ cos φdr + r cos θ cos φdϑ − r sin ϑ sin φdφ] , (3.147b) e 2 = a(t) [sin ϑ sin φdr + r cos θ sin φdϑ + r sin ϑ cos φdφ] , (3.147c) e3 = a(t) [cos ϑdr − r sin ϑdϑ] . (3.147d) As we have shown before, this tetrad will give W[µν] = 0 for any TG theory. If we perform the Lorentz transformation (3.135) we find that the new tetrad can be written as follows e ′0 = N(t)dt, e ′1 = a(t)dr, Using these translator generators and the rotational ones (see equation (3.130)), we can solve the symmetric conditions (3.122), which gives us the spherically symmetric tetrad (3.133) and some additional conditions on the functions, namely e ′2 = a(t)rdϑ, e ′3 = a(t)r sin ϑdφ, (3.148) C1 = N(t), while having a non-zero spin connection with the same components as (3.137). Another interesting result is that the tetrad (3.147) can be transformed to Cartesian coordinates (t, x, y, z), and gives us that the tetrad in the Weitzenböck gauge in Cartesian coordinates reads 0 e = N(t)dt, 1 e = a(t)dx, 2 e = a(t)dy, C4 = a(t) , χ C5 = rχa(t), √ C6 = ∓ k r 2 a(t), C2 = C3 = 0, (3.153) where k ∈ {0, ±1}, giving us that the final expression of the symmetric tetrad satisfying cosmological symmetries in the Wetizenöck gauge being equal to e0± = N(t)dt, 3 e = a(t)dz, (3.149) e1± = a(t) h sin ϑ cos φ dr + r(χ cos θ cos φ ± (3.154a) √ kr sin φ)dϑ i √ − r sin ϑ(χ sin φ ∓ kr cos ϑ cos φ)dφ , h sin ϑ sin φ √ 2 dr + r(χ cos θ sin φ ∓ kr cos φ)dϑ e± = a(t) χ i √ + r sin ϑ(χ cos φ ± kr cos ϑ sin φ)dφ ,   √ cos ϑ dr − rχ sin ϑdϑ ∓ kr 2 sin 2 ϑdφ . e3± = a(t) χ (3.154b) or equivalently e A µ = diag(N(t), a(t), a(t), a(t)). (3.151) (3.150) The above tetrad in Cartesian coordinates is diagonal and is valid in the Weitzenböck gauge. Since the form of this tetrad is very simple, we will largely use this tetrad in the cosmology part in section 6 in a more practical setting. Furthermore, since this tetrad respects cosmological symmetries, we will also have W[µν] = 0 for any TG theory. It is also useful to 30 χ Rep. Prog. Phys. 86 (2023) 026901 Review √ Here, χ = 1 − kr 2 . It should be noted that k = 1 (k = −1) represents a closed (open) Universe whereas k = 0, a flat Universe. As expected, when k = 0, we have χ = 1 and then we recover the flat FLRW symmetric tetrad (3.147). Note that the above tetrad is real only if either k = 0 or k = 1, so that, the tetrad becomes complex for the negatively curved k (see section 3.6.3.3 for another tetrad which is real for k ⩾ 0). In this case we have two different tetrads depending on the sign chosen. By taking the following Lorentz transformation  1 0 0 0 0 sin ϑ cos√ φ sin ϑ sin√ φ cos ϑ  A ,  Λ± B=  0 χ√cos ϑ cos φ ± k r sin φ χ cos ϑ sin√ φ ∓ k r cos φ −χ sin ϑ √ 0 ± k r cos ϑ cos φ − χ sin φ χ cos φ ± k r cos ϑ sin φ ∓ k r sin ϑ  3.6.3.3. Real tetrad for k ⩽ 0—SO(1,3). the tetrads (3.154) transform as follows ′0 e± = N(t)dt, ′1 e± = a(t) dr, χ Following the same algorithm mentioned in the previous sections, there is another non-trivial translator generators satisfying cosmological symmetries, namely ′2 e± = a(t)rdϑ, ′3 e± = a(t)r sin ϑdφ, (3.156)  0 1 ± λ (X1 ) = ±  0 0  0 0 ± λ (X2 ) = ±  1 0  0 0 ± λ (X3 ) = ±  0 1 which then induces a non-zero spin connection with components, √ ω±′1 2ϑ = −ω±′2 1ϑ = −χ, ω±′1 2φ = −ω±′2 1φ = ± k r sin ϑ, √ (3.157a) ω±′1 3ϑ = −ω±′3 1ϑ = ∓ k r, √ 1 ω±′1 3φ = −ω±′3 1φ = −χ sin ϑ, ω±′2 3r = −ω±′3 2r = ± k , χ ′2 ′3 ω± 3φ = −ω± 2φ = − cos ϑ. (3.157b) As expected, the Lorentz transformation (3.155) becomes the same one as the one performed in equation (3.135) for the flat case, and then the spin connection (3.157) is also the same as the one found in equation (3.137) when k = 0. For the symmetric tetrad (3.154) in the Weitzenböck gauge or equivalently, by considering the tetrad in equation (3.156) with the spin connection in equation (3.157), we find that the non-zero components of the torsion tensor T α µν become √ T i tj = NHδ i j , Tr ϑφ = ∓2 kr 2 sin ϑχ, √ Tϑ rφ k Tφ rϑ = − 2 = ∓2 csc ϑ. (3.158) χ sin ϑ 6k − 6H 2 , a2 B=− 6Ḣ − 18H 2 N 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0  0 0 , 0 0  0 0 , 0 0  1 0 . 0 0 (3.160) Then, the spherically symmetric tetrad (3.133) for this case becomes √ r e0± = ±N(t)χdt ∓ a(t) −k dr, χ (3.161a) "   N(t) √ e = a(t) sin ϑ cos φ dr − −krdt + r cos ϑ cos φdϑ a(t) # 1 − r sin ϑ sin φdφ , As expected, the flat case k = 0 coincides with the torsion found before in equation (3.151) and if k = −1, the torsion tensor becomes complex, meaning that one needs to be careful on interpreting the physical quantities for the negatively spatial curvature. Even though the torsion tensor is complex, it is important to mention that some important scalars as the torsion scalar T and the boundary term B (see equations (4.158) and (4.156), respectively) become T= (3.155) " (3.161b)   N(t) √ e = a(t) sin ϑ sin φ dr − −krdt + r cos ϑ sin φdϑ a(t) # 2 + r sin ϑ cos φdφ , (3.161c)     N(t) √ e3 = a(t) cos ϑ dr − −krdt − r sin ϑdϑ , a(t) (3.161d) (3.159) are real for this tetrad for any k. Therefore, any action constructed from these scalars will be real for this tetrad. 31 Rep. Prog. Phys. 86 (2023) 026901 where again χ = of the form of 1 − kr 2 and the functions were taken to be r√ −k a(t), χ C5 = ra(t), C6 = 0, C1 = ±χN(t), C4 = a(t), √ Review C2 = ∓ Let us finish the cosmological symmetry section by emphasizing again that the tetrads with zero spin connection components, namely equations (3.154) and (3.161), which are valid for k ∈ {0, ±1} will solve all the antisymmetric field equations for any modified teleparallel theory of gravity. Equivalently, the tetrad-spin connection pairs given in equations (3.156), (3.157) and (3.164), (3.165) also satisfy all the antisymmetric field equations for any teleparallel theory. Then, we will use these tetrads in forthcoming sections to study FLRW cosmology. √ C3 = − −kr N(t), (3.162) to fulfill the symmetry condition n equation (3.122). As in the previous case, this symmetric tetrad can become the flat FLRW symmetric tetrad (3.147) but as it happened before, the above tetrad is complex for k = 1. Similarly as we did in the previous section, we can perform the following Lorentz transformation 3.6.4. Maximally symmetric spacetimes—ISO(1,3), SO(1,4), SO(2,3). A spacetime with the maximum number of symA Λ± B  metries can be obtained by taking the following generators can be written as  √ √ √ ±χ −kr sin ϑ cos φ −kr sin ϑ sin φ −kr cos ϑ √ ± −kr χ sin ϑ cos φ χ sin ϑ sin φ χ cos ϑ    = , cos ϑ cos φ cos ϑ sin φ − sin ϑ   0 0 − sin φ cos φ 0 X0 = χ∂t − krχTk (t)∂r , (3.163) XX = r sin ϑ cos φ∂t + χ 2 sin ϑ cos φTk (t)∂r to rewrite the tetrad (3.161) as ′0 e± = N(t)dt, ′1 e± = a(t) dr, χ cos ϑ cos φTk (t) sin φTk (t) ∂ϑ − ∂φ , r r sin ϑ + ′2 e± = a(t)rdϑ, ′3 e± = a(t)r sin ϑdφ, (3.164) + where χ = √ ω±′1 2ϑ = −ω±′2 1ϑ = −χ, ω±′1 3φ = −ω±′3 1φ = −χ sin ϑ, (3.165b) (3.168c) sin ϑTk (t) ∂ϑ , (3.168d) r 1 − kr 2 and we have defined   tanh t for k = 1, Tk (t) := t for k = 0, .   tan t for k = −1 (3.169) The metric reproduced by the above tetrad becomes The nonvanishing components of the torsion tensor Tλ µν for the tetrad-spin connection pair (3.164) and (3.165) (or equivalently with the Weitzenböck tetrad (3.161)) can be expressed as  2 ds = dt 2 − Ck2 (t)   dr 2 2 2 2 2 , + r dϑ + sin ϑdφ 1 − kr 2 (3.170) where (3.166a)   cosh t Ck (t) := 1   cos t Note that if k = 1, the above torsion tensor is complex. Furthermore, the situation is slightly different than the previous section since the the torsion scalar and boundary term (again see equations (4.158) and (4.156)) √ 12 −kH 6k T=− + 2 − 6H 2 , a a √ 12 −kH 6Ḣ B=− − − 18H 2 a N sin φTk (t) cos ϑ sin φTk (t) ∂ϑ + ∂φ , r r cos ϑ XZ = r cos ϑ∂t + χ 2 cos ϑTk (t)∂r − √ 1 √ ω±′0 1r =ω±′1 0r = − −k , ω±′0 2ϑ = ω±′2 0ϑ = − −kr, χ √ ′0 ′3 ω± 3φ =ω± 0φ = − −kr sin ϑ, (3.165a)  1√  i Ti 0j = H + −k N δ j . a (3.168b) XY = r sin ϑ sin φ∂t + χ 2 sin ϑ sin φTk (t)∂r but now with a non-zero spin connection which has the following components, ω±′2 3φ = −ω±′3 2φ = − cos ϑ. (3.168a) for k = 1, for k = 0, . for k = −1 (3.171) Thus, k = 0 represents Minkowski and k = 1 (k = −1) deSitter (anti de-Sitter) spacetimes. The Minkowski case is represented by the ISO(1, 3) group whereas for the de-Sitter (anti de-Sitter) the group is SO(2, 3) (SO(1, 4)). It is easy to notice that for Minkowski, the tetrad in the Weitzenböck gauge (3.147) with N(t) = a(t) = const. = c, is the solution of the symmetry condition under the generators (3.168), which gives us the following tetrad (3.167) are complex for k > 0, meaning that any action constructed from this tetrad will be complex for the positively curved case. 32 Rep. Prog. Phys. 86 (2023) 026901 e0 = c dt, Review (3.172a) The relation between the different parametrizations is given by (3.172d) One can also perform a coordinate transformation and find that in Cartesian coordinates (t, x, y, z), the tetrad satisfying Minkowski symmetries in the Weitzenböck gauge becomes a diagonal one, namely e A µ = c diag(1, 1, 1, 1). (3.173) where is found using the inverse formula (3.36) (which can also be used for the non-trivial tetrads), and likewise from the transformation of the determinant of the tetrad, given by Finally, it is important to mention that for de-Sitter and anti de-Sitter, there are not symmetry solutions to the symmetry condition (3.122) under the generators (3.168). This means that there are no de-Sitter or anti de-Sitter symmetries in TG. Then, the metric can be de-Sitter or anti de-Sitter and then it can respect the symmetries but the teleparallel connection cannot respect this type of symmetry. ẽ = C3 (C − 2XD)e. g̃ = A3 (A − 2XB)g = C 6 (C − 2XD) 2 g, T̃ A µν = CT A µν + 2∂[µ Ce A ν] + 2η AB EB ρ  × ∂ρ ϕ∂[µ D∂ν] ϕ + D∂[ν ϕ∇µ] ∂ρ ϕ , K̃ABµ = KABµ + 2E[A α EB] β ◦ × ◦ ◦ ◦ ◦ ◦ D∇µ ∇α ϕ∇β ϕ D 2 ∇µ ϕ∇α ϕ∇β X + C − 2XD C(C − 2XD) ◦ ◦ ◦ ◦ ◦ ◦ C,X Dgαµ ∇β ϕ∇γ ϕ∇γ X D,X ∇α X∇β ϕ∇µ ϕ + + C(C − 2XD) C ! ◦ ◦ C,X ∇α Xgβµ C,ϕ ∇α ϕgβµ + , (3.183) + C C − 2XD (3.175) with two free functions A, B of the scalar field and its kinetic term which constitute a special case of a more general class of scale transformations of the teleparallel affine connection [81]. Note that the latter is independent of the teleparallel spin connection ω A Bµ , and can thus be obtained directly from the transforma◦ tion of the Levi-Civita spin connection ω A Bµ using the relation (3.176) If the metric is defined by a tetrad e A µ , it is more convenient to introduce a different parametrization of disformal transformations. In this case one may define [80] ẽ A µ = C(ϕ, X)e A µ + D(ϕ, X)∂µ ϕ∂ν ϕgνρ e A ρ , (3.182) and contortion (3.174) where A is a free function of the scalar field. If one allows also for a dependence on first derivatives of the scalar field, the most general class of transformations is of the disformal type [79] 1 X = − gµν ∂µ ϕ∂ν ϕ. 2 (3.181) in terms of the original parameter functions A, B. It is straightforward to derive the corresponding transformation rules for other relevant quantities in the teleparallel geometry, such as the torsion, The concept of conformal and disformal transformations arises from the question which is the most general possibility to construct a metric g̃µν from a given metric gµν and a scalar field ϕ. If one demands that the new metric depends only on the values of the old metric and scalar field, but not on their derivatives, one finds that the most general form of the new metric is given by the conformal transformation [78] g̃µν = A(ϕ, X)gµν + B(ϕ, X)∂µ ϕ∂ν ϕ, (3.180) It is a remarkable fact that the transformation of the metric determinant takes a similarly simple form 3.7. Conformal and disformal transformations g̃µν = A(ϕ)gµν , (3.178) where here, and in the remainder of this section, we will omit the arguments ϕ, X for brevity’s sake. Note that we recover the class of conformal transformations for C,X = D = 0, where the subscript denotes the derivative wrt the argument X. In order to obtain an invertible transformation, one must demand C 6= 0 and C − 2XD 6= 0. This can also be seen from the transformation behavior of the inverse tetrads, which is given by   1 D ẼA µ = EA µ − gµν EA ρ ∂ν ϕ∂ρ ϕ , (3.179) C C − 2XD e 2 = c [sin ϑ sin φdr + r cos θ sin φdϑ + r sin ϑ cos φdφ] , (3.172c) e3 = c [cos ϑdr − r sin ϑdϑ] . B = 2D(C − XD), A = C 2, e1 = c [sin ϑ cos φdr + r cos θ cos φdϑ − r sin ϑ sin φdφ] , (3.172b) ◦ ◦ 0 = ω̃ A Bµ − ω A Bµ ≡ ω̃ A Bµ − ω A Bµ + K̃ A Bµ − K A Bµ . (3.184) (3.177) while the spin connection is left unchanged, ω̃ A Bµ = ω A Bµ , since any transformation preserving its flatness and antisymmetry could be absorbed into a local Lorentz transformation. To obtain the corresponding forms given only with spacetime indices, one must take into account the transformation of the 33 Rep. Prog. Phys. 86 (2023) 026901 Review ´ √ ◦ where the variation δ[ d4 x −gG]/δgµν naturally vanishes as a boundary term. In this setup, the Gauss–Bonnet invariant is derived from first principles through the Gauss–Bonnet theorem which connects the curvature of surfaces in differential geometry with their Euler characteristic in topology. The teleparallel equivalent of the Gauss–Bonnet (TEGB) invariant is tetrad and its inverse, which relate these different representations. Explicit formulas can be found in [82]. A more compact form of the disformal transformations can be obtained by using the language of differential forms. For this purpose one defines the one-forms [83] π A = Dϕ,A = dϕ,A − ω B A ∧ ϕ,B , ψ A = ϕ,A dϕ, ◦ ◦ ◦ π A = Dϕ,A = dϕ,A − ω B A ∧ ϕ,B , ◦ found by transforming the Gauss–Bonnet term G through the contortion tensor to its teleparallel equivalent T G . This procedure will produce a dynamical scalar T G as well as a total divergence term BG which together will sum to the Gauss– (3.185) using the abbreviation ◦ µ ϕ,A = EA ∂µ ϕ. Bonnet scalar G = TG + BG [85, 86]. This is the fundamental ◦ (3.186) difference between the scalar G and T G . A teleparallel analogue to the Gauss–Bonnet theorem may lead to a separate scalar invariant that does not depend on the results of standard gravity. The TEGB term was first derived in [87] using the language of forms, since the coordinate description leads to much more cumbersome equations. In standard gravity, the Gauss–Bonnet invariant is expressed in this context as In terms of these one-forms, the disformal transformation (3.177) takes the simple form ẽ A = Ce A + Dψ A , (3.187) while the corresponding dual vector fields obey the transformation   1 D B BC ẼA = (3.188) δ − ϕ,A ϕ,C η EB , C A C − 2XD ◦ ◦ such that the variation of which is obtained from the corresponding formula in equation (3.179). Finally, we note that the torsion two-form transforms as ◦ S ◦ := G (3.189) while for the contortion we have K̃AB = KAB + 2 ◦ ◦ ˆ ◦ G + Sm , C C C Γ AB = Γ BA + K BC , 2 Dϕ,[A π B] D X,[A ϕ,B] dϕ − − C − 2XD C(C − 2XD) × 1 2κ2 (3.192) (3.193) vanishes, where κ2 = 8πG (as described in equation (1.7)) is the gravitational constant (whose value is not relevant here), and the exterior (wedge) product is again used. Using the relation between the Levi-Civita and teleparallel connections in equation (3.67), namely T̃ A = CT A + C,ϕ dϕ ∧ e A + C,X dX ∧ e A + Dπ A ∧ dϕ + D,X dX ∧ ψ A , ◦ G = ϵabcd Rab ∧ Rcd , (3.194) together with the following properties C,X Dg−1 (dϕ, dX)ϕ,[A eB] D,X X,[A ϕ,B] dϕ + C(C − 2XD) C ! C,X X,[A eB] C,ϕ ϕ,[A eB] + + . (3.190) C C − 2XD • DRAB = 0—The covariant derivative of the Lorentz ◦◦ curvature, DRAB = dRAB + ω AC ∧ R CB + ω BC ∧ RAC , vanishes which gives the second Bianchi identity in the language of differential forms [88]; Similar expressions may also be derived for the transformation behavior of the one-forms ψ A and π A introduced above [83]. • Dea = 0—The Levi-Civita Lorentz covariant derivative is torsionless and satisfies metricity, and so the tetrad exterior derivative will identically vanish; − ◦◦ ◦ ◦ 3.8. The teleparallel Gauss–Bonnet invariant In (3 + 1)−dimensions, the Gauss–Bonnet term is a topological invariant quantity that does not contribute dynamically since it is a total divergence term in the gravitational action. For a standard gravity scenario, based on the Levi-Civita connection, this boundary term takes on the form [84] ◦ ◦ ◦ ◦ ◦ ◦ G := R 2 − 4Rµν Rµν + Rµνκλ Rµνκλ , ◦ • RAB + DKAB = RAB − KAC ∧ K CB —This describes how the Ricci tensors as calculated with both connections are related together which can be observed from equation (3.107). The Levi-Civita Lorentz derivative term can be expressed ◦ in terms of the teleparallel connection through DKAB = DKAB − 2KAC ∧ K CB which directly expresses the LeviCivita Ricci tensor in terms of teleparallel quantities as ◦ RAB = RAB + KAC ∧ K CB − DKAB . (3.191) 34 (3.195) Rep. Prog. Phys. 86 (2023) 026901 Review These properties combine to re-express the Levi-Civita based Gauss–Bonnet invariant in equation (3.192) as [87, 89] STEGB := 1 2κ2 teleparallel connection through the scalars in equation (3.202) which contain a teleparallel scalar and a total divergence term. As in standard gravity, these scalars can be used to construct theories of gravity [86, 90, 91], and may also lead to a purely TEGB invariant constructed purely within TG. Another important aspect of the Gauss–Bonnet term formulated in ˆ (TG − dBG ) + Sm , (3.196) ◦ where the integrands can be expressed as TG := TG e1 ∧ e 2 ∧ e3 ∧ e4 ,   ◦ ◦ BG := ϵABCD 2KAB ∧ R CD + KAB ∧ DK CD , equation (3.201) is that since G and BG are boundary terms, this means that T G also would be a boundary term. Another point to remark is that one can also construct theories by defining the following invariants (3.197a) (3.197b) in which the TEGB Lagrangian is given by µνσλ αβ µνσλ α TG = δαβγϵ K χµ Kχβ ν Kγ ξσ Kξϵ λ + 2δαβγϵ K µ Kγ χν Kχϵ ξ Kξ σλ µνσλ αβ + 2δαβγϵ K µ Kγ χν Dλ Kχϵ σ . (3.198) In terms of coordinates rather than differential forms, these scalar invariants can then be expressed as [85] ◦ G = TG + BG , ABCD I TG1 := δIJML K EA KEJB KMFC KFLD , (3.204a) ABCD IJ TG2 := −2δIJML K A KMEB KEFC KFLD , (3.204b) ABCD IJ TG3 := 2δIJML K A KMEB KELF KFCD , (3.204c) ABCD IJ ABCD TG4 := 2δIJML K A KMEB KELC,D δIJML , (3.204d) such that TG = TG1 + TG2 + TG3 + TG4 , but these TGi alone would not be boundary terms anymore and would have nontrivial contributions linearly in an action. (3.199) where the boundary term is given by [85] i 1 h µνσλ αβ  γ 1 BG = ∂µ eδαβγϵ K ν K ξσ Kξϵ λ − R̊γϵ σλ , (3.200) e 2 3.9. Analogue with other branches of physics TG offers a new perspective on interpreting gravitational interactions that revisits the foundations on which gravity is built. Similar to the introduction of curvature as the mediator of gravitation in GR, attempts have been made in other branches of physics to utilize this description, and its geometric baggage, in advancing or uniting these field descriptions. One important direction in which the mechanics of geometric gravity has been used to describe other phenomena is that of lattice structures [92–94] which can be applied within the trinity of geometric structures, namely Riemann curvature [95], torsion [96–98] and non-metricity. In the context of continuum mechanics, the geometric tools of gravity have been used to describe various effects in continuum mechanics. The different geometric tools available within the trinity of gravity can describe different types of possible deformations within these structures as shown in figure 6, where deformations are made to an analogous Minkowski crystal. In figure 6 curvature, torsion and metricity are shown to produce distinct and non-overlapping changes to crystal geometry. For continuum mechanics, deformations can be studied through this prism of a trinity of geometric tools. For torsional geometry, the tools being developed can be used to describe dislocations which are crystallographic defects, or irregularities, in the crystal structure [99]. Dislocation movements appear as microstructures glide over each other producing geometric information that can be interpreted through the tensorial formalism associated with gravitational theories. The creation and movement of a large number of dislocations can lead to plastic deformations, while more generally, the number and arrangement of dislocations can effect with Dλ Kχϵ σ = ∂λ Kχϵ σ + (Γ̊ + K)χ βλ Kβϵ σ + (Γ̊ + K)ϵ βλ Kχβ σ − (Γ̊ + K)β σλ Kχϵ β . This term is a total divergence term. Thus, the total Levi-Civita Gauss-Bonnet invariant can be written in terms of the teleparallel connection through the scalars in equation (3.202) which contain a teleparallel scalar and a total divergence term. As in standard gravity, these scalars can be used to construct theories of gravity [86, 90, 91], and may also lead to a purely TEGB invariant constructed purely within TG. Another important aspect of the Gauss-Bonnet term formulated in equation (3.201) is that TG is a topological invariant in 4 dimensions (as G̊) and BG is a boundary term in all dimensions. also would be a boundary term.  TG = KIEA KEJB KMFC KFLD − 2KIJA KMEB KEFC KFLD  ABCD + 2KIJA KMEB KELF KFCD + 2KIJA KMEB KELC,D δIJML . (3.201) In terms of coordinates rather than differential forms, these scalar invariants can then be expressed as [85] ◦ G = TG + BG , (3.202) where the boundary term is given by  ABCD 1  , BG = ∂A eKIJB KKLC,D + KMCD KKLM δIJML e (3.203) which is a total divergence term. Thus, the total LeviCivita Gauss–Bonnet invariant can be written in terms of the 35 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 6. Crystalline structure and its analogy with curvature, torsion and non-metricity. the properties associated with a material body. Thus, through the lens of geometry the impact of dislocations can be studied through the geometric tools being developed in gravitational physics. The dislocations in these structures take on the form of particle-like bodies in that their behavior can be modeled in this way. As identified in [100], this phenomenon can be described by a deformation field is represented ux = 0, uy = 0, uz = b y arctan , 2π x In [101], torsion is first related with dislocations in that their density and dynamics could be mapped by the torsion tensor. In figure 7, the two main types of dislocations are represented, namely edge and screw type dislocations. As shown in the axis defined in the figure, we take 1, 2 and 3 to point to the right, backwards and up respectively. Hence, an edge dislocation can be interpreted as the 1–3 plane being shifted in the 1-axis, and so giving a contribution to T113 component of the torsion tensor. Similarly, screw dislocations can be indicated as the 1– 2 plane being shifted in the 3−axis direction, and thus giving a contribution to T312 . In the context, torsion then becomes a measure of dislocation density for a vanishing lattice spacing scenario. There has been a number of important works that build on this approach to using torsion for dislocation theory [97, 102–105]. On the other hand, it may also be the case that the torsion tensor can be used on a discrete space which may better model certain crystal interactions. Lattice discretizations using the TG framework has also been applied using Regge calculus as in [106, 107] where dislocations are added with the curvature associated disclinations to form a fuller theory of discrete lattice structures using differential geometry. Another approach was attempted in [108] where the usual Regge calculus is used within a torsional framework in which the tetrads are used to replace the metric. This approach is then used to replace the Riemann scalar (3.205) which illustrates a screw dislocation in Cartesian coordinates, and where b is the maximal displacement between the microstructures, and u = (ux , uy , uz ) is the dislocation velocity. This is a solution to the Navier–Stokes equationq and produces 2 a longitudinal length contraction by a factor 1 − vc 2 . The s q 2 dislocation also has an associated energy of E0 / 1 − vc 2 (E0 s being the rest energy of the dislocation). The transverse speed of sound is represented by cs , which when identified with the speed of light, renders the motion analogous to that of an electric field undergoing contractions or expansions. In this way, propagating dislocation solutions behave exactly as relativistic particles. 36 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 7. Crystal dislocations are shown against a regular crystal with no dislocation (left), and where screw (middle) and edge (right) dislocations are represented. description with a TG approach. More recently, a more direct quantum gravity approach was used in [109] where the Abelian gauge group of translations is used immediately as the lattice variable without the use of additional structures. In closing, TG has reached into a number of seemingly unrelated topics but which share a geometric connection through which the mechanics developed in the theory can be more broadly applied. One other example of this is the body of work related to Cosserat mechanics where the geometric manifold is constrained to not express dislocations but only rotations which is a long studied topic in continuum mechanics [110]. This has been strongly connected to the mechanics of TG by associating the tetrad again with the manifold deformations and using the torsion tensor to express actions on the manifold [111–118]. Another would be in condensed matter media where low-energy fermions act like an effective and non-relativistic torsional medium [119–121]. While the governing equations are different to those of gravity, it is another application of this geometric outlook in another setting. On the other hand, there has not been much progress of an analogous interpretation of non-metricity in terms of crystal structures, which may be an interesting area to focus future works on the topic. after which we then analyze the properties of this theory and its dynamical equivalent to GR. Finally, we discuss some quantum gravity attempts in section 4.4 where we also highlight some future possibilities for the theory. 4.1. General properties of TG Before we come to a detailed study of specific TG theories and their properties, we discuss a number of general properties which can be shown to hold for any TG theory. The starting point for these considerations is a generic teleparallel action and its field equations, which we derive in section 4.1.1. We then derive a number of properties from the local Lorentz invariance of the matter action in section 4.1.2, as well as the gravitational action in section 4.1.3. In section 4.1.4, we show how these properties appear within the Palatini approach. They are further used in section 4.1.5, where we discuss the energymomentum conservation. The latter is closely related to the Bianchi identities, which we discuss in section 4.1.6. Finally, we study the premetric approach in section 4.1.7. In TG one most commonly assumes an action which of the form 4.1.1. Action and field equations. 4. The teleparallel formulation of gravity and its GR equivalent STG := Sg [e, ω] + Sm [e, χ], In this section, we explore the theoretical foundations on which we can construct TG theories. Ultimately, this is the goal of gravitational physics, namely, producing theories of gravity on which to probe Nature. Firstly, we describe some general properties of teleparallel theories of gravity in section 4.1, such as the local Lorentz structure of the gravitational and matter contributions together with other symmetry and conservation considerations. In section 4.2, we expand on the 3 + 1 decomposition of teleparallel theories and the procedure by which this can be developed. The TEGR is then extensively described in section 4.3 where we derive this from first principles depending on only a few physical assumptions, where the gravitational part Sg of the action depends on the tetrad e A µ and the spin connection ω A Bµ , while the matter part depends on the tetrad e A µ and arbitrary matter fields χI , but not on the spin connection; see [53, 122] for an extensive discussion. In brief the reason for this is that we assume that the hypermomentum vanishes which is one way of imposing that this coupling does not occur. If it were to depend on spin then we would be effective introducing a second matter tensor which would result from the variation of the matter Lagrangian wrt the spin connection. The variation of the matter part of the action, after integration by parts to remove any derivatives acting on the field variations, can thus be written in the form 37 (4.1) Rep. Prog. Phys. 86 (2023) 026901 Review ˆ δSm = d4 x e (ΘA µ δe A µ + ΩI δχI ), to the vector field Xµ = YA Bµ ξ A B , this takes the form ˆ  ◦ δω Sg = d4 x e ∂µ YA Bµ + Γµ νµ YA Bν − ω C Aµ YC Bµ  (4.10) + ω B Cµ YA Cµ ξ A B . (4.2) where ΩI = 0 are the matter field equations and ΘA µ is the energy-momentum tensor (1.10). The corresponding variation of the gravitational action takes the form ˆ δSg = − d4 x e (WA µ δe A µ + YA Bµ δω A Bµ ), (4.3) Finally, recalling that the variation ξ A B must be antisymmetric, the resulting field equations therefore take the form ◦ with tensors WA µ and YA Bµ arising from the variation and integration by parts, whose explicit form depends on the particular theory under consideration. The expression for WA µ will be shown for a number of theories in sections 4.3 and 5; see (4.165) for a simple example. We do not show YA Bµ here for brevity, since it turns out to be redundant in the derivation of the field equations, so that the latter are fully determined from WA µ alone. We will show this explicitly in the remainder of this section and in section 4.1.3. From the variation wrt the tetrad one obtains the field equations µ µ WA = Θ A . 0 = −W̃AB = ∂µ Y[AB]µ + Γµ νµ Y[AB]ν − ω C[A µ YC B]µ + ω [B Cµ YA]Cµ . This equation is more conveniently expressed in spacetime indices. For this purpose, note that the first term on the RHS of the field equation (4.11) can be rewritten using ∂µ Y[AB]µ = ∂µ Y[ρσ]µ e A ρ e B σ + Yρσµ ∂µ e[A ρ eB] σ + Yρσµ e[A ρ ∂µ eB] σ . Θµν = e A µ gρν ΘA ρ , (4.4) −ω C[A µ YC B]µ = ω [A|C| µ YC B]µ = ω [A Cµ Y|C|B]µ , Yρσµ ∂µ e[A ρ eB] σ + ω [A Cµ Y|C|B]µ = Γν ρµ Yρσµ e[A ν eB] σ , (4.14a) (4.5) Yρσµ e[A ρ ∂µ eB] σ + ω [B Cµ YA]Cµ = Γν σµ Yρσµ e[A ρ eB] ν . (4.14b) (4.6) The derivation of the field equations for the spin connection, however, is less trivial, since by definition it must be flat, R α βµν = 0, and metric-compatible, ∇α gµν = 0. Different possibilities exist in order to maintain this property during the variation procedure [123, 124]. One possibility is to explicitly restrict the variation of the spin connection such that its curvature and non-metricity are unaffected. This is the case if and only if the variation obeys the form δω A Bµ = Dµ ξ A B = ∂µ ξ A B + ω A Cµ ξ C B − ω C Bµ ξ A C , Finally, one is left with the task of transforming the second term in the field equation (4.11). Here, we can use the relation ◦ ◦ Γ νµ = Γ µν = Γ µν − K µν = Γ µν = Γ νµ − T µν , (4.15) µ µ µ µ µ µ µ which consists of the following three steps. The first equality follows from the fact that the Christoffel symbol components ◦ (4.7) Γµ νρ are symmetric in their lower two indices. In the second equality they are replaced by the teleparallel connection coefficients and the contortion using the definition (2.17). In the third equality the second term is omitted since the trace Kµ µν of the contortion over its first two indices vanishes due to antisymmetry in these indices. Finally, in the fourth and last equality the lower two indices of the teleparallel connection coefficients Γµ νρ are switched by introducing a torsion tensor. This can also be obtained directly using where ξ A B are the components of an antisymmetric matrix, ξ (AB) = 0, where the index has been raised with the Minkowski metric. The variation δω Sg of the gravitational action wrt the spin connection, while keeping the tetrad e A µ fixed, then reads [125] ˆ δω Sg = − d4 x e YA Bµ (∂µ ξ A B + ω A Cµ ξ C B − ω C Bµ ξ A C ). (4.8) ◦ Γ νµ = Γ νµ − K νµ = Γ νµ − T µν , After integration by parts, and applying the covariant divergence formula   ◦ ◦ ∂µ (eXµ ) = e∇µ Xµ = e ∂µ Xµ + Γµ νµ Xν , (4.13) the derivatives acting on the tetrad can be combined with the spin connection to form the connection coefficients so that the field equations take the form Wµν = Θµν . (4.12) After transforming the third term in (4.11) by using the antisymmetry of the spin connection, from which it follows Another, more common form of the field equations is obtained by transforming the first index into a spacetime index with the tetrad, while lowering the second index, to obtain Wµν = e A µ gρν WA ρ , (4.11) µ µ µ µ µ (4.16) by taking the trace over the outer indices of the contortion tensor (2.19). Combining the partial derivative and the connection coefficients into a teleparallel covariant derivative, the connection field equation finally reads (4.9) 38 Rep. Prog. Phys. 86 (2023) 026901 0 = −W̃µν = ∇ρ Y[µν]ρ − T σ σρ Y[µν]ρ . Review   ∇µ Y[ρσ]µ + 2Kµ ων r̃[ρσ][ων] + 2Kν ων r̃[ρσ][µω] (4.17) + T ω µν ∇ω r̃[ρσ]µν = 0. Note that by construction, it is antisymmetric in its two indices. Another possibility to impose the properties of the spin connection is by introducing Lagrange multipliers [123, 124]. In this case one adds an additional term to the action, which is given by ˆ SLM :=   d4 x e q̃(AB)µ ωABµ + r̃A Bµν R A Bµν , Expanding the contortion tensor in terms of its definition (2.19), and applying the product rule, this equation becomes ∇µ Y[ρσ]µ − 2T ω ωµ ∇ν r̃[ρσ][µν] − 3∇[ω T ω µν] r̃[ρσ][µν] = 0. (4.25) (4.18) where q̃ABµ and r̃A Bµν are Lagrange multipliers. These equations are obtained by taking a variation wrt these Lagrange multipliers, and then enforcing that the spin connection is flat, Rµ νρσ = 0, and metric-compatible, Qµνρ = 0. Now the spin connection field equation is obtained by the variation ˆ δω SLM = On the last term, one can use the first Bianchi identity (2.16) for a flat connection, which yields 3∇[ω T ω µν] = −3T ω τ [ω T τ µν] = T ω ωτ T τ µν + T ω µτ T τ νω − T ω ντ T τ µω = T ω ωτ T τ µν , (4.26)  d4 x e − 2r̃AB[µν] Dν δωABµ + q̃(AB)µ δωABµ  − YABµ δωABµ , (4.19) where the last two terms cancel each other due to symmetry. This leads to the equation ∇µ Y[ρσ]µ − 2T ω ωµ ∇ν r̃[ρσ][µν] − T ω ωτ T τ µν r̃[ρσ][µν] = 0. (4.27) wrt the full spin connection δω A Bµ . Integration by parts yields ˆ δω SLM =  ◦ d4 x e 2∇ν r̃ρσ[µν] + 2Kρ ων r̃ ωσ[µν]  + 2Kσ ων r̃ρω[µν] + q̃(ρσ)µ − Yρσµ e A ρ e B σ δωABµ Comparing this with the antisymmetric equation (4.22), we see that it matches the covariant derivative of the Lagrange multiplier r̃[ρσ][µν] , contracted with the torsion vector T ω ωτ . Performing the same contraction on the antisymmetric equation (4.22) yields (4.20a) ˆ =  d4 x e 2∇ν r̃ρσ[µν] − 2Kµ ων r̃ρσ[ων] − 2Kν ων r̃ρσ[µω]  + q̃(ρσ)µ − Yρσµ e A ρ e B σ δωABµ , (4.20b)   T ω ωτ 2∇ν r̃[ρσ][τ ν] + T τ µν r̃[ρσ][µν] − Y[ρσ]τ = 0, (4.28) where we have once again used the definition (2.19) of the contortion in terms of the torsion. Taking the sum of the last two equations, we find that the Lagrange multiplier cancels, and the remaining equation reproduces the spin connection equation (4.17). This equation has six independent components, and so it determines the six free components of the spin connection, which are left after fixing the metric compatibility and flatness constraints. The remaining components of equation (4.22) determine the Lagrange multiplier. Hence, we have shown that the method of Lagrange multipliers yields the same field equations as the constrained variation which we used earlier in this section. We finally remark that a priori, there is no relation between the field equations obtained by variation wrt the tetrad (4.6) and the spin connection (4.17). However, this changes if one imposes a number of conditions, as will be shown below. so that the field equations read 2∇ν r̃ρσ[µν] − 2Kµ ων r̃ρσ[ων] − 2Kν ων r̃ρσ[µω] + q̃(ρσ)µ − Yρσµ = 0. (4.24) (4.21) Here the contribution which is symmetric in ρ and σ simply determines the value of the Lagrange multiplier q̃(ρσ)µ , and so it does not contribute to the field equations for the remaining fields. One is thus left with the antisymmetric part 2∇ν r̃[ρσ][µν] − 2Kµ ων r̃[ρσ][ων] − 2Kν ων r̃[ρσ][µω] − Y[ρσ]µ = 0. (4.22) In order to eliminate the remaining Lagrange multipliers from this equation, one may take its divergence with the teleparallel connection ∇µ , and use the relation We first turn our attention to the matter action Sm , which does not depend on the spin connection, following the assumed decomposition (4.1). We demand that the matter action is invariant under infinitesimal and local Lorentz transformations λBA with λ(AB) = 0, which act on the tetrad following the transformation (2.38), and which we assume to act trivially on matter 4.1.2. Local Lorentz invariance of the matter action. 2∇µ ∇ν r̃[ρσ][µν] = 2∇[µ ∇ν] r̃[ρσ]µν = −T ω µν ∇ω r̃[ρσ]µν , (4.23) which follows from the fact that the teleparallel connection has vanishing curvature, but nonvanishing torsion, to obtain the equation 39 Rep. Prog. Phys. 86 (2023) 026901 Review fields, δλ χI = 0. The resulting variation of the matter action is therefore given by ˆ ˆ 4 µ A B δλ Sm = d x e ΘA λB e µ = d4 x e Θµν λµν , (4.29) Taking into account the antisymmetry of λAB , and comparing with the field equations (4.6) and (4.17), we see that local Lorentz invariance imposes that they are related by where and so the antisymmetric part of the tetrad field equation agrees with the spin connection equation [125]. This reflects the fact that the spin connection is a pure gauge DoF, which is not restricted by an independent field equation. The linear dependence between the field equations exactly accounts for this gauge freedom. Hence, in the remainder of this Review we will make use of this fact, and omit deriving the spin connection field equations explicitly, as they are redundant. As for the matter action, we can also obtain an equivalent description of local Lorentz invariance of the gravitational action by demanding that the gravitational action is constructed from the metric (2.21), the torsion (2.27) and its covariant derivative. To see this, note first that the general variation of the metric and the torsion tensor can be obtained from those of the tetrad and the spin connection as λµν = gνρ EA µ e B ρ λBA . W[µν] = W̃µν , (4.30) From the antisymmetry of λµν it thus follows that the antisymmetric part of the energy-momentum tensor vanishes, Θ[µν] = 0. Note that this holds both on-shell and off-shell, i.e. independently of the matter field equations, since these have not entered the derivation. It does, however, depend on the assumption that there is no direct coupling between the matter fields and the spin connection. Given such a coupling, one would obtain a contribution from the spin current [126]. We also remark that the condition of local Lorentz invariance of the matter action is equivalent to demanding that the tetrad enters the matter action only through the metric and its derivatives (which appear through the Christoffel symbols in covariant derivatives of the matter tensor fields χI ). To see this, one uses the symmetry of the energy-momentum tensor to write the variation of the matter action wrt the tetrad as ˆ ˆ 4 µ A δe Sm = d x e ΘA δe µ = d4 x e Θνµ ηAB e B (ν δe A µ) ˆ 1 (4.31) = d4 x e Θµν δgµν , 2 (4.34) δgµν = 2ηAB e A (µ δe B ν) ,  δTρ µν = EA ρ 2δω A B[µ e B ν] + 2D[µ δe A ν] − T σ µν δe A σ . (4.35) Two properties follow immediately from these variations. First, note that for a local Lorentz transformation given by the relations (2.38) and (2.39), both δgµν and δTρ µν vanish. This can easily be seen by direct calculation, which yields which shows that it is fully described by the variation of the metric. Hence, the matter action does not depend on any other DoF arising from the tetrad besides the ones which determine the metric. The converse is obvious: if the tetrad enters into the matter action only through the metric, then local Lorentz invariance follows from the fact δλ gµν = 0 that the metric is invariant under local Lorentz transformations. δλ gµν = 2ηAC e A (µ e B ν) λC B = 2e A µ e B ν λ(AB) = 0, (4.36) as well as h  δλ Tρ µν = EA ρ 2D[µ λBA e B ν] i − 2EC σ D[µ eC ν] λBA e B σ − 2e B [ν Dµ] λBA = 0. (4.37) 4.1.3. Local Lorentz invariance of the gravitational action. Secondly, one finds that the variation of the spin connection can be expressed in terms of the variations of the torsion and the tetrad. This relation can be derived in multiple steps. First, it follows from variation of the tetrad postulate (2.25) that the variation of the spin connection takes the form   δω A Bµ = e A ρ EB ν δΓρ νµ − ∇µ (EC ρ δeC ν ) , (4.38) Next, we turn our attention to the gravitational part Sg of the action. Demanding invariance under local Lorentz transformations, as discussed for the matter action above, we must here also take into account the transformation (2.39) of the spin connection. It then follows that the total induced variation of the action is given by ˆ h  δλ Sg = − d4 x e WA µ λBA e B µ − YA Bµ ∂µ λBA + ω A Cµ λC B i − ω C Bµ λ A C . in terms of the variations of the tetrad and the teleparallel affine connection. The coefficients of the latter can then be decomposed using the relation (2.17), where the disformation vanishes, to obtain (4.32) ◦ δΓρ νµ = δ Γρ νµ + δKρ νµ , After integration by parts, we then find the expression ˆ  ◦ δλ Sg = − d4 x e WA µ e B µ + ∂µ YA Bµ + Γµ νµ YA Bν  − ω C Aµ YC Bµ + ω B Cµ YA Cµ λBA . (4.33) (4.39) where the variation of the Levi-Civita connection coefficients is given by  ◦ ◦ ◦ ◦ 1 δ Γρ νµ = gρσ ∇ν δgσµ + ∇µ δgνσ − ∇σ δgνµ . 2 40 (4.40) Rep. Prog. Phys. 86 (2023) 026901 Review ˆ The variation of the contortion follows from its definition (2.19) and reads δSg = − 1 δKρ νµ = δTρ µν − gρσ gτ (µ δT τ ν)σ + gρσ T(µν) τ δgστ 2 − T σ (µ ρ δgν)σ . (4.41) ˆ =− d4 x e h  d4 x e UA µ δe A µ + Zρ µν EA ρ 2δω A B[µ e B ν] i + 2D[µ δe A ν] − T σ µν δe A σ (4.46a) h Taking the sum of the latter two expressions, one finds that the variation of the teleparallel affine connection is given by UA µ − EA ν Tµ ρσ Zν ρσ − 2EA ρ Kσ ρν Zσ µν  i ◦ + 2EA ρ ∇ν Zρ µν δe A µ + 2EA ρ e B ν Zρ µν δω A Bµ (4.46b) 1 δΓ νµ = gρσ (∇ν δgσµ + ∇µ δgνσ − ∇σ δgνµ ) 2 1 + δTρ µν − gρσ gτ (µ δT τ ν)σ . (4.42) 2 ˆ ρ =− Finally, substituting the metric variation δgµν in terms of the tetrad variation given by the expression (4.35), we have expressed the variation of the teleparallel connection in terms of the variations of the tetrad and torsion. With the help of the relations derived above, it is possible to rewrite the variation in equation (4.3) of the gravitational action in terms of the variations of the tetrad and the torsion. By successive substitution and integration by parts, one finds [125, 127] ˆ    δSg = − d4 x e WA µ δe A µ + Yρ νµ δΓρ νµ − ∇µ (EC ρ δeC ν ) d4 x e h  UA µ − 2EA ρ T σ σν Zρ µν + 2EA ρ ∇ν Zρ µν δe A µ i + 2EA ρ e B ν Zρ µν δω A Bµ , (4.46c) where we have performed integration by parts, and omitted antisymmetrization in the last two indices of Zρ µν , since it is already antisymmetric due to the antisymmetry of the torsion in the variation (4.45). By comparison with the action variation (4.3), one thus finds the relation WA µ := UA µ − 2EA ρ T σ σν Zρ µν + 2EA ρ ∇ν Zρ µν , YA Bµ := EA ρ e B ν (Zρ µν − Zνµ ρ ), (4.47) between the occurring terms, where the antisymmetrization in the last term is due to the fact that the spin connection is, by (4.43a) definition, antisymmetric in its two Lorentz indices. One easily checks that the relations (4.44) and (4.47) are indeed each ( ˆ h other’s inverses. = − d4 x e Wν µ + Tσ σρ (2Yµ (νρ) + Yνρ µ ) Writing the variation of the action in the form shown in equation (4.45), the local Lorentz transformation (4.32) of the ) i h1 i gravitational part of the action can now simply be written as +∇ρ (2Yν [µρ] + Yρµ ν ) EA ν δe A µ + Yµν ρ − Yρ µν δTρ µν 2 ˆ ˆ δλ Sg = − d4 x e UA µ λBA e B µ = d4 x e Uµν λµν , (4.48) (4.43b) ˆ =− d4 x e (UA µ δe A µ + Zρ µν δTρ µν ), and thus it is of the same form as the corresponding local Lorentz transformation (4.29) of the matter action. Hence, we see that the gravitational action is invariant under arbitrary local Lorentz transformations if and only if U[µν] = 0, since λµν is antisymmetric by definition of a local Lorentz transformation. With the relation (4.44), this translates to (4.43c) where the variation terms are given by h UA µ := Wν µ + Tσ σρ (2Yµ (νρ) + Yνρ µ ) i + ∇ρ (2Yν [µρ] + Yρµ ν ) EA ν , 1 Zρ µν := Yµν ρ − Yρ [µν] , 2 0 = U[µν] = W[µν] − T σ σρ Yµν ρ + ∇ρ Yµν ρ = W[µν] − W̃µν , (4.49) (4.44) where we have applied the antisymmetry of the torsion tensor in its last two indices. To derive the converse relation, one starts from the variation of the action in the equivalent form ˆ δSg = − d4 x e (UA µ δe A µ + Zρ µν δTρ µν ). (4.45) and thus reproduces the Lorentz invariance condition (4.34) which we derived from the variation in equation (4.32). Following the same routine as we did for the variation (4.31) of the matter action, the gravitational action is locally Lorentz invariant if and only if its variation can be written as   ˆ 1 µν 4 √ µν ρ δSg = − d x −g U δgµν + Zρ δT µν , (4.50) 2 Inserting the variation (4.35) of the torsion into this variation, one obtains thus also expressing the determinant of the tetrad through that of the metric, which means that the action can be fully 41 Rep. Prog. Phys. 86 (2023) 026901 Review expressed through the metric and the torsion tensor (and their covariant derivatives). Note that this is equivalent to the statement that the gravitational action can be completely expressed in terms of variables which are invariant under local Lorentz transformations [50, 124, 128]. thereby reproducing the field equation (4.17) for the spin connection (see [124] for a detailed derivation). By taking a variation wrt the metric one obtains the equation Uµν + 2∇ρ q̃ρ(µν) + 2q̃σ(µν) Kρ σρ = Θµν , which now contains the Lagrange multiplier q̃ρ(µν) . This must be eliminated by substituting its value from the previously found in equation (4.53), which contains the Lagrange multiplier r̃µ νρσ ; the latter is eliminated using the same procedure as above, finally leading to the field equation The aforementioned considerations regarding the local Lorentz invariance open the possibility for yet another formulation of TG theories, which is known as the Palatini approach [129]. This approach is similar to writing the gravitational action and its variation in the form on equation (4.50), using the metric gµν instead of the tetrad as one of the fundamental field variables. For the second field variable, the teleparallel affine connection with coefficients Γµ νρ is chosen. This allows one to write the variation of the gravitational part of the action in the form 4.1.4. Palatini approach. ˆ δSg = − √ d4 x −g  Uµν − ∇ρ Y(µν)ρ + Y(µν)ρ T τ τ ρ = Θµν , ˆ  1 µν U δgµν + Yµ νρ δΓµ νρ , (4.51) 2 √ (r̃µ νρσ Rµ νρσ + q̃µνρ Qµνρ ) −gd4 x, Another important relation can be derived from the demand that the action is invariant under diffeomorphisms. Infinitesimally, these are generated by vector fields X = Xµ ∂µ , and the induced change on any tensor field is given by its Lie derivative. For the tetrad and spin connection these Lie derivatives are given by the relations (3.120). For the matter action this yields the variation ˆ   δX Sm = d4 x e ΘA µ (Xν ∂ν e A µ + ∂µ Xν e A ν ) + ΩI LX χI . 4.1.5. Energy-momentum conservation. (4.52) into the action, where r̃µ νρσ and q̃µνρ are Lagrange multipliers, which is similar to the corresponding term (4.18) in the tetrad and spin connection formulation. The equations obtained by taking a variation wrt these Lagrange multipliers, which then yields the conditions that the connection represented by Γµ νρ is flat, Rµ νρσ = 0, and metric-compatible, Qµνρ = 0. The remaining field equations are obtained by variation wrt the metric and to the affine connection, essentially following the same procedure as shown in section 4.1.1 using the tetrad and spin connection as fundamental variables. Variation wrt the affine connection, and raising one index for convenience, yields the equation (4.57) One then imposes that the matter field equations ΩI = 0 are satisfied. Hence, it must be emphasized that everything which follows holds only on-shell. Integrating by parts, as well as a few transformations of the resulting terms, lead to ˆ δX Sm = d4 x e ΘA µ (Xν ∂ν e A µ + ∂µ Xν e A ν ) (4.58a) 2∇σ r̃µν[ρσ] − 2r̃µν[ρσ] K τ στ − 2r̃µν[στ ] Kρ στ − 2q̃ ρ(µν) −Y µνρ = 0. (4.56) which now contains only the physical fields. In order to relate this equation to the field equations in the tetrad and spin connection formulation, one substitutes Uµν from the relation (4.44), keeping only its symmetric part, as mandated by the Lorentz invariance condition (4.49), and finally also uses the connection field equation (4.17), to reproduce the tetrad field equation (4.6). Hence, the Palatini approach yields the same field equations as the more common approach we discussed before, albeit using different fundamental fields. while the variation of the matter action is expressed in the form (4.31) through the symmetric energy-momentum tensor Θµν and the metric variation δgµν . Instead of a priori assuming that the connection is flat and metric-compatible, one imposes these conditions by introducing an additional term SLM = (4.55) ˆ = (4.53) h i ◦ d4 x e ΘA µ ∂ν e A µ − Γρ µρ ΘA µ e A ν − ∂µ (ΘA µ e A ν ) Xν (4.58b) As in the tetrad and spin connection formulation, one finds that the symmetric part of this equation only determines the value of the Lagrange multiplier q̃ρ(µν) , so that one only retains the antisymmetric part of the equation in order to derive a field equation for the physical fields. Eliminating also the Lagrange multiplier r̃µ νρσ by taking the covariant divergence of the antisymmetric equation, one derives the connection field equation from which follows the energy-momentum conservation ∇ρ Y[µν]ρ − Y[µν]ρ T τ τ ρ = 0, ∇ν Θ ˆ = ◦  ◦ d4 x e Γρ νµ Θρ µ − Γρ µρ Θν µ − ∂µ Θν µ Xν ˆ =− ◦ (4.54) 42 ◦ d4 x e X ν ∇ µ Θ ν µ , µν = 0, (4.58c) (4.58d) (4.59) Rep. Prog. Phys. 86 (2023) 026901 Review ˆ since the vector field is arbitrary. A crucial step in this derivation is the substitution ◦ 1 ρσ ρ µ µ Γ νµ Θρ = g (∂ν gσµ + ∂µ gνσ − ∂σ gνµ )Θρ 2 1 = Θµσ (∂ν gσµ + ∂µ gνσ − ∂σ gνµ ) 2 − (4.60a) Comparing with the expression (4.10), and using the fact that the spin connection is antisymmetric, we see that this is simply ˆ ˆ − d4 x e W̃AB ωABτ X τ = − d4 x e W[AB] ωABτ X τ , (4.65) (4.60b) 1 = Θµσ ∂ν gσµ 2 (4.60c) = Θµσ ηAB e B µ ∂ν e A σ (4.60d) = Θ A σ ∂ν e A σ , (4.60e)  d4 x e YC Bµ ω C Aµ − YA Cµ ω B Cµ − ∂µ YA Bµ  ◦ − Γρ µρ YA Bµ ω A Bν Xν . (4.64) where we used the local Lorentz invariance, which implies the relation (4.34) between the tetrad and spin connection field equations. Inserting this in the variation (4.61), we thus find the expression ˆ δX Sg = − [µν] which makes explicit use of the symmetry Θ = 0 of the energy-momentum tensor. This shows that the Lorentz invariance of the matter action is a necessary ingredient in deriving the conservation equation (4.59). Note that energymomentum conservation is a consequence of the invariance of the matter action under diffeomorphisms, and that its common form (4.59) is a consequence of coupling to the metric only, which is independent of the choice of the gravitational action. h i d4 x e WA µ (Xν ∂ν e A µ + ∂µ Xν e A ν ) + WAB ωABτ X τ , (4.66) where we omitted the antisymmetrization brackets on the last term, since it is contracted with the antisymmetric spin connection. Once again, we perform integration by parts to obtain ˆ  δX S g = − d4 x e W A µ ∂ν e A µ − W A µ ∂µ e A ν − ∂µ W A µ e A ν  ◦ − Γρ µρ WA µ e A ν + WAB ωABν Xν . (4.67) Also, for the gravitational action Sg the invariance under diffeomorphisms is imposed. In this case the induced variation is given by ˆ h δX Sg = − d4 x e WA µ (Xν ∂ν e A µ + ∂µ Xν e A ν ) i + YA Bµ (Xν ∂ν ω A Bµ + ∂µ Xν ω A Bν ) . 4.1.6. Bianchi identities. Note that the second and third term can be combined by transforming a Lorentz index into a spacetime index, which yields ˆ δX S g = − (4.61) This variation can be decomposed in a number of steps. We start with the second part of this term, which contains the contribution from the spin connection variation YA Bµ . After integration by parts this contribution reads ˆ  − d4 x e YA Bµ ∂ν ω A Bµ − YA Bµ ∂µ ω A Bν − ∂µ YA Bµ ω A Bν  ◦ (4.62) − Γρ µρ YA Bµ ω A Bν Xν ,  ◦ d4 x e WA µ ∂ν e A µ − ∂µ Wν µ − Γρ µρ WA µ e A ν  + WAB ωABν Xν . (4.68) This can also be done with the remaining Lorentz indices, so that one obtains ˆ  ◦ δX Sg = − d4 x e Wρ µ EA ρ ∂ν e A µ − ∂µ Wν µ − Γρ µρ Wν µ  (4.69) + Wρµ e A ρ e B µ ωABν Xν . Now the first and the last term can be combined to yield the teleparallel affine connection ˆ   ◦ δX Sg = − d4 x e Wρ µ Γρ µν − ∂µ Wν µ − Γρ µρ Wν µ Xν . where the last term originates from the derivative of the determinant e of the tetrad. Note that the first two terms constitute an antisymmetric derivative of the spin connection. Using the fact that its curvature (2.26) vanishes, one thus finds the expression ˆ  − d4 x e YA Bµ ω A Cµ ω C Bν − YA Bµ ω A Cν ω C Bµ − ∂µ YA Bµ ω A Bν  ◦ − Γρ µρ YA Bµ ω A Bν Xν . (4.63) (4.70) After splitting the teleparallel connection in the first term into a contortion and Levi-Civita connection ˆ  ◦ δX Sg = − d4 x e Wρ µ Kρ µν + Wρ µ Γρ µν − ∂µ Wν µ  ◦ − Γρ µρ Wν µ Xν , (4.71) After renaming indices, one can extract a common factor to obtain 43 Rep. Prog. Phys. 86 (2023) 026901 Review where the quantities EABµν ρ and TABµν ρσ depend on the choice of the constitutive relation. The full variation of the gravitational part of the action is then expressed as ˆ δSg = (ΥA ∧ δe A + ΠA ∧ δT A ), (4.79) the remaining terms combine into a covariant derivative ˆ δX S g = −   ◦ d4 x e Wρ µ Kρ µν − ∇µ Wν µ Xν , (4.72) using the fact that the Levi-Civita connection is symmetric. The resulting equations ◦ Wρν Kρν µ − ∇ν Wµν = 0, in terms of a three-form and a two-form 1 ΥA := ΥAµνρ dxµ ∧ dxν ∧ dxρ , 6 (4.73) are the Bianchi identities for a general TG theory. Note that although we have used the relation (4.34) between the tetrad and spin connection field equations, which follows from the Lorentz invariance of the gravitational action, we have not imposed that either of these field equations hold. Hence, the Bianchi identities (4.73) hold both on-shell and off-shell, i.e. they are geometric identities. Using the antisymmetry K(µν)ρ = 0 of the contortion, one can decompose them into their symmetric and antisymmetric parts ◦ ∇ν W (µν) = 0, ◦ W[ρν] Kρν µ − ∇ν W[µν] = 0. whose components are given by 1 ΥAµνρ := T B αβ EBAγδ τ ϵ αβγδ ϵτ µνρ = 3T B [µν EBAρτ ] τ , 8 (4.81a) 1 1 1 ΠAµν := HAµν − T B αβ TBAγδ ωτ ϵ αβγδ ϵωτ µν = HAµν 2 16 2 3 B ωτ + T [µν TBAωτ ] . (4.81b) 2 (4.74) This variation is complemented with the variation of the matter action, written as ˆ δSm = ΣA ∧ δe A , (4.82) Together with a corresponding decomposition W(µν) = Θµν , W[µν] = 0, 1 ΠA := ΠAµν dxµ ∧ dxν , 2 (4.80) (4.75) in terms of the energy-momentum three-form of the field equation (4.6), which follows from the symmetry Θ[µν] = 0 of the energy-momentum tensor. We see that inserting the field equations into the Bianchi identity (4.74) yields the energy-momentum conservation (4.59). Hence, we find that the latter is indeed imposed by the field equations. Thus, for any TG theory we can always write its field equations as the symmetric and antisymmetric Wµν (which is the variation of the action wrt the tetrad). This then accounts for both the tetrad and spin connection field equations. 1 ΣA := ΣAµνρ dxµ ∧ dxν ∧ dxρ . 6 (4.83) This approach is related to the preceding discussion through the torsion variation δT A = Dδe A + δω A B ∧ eb , (4.84) from which follows that one can equivalently rewrite the variation (4.79) of the gravitational action as [125, 132] ˆ   δSg = (ΥA − DΠA ) ∧ δe A − ΠA ∧ e B ∧ δω A B . (4.85) To conclude the preceding discussion of the general properties of TG theories, we finally show how these can be found from the premetric point of view [125, 130, 131]. For this purpose, we write the gravitational part Sg of the general action in yet another form 4.1.7. Premetric approach. where we introduced the gravitational excitation two-form By comparing with the variation in equation (4.3), one therefore obtains the identifications   1 1 WA µ = ϵµνρσ Dν ΠAρσ − ΥAνρσ , 2 6 1 (4.86) YA Bµ = ϵµνρσ ΠAνρ e B σ . 2 1 HA = HAµν dxµ ∧ dxν . 2 Similarly, comparing the variations in equations (4.2) and (4.82), one obtains the relation Sg = 1 2 ˆ T A ∧ HA = − 1 8 ˆ d4 x e T A µν HAρσ ϵµνρσ , (4.76) (4.77) 1 ΘA µ = ΣAνρσ ϵµνρσ . 6 Any TG theory can be defined by specifying a constitutive relation, i.e. a functional dependence of HAµν on the tetrad e A µ and the torsion tensor T A µν . It follows that its variation can be expressed in the form 1 δHAµν = EABµν ρ δe B ρ + TABµν ρσ δT B ρσ , 2 (4.87) With these identifications in place, we see that the Lorentz invariance of the matter action, which is manifest in the symmetry of the energy-momentum tensor, takes the form (4.78) Σ[A ∧ eB] = 0, 44 (4.88) Rep. Prog. Phys. 86 (2023) 026901 Review while the Lorentz invariance of the gravitational action implies Υ[A ∧ eB] + Π[A ∧ TB] = 0. We finally remark that in other, curvature-based gravity theories, similar formal analogies with electrodynamics or Yang–Mills theories can be constructed. This is the case, e.g. for MacDowell–Mansouri gravity [134] or Chern–Simons gravity [135]. (4.89) Further, these two relations can be used to derive the field equation   D Π[A ∧ eB] = DΠ[A ∧ eB] + Π[A ∧ TB] = 0, (4.90) 4.2. The ADM formalism A very useful tool to study gravity theories is the ADM formalism, named after Arnowitt, Deser and Misner. Before discussing its application to particular TG theories, we now discuss a few general aspects of its use in this context. In section 4.2.1, we discuss the 3 + 1 decomposition of the teleparallel geometry into space and time components. This leads to the introduction of lapse and shift variables in section 4.2.2. The question of Lorentz gauge fixing is discussed in section 4.2.3. In section 4.2.4, we introduce canonical variables, and discuss their irreducible decomposition in section 4.2.5. which is obtained from variation wrt the spin connection, from the antisymmetric part of the tetrad field equation DΠA − ΥA = ΣA . (4.91) Finally, the covariant energy-momentum conservation takes the form ◦ DΣA = 0, (4.92) while the Bianchi identities for the gravitational sector read ◦ D (DΠA − ΥA ) = 0. In order to perform a Hamiltonian analysis, one usually starts with a split of all dynamical variables in their space and time components. Such a decomposition is done relative to a foliation of spacetime with timelike hypersurfaces, i.e. one identifies the spacetime manifold M, which is assumed to be globally hyperbolic, with a direct product of the form R × Σ. This identification allows essentially two different possibilities to split the dynamical variables into space and time components, which we will write out explicitly for the tetrads and their duals: 4.2.1. 3 + 1 decomposition of teleparallel geometry. (4.93) The main advantage of this premetric approach lies in its formal analogy to electrodynamics and the possibility to write the gravitational field equations in a form which is reminiscent of gauge theories. This becomes apparent by writing the tetrad field equation (4.91) in the form DΠA = ΥA + ΣA . (4.94) The RHS of this equation can be interpreted as the total energy-momentum current, which contains both a gravitational contribution ΥA and a matter contribution ΣA . This total current is covariantly conserved wrt the teleparallel connection, 0 = D 2 ΠA = D(ΥA + ΣA ), (a) In the first approach, one uses coordinates (xµ ) = (x0 = t, xi ), where lowercase Latin letters denote spatial indices following the convention 1.3, which are adapted to the foliation, such that the time coordinate t parametrizes the different spatial slices, while (xi ) are spatial coordinates on each spatial slice Σt ⊂ M. In these coordinates the tetrad naturally decomposes as (4.95) since D 2 = 0 due to the flatness of this connection. Together with the torsion relations De A = T A , DT A = 0, e A = e A µ dxµ = e A 0 dt + e A i dxi , while its inverse decomposes as (4.96) EA = EA µ ∂µ = EA 0 ∂0 + EA i ∂i . one easily recognizes the similarity to the equations dA = F, dF = 0, d ⋆ F = J, dJ = 0, (4.98) (4.99) Notable works using this decomposition are [136–149]. (b) Another approach does not make use of adapted coordinates, but employs the spacetime metric gµν . In this approach, one starts from the unique normalized, future pointing, hypersurface-orthogonal vector field nµ ∂µ , which satisfies (4.97) in Lorentz–Maxwell electrodynamics, where A is the electromagnetic vector potential, F is its field strength and J is the conserved electromagnetic current density. Also here more general theories of electrodynamics, such as effects of media, can be formulated by replacing the excitation ⋆F with a more general constitutive relation [133]. This analogy is particularly visible in the case that the gravitational excitation HA depends linearly on the torsion TA , in which case ΠA = HA [130]. In the non-linear case, this relation becomes more involved; a number of constitutive relations for various modified TG theories are given in [125]. nµ nµ = gµν nµ nν = 1, Xµ nµ = gµν Xµ nν = 0, (4.100) for any vector field Xµ ∂µ which is tangent to the spatial hypersurfaces. The metric then decomposes into the form gµν = nµ nν − hµν , 45 (4.101) Rep. Prog. Phys. 86 (2023) 026901 Review where n A = e A µ nµ = e A ⊥ is the expression of the unit normal vector field in the tetrad basis. A crucial insight in the transition to the new variables is the fact that the latter can be fully expressed in terms of the spatial tetrad components e A i , without using the time components e A 0 . To see this, note first that the condition that nµ is normal to the spatial hypersurfaces can be expressed as where hµν restricts to a Riemannian metric on every spatial hypersurface and nµ hµν = 0. Raising one index, one obtains two projectors δνµ = (P⊥ )µν + (P∥ )µν , (P⊥ )µν = nµ nν , (P∥ )µν = −hµν , (4.102) where it is convenient to write the indices directly on top of each other, as for the Kronecker symbol δνµ . These projectors can be applied to any covariant or contravariant index 0 = g(n, ∂i ) = nµ gµi = n A ηAB e B i = nA e A i = ni , from which likewise follows Vµ = (P⊥ )µν Vν + (P∥ )µν Vν = nµ nν Vν − hµν Vν , (4.103a) Vµ = (P⊥ )νµ Vν + (P∥ )νµ Vν = nµ nν Vν − hνµ Vν . n0 = nA e A 0 = nA (Nn A + Ni e A i ) = N. nµ Vµ = V⊥ , The second condition is the normalization, which in the tetrad basis simply reads 1 = n A nA = ηAB n A n B , −hµν Vν = Vµ̄ , (4.104) 1 n A = ± ϵ A BCD εijk e B i eC j eD k , 6 (4.106) Notable works using this decomposition are [150–154]. Finally, one can analogously decompose also the teleparallel spin connection ω A Bµ . However, this step can be avoided by working in the Weitzenböck gauge, which is always possible, as we shall see below. Before performing the Hamiltonian analysis, it turns out to be more convenient to introduce a different set of variables in order to replace the temporal tetrad components e A 0 or e A ⊥ introduced in the 3 + 1 decomposition. For this purpose one writes the components of the unit normal vector field in the form 4.2.2. Lapse and shift variables. n0 = 1 , N ni = − Ni , N (4.107) introducing the lapse N and the components N i of the shift vector field. They satisfy the relation ∂0 = Nnµ ∂µ + Ni ∂i , (4.108) and therefore relate the two different possibilities to split tensor components in the previous section. Using these variables, it is possible to express the temporal tetrad components in the form e A 0 = Nn A + Ni e A i , (4.113) where εijk denotes the totally antisymmetric tensor field defined by the induced metric hij = ηAB e A i e B j on the spatial hypersurfaces. The sign is finally fixed by the demand that the normal vector field is future pointing, which means positive lapse, N > 0, following the decomposition (4.107). This can be conducted by realizing that the four Lorentz vectors (n A , e A i ) with i = 1, 2, 3 are linearly independent; this follows from the fact that only if the e A i are linearly independent, the metric hij is non-degenerate, while the orthogonality condition (4.110) implies that also nA is linearly independent. Hence, these four Lorentz vectors constitute a basis of Minkowski space. Assuming that the three spatial tetrad components e A i are related to an oriented coordinate basis ∂i of the spatial hypersurfaces, then translates the condition that nA is future pointing into a condition on the orientation of the basis (n A , e A i ). This basis is oriented if the determinant (4.105) while its dual decomposes as EA µ = nµ EA ⊥ + EA µ̄ . (4.112) and so involves only the Minkowski metric. These two conditions uniquely determine nA up to a sign difference, and leads to the expression [141] where a bar on top of an index is a shorthand notation to indicate that the corresponding index has been contracted with a spatial projector. Hence, for the tetrad one has the decomposition e A µ = nµ e A ⊥ + e A µ̄ , (4.111) (4.103b) A common notation for the obtained projected components is given by nµ Vµ = V⊥ , −hµν Vν = Vµ̄ , (4.110) (4.109) 46 e = ϵABCD e A 0 e B 1 eC 2 eD 3 (4.114a) = NϵABCD n A e B 1 eC 2 eD 3 (4.114b) 1 = ± NϵABCD ϵ A EFG εijk e B 1 eC 2 eD 3 eE i eF j eG k 6 (4.114c) = ∓Nεijk hi1 hj2 hk3 (4.114d) = ∓Nε123 (4.114e) √ = ∓N h, (4.114f ) Rep. Prog. Phys. 86 (2023) 026901 Review is positive, and so the lower sign must be chosen. This determines the normal vector to be given by 1 n A = − ϵ A BCD εijk e B i eC j eD k . 6 (a) Using Lagrange multipliers: As discussed in section 4.1.1, one possibility to include the spin connection into the teleparallel action and enforce its flatness and compatibility with the metric is by introducing Lagrange multipliers, which impose the conditions of vanishing curvature and non-metricity. In the Hamiltonian formalism, these lead to constraints on the spin connection, thus implementing its flatness and metric compatibility. Further primary constraints arise from the Lorentz gauge freedom, linking the spin connection and the tetrad through their common change under Lorentz transformations, which leave the action invariant. Finally, also for the Lagrange multipliers, constraints arise which implement the condition that their canonical momenta vanish identically. This large number of variables and associated constraints therefore makes this approach rather cumbersome. It is used in [136, 151, 152]. (b) Integrating the spin connection: Another approach to treat the spin connection, which does not necessitate the use of Lagrange multipliers, arises from the fact that locally it may be integrated in the form (2.35) and thus expressed by a local Lorentz transformation Λ A B . Instead of ω A Bµ , one thus considers Λ A B as the fundamental field variable which implements the Lorentz gauge freedom. As a consequence, the derived connection is flat and metric-compatible by construction, and therefore does not have to be enforced by constraints. Together with the absence of Lagrange multipliers, the number of variables and associated constraints is therefore significantly reduced compared to the aforementioned approach. Nevertheless, constraints still appear from the gauge invariance of the teleparallel action under local Lorentz transformations, and these constraints intertwine the conjugate momenta of the tetrad and spin connection DoF. However, this approach is not common for the Hamiltonian analysis. (c) Fixing the gauge: Finally, it is also possible to eliminate the spin connection completely prior to performing the Hamiltonian analysis. This can be achieved starting from the preceding approach which integrates the spin connection locally to a variable Λ A B , which constitutes a local Lorentz transformation, by making a change of variables according to (4.115) Together with the relation (4.109), it is thus possible to express the time component e A 0 of the tetrad through its spatial components, the lapse and the shift. In summary, this means that the 16 components e A µ of the tetrad are equivalently expressed by the 16 variables Ni , N, e Ai. (4.116) This is the ADM decomposition of the tetrad in TG. With its help, the inverse tetrad is written as EA 0 = 1 nA , N EA i = −ηAB hij e B j − Ni nA , N (4.117) where hij denotes the inverse of the spatial metric hij = −ηAB e A i e B j . Further, for the metric this yields the usual ADM decomposition g00 = N 2 − Ni Nj hij , g0i = −Nj hij , gij = −hij , (4.118) using the fact that ni = 0, together with the inverse metric g00 = 1 , N2 g0i = − Ni , N2 gij = Ni Nj − hij . N2 (4.119) Transforming the indices of the metric to Lorentz indices with the help of the tetrad, this yields the helpful relations ηAB = nA nB − ηAC ηBD hij eC i eD j , (4.120) η AB = n A n B − hij e A i e B j . (4.121) and These allow us to express the orthogonal and parallel projectors (4.102) as (P⊥ )BA = n A nB , (P∥ )BA = −ηBC hij e A i eC j , ẽ A µ = (Λ−1 ) A B e B µ , (4.122) Λ̃ A B = Λ A B , (4.123) which is equivalent to the change of ADM variables given by in the tetrad basis, which we will use later for the decomposition of further quantities into the spatial and normal components. Ñ = N, Ñi = Ni , ẽ A i = (Λ−1 ) A B e B i , Λ̃ A B = Λ A B . (4.124) One finds that the new variable ẽ A µ simply represents the Weitzenböck tetrad which is associated to the original tetrad e A µ . Since the Weitzenböck tetrad is uniquely determined from the original tetrad (up to a global Lorentz transformation) by the condition that Λ A B relates the original spin connection ω A Bµ to the vanishing Weitzenböck spin connection, it follows that the new variable ẽ A B is invariant As discussed in section 4.1, the covariant formulation of TG generically involves a Lorentz gauge freedom, which is implemented via the flat, metriccompatible spin connection ω A Bµ . Since the latter represents a pure gauge DoF, the question arises how to treat this variable in the Hamiltonian formalism. The treatment of the spin connection can essentially be classified into three approaches: 4.2.3. Lorentz gauge fixing. 47 Rep. Prog. Phys. 86 (2023) 026901 Review under any further local Lorentz transformation Λ̂ A B , which acts on the original variables as e ′A µ = Λ̂ A B e B µ , Λ ′A B = Λ̂ A C ΛC B . of the metric and teleparallel affine connection through its torsion, as we have shown in section 4.1.3, it follows that the action depends only on the gauge-invariant variable ẽ A µ , while the pure Lorentz gauge variable Λ̃ A B does not contribute to the action. Its canonical momenta therefore vanish identically, and it does not need to be considered further in the Hamiltonian approach, so that one is left with working with ẽ A µ as the only teleparallel field variable. Note that this approach is equivalent to imposing the Weitzenböck gauge ω A Bµ ≡ 0 from the beginning [148, 155], and thus to working in the non-covariant, pure tetrad formulation of TG. This is employed in [137–150, 153, 154]. (4.125) Note that here the Lorentz transformation Λ̂ A B is applied only to the first index of the variable Λ A B , but not to the second. This is due to the fact that the first index of Λ A B represents the variable, non-Weitzenböck gauge, in which the tetrad is given by e A µ , while the second index refers to the fixed Weitzenböck gauge, from which the spin connection ω A Bµ is obtained through the transformation Λ A B . This can also be seen by explicitly calculating the transformation ω ′A Bµ = Λ ′A C ∂µ (Λ ′−1 )C B h i = Λ̂ A D ΛD C ∂µ (Λ−1 )C E (Λ̂−1 )E B In the following, we will follow the latter, gauge-fixed approach, and use the Weitzenböck tetrad ẽ A µ and its ADM decomposition as the only teleparallel field variable. For convenience, we will drop the tilde on all variables, hence assuming the Weitzenböck gauge from the beginning. (4.126a) (4.126b) h i = Λ̂ A D (Λ̂−1 )E B ΛD C ∂µ (Λ−1 )C E + ∂µ (Λ̂−1 )D B In the following we will assume that the gravitational dynamics is obtained from a teleparallel action of the form (4.1), where we further assume that the gravitational action Sg is of first derivative order in the gravitational field variables. From the Lorentz invariance of the gravitational action it then follows that it is constructed from the metric gµν and the torsion tensor Tµ νρ only, while derivatives thereof are excluded, as they would introduce at least second order derivatives of the tetrad. Hence, the action depends only on the tetrads e A µ and their first order derivatives ∂µ e A ν , where the latter enter only through the torsion. Splitting these torsion into space and time components, 4.2.4. Canonical variables. (4.126c) = Λ̂ A D (Λ̂−1 )E B ω D Eµ + Λ̂ A D ∂µ (Λ̂−1 )D B , (4.126d) which is the behavior (2.34) of a spin connection under local Lorentz transformations. From the change of variables (4.123) one can see that these transformations cancel each other for ẽ A µ , as can be seen by explicitly calculating ẽ ′A µ = (Λ ′−1 ) A B e ′B µ g = (Λ−1 ) A B (Λ̂−1 ) B C Λ̂C D eD µ g = (Λ−1 ) A B e B µ g = ẽ A µ . (4.127) Hence, Lorentz transformations act exclusively on Λ̃ A B , so that the new variables (ẽ A µ , Λ̃ A B ) provide a split into a gauge-invariant and a pure gauge variable. Further, one finds that the metric and teleparallel affine connection, which enter TG actions, are expressed in the new variables as T A 00 =0, T A ij =∂i e A j − ∂j e A i , e A0, (4.129a)   = (Λ−1 )D A ẼD µ ∂ρ (Λ A B ẽ B ν ) + Λ A B ∂ρ (Λ−1 ) B C ΛC E ẽE ν (4.129b) = (Λ−1 )D A ẼD N, ∂0 e A i , ∂i e A 0 , ∂i e A j , (4.131) Ni , e Ai, ∂0 e A i , ∂i N, ∂ i Nj , ∂i e A j . (4.132) The time derivatives (or velocities) of the spatial tetrad components are conventionally denoted  E Λ A B ∂ρ ẽ B ν + ẽ B ν ∂ρ Λ A B − ∂ρ Λ A E ẽ ν (4.129c) = ẼA µ ∂ρ ẽ A ν , e Ai, while the time derivative ∂0 e A 0 does not contribute. Equivalently, one may use the relation (4.109) to replace e A i by the ADM variables (4.116), so that the action is expressed by the fields as well as  µ (4.130) it follows from the fact that the torsion is antisymmetric in its last two indices, that the only terms which enter the action are gµν = ηAB e A µ e B ν = ηAB Λ A C Λ B D ẽC µ ẽD ν = ηAB ẽ A µ ẽ B ν , (4.128)   Γµ νρ = EA µ ∂ρ e A ν + Λ A B ∂ρ (Λ−1 ) B C eC ν T A 0i = −T A i0 = ∂0 e A i − ∂i e A 0 , v A i = ė A i = ∂0 e A i . (4.133) Variation of the gravitational action Sg wrt the velocities defines their conjugate momenta πA i as ˆ δė Sg = d4 x πA i δ ė A i . (4.134) (4.129d) where the latter confirms the previous statement that ẽ A µ represents the tetrad in the Weitzenböck gauge. Since any Lorentz invariant TG action can fully be expressed in terms 48 Rep. Prog. Phys. 86 (2023) 026901 Review The virtue of this decomposition becomes apparent, for example, in calculating the product πA i v A i , which enters the Hamiltonian. It follows from the irreducibility of the decomposition above that this term reduces to Note that by definition, they are tensor densities, i.e. they include the density factor e given by the determinant of the tetrad. It follows that also the Hamiltonian H = πA i v A i − L, (4.135) V T decomposition. As argued in section 4.2.4, varying the action of a generic TG theory wrt the velocities ė A i leads to the canonical momenta πA i , which are conjugate to the spatial tetrad components e A i . These momenta are obtained as functions which depend on the teleparallel field variables and their time derivatives. In order to proceed with the Hamiltonian analysis, this relation needs to be inverted and the velocities determined as functions of the momenta. This task can become rather involved, since the relation between momenta is, in general, non-trivial. However, it can be simplified by choosing a decomposition of the momenta and velocities into irreducible components under the rotation group acting on the spatial hypersurfaces. For the momenta, one defines the objects A i π = n πA , Aij π = e A k hk[i πA j] , 1 S π ij = e A k hk(i πA j) − hij e A k πA k , 3 1 T π = e A i πA i , 3 = ηBC e A k hk(i πA j) eC (i v B j) . (4.136a) (4.136b) Aij A S S TT since cross-terms arising from the contraction of symmetric and antisymmetric tensors cancel, and we obtain the parallel projector (4.122). Finally, the remaining term is now simply the orthogonal part (4.136d) V T π i v i = n A nB πA i v B i = (P⊥ )BA πA i v B i , (4.143) so that their sum indeed yields πA i v B i . An alternative approach to this result is by introducing the notation π A i = nA π i , V V A i A π A = ηAB e B j π ij , T T S S π A i = ηAB e B j π ij , π A i = ηAB e B j hij π (4.144) and (4.138a) A v ij = ηAB e B [i v A j] , (4.138b) 1 v ij = ηAB e B (i v A j) − ηAB hij hkl e B l v A k , 3 (4.138c) 1 v = ηAB hij hkl e B l v A k , 3 (4.141c) π v ij + π ij v ij + 3π v = ηBC hki e A k eC i πA j v B j = (P∥ )BA πA i v B i , (4.142) (4.136c) v i = nA v A i , T (4.140) Together with the second term in the decomposition (4.140) this yields A similar decomposition is introduced for the velocities v A i = ė A i , and which can be defined as S TT 2 = ηBC e A k hk(i πA j) eC (i v B j) + ηBC hjk e A i eC k πA i v B j 3 1 1 − ηBC hij e A k πA k eC (i v B j) − ηBC hij hlm eC m v B l e A k hk(i πA j) 3 3 (4.141b) where the letters V, A, S, T are chosen to indicate the vector, antisymmetric, symmetric trace-free and trace components. From these irreducible components, the original components of the momenta are recovered as h V A  i S T πA i = ηAB n B π i + e B j π ij + π ij + e B j hij π . (4.137) T S S which can be seen explicitly as follows. First, note that the last two terms in the decomposition (4.140) combine as   1 TT S S π ij v ij + 3π v = e A k hk(i πA j) − hij e A k πA k 3   1 × ηBC eC (i v B j) − ηBC hij hlm eC m v B l 3 1 + ηBC hjk e A i eC k πA i v B j (4.141a) 3 4.2.5. Irreducible Vi A A πA i v A i = π i v i + π ij v ij + π ij v ij + 3π v, like the Lagrangian L, is a scalar density. TA T TA T v i = n A v i, A AA v i = e A k hkj v ij , S SA v i = e A k hkj v ij , v i = e A i v, (4.145) for the terms in the decompositions (4.137) and (4.139). Using this notation, the decomposition (4.140) equivalently reads V T A A S S T T πA i v A i = π A i v A i + π A i v A i + π A i v A i + π A i v A i , (4.138d) (4.146) which is proven in a similar way. Another advantage of the irreducible decomposition is that the different components in the relations of the momenta and velocities generically decouple, which simplifies inverting these relations. This will be shown explicitly when we discuss the application of the ADM formalism to specific TG theories. and from which the original velocity components are recovered as   S T A kj A A AT v ij + v ij + e A i v. v i = n v i + e kh (4.139) 49 Rep. Prog. Phys. 86 (2023) 026901 Review that ultimately produces the Ricci scalar associated with standard gravity resulting in 4.3. The teleparallel equivalent of GR In the context and circumstances of the points raised in this section thus far, a gravitational analogue to GR can be written down that is dynamically equivalent. It was Einstein who first proposed the TEGR [156], which was initially part of his attempt to unify gravitational and electromagnetic interactions. An attempt that ultimately failed but which produced an independent theoretical formalism of gravitation that was dynamically equivalent to GR. In practical terms, this means that the field equations and so the classical predictions of the theory, would be equivalent. However, given the differences in their action formulations, these theories are not bound to the same fate when quantum considerations are taken into account, among other areas where it is the action rather than the dynamical equations that play the more important role. ◦ 0 = R = R + P, where R = gλµ Rλµ , and  2 P = gλµ Pρλρµ = ∂ρ eTµρµ + Kρσµ Kµσρ − Kρσρ Kµσµ , e (4.154) where the following identities were used Kµρµ = Tµµρ ,  2 2 B := ∂ρ (eTµ µ ρ ) ≡ − ∂ρ eTµρµ , e e In GR, the Einstein–Hilbert action is ultimately dependent on a single dynamical variable, namely the metric tensor [2]. This occurs through the metric tensor itself and the Riemann tensor expressed through ◦ρ ◦ρ ◦ ◦ ◦ which is the original definition of the torsion scalar, namely which gives a measure of the curvature associated with the Levi-Civita connection. If this is exchanged with the teleparallel connection, the resulting Riemann tensor 1 1 T = Tρσµ Tρσµ + Tµσρ Tρσµ − Tρρσ Tµµσ . 4 2 ◦ R = −P = −T + B, will identically vanish due to the connection being curvatureless property. Using equation (2.17) the two connections can be related together through the contortion tensor Kρµν with the relation (4.149) (4.150) where  ◦ρ ◦ρ ◦ρ ρ ρ σ σ R λνµ = − K λµ,ν − K λν,µ + Γ σν K λµ − Γ σµ K λν ◦ ◦ + Γσλµ Kρσν − Γσλν Kρσµ + Kρσν Kσλµ − Kρσµ Kσλν =: −Pρλνµ .  (4.151) It is important to note that the Pρλνµ tensor is expressed only in terms of the teleparallel connection. Naturally, this will lead to a relation for the Ricci tensor ◦ 0 ≡ Rλµ = Rρλρµ = Rρλρµ + Pρλρµ , (4.159) which guarantees the dynamical equivalence between the Einstein–Hilbert and TEGR action (2.52) constructed from the torsion scalar. In this way, GR and TEGR must produce identical field equations, which may only be distinguishable in appearance rather than in any dynamical features. The reason a divergence term arises is because both the GR and TEGR Lagrangians depend on the first and second derivatives of the metric and tetrad respectively, however, the second derivative elements in the tetrad instance simplify to a total divergence term. In fact, the boundary term B is related to the origin of the Lovelock theorem in that it forces even minor modifications of the Einstein–Hilbert action to produce high order field equations. Thus, using the tetrad as the fundamental variable of the theory, instead of the metric, gives the possibility of obtaining second order field equations using more general actions from a local Lagrangian, in four dimensions, which is consistent with the Lovelock theorem. On the boundary term, its definition has a noteworthy feature which is based on a well known theorem that related partial and covariant derivatives using the metric tensor determinant, namely which means that the two forms of the Riemann tensor can be related through [36] ◦ (4.158) Hence, the Ricci scalar differs by a boundary term wrt the torsion scalar and then, it can be written as Rρλνµ = Γρλµ,ν − Γρλν,µ + Γρσν Γσλµ − Γρσµ Γσλν ≡ 0, (4.148) 0 ≡ Rρλνµ = Rρλνµ + Pρλνµ , (4.156) 1 1 Kρσµ Kµσρ − Kρσρ Kµσµ = Tρσµ Tρσµ + Tµσρ Tρσµ 4 2 (4.157) − Tρρσ Tµµσ , ◦ ◦ (4.155) while the remainder can be simplified to R λνµ = Γ λµ,ν − Γ λν,µ + Γρσν Γσλµ − Γρσµ Γσλν , (4.147) Γρµν = Γρµν + Kρµν , T σµν = −T σνµ . The first term in P is a total divergence term written as 4.3.1. The TEGR action and field equations. ◦ρ (4.153)  ◦ √ 1 µ µ ∇µ A = √ ∂µ −gA , −g (4.152) 50 (4.160) Rep. Prog. Phys. 86 (2023) 026901 Review eAβ gµα ΘAµ . These field equations organically admit all solutions that appear in GR since they are dynamically equivalent to each other. This can be seen after noticing that the Einstein tensor can be written in term of teleparallel quantities as follows  ◦ 1 Gαβ = − TB νβ SB ν α − ω B βν SB ν α − gµα eA β ∂ν (eSA µν ) e  T (4.168) − gαβ . 2 where Aµ is an arbitrary tensor. Using the tetrad determinant relation in equation (2.54), this can be directly used to rewrite the boundary term definition [157, 158] in the following way  ◦ B = 2∇µ Tρρµ , (4.161) from which the total divergence nature of the term becomes much more evident. In terms of making calculations, this expression of the boundary term can be useful in determining the value of the quantity. However, we highlight that despite the calculation relying on the Levi-Civita connection, it does not mean that this is not a genuine torsion term, its just an algebraic way to re-express the term in this way. On the other hand, an interesting property of the torsion scalar is that it can be written in a so-called superpotential format in which [159] 1 T = Sρσµ Tρσµ , 2 Also, these equations are invariant under local Lorentz transformations and diffeomorphisms since each of the terms in the torsion scalar in equation (1.6d) are invariant under these symmetries. The tetrad and spin connection represent independent DoF and thus are determined through extra equations. In the case of TEGR the spin connection field equations are found to be identically satisfied [43] which is not the case for more general theories as will be explored in section 5. This is consistent with the spin connection implementation in TEGR where it represents DoF associated with the Lorentz group and so cannot contribute to the number of dynamical equations for an arbitrary gravitational system [123, 151, 160]. It is interesting to note that these field equations can be reformulated as a current equation using the gauge current JAµ (see section 4.3.2 for more details), which is derived in appendix A.2, namely [161, 162] (4.162) where the superpotential takes on the form Sρσµ = Kσµρ − δρσ Tν ν µ + δρµ Tν ν σ , (4.163) which observes the anti-symmetry Sρσµ = −Sρµσ . Consider the following action for TEGR in which the matter contribution is also included (to the gravitational component in equation (2.52)) and the boundary term is already removed since it does not contribute dynamically ˆ ˆ 1 STEGR := − 2 d4 xe T + d4 xeLm , (4.164) 2κ 1 e−1 ∂σ (eSAµσ ) − JAµ = κ2 ΘAµ , 2 where in this case JAµ becomes the Noether energymomentum density of gravitation [28, 161, 163]. Using this formulation of the field equations [164], the vacuum field equations can be written with the gauge current as a source term e ∂σ (eSAµσ ) = JAµ , (4.170) 2 where κ2 = 8πG as defined in equation (1.7), and the negative sign for the gravitational sector appears so that the exact same dynamics are obtained when compared with the Einstein– Hilbert action. Using the variation identities in appendix A, and the energy-momentum definition in equation (1.10), we can write the TEGR field equations as so that the gravitational field can take on a Noether gauge current structure. Also, the anti-symmetry of the superpotential gives rise directly to the conservation equation e  ∂µ JAµ + eΘAµ = 0, (4.171) 2 1 WAµ = e−1 ∂σ (eSAµσ ) − T σν A Sσνµ + EAµ T + ω BAν SBνµ 2 = κ2 ΘAµ , (4.165) after some lengthy simplification. By contracting with eAβ gµα , we can rewrite the field equations as A e β gµα e −1 ∂σ (eSA µσ which can be shown to produce the TEGR field equations. The motivation for the particular choice of TEGR action (4.164) was to produce a GR equivalent theory using torsional terms. However, we can also arrive at this point using a perspectives from gauge theory as our motivation. As discussed in section 3.1, TG can be written as the gauge theory of translations and the field strength is the torsion tensor. As with any gauge theory, the action is constructed from the trace of the torsion tensor squared, and since there are three possible contractions of the torsion tensor, it is natural to consider the action [36] ˆ   1 STEGR = 2 tr T ∧ ⋆T , (4.172) 2κ 1 ) − T σνβ Sσν α + gαβ T + ω Bβν Sβ ν α 2 = κ2 Θαβ . (4.166) The field equation components satisfy both symmetric and antisymmetric equations W(µν) = Θµν , and W[µν] ≡ 0, (4.169) (4.167) where we stress that the antisymmetric equations W[µν] for TEGR are actually identically satisfied W[µν] ≡ 0 due to the Bianchi identities (see section 4.1.6), and where the energymomentum tensor is expressed as Θαβ = Θβα = eAβ ΘAα = 51 Rep. Prog. Phys. 86 (2023) 026901 Review where T = (1/2)TA µν PA dxµ ∧ dxν is the torsion 2-field and PA = ∂A is the translation generators (see section 3.1). The last term appearing in the integral is the dual 2-form of the torsion tensor, i.e. ⋆T = (1/2) ⋆ TA µν PA dxµ ∧ dxν with ⋆TA µν ≡ (e/2)ϵµναβ SAαβ . This is in analogy to Yang–Mills theory where a gauge theory of the non-abelian symmetry group is considered and its action is constructed from the trace of the field strength. It turns out that after some identities, one can show that the above action is identical to the TEGR action (4.164). This means that even without the a priori knowledge of GR, one can still formulate TEGR using information coming from gauge theory and produce the dynamics of GR using the standard formulation of gauge theory. term in the TEGR field equation (4.174) to form a Fock– Ivanenko derivative, ∂ν (eSA µν ) − ω B Aν eSA µν = Dν (eSA µν ). This allows for us to write the TEGR field equations in the form Dν (eSA µν ) − κ2 e tA µ = κ2 eΘA µ . tA eWA Here the two terms on the LHS are given by JA µ := 1 ∂L 1 = 2 EA λ SB νµ T B νλ κ2 e ∂e A µ κ  1 µ B µν − EA T + ω A ν SB . 2 (4.176) (4.181) 1 B ω Aν SB µν , κ2 (4.183) The number and nature of the DoF of any field theory is a core question by which to assess its consistency as well as the wellposedness of the Cauchy problem13 . These are central to our understanding of any formulation of gravity since hidden DoF can infiltrate every numerical problem by introducing some uncertainty in the necessary number of initial conditions to determine a system. In field theories in which higher-order field equations appear, Ostrogradsky instability is well known to lead to ghost-like DoF [168] which has created a favorable environment for second-order formulations of gravity. However, even in second-order theories of gravity, its important to identify whether DoF are propagating or non-propagating since the latter would not have an effect on observations. This is not a trivial task and remains an open question in many formulations of gravity. 4.3.3. The DoF of TEGR. One finds that the expression JA µ formally appears in analogy to the matter energy-momentum tensor ΘA µ . However, there is a fundamental difference between these two objects. It is well known that ΘA µ is a tensor, and as such can only satisfy a covariant conservation equation ◦ EA SB  1 µ T νλ − EA T , 2 νµ B is attributed to inertial effects. However, as can be seen from the expression on the RHS, this term is not invariant under local Lorentz transformations, and even vanishes in the Weitzenböck gauge. It follows from this lack of Lorentz invariance that it cannot be an observable quantity in a theory in which Lorentz invariance is preserved, so that all physically observable quantities must also be invariant under local Lorentz transformations; hence, and the physical relevance of the split into iA µ and JA µ is questionable [36, 123, 166]. and ◦ λ iA µ = JA µ − tA µ = (4.174) ∂L 1 = Kµν A − EA µ Tρ ρν + EA ν Tρ ρµ , e ∂(∂ν e A µ ) (4.175)  where also in this case the first term vanishes since SA µν is antisymmetric it its last two indices, while the Fock–Ivanenko derivatives commute due to the flatness of the teleparallel connection. It is common to interpret tA µ as the energy-momentum tensor of gravity, while the difference (4.173) SA µν = 1 = 2 κ Dµ Dν (eSA µν ) − κ2 Dµ (etA µ ) = −κ2 Dµ (etA µ ) = 0, (4.182) of a first order Lagrangian, is the fact that they can be written in the so-called potential form ∂ν (eSA µν ) − κ2 eJA µ = κ2 eΘA µ . µ is a tensor. In the absence of matter, it satisfies the covariant conservation equation A direct consequence of the fact that the field equation (4.165), being a special case of the general field equation (4.4) arise as the Euler– Lagrange equations arising from the action (4.164) ∂L ∂L δLm − = ∂ν = A = eΘA µ , ∂(∂ν e A µ ) ∂e A µ δe µ (4.180) In this equation, 4.3.2. The gravitational energy problem. µ (4.179) ◦ ν µ B µ µ µ Dµ (eΘA ) = e(∂µ ΘA − ω Aµ ΘB + Γ νµ ΘA ) = 0. (4.177) The quantity JA µ , in contrast, is not a tensor, which can already be seen from the explicit appearance of the spin connection ω A Bµ in its definition (4.176). It follows from the antisymmetry SA (µν) = 0 of the superpotential that in the absence of matter it satisfies the conservation equation [165] ∂µ ∂ν (eSA µν ) − κ2 ∂µ (eJA µ ) = −κ2 ∂µ (eJA µ ) = 0. (4.178) 13 The Cauchy problem is related to the existence and uniqueness of solutions In order to compensate for this difference, note that the last term in the expression (4.176) can be combined with the first to the field equations given boundary conditions (for more information see [167]). 52 Rep. Prog. Phys. 86 (2023) 026901 Review The most popular way to count the number of DoF in TEGR is to use a Hamiltonian analysis approach (see section 4.2) which gives the total number of localized DoF associated with a field theory [59]. In this setup, the DoF are associated with the pairs of conjugate variables, namely the dynamical system generalized coordinates and their conjugate momenta. The total number of DoF can then be calculated by finding the number of irreducible first-class constraints a and second-class constraints b and using the formula Number of DoFs = 2N − 2a − b , 2 years. Given the expanded framework of alternative formulations of GR and modified theories of gravity, the amount of work in the literature has drastically increased with several promising avenues of research. The added contribution of possible detections through experimental and observational advances in recent years has also highlighted the vital nature of this subject and brought to the fore questions of quantum gravity properties in many reformulations of gravitation. Ultimately quantum gravity phenomena may be expressed in a variety of possibles ways. The recent detection of GW makes the likelihood of astrophysical or cosmological signatures of quantum gravity a real possibility with the next generation of GW observatories [171]. Another crucial outlet for quantum phenomena in a gravitational context is through gamma rays which may reveal some of the effects of quantum gravity corrections within classical theories of gravity such as energy-dependent dispersion relations [172]. Quantum gravity could also have a strong impact on our measurements of beyond the standard model of particle physics which has mostly been directed at neutrino physics in the realm of gravitation [173]. TEGR has a number of attractive properties which make it might be more amenable to a quantization procedure. For instance, the strong equivalence principle has been shown not to be a fundamental requirement of the theory (as explained further in section 3) unlike in GR where the theory necessitates the principle a priori. The strong equivalence principle continues to satisfy all observational constraints [174], while on the other hand, the fundamental nonlocality of quantum theory may eventually require the violation of this principle where in TG may offer an alternative avenue that largely preserve the dynamics of GR. Another crucial aspect to quantizing gravity is the fact that the spin connection is solely related to the local Lorentz transformation while the teleparallel connection includes no inertial effects and so represents a genuine gravitational connection [175, 176]. In GR the connections share these properties and so pose a serious issue for such procedures. TG offers a scenario that is more akin to the regular field theory approach to quantum gravity. While not enough in and of itself, these properties may help in the formation of a quantum theory of gravity in TEGR where some of the core issues of GR are eliminated ab initio. TEGR also has a number of other properties which makes it more malleable for a quantum setting such as its likeness to Yang–Mills theories [177] giving TG a strong similarity to a particle physics theory. Another notable property is that TG naturally has a GibbonsHawking-York boundary term embedded in its action giving the TEGR Hamiltonian a more well-defined expression [178]. Despite all this further work is needed on both the theoretical development in terms of quantum predictions from TEGR [179] as well as more information on the what observational signatures to expect from a quantum theory of gravity. Another interesting proposed route for a quantum TEGR approach comes from [180], where the Poincaré group is replaced with the de Sitter group which both retain the Lorentz group of symmetries as a subgroup. This approach has some attractive features which naturally define a particular length scale l which may be related to the cosmological constant (4.184) where 2 N is the number of canonical coordinates [169]. In this terminology, each first-class constraint eliminates a DoF while it takes two second-class constraints to achieve the same result. This is consistent with classical mechanics where it is known that first-class constraints and their associated gauge fixing conditions form a system of second-class constraints. TEGR is dynamically equivalent to GR and so produces the same classical predictions in their field equations such as metric solutions and energy conditions. However, these theories stem from starkly different Lagrangians as evidenced in section 4.3.1 where the Ricci scalar in equation (4.159) is separating into a contributing second-order torsion scalar and a total divergence term. The result is a drastically different Einstein–Hilbert and TEGR Lagrangian which naturally leads to the question of how many DoF appear. Another important property of TEGR is its dependence on the tetrad rather than the metric tensor which has a significant impact on the Hamiltonian analysis implementation. The formulation of the Hamiltonian structure of TEGR has had a varied history in the literature. In [170] the TEGR Hamiltonian was formulated using a strategy in which an auxiliary field representing a 3 + 1 decomposition of the torsion tensor was used to reduce the over-all order the Euler– Lagrange equations [37, 136, 137, 140]. The approach is altered in [151] where curvature is eliminated through the use of Lagrange multipliers and the Weitzenböck gauge is not immediately assumed. This larger set of dynamical variables leads to a more intricate analysis that eventually produces two physical DoF corresponding to a massless graviton. The canonical formulation that makes up the Hamiltonian analysis has also been put in the language of one-forms in [142, 143] where it is shown that the constraint algebra is indeed closed. The previous works on the topic of the number of DoF of TEGR left some questions open which were then tackled in [144] where an exhaustive study of the first- and second-class constraints is conducted resulting in TEGR having the same number of DoF as GR in n−dimensions, namely n(n − 3)/2. In the context of TEGR, this point is again independently verified in [154] where the Hamiltonian analysis is revisited. It is worth noting that both analyses were done in the Weitzenböck gauge. 4.4. Outlook toward quantum TG and other dimensions The prospect of a viable quantum theory of gravity has come into sharp focus in the context of community efforts in recent 53 Rep. Prog. Phys. 86 (2023) 026901 Review (Λ = 3/lpl2 ) [181, 182]. The length scale in this scenario may coincide with the Planck length scale lpl . This is different to the scenario in the Poincaré group where no invariant length scale is possible at all. Thus, by coupling the field theoretical approach of TG with a special relativity that replaces the Poincaré with the de Sitter group provides an interesting regime to produce a field theoretic theory of gravity that has a naturally associated length scale (possibly relating the cosmological constant with the Planck scale). The literature already contains a number of intriguing works on quantum aspects of TG. For instance, in [183] the first quantization suggests a loop quantum gravity nature to the discretization of TG similar to the development of regular loop quantum gravity from the Einstein–Hilbert metric description. Renormalization is tackled in [163] where the resemblance of renormalization to the process of finding an appropriate spin connection conjugate to a tetrad choice is pointed out and used to highlight the possibility of defining local energy and momentum densities in TG. Another important contribution to the literature on the quantization of TG is the series of works in [184–187] where projective techniques are used to construct kinematic quantum states of TEGR. However, the work needs more development to fully quantize the theory. More generally, the quantum regime has also been studied in the context of black hole solutions of TEGR such as [188, 189] where the well known Weyl quantization procedure is applied to the Schwarzschild and Kerr solutions, as well as [190] in which an initial second quantization procedure is taken on the Schwarzschild solution. Of note as well is some initial work that has started in handling brane cosmology, namely [191, 192], but it is still in very early stages when compared with standard gravity. proposed teleparallel theories that have been studied in the literature. 5.1. Foundations of modified teleparallel theories and how to make them fully invariant One of the main motivations of this Review, has been the fact that TG gained a lot of interest in the astrophysics and cosmology community in the last few years. However, since it is a not so well studied framework (yet), a lot of ambiguities and misunderstandings have arisen through the years; especially in the modifications sector of the theory. TG, being a theory built on the tangent space, must be invariant under general coordinate transformations, as well as under local Lorentz transformations, because special relativity should be recovered in locally inertial frames [221]. As already known from special relativity, local Lorentz transformations define two classes of frames: the inertial frames, being the ones in which inertial effects vanish, and the noninertial ones, where inertial effects are present. Quantitatively, these frames can be distinguished by a purely inertial connection, also known as spin connection, which at inertial frames vanishes, while at non-inertial frames is different from zero. Before going in more depth, let us briefly explain the issue of local Lorentz invariance. The first formulation of teleparallel theories assumed that the spin connection was always zero, and then the torsion tensor would only depend on the tetrads, namely Tλ µν = EA λ (eA ν,µ − eA µ,ν ). (5.1) As we have explained in section 2.2.3, the above torsion tensor is a particular one which is computed in the so-called Weitzenböck gauge (where the spin connection vanishes). Now, let us take a local Lorentz transformation only in the tetrads e ′ A µ = Λ A B e B µ . By doing this transformation, the above torsion tensor transforms as 5. Modified TG theories This section is devoted in the so-called extensions or modifications of TG. The motivation is the same as in GR: the difficulty of finding a consistent quantum theory of gravity, the nature of singularities, but mostly the unknown, dark Universe led cosmology to the pursuit of a better description for the gravitational interactions. There have been a great many proposals for modifying the TEGR and in this section we will try to present the features of most of them. In figure 8 one can see a bird’s eye view of the spectrum of theories, deviating from GR. Together with the different ways of modifying GR, some of the most important example theories are depicted as well. The figure is drawn by thinking on breaking some of the conditions in the Lovelock’s theorem. It turns out that some parts of the figure are connected. For example, some theories which add invariants can be rewritten as scalar-tensor theories. Furthermore, theories can be part of multiple branches of the figure. An example of this could be teleparallel theories, which is a framework beyond GR which modifies the geometry. As we will see in this section, modified teleparallel theories can also break other conditions in the Lovelock’s theorem, as it happens in teleparallel scalar-tensor theories (see section 5.8) where besides modifying the geometry, one adds a scalar field in the Lagrangian. In what follows, we will summarize different T ′λ µν = Tλ µν + ΛA B EB λ (eC ν ∂µ Λ A C − eC µ ∂ν Λ A C ), (5.2) where one immediately notices that this quantity is noncovariant under local Lorentz transformations. Further, the torsion scalar in this particular gauge will transform as [222] 8 T ′ = T + ∂µ (e η CB EB µ ∂A Λ A C ). e (5.3) From the above equation, we get that in the old version of TG where one only has the tetrads as fundamental variables and the spin connection is always zero, TEGR is a pseudo-local Lorentz invariant theory. Since this extra term only appears as a boundary term in the above equation, independently of the choice of the spin connection (or local Lorentz transformation), the manifestation of breaking the local Lorentz invariance does not appear in TEGR (at least at the level of the field equations—since it is just GR). For more information, please check sections 2.2.3 and 3.3. However, the situation is very different when one modifies the action and studies modified teleparallel theories of gravity. In these theories, for example 54 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 8. Representation of some possible ways of modifying GR through breaking the Lovelock’s theorem along with some examples. modified theories (and specifically f (T) gravity, as we will see below) were not local Lorentz invariant, and since they were (still) ignoring the fact that a nonvanishing spin connection could solve the problem, they introduced the notion of ‘good’ and ‘bad’ tetrads [223]. Good were those tetrads that led to non-trivial solutions of the field equations (of f (T) gravity), with a vanishing spin connection, and bad were those tetrads were the theory trivially reduced to TEGR. As expected, these notions tend to become obsolete since the realization that teleparallel theories are local Lorentz invariant, but for old times’ sake, it is nice to keep a part of them and slightly modify it. For this reason, and since it concerns all theories built on teleparallel geometry, we thought it would be better to discuss what a good tetrad-spin connection pair would be, before we dive into the realm of modified teleparallel theories. In section 3.6 we defined the concept of symmetric tetrads, that are specific tetrads satisfying certain symmetries related to teleparallel geometries. These tetrads are important to study astrophysical systems and cosmology since in them, one always has certain symmetries associated to the physical situation, such as spherical or axial symmetry. In TG, one has the additional complication that these type of tetrads do not necessarily solve the antisymmetric part of the field equations of a certain theory. Moreover, one also needs to be careful that the choice of the symmetric tetrad is compatible with the choice of the spin connection. Due to the Lorentz transformations, one can always choose a specific gauge where the spin connection vanishes (Weitzenböck gauge), but one can also find a pair of tetrad with a non-zero spin connection which respects the symmetries and also solves the antisymmetric field equations of a theory as we will see later in this section. Then, it is convenient to define a subclass of symmetric tetrads being compatible with the choice of the spin such as in f (T) gravity, the boundary term appearing in the above equation is not a boundary term anymore, meaning that theory will depend on Λ A B . This means that when one is modifying TEGR, the theories would break local Lorentz invariance if one assumes that the spin connection vanishes and construct theories with the torsion tensor (5.1) only depending on the tetrads. This problem was not well understood until the authors in [44] noticed that the problem of breaking the local Lorentz invariance was only related to choosing the particular Weitzenböck gauge. This can be easily seen by noticing that the torsion tensor (1.6a) transforms covariantly under local Lorentz transformations if we take the simultaneous transformations in the tetrads and the spin connection (see equations (2.32) and (2.34)), which indeed gives us that T ′λ µν = Tλ µν . In conclusion, are teleparallel theories fully invariant (diffeomorphisms and local Lorentz)? The answer is yes, they are if we perform the simultaneous local Lorentz transformations for the tetrad and the flat spin connection. In addition, this quantity is a tensor and thus transforms invariantly under diffeomorphisms. This means that, any action (and consequently field equations) constructed from the torsion tensor will be fully invariant. The spin connection in GR describes both gravitational and inertial effects, and because of the equivalence principle, there exists a local frame, where this connection vanishes. However, the picture in TG becomes different: gravitational effects are described by a translational gauge potential, since TG is a gauge theory of the translation group (see discussion in section 3.1), while the spin connection represents purely inertial effects. That is why, when one sets the spin connection to zero local Lorentz invariance is broken. Much of the work on TG has been performed in the Weitzenböck gauge. When the community realized that 55 Rep. Prog. Phys. 86 (2023) 026901 Review connection satisfying the symmetries. We will label the ‘good tetrad-spin connection pair’ as the pair satisfying the following three conditions: 5.2. New general relativity We start the discussion on modifications of TEGR with a class of theories, dubbed NGR [34]. This was the first modification of TEGR considered in the literature and it was formulated in 1979 by Hayashi and Shirafuji. In their original use of the name, NGR refers to a one-parameter subclass of a larger, three-parameter class of models, while nowadays the term NGR is also used to refer to the whole three-parameter class. In this section we will discuss the structure, action and field equations of NGR in section 5.2.1, as well as its Hamiltonian analysis and the number of DoF in section 5.2.2. (a) The tetrad must be a symmetric tetrad compatible with the choice of the spin connection. This means that it must fulfill equation (3.122) (Weitzenböck gauge) or equation (3.120) (non-zero spin connection). (b) It must solve the antisymmetric field equations without constraining the underlying theory. This definition was put together in order to match and extend the first ideas related to TG in different symmetries [223], where the authors defined the good tetrads. These first ideas were introduced in the non-invariant theory, so it was not really clear that the spin connection plays an important role in TG. Let us here explain all the conditions in more detail. The first condition ensures that the good tetradspin connection pair satisfy the underlying symmetries of the problem. A priori, there is no fundamental reason to assume that the connection must also respect the symmetries of the physical situation, but, it is a sensible choice to impose this condition to then have the torsion tensor (and field equations) satisfying the same symmetries as the tetrad (or the metric). The second condition is the most important characteristic of the good pair, meaning that they are some specific pair which solve the antisymmetric field equations of a certain set of equations of a TG. For practical reasons, these are the pairs that one needs to use to compute the field equations for different physical situations with a certain type of symmetry. It can also be inferred that the pair depends on the theory studied and it is possible to find that a good tetrad-spin connection pair in one theory is not a good tetrad-spin connection pair in another theory. It is of great importance though, that a good pair does not constrain the theory itself. If for example, the antisymmetric field equations, say in f (T) gravity, yield d 2 f/dT 2 = 0, the associated tetrad-spin connection pair would not be considered as a good one. This condition is ignored in many studies concerning TG. Then, a good tetrad-spin connection pair satisfies the symmetry conditions that are compatible with either a zero spin connection or a non-zero spin connection and also solve the antisymmetric part of the field equations of the theory. Thus, in any modified TG, one would first need to find a symmetric tetrad (with or without zero spin connection) satisfying a certain the symmetry of the physical situation and then find the good tetrad-spin connection pair which solve the antisymmetric field equations of this theory, with the important condition that the symmetric field equations cannot be the trivial ones (GR + Λ). To match the existing literature, apart from the good tetrad-spin connection pair, we will refer to good tetrads as the ones that satisfy the conditions above and are computed in the Weitzenböck gauge. In the following, we will give a comprehensive review of the most important modified teleparallel theories explored in the literature, along with their most important theoretical descriptions. A straightforward generalization of the TEGR action (4.164) can be obtained by realizing that the torsion scalar can be decomposed into a sum of three terms. Two different decompositions of this type are commonly used, for which we introduce the notation 5.2.1. Action and field equations. 1 3 2 2 1 T = T1 + T2 − T3 = Taxi + Tten − Tvec . 4 2 2 3 3 (5.4) Here the terms in the first parametrization read T1 := Tµνρ Tµνρ , T2 := Tµνρ Tρνµ , T3 := Tµ µρ Tν νρ . (5.5) Alternatively, one may use the decomposition of the torsion tensor into parts which transform under irreducible representations of the Lorentz group, and which can be written in two equivalent ways. First, one may define tσµν := 1 aµ := ϵµνσρ Tνσρ , 6 (5.6a) vµ := T σ σµ , (5.6b) 1 1 1 (Tσµν + Tµσν ) + (gνσ vµ + gνµ vσ ) − gσµ vν , 2 6 3 (5.6c) where ϵµνσρ represents the totally antisymmetric Levi-Civita tensor associated to the metric gµν . Note that the tensor part tσµν is symmetric in its first two indices, while its totally symmetric part and any trace vanish tαµν = tµαν , tαµν + tναµ + tµνα = 0, = tµα α = 0. t αµ α = tα αµ (5.7) In terms of these one finds the scalar invariants14 Taxi := aµ aµ = 1 (2Tσµν Tµσν − Tσµν T σµν ) , 18 Tvec := vµ vµ = T σ σµ Tρ ρµ , Tten := tσµν tσµν = (5.8a) (5.8b)  1 1 Tσµν T σµν + Tσµν Tµσν − T σ σµ Tρ ρµ . 2 2 (5.8c) 14 Note a sign mistake in the definition of T 56 axi in reference [46, equation (12)]. Rep. Prog. Phys. 86 (2023) 026901 Review Another possibility is to start from a decomposition of the torsion tensor in the form axi vec vec Tµ νρ = T µ νρ + T µ νρ + T µ νρ , the Lagrangian (5.14) reduces to that of TEGR. It was recently shown that the Lagrangian (5.14) can be renormalized at one-loop order without introducing higher-order terms [224]. However, the theory with the three parameters contain ghosts. The field equations are easily derived by variation of the action (5.14) wrt the tetrad and read (5.9) where the three components are given by axi T µνρ := ϵµνρσ aσ , vec 1 T µνρ := (gµν vρ − gµρ vν ), κ2 Θµν = 3 ten 2 T µνρ := (tµνρ − tµρν ).  ◦ + 2∇ρ c1 Tνµρ + c2 T[ρµ]ν + c3 T σ σ[ρ gµ]ν   1 + c1 Tρσ µ Tνρσ − 2T[ρσ]ν − c3 T σ σρ Tρ µν + 2T(µν) ρ 2  1  + c2 Tµ ρσ (2Tρσν − Tνρσ ) + Tρσ µ 2T[ρσ]ν − Tνρσ , 2 (5.10) 3 These tensors vanish when mutually contracted with each other due to the symmetries of the torsion tensor, and yield the scalar invariants in the form 1 axi axi Taxi = − T µ νρ T µ νρ , 6 3 vecµ vec νρ Tvec = T νρ T µ . 2 Tten = (5.18) 3 tenµ ten νρ T νρ T µ , 4 in terms of the torsion tensor and the energy-momentum tensor (1.10), and can equivalently be expressed through the decomposition (5.6) as (5.11)  Hence, the different parametrizations are related by Taxi = 1 (2T2 − T1 ), 18 1 Tten = (T1 + T2 − T3 ), 2 4 2 1 ρ a a(ρ gµν) − ϵναβγ a α tµ βγ − ϵµνρσ aρ vσ 2 9 9  ◦ 1 − ϵµνρσ ∇ρ aσ 3  ◦ 1 ρ 4 v v(ρ gµν) + tµ[ρν] vρ + 2gµ[ν ∇ρ vρ] + cvec 2 3  1 ρ σ − ϵµνρσ a v 2  ◦ 2 4 + cten tα[βγ] t αβγ gµν − tµ[ρσ] tν ρσ + 2∇ρ tµ[νρ] 3 3  1 2 ρ α βγ . (5.19) − tν[µρ] v + ϵµαβγ a tν 3 2 2 κ Θµν = caxi Tvec = T3 , (5.12) or conversely by 2 4 T1 = Tvec + Tten − 6Taxi , 3 3 T3 = Tvec . 1 2 T2 = Tvec + Tten + 6Taxi , 3 3 (5.13) With these definitions in place, one generalizes the TEGR action (4.164) to an arbitrary linear combination of the three scalar invariants, hence LNGR := c1 T1 + c2 T2 + c3 T3 + Lm = caxi Taxi + cten Tten + cvec Tvec + Lm . The energy-momentum tensor Θµν is symmetric by assuming a minimal, Lorentz invariant matter coupling as discussed in section 3.5. The field equations are then decomposed into their symmetric and antisymmetric parts, which are given by (5.14) The relation between the different constants follows from the relations (5.12) and (5.13) to be caxi := 6(c2 − c1 ),  ◦  κ2 Θµν = ∇ρ (2c1 + c2 )T(µν)ρ + c3 (T σ σρ gµν − T σ σ(µ gν)ρ ) 1 + (c1 T1 + c2 T2 + c3 T3 ) gµν 2  + c1 Tρσ (µ Tν)ρσ − 2T[ρσ]µ Tρσ ν  1 + c2 Tρσ (µ Tν)ρσ + 2T[ρσ]µ Tρσ ν − Tµ ρσ Tνρσ 2 − c3 T σ σρ T(µν) ρ , (5.20) 4 2 cten := c1 + c2 , 3 3 1 2 cvec := c1 + c2 + c3 , 3 3 (5.15) or equivalently 1 1 c1 := cten − caxi , 2 18 1 c3 := cvec − cten . 2 1 1 c2 := cten + caxi , 2 9 and ◦ ◦ 1 0 = (2c2 + c3 )∇ρ Tρµν − (2c1 − c2 )∇ρ T[µν]ρ 2 1 + (2c1 − 3c2 − c3 )Tρσ [µ Tν]ρσ , (5.21) 2 (5.16) Obviously, for the particular values 1 1 c1 = , c2 = , 4 2 2 cvec = − , 3 c3 = −1 1 (c1 T1 + c2 T2 + c3 T3 ) gµν 2 ⇔ 3 caxi = , 2 2 cten = , 3 using the geometric identity ◦ (5.17) 3∇[ρ Tρ µν] + Tρ µν T σ σρ − Tρσ [µ Tν]ρσ = 0, 57 (5.22) Rep. Prog. Phys. 86 (2023) 026901 Review in the second equation. Recall that in the covariant formulation of TG [43] we use here, in which an arbitrary, flat spin connection is assumed, the antisymmetric part (5.21) is equivalently obtained by variation wrt the spin connection, and vanishes identically if the parameter values satisfy whether their time evolution is uniquely defined, such that the theory has a well-defined Cauchy problem, and whether there is a mismatch between the linearized and the full, nonlinear theory [228]. Specifically, in [228] the authors found that the gauge symmetry that appears in the quadratic Lagrangian, realized as a massless 2-from field, i.e. Kalb-Ramond field, is necessarily broken at cubic order, signaling a strong coupling problem around Minkowski spacetime. This means that the Minkowski solution is very unstable against perturbations and thus it will be impossible to be reached from an aritrary point in phase space. For NGR in its most general form this is still an open question. The most comprehensive approach to answering this question is the Hamiltonian analysis (see [229] for a recent review). Already, for the original one-parameter NGR [34], possible issues concerning a non-unique time evolution, and hence an ill-defined Cauchy problem, have been noted, with the help of a simple example [230]. Consider the diagonal tetrad e A µ = diag(1, 1, 1, 1) and nonvanishing spin connection ω A Bµ dxµ = Ω A B (t)dt, where the spatial components ΩI J (t) = 0 vanish, with I, J = 1, . . . , 3. One finds that this spin connection is indeed flat, R A Bµν = 0, while the only nonvanishing torsion components are given by 2c1 − c2 = 2c2 + c3 = 0 ⇔ cten + cvec = 9cten − 4caxi = 0, (5.23) and are hence a multiple of the TEGR values (5.17). The theory based on general form of the Lagrangian (5.14) discussed above is known as (the three-parameter class of) NGR [34]. It can also be obtained from the premetric approach [130]. However, as mentioned in the introduction of this section, in the original work the name NGR is used to refer to the more restricted one-parameter class. This smaller class is obtained by fixing cvec and cten to their TEGR values (5.17), and allowing only caxi to deviate, so that the parameters take the form 1 ϵ 1 ϵ − , c2 = + , 4 18 2 9 2 2 cten = , cvec = − , 3 3 c1 = c3 = −1 ⇔ 3 caxi = ϵ + , 2 (5.24) with a free parameter ε. This restricted class has the interesting and distinguishing property that observational discriminators, such as the parameterized post-Newtonian (PPN) limit, agree with TEGR, and hence with GR [34, 225]. Its field equations take the form  ◦ 1 ρ 4 κ2 Θµν = Gµν + ϵ a a(ρ gµν) − ϵναβγ a α tµ βγ 2 9  ◦ 2 1 − ϵµνρσ aρ vσ − ϵµνρσ ∇ρ aσ , 9 3 (5.25) T0 0I = −T0 I0 = Ω0 I . (5.26) Note that, in particular, the axial torsion (5.6a) vanishes, due to the contraction with the totally antisymmetric tensor. Further, the metric gµν obtained from the diagonal tetrad e A µ is simply the Minkowski metric, and so the curvature of the Levi-Civita connection vanishes. Hence, this tetrad and spin connection constitute a solution to the field equations (5.25). Since the functions Ω0 I (t) can be chosen arbitrarily, it follows that their time evolution, and hence also the time evolution of the nonvanishing torsion components, is not determined by the field equations and initial conditions. While this has no observable consequences if the matter couples only to the metric, it does lead to an undetermined influence on matter if one allows for a direct coupling to the spin connection. Note that similar statements can be made also about other TG theories, such as f (T) gravity [231, 232]. In order to obtain a well-defined Cauchy problem also for the torsion, different modifications of NGR have been suggested. The most straightforward extension is to consider the full action (5.14), with three free parameters, where the deviation from their TEGR values must be chosen sufficiently small so their PPN limit is compatible with Solar System observations [230]. Other possibilities are given by adding higher order terms [233] or parity-violating terms [234] to the action. However, it has been pointed out that the non-unique time evolution occurs only for particular solutions and is not generic [235], and that tetrads which are related by the remaining partial Lorentz invariance which does not alter the axial torsion may be regarded as physical equivalent [236]. In order to address the aforementioned issues, and assess the consistency of NGR, various approaches toward a full Hamiltonian description have been carried out. For the most general, three-parameter class of NGR governed by the showing that they reduce to GR in the absence of axial torsion aµ . Note that the covariant, three-parameter approach to NGR discussed above differs in two further aspects from the originally proposed theory [34]. In the latter, a non-covariant formulation is adopted, in which the spin connection is assumed to vanish identically, and the tetrad is the only dynamical field. Further, a different matter coupling is assumed, in which Dirac fields couple to the (vanishing) teleparallel spin connection instead of the Levi-Civita connection. As a consequence, their energy-momentum tensor receives an antisymmetric contribution, which acts as a nonvanishing source for the antisymmetric equation (5.21). Hence, in this original approach is was assumed that the parameter values strictly differ from the condition (5.23), since otherwise the theory would not have allowed for fermionic fields. An important question which arises in the context of modifications of GR is the appearance and nature of additional DoF. Of particular importance is the question whether such new DoF are pathological, such as to behave as ghosts or tachyons [226, 227], 5.2.2. Hamiltonian analysis and DoF. 58 Rep. Prog. Phys. 86 (2023) 026901 Review i h N T 2 B m B B l k S = √ π − (2c1 + c2 + 3c3 ) ∂k (Nn N e m ) − T kl N eB . 3 h Lagrangian (5.14), the canonical momenta defined by the variation (4.134) are given by [147] T √ πA i = (5.32d)  h B i j T 0j M A B + T B kl Mi A k B Nl + 2Nhik N  ×(c2 nB eA l + c3 nA eB l ) , (5.27) Applying the same decomposition to the LHS of the velocity equation (5.30), and substituting the velocity components (4.138), one finds that the matrix (5.28) greatly simplifies, and the equation (5.30) decomposes into the components using the abbreviation Mi A j B = −2(2c1 hij ηAB − (c2 + c3 )nA nB hij + c2 eA j eB i + c3 eA i eB j ). Ni eA = ηAB h e j 6= EA = eA + nA . N ij B i i (5.29) A = −2AA v mp , S S Smp = −2AS v mp , (5.33) AS = 2c1 + c2 , (5.34) or equivalently, 2 1 AA = cten − caxi , 2 9 AV = cten + cvec , 3 AS = cten , 2 (5.35) AT = 3cvec . If all of these constants are nonvanishing, one can solve all velocity components for the corresponding momentum components. In general, however, this is not the case, as certain constants may vanish, depending on the values of the parameters c1,2,3 , thereby leading to the appearance of primary constraints, which are given by (5.30) V   N SA = √ πA i + ∂k (Nn B + Nm e B m ) − T B kl Nl Mi A k B h − 2NT B kl hik (c2 nB eA l + c3 nA eB l ), (5.31) ⇒ AV = 0 i V πi 0 ≈ Ci = √ − 2c3 T B kl hik eB l , h (5.36a) A AA = 0 and does not contain any velocities. In order to solve the equation (5.30) for the velocities v A i , it is necessary to invert the object in equation (5.28), viewed as a matrix. This can be achieved by applying the irreducible decomposition shown in section 4.2.5 to both sides of the equation (5.30). On the RHS, proceeding in analogy to the momentum decomposition (4.136) yields the decomposition A π mp 0 ≈ Cmp = √ − 2c2 hlm hpk T B kl nB , (5.36b) h ⇒ S AS = 0 ⇒ AT = 0 ⇒ S π mp 0 ≈ Cmp = √ , h (5.36c) T V N V S i = √ π i − 2Nc3 T B kl hik eB l + 2(2c1 + c2 + c3 ) h   × ∂k (Nn B + Nm e B m ) − T B kl Nl nB hik , (5.32a) S T S AV = 2c1 + c2 + c3 , AA = 2c1 − c2 , AT = 2c1 + c2 + 3c3 , where the term SA i on the RHS is defined as A mp T A mp where the constants AV,A,S,T are given by Further, from the decomposition (4.130) of the torsion follows that the velocities enter the momenta (5.27) only via the component T A 0i in the first term, while the remaining terms depend only on the lapse N, shift N i , spatial tetrad components e A i and their spatial derivatives. Hence, in order to solve for the velocities v A i = ∂0 e A i in terms of the momenta, it is convenient to isolate the term which is linear in the velocities on one side of the equation, which then takes the form M i A j B v B j = SA i , T S = −2AT v, Note that eA i is not the inverse tetrad, but defined from the tetrad by raising and lowering indices with the Minkowski and induced metrics, η AB and hij , so that i V S i = −2AV v i , (5.28) T π 0≈C= √ , h (5.36d) which defines the constraint expressions CA i in terms of their irreducible decomposition. The different possible cases can most easily be visualized by introducing normalized parameters c̃i and polar coordinates (α, β) using the definition N A = √ π mp − 2Nc2 hlm hpk T B kl nB − 2(2c1 − c2 ) h   × ∂k (Nn B + Ns e B s ) − T B kl Nl eB [m hp]k , (5.32b) c̃i = q ci c12 + c22 + c32 c̃3 = cos α.   N S Smp = √ π mp − 2(2c1 + c2 ) ∂k (Nn B + Ns e B s ) − T B kl Nl h   1 × eB (m hp)k − hpm eB k , (5.32c) 3 , c̃1 = sin α cos β, c̃2 = sin α sin β, (5.37) These coordinates parametrize a sphere, one hemisphere of which is shown in figure 9; the other hemisphere is obtained by a point reflection at the origin. A different parametrization, based on the axial-vector-tensor decomposition of the torsion, is given in [237]. The same result is presented in the language S 59 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 9. Primary constraints of NGR, depending on the values of the parameters in the action. The radial axis shows 0 ⩽ α ⩽ π2 , while the polar axis shows 0 ⩽ β ⩽ 2π, using the definition (5.37). The red line AV = 0, except for the two points where it meets the plot boundary α = π2 , constitutes the one-parameter class of NGR, with TEGR being marked by the dot at the intersection of the red and black lines, AV = AA = 0. Reprinted figure with permission from [147], Copyright (2019) by the American Physical Society. which depends on the constants λ, ν, σ. Note the appearance of a parity-violating term vµ aµ , governed by the parameter σ. In the absence of this term, choosing σ = 0, the theory reduces to NGR with the parameters given by   1 λ λ c1 = 1 − − ν , c2 = , c3 = −1, (5.39) 2 2 2 of differential forms in [149] and in the premetric approach in [146]. Several particular subclasses of NGR have been studied, which can be characterized by their primary constraints. The most prominent example is TEGR given by the parameter values (5.17), which imply AV = AA = 0; this is discussed in section 4.3.3. A toy model with c1 = 1/2, c2 = c3 = 0 with no primary constraints has been studied in [141] in the language of differential forms, in order to present the general formalism of Hamiltonian analysis in TG. The one-parameter class of NGR, with parameters (5.24), has been studied for ε 6= 0, so or equivalently,   3 2 1 caxi = 3 λ + ν − 1 , cten = (1 − ν), cvec = (ν − 2). 2 3 3 (5.40) V that the only primary constraint Ci ≈ 0 arises from AV = 0, in [150]. Finally, also an extension of NGR by a parity-violating term has been considered, whose Lagrangian is given by [238] L=   λ 1 λ 1 − − ν T1 + T2 − T3 + 6σvµ aµ , 2 2 2 Further restricting to ν = 0, the theory reduces to the oneparameter NGR values (5.24) with ϵ = 2(λ − 1)/9, and giving TEGR for λ = 1. This class of theories avoids the undetermined evolution of the torsion mentioned above if the parameters satisfy the inequality (5.38) ν(1 − ν − λ) − 2σ 2 6= 0. 60 (5.41) Rep. Prog. Phys. 86 (2023) 026901 Review V For pure NGR, where σ = 0, this condition reduces to ν 6= 0 and 1 − ν − λ 6= 0. By comparison with the conditions for the appearance of primary constraints, one finds that this corresponds to AV 6= 0 and AA 6= 0, so that no vector and antisymmetric constraints appear. Further, consistency of the PPN limit with Solar System observations, |ν| ⩽ 1 is required, so that also AS 6= 0 and AT 6= 0. Hence, this corresponds to the generic class of NGR theories, in which there are no primary constraints. Further studies in the Hamiltonian approach to NGR have so far been restricted to such special cases. In the case that none of the constants AV,A,S,T vanish, so that none of the primary constraints (5.36) appear. In this case all velocities can be expressed in terms of canonical momenta, and the full Hamiltonian is identical to the kinematic Hamiltonian (4.135) and reads [147] √  V V i  Ci C A A Cij Cij S S Cij Cij T constraint Ci ≈ 0. It then follows that the kinematic Hamiltonian is given by [150] (     √ 2 9 π[ij] √ − 1 − ϵ nA T A ij H0 = N h 8ϵ 9 h  [ij]    ˜ i πA i π 2 n A∇ Aij √ − 1 − ϵ nA T − T̃ − √ × 9 h h ) (ij) i A 2 π(ij) π (πA e i ) − + 4h 8h ˜ i πA i ) + ∇ ˜ i [πA i (Nn A + Nj e A j )], − Nk (T A jk πA j + e A k ∇ (5.46) to which a Lagrange multiplier term V  √ H1 = Vi π i + 2 hT B kl hik eB l ,  3C n ∇i π A  − − T̃ − √ 4AS 4AT h     ˜ i πA i Nn A + Nj e A j , ˜ i πA i + ∇ − Nk T A jk πA j + e A k ∇ H=N h 4AV − 4AA − A˜ 2 (5.47) i must be added, in order to implement the primary constraint V Ci ≈ 0. In order to obtain the full Hamiltonian, further terms must be added, which arise from the time evolution of the V (5.42) constraint Ci , given by its Poisson bracket with the kinematic Hamiltonian [150]. where we used the abbreviation 5.3. f(T) gravity T̃ = c1 ηAB T A ij T B kl hik h jl + c2 eA i eB j T A kj T B li hkl + c3 eA i eB j hkl T A ki T B lj , A relatively recent modification of TEGR is the so-called f (T) (5.43) ◦ theory of gravity. In the same spirit as f(R) gravity, that is a straightforward generalization of the Einstein–Hilbert action, f (T) theory was proposed almost more than a decade ago as a generalization to the TEGR action, by Ferraro and Fiorini [38]. In this section, we will study its action and field equations together with its behavior under conformal and disformal transformations. ˜ i denotes the purely spatial Levi-Civita covariant while ∇ derivative of the induced metric hij , and eA i are the spatial tetrad components (5.29). Note that in this case none of the expressions CA i given by the RHS of the expressions (5.36) are constraints, i.e. they are not imposed to vanish; they simply turn out to be convenient for expressing the Hamiltonian. Furthermore, note that all derivatives of the lapse N and shift N i are gathered into a total divergence term, while they appear as Lagrange multipliers in the remaining terms. Hence, one finds that they lead to the further constraints V V Ci Ci 4AV − A A Cij Cij 4AA − S S Cij Cij 4AS T − 5.3.1. Action and field equations. reads Sf(T) := 3C n ∇ i πA − T̃ − √ ≈ 0, (5.44) 4AT h 2 A˜ i ˆ d4 x e f(T) + Sm , (5.48) where Sm is the action related to the matter sector defined as in equation (1.9). Let us take a step back and consider the geometric trinity of gravity discussed in section 2.3. There we saw from equations (2.49) and (2.50), that the relation between the Ricci, the torsion and the non-metricity scalars is and ˜ i πA i ≈ 0, T A jk πA j + e A k ∇ 1 2κ 2 The action of f (T) theory (5.45) ♢ ◦ which correspond to the Hamiltonian and momentum constraints in GR. Further, it follows that imposing these constraints leads to a vanishing Hamiltonian on the constraint surface, up to a total divergence, as it is also the case in GR. In the case that any of the constants AV,A,S,T vanishes, the corresponding term in the Hamiltonian (5.42) would diverge, and so it must be replaced, taking into account the arising primary constraints. The most straightforward case in which this can be seen is the non-trivial (ε 6= 0) one-parameter NGR (5.24). In this case one has AV = 0, leading to the R = −T + B = −Q − B♢ , (5.49) Q where we defined the two boundary terms, B and B♢ as Q ◦ 2 B = 2∇µ Tµ = ∂µ (eTµ ), e ◦ ♢ ◦ ♢ B♢ = ∇ν Qµ µν − ∇µ Qµν ν . Q 61 (5.50a) (5.50b) Rep. Prog. Phys. 86 (2023) 026901 Review Figure 10. Equivalency between GR, TEGR and STEGR, but not equivalency between their modifications. It is easy to conclude from equation (5.49) that since the total derivatives, B and B♢ will not contribute in the consider the respective boundary term to be a variable of the ♢ arbitrary function, meaning f(T, B) and f(Q, B♢ ) theories. Q Q action integral, the field equations, resulting from their variations, will be equivalent in all three theories, GR, TEGR and STEGR, even though the geometric background is different in each one of them. The same does not happen though when one generalizes to arbitrary functions of these scalars [239] and this is described in the figure 10. Recently, a different decomposition of the curvature scalar has been utilized, ◦ ◦ ◦ R = G + B, As already discussed in a previous section, modified teleparallel theories employ as their field variables both the tetrad and the spin connection, meaning that both of them should be described by respective field equations. These two variations are closely connected, since the field equations of the spin connection are the antisymmetric part of the field equations of the tetrad. Specifically, varying the action (5.48) wrt the tetrad e A µ we get its field equations that read (5.51) ◦  ◦ ◦ ◦ ◦ ◦ and G = gµν Γλ νσ Γσ λµ + Γσ µν Γλ λσ B= √ √ ∂ν (∂µ (ggµν )/ −g) / −g, that are pseudo-scalars (in contrast with T and B) and a new generalized theory was proposed [55] that has the form ˆ ◦ ◦ √ 1 S ◦ ◦ := (5.52) d4 x −gf (G, B ) + Sm , 2 f(G,B) 2κ 1 − fT ∂ν (eSA µν ) − SA µν ∂ν fT + fT T B ν A SB νµ − fT ω B Aν SB νµ e 1 − f E A µ = κ 2 ΘA µ , (5.53) 2 where ◦ where f T is the first derivative of f wrt T, SA µν is the superpotential, ω B Aν the spin connection and ΘA µ is the energy-momentum tensor of the matter fields; details regarding the variations can be found in the appendix A. Moreover, varying the action (5.48) wrt the spin connection [123, 125, 241] we get ◦ with f being an arbitrary function of G and B . ◦ In greater detail, f(R) theories result in fourth order field equations and that is because, the Ricci scalar consists of second order derivatives of the metric. On the other hand, both the torsion and the non-metricity scalar are built of only first derivatives of the tetrad and the metric respectively, yielding second order field equations. Actually, because of the simil- δω Sf(T) = S[AB] ν ∂ν fT = 0. (5.54) It is worth noticing that the RHS vanishes, because the energymomentum tensor, ΘA µ , has no antisymmetric part. This occurs since we have assumed that matter is not coupled to the spin connection. As has already been discussed in a general framework in section 4.1.3, equations (5.53) and (5.54) are related, in the sense that the second is the antisymmetric part of the first. Let us see why: it is helpful to write down the LHS of equation (5.53) as ♢ arity of these two scalars, f(Q) and f (T) theories are indistinguishable at cosmological background level [239]; that is why if one wants to look for differences between the two theories, they have to study perturbations [240]. If one wants to imitate ◦ f(R) theory in the teleparallel geometries, they would have to 62 Rep. Prog. Phys. 86 (2023) 026901  WAB := −ηBC e µ C Review 1 fT ∂ν (eSA µν ) + SA µν ∂ν fT − fT T D ν A SD νµ e  1 D νµ µ +fT ω Aν SD + f EA (5.55a) 2 ◦ 1 = −SAB ν ∂ν fT − fT GAB − ηAB (f − TfT ), 2 which would mean breaking of the Lorentz symmetry. However, there are known backgrounds with no new DoF in linear perturbations, thus rendering their results controversial. Some years later, a new perspective was given to the field [145, 247–249], by claiming that there exist only three DoF in a four-dimensional spacetime, and thus making f (T) theor◦ (5.55b) ies comparable to f(R) in this regard. However, it was shown in [154], that the Poisson brackets of the Lorentz constraints are incomplete, and as a consequence their main conclusion is wrong. Even though, the last study that came our recently, [154], points out several ‘mistakes’ of the previous approaches, it does not seem to lack problems itself. The authors claim, among other things, that the total number of DoF in 4 dimensions is five. Yet, the matrix of Poisson brackets of Lorentz constraints does not seem to be non-degenerate, since it contains a non-trivial term [245]. They claim that this non-trivial term proves that the constraints are second class, however, since the matrix could be degenerate, they could be a combination of first and second class. Last but not least, apart from pure f (T) theories, there have been similar attempts to generalize TEGR as well, that could be considered as special cases. One of them is the socalled unimodular gravity, where one considers the determinant of the tetrad to be constant. The Hamiltonian formulation of unimodular TEGR was studied in [140], reconstruction of unimodular f (T) models was studied in [250], inflationary cosmology was studied in [251] and holographic dark energy models of f (T) were studied in [252]. ◦ where GAB is the symmetric Einstein tensor of the LeviCivita connection expressed with tetrads. Note that we used ◦ equation (4.168) to obtain an Einstein term GAB . Since the last term is symmetric as well, the only antisymmetric part of the tetrad field equations is the first term, W[AB] = −S[AB] ν ∂ν fT , (5.56) which coincides with equation (5.54). This is imposed by the local Lorentz invariance and indicates the fact that the spin connection is nothing more but a pure gauge DoF, and not a Stückelberg field. Apart from the local Lorentz invariance of f (T) theories, there have been other controversial issues as well. These theories apparently suffer from strong coupling problems, since in some backgrounds, the kinetic term of some physical DoF disappears at quadratic order, and thus perturbation theory, at linear order, breaks down at these backgrounds. Specifically, in [231] the authors perturb a trivial tetrad and they find explicitly a new mode at fourth order in perturbations, that goes beyond the Minkowski background. Apart from that, cosmological perturbations up to second order [242, 243] show that no extra modes appear around a flat Friedmann background, signalling again a strong coupling problem. In [244] though, the authors performed perturbations in non-flat cosmological backgrounds; it turns out that the tensor and vector modes are well determined just like in the flat case. Unlike the flat case though, the scalar perturbations are fully determined by the perturbed equations at linear order. Unfortunately, this does not mean that the strong coupling is not present in non-flat backgrounds, since some of these scalar modes are still non-propagating. It could be seen as progress though, toward limiting the strongly coupled behavior in such backgrounds. What’s more, it could mean that the theory is not ill-behaved, but the way we approach the perturbation analysis is not the ideal one, thus leading to strong coupling problems. Up to now, there have been three different Hamiltonian analyses of f (T) gravity in the literature. There is indeed a consensus that there are extra DoF, however, the number of them is still under investigation. A recent review on these analyses is [229], while in [245] the authors review the open problems in f (T) gravity. In the first paper that appeared almost immediately after the introduction of f (T) theory, [246], the authors claim to find three new dynamical modes, meaning that in total there are five DoF in 3 + 1 dimensions. Nevertheless, it is claimed [154, 245] that, not only their arguments lack justification, but also part of the results are incorrect. In particular, in many cases it seems that all Lorentz constraints become second class, 5.3.2. Conformal and disformal transformations in f(T) gravity. As discussed in section 3.7, if one wants to construct a quantity, say a metric g̃µν from a given metric gµν and a scalar field ϕ, they come across with the notion of conformal and disformal transformations. The conformal transformation of the metric is given by g̃µν = Ω−2 (ϕ)gµν , g̃µν = Ω 2 (ϕ)gµν , (5.57) where Ω 2 = A from section 3.7. Under the same conformal transformation, the tetrad and its inverse must transform as [253–255] ẽ A µ = Ω(ϕ)e A µ , ẼA µ = Ω−1 (ϕ)EA µ . (5.58) Respectively, the volume element e transforms as ẽ = Ω4 (ϕ)e. (5.59) Having the above relations in mind, we can easily calculate how the rest of the teleparallel quantities transform under (5.58). Hence, the torsion tensor transforms as T̃A µν = ΩTA µν + 2∂[µ Ω e A ν] , (5.60) and the contortion tensor as K̃ABµ = KABµ + 2E[A α EB] β gβµ 63 Ω,ϕ ◦ ∇α ϕ. Ω (5.61) Rep. Prog. Phys. 86 (2023) 026901 Review The torsion scalar becomes where U(ψ) = V(ϕ)/F(ϕ) 2 . As can be seen from equation (5.70), conformal transformations cannot lead to the Einstein frame in f (T) gravity because of the fact that the torsion tensor is not conformally invariant. Further manipulation of the second term in equation (5.68) yields T̃ = Ω−2 T + 4Ω−3 gµν Tµ (∂ν Ω) − 6Ω−4 gµν (∂µ Ω)(∂ν Ω), (5.62) from which we can derive the inverse transformation that reads T = Ω 2 T̃ − 4Ωg̃µν T̃ν (∂µ Ω) − 6g̃µν (∂µ Ω)(∂ν Ω). 1 ∂µ F = ∂µ (ln F), F (5.63) and by substituting it back and integrating by parts we get Furthermore, the boundary term transforms as ˜◦ B = Ω 2 B̃ − 4ΩT̃µ ∂µ Ω − 18∂˜µ Ω + 6Ω□Ω. (5.64) Sf(T) = After introducing how the teleparallel quantities behave under conformal transformations, let us proceed by studying what happens with f (T) gravity. If one introduces two auxiliary fields χ and ϕ we can rewrite the gravitational part of the action (5.48) as ˆ ˆ 1 1 4 Sf(T) = d x e f(T) = − d4 x e [χ(T − ϕ) − f(ϕ)] . 2κ 2 2κ 2 (5.65) ˆ   ˜◦ ˜◦ ψ 1 d4 xẽ −T̃ − √ B̃ + g̃µν ∇µ ψ ∇ν ψ − U(ψ) , 2 3 (5.72) ẽ A µ = C(ϕ, X)e A µ + D(ϕ, X)gνρ e A ρ (∂µ ϕ)(∂ν ϕ). ẼA µ = ◦ scalar field, as is the case for f(R) gravity. Applying the conformal transformation (5.58) in equation (5.66) the action transforms into ˆ  1 Sf(T) = 2 d4 xẽ −F(ϕ) Ω−2 T̃ − 4Ω−3 g̃µν T̃ν (∂µ Ω) 2κ   −6Ω−4 g̃µν (∂µ Ω)(∂ν Ω) − Ω−4 V(ϕ) . (5.67)   1 D EA µ − gµν EA ρ (∂ν ϕ)(∂ρ ϕ) , C C − 2XD ẽ = C3 (C − 2XD) e. g̃µν = C 2 gµν + 2D(C − XD)(∂µ ϕ)(∂ν ϕ). ˆ (5.76) The disformal transformations of the rest of the teleparallel quantities are given in section 3.7. Considering now the scalar-torsion representation of f (T) gravity, given in equation (5.65) and setting (for simplicity) ϕ = f ′ (ϕ) the action is rewritten as  Sf(T) = − (5.69) 1 2κ2 ˆ d4 xe (ϕT − V(ϕ)) , (5.77) where V(ϕ) = f(T) − ϕT. Using the first of equation (5.49) to replace T in the above action we can see, as previously, that the ´ term ϕB, integrated by parts, will become −2 d4 xeTµ ∂ µ ϕ. This term takes the following form under disformal transformations the action (5.68) becomes 1 2κ 2 (5.75) One can thus immediately obtain the transformation of the metric, that reads If we introduce a new scalar field ψ so that Sf(T) = (5.74) and the volume element as V(ϕ) 3F ′ (ϕ) 2 µν g̃ (∂µ ϕ)(∂ν ϕ) − . 2F(ϕ) 2 F(ϕ) 2 (5.68) √ dψ √ F ′ (ϕ) = 3 ⇒ ψ = 3 ln F(ϕ), dϕ F(ϕ) (5.73) The inverse tetrad transforms as with V(ϕ) = ϕf ′ (ϕ) − f(ϕ). This formulation produces a scalar-tensor type of theory without a kinetic term for the + 1 2κ 2 because of equation (5.69). We thus see that a conformal transformation of f (T) gravity leads to TEGR with a non-minimal coupling of a phantom scalar field, meaning that its kinetic term has the opposite sign, to the boundary term B̃. More details on conformal transformations in f (T) gravity can be found in [253, 256] and references therein. Before we move to other theories, it is interesting to see what happens if one allows the new tetrad in equation (5.58) to depend on first derivatives of the scalar field as well. Can we get rid of the above symmetry-breaking term? In this case, the transformations are called disformal and have the form Varying this action wrt χ yields ϕ = T which shows that the above action is equivalent to (5.48), unless f ′ ′ (T) ≡ fTT = 0; in this case, the theory reduces to TEGR. Varying the action wrt the other scalar ϕ it yields χ = −f ′ (ϕ). Setting F(ϕ) = −f ′ (ϕ), we can rewrite the above action as ˆ 1 Sf(T) = d4 x e [−F(ϕ)T − V(ϕ)] , (5.66) 2κ 2 Setting Ω 2 = F(ϕ) the above action becomes  ˆ 2 µν 1 4 Sf(T) = d xẽ −T̃ + g̃ T̃ν ∂µ F(ϕ) 2 2κ F(ϕ) (5.71)   ˜◦ ˜◦ 2 1 d4 x ẽ −T̃ + T̃µ ∂µ F + g̃µν ∇µ ψ ∇ν ψ − U(ψ) , F 2 (5.70) 64 Rep. Prog. Phys. 86 (2023) 026901 T̃µ ∂˜µ ϕ = Review  D ∂µ C µ 1 − 2X Tµ ∂ µ ϕ + 3 ∂ ϕ 2 C C (C − 2D)  ◦  D α β ◦ + ∂ ϕ∂ ϕ∇α ∂β ϕ + 2X□ϕ . C (5.78) 1  full generality in section 4.1.3 and also discussed in the previous section of f (T) gravity, the antisymmetric part of the tetrad’s field equations will coincide with the equations of motion of the spin connection. We will focus on specific applications of this theory, as well as on constraints of its functional form in the following sections. In particular, in section 6.2.3 we will study its cosmological solutions, in section 7.7.2 we will see the matter density perturbation equation, its polarization modes will be presented in section 8.3 and precision cosmology will be discussed in section 10.3.3. Conformal transformations of f(T, B) gravity are studied in [256]; the authors show that generalized non-minimally coupled ‘teleparallel dark energy models’ (see section 5.8.1) are conformally equivalent to f(T, B) gravity. One particular interesting case is when f(T, B) = −T + F(B) since this theory is conformally equivalent to (5.114) with the coupling function A(ϕ) = β 2 (1 + 2βϕ√3 ) 2 . This theory only has a non-minimally coupling between the torsion scalar and the scalar field and the kinetic term has the correct sign (canonical); this will be studied in detail in section 5.8.1.2. The existence and the stability of specific relativistic solutions, such as de Sitter and scaling solutions, are studied in [257]; the authors claim that if one uses a Lagrange multiplier, then f(T, B) gravity becomes equivalent to GR plus a minimally coupled non-canonical scalar field. In [258] the authors consider two specific models of the theory, one exponential and one power-law, and they find solving the field equations numerically, that both of them are capable of reproducing the late-time acceleration of the Universe. In [259, 260] the authors constrain the function form of f using energy conditions. Thermodynamics is used in [261, 262], in order to constrain and cosmologically reconstruct f(T, B) models. The H 0 tension was studied in [263], together with cosmologically viable models; weak field limit tests were studied in [264, 265] and stability analysis in [266]. Further, bouncing cosmological solutions are considered in [267], Noether symmetries in [268]. In [269] the minisuperspace quantization is considered in cosmology. Last but not least, some other approaches of Replacing this together with equation (5.75) into the action we readily see that it is not possible to remove the symmetrybreaking terms with disformal transformations. More details on disformal transformations in f (T) gravity can be found in [82, 83] and references therein. 5.4. f(T, B) gravity Another very interesting and well-studied generalization of TEGR is the so-called f(T, B) gravity, where apart from the torsion scalar one considers the boundary term B, equation (5.50a), that is the difference between the Ricci scalar, computed from the Levi-Civita connection and the torsion scalar, equation (5.49). The action of this theory reads Sf(T,B) := 1 2κ 2 ˆ d4 x e f(T, B) + Sm . (5.79) This theory was firstly proposed in [157] by Bahamonde et al, and the motivation was, mostly, to fully incorporate the ◦ dynamics of f(R) theory. However, as one could immediately ◦ realize, f(R) is just a limiting case of equation (5.79) (only when the specific combination −T + B appears as an argument to the function), and thus f(T, B) has offers much richer phenomenology. As in any teleparallel theory, since the torsion scalar depends both on the frame field and the spin connection, f(T, B) theory will result non-trivial equations for both of them. Thus, the antisymmetric part of the tetrad field equations, will coincide with the field equations for the inertial connection. This means that, if one chooses to work on the Weitzenböck gauge, where the spin connection vanishes, the respective field equations will be solely solved by the tetrad. Let us go a step further, in order to examine quantitatively the above picture. Varying the action (5.79) wrt the tetrad we obtain its field equations that read ◦ the form f(T, R) have been considered in [270, 271], however, they are considered controversial because of the mixed framework they use, i.e. both curvature and torsion are non-zero. A complete analysis about f(T, B) cosmology will be provided in sections 6.2 and 10.3.3. ◦ ◦ ◦ 1 EA µ □fB − EA ν ∇µ ∇ν fB + BfB EA µ − (∂ν fB + ∂ν fT ) SA µν 2 1 µν B − fT ∂ν (eSA ) + fT T ν A SB νµ − fT ω B Aν SB νµ e 1 µ − fEA = κ 2 ΘA µ , (5.80) 2 5.5. Extensions of NGR Following section 5.2, NGR is described by the Lagrangian density (5.14). There are actually two more quadratic scalars that one can construct from the torsion tensor [46] P1 := uµ aµ , where f B and f T are partial derivatives of f wrt the boundary term and the torsion scalar. Details on the variations can be found in the appendix A. As one would expect, by choosing the form of the theory to be f(T, B) = f(−T + B), the above P2 := ϵµνρσ tλµν tλ ρσ , (5.81) however, they are both parity violating. The interested reader should read [234] where the authors argue that parity violating terms play a significant role in the wellposedness of the Cauchy problem. It is not so clear thought that such terms can play a fully consistent role in gravitational theories. ◦ equations coincide with the equations of motion in f(R) gravity, expressed in the tetrad formulation. As already shown in 65 Rep. Prog. Phys. 86 (2023) 026901 Review Theories beyond NGR have been proposed as well. In [46], Bahamonde et al consider the theory which assumes the Lagrangian density L = f(Taxi , Tten , Tvec ), contractions of the torsion tensor and still obtain second order field equations. However, as we will discuss in section 5.8.2 we do not know how physical such theories would be. In [275] the authors considered a generalization of NGR by adding nine functions of the d’Alembertian operator, showing that it can accommodate the ghost- and singularity-free structure that was realized in the metric theories [276–278]. Furthermore, in [279] the authors study the extensions of NGR as a reduction of the five-dimensional Kaluza–Klein theory and in [280] the equivalence between NGR and a Weyl geometry is presented. (5.82) i.e. an arbitrary function of the irreducible parts of the torsion tensor. One can also use the alternative decomposition of the torsion scalar, described by the equation (5.4), and thus consider theories of the form f(T1 , T2 , T3 ). The equations of motion for the above theory (5.82) in vacuum are given by EA µ f + δf = 0, δe A µ (5.83) 5.6. Gauss–Bonnet theories The Gauss–Bonnet invariant, depending only in higher order curvature terms where δf δf δf δf = A |axi + A |ten + A |vec , δe A µ δe µ δe µ δe µ ◦ (5.84) and each of the terms on the RHS are ◦ (5.88) (5.89) where T G is the TEGB term and BG its boundary term, as shown in section 3.8. Theories that have been studied in the literature have the form ˆ 1 Sf(T,TG ) := d4 x e f(T, TG ) + Sm . (5.90) 2κ 2 (5.85b) The equations of motion in the Weitzenböck gauge are presented in [87] by Kofinas and Saridakis. In [281] cosmological solutions using Noether symmetries were found. More general theories such as f(T, B, TG , BG ) were studied in [85]. In curvature-based gravity, the Gauss–Bonnet term is intrinsically related to the Lovelock theorem which defines the conditions by which an action for gravity can be written such that it has field equation derivatives that are at most second order in the metric. Naturally, the scalar is also related to the teleparallel analogue of this theorem in TG. The Lovelock analogue in TG as well as cosmological solutions are presented and studied in [282, 283]; while couplings with a scalar field and its kinetic term are considered in [284]. Varying the action (5.90) we get the following field equations + Tµ AB v B + vµ vA − vµ ω ρ Aρ + v B ω µ AB  +2TB ρµ + Tρ B µ − Tµ B ρ − e−1 ∂ν [e fTten (−2TA µν + Tµν A − Tνµ A − vµ EA ν +vν EA µ )] . (5.85c) The theory expressed in the Lagrangian in equation (5.82) is manifestly covariant and invariant under local Lorentz transformations, since the spin connection is nonvanishing. Alternatively, one could work on those frames where the spin connection vanishes (Weitzenböck gauge) but in that case they should employ only with good tetrads [223, 272–274]; more details can be found in section 5.3. In the same spirit, inclusion of parity violating terms and the boundary term was proposed [46], in theories of the form 1 − fT ∂ν (eSA µν ) − SA µν (∂ν fT ) + fT TB ν A SB νµ e 1 MBCD − fEA µ − fTG δIJKL ED µ KIJ M KK EB ∂A KEL C 2  1  + ∂β ηAL (YB[LH] − YH[LB] + YL[BH] )EH β EB µ 2e 1 + TIAB EH µ (YB[IH] − YH[IB] + YI[BH] ) 2e = κ2 Θ A µ , (5.91) (5.86) as well as higher order contractions of the torsion tensor such as S2 := tλµν aλ vµ aν . ◦ ◦  δf |ten = fTten −2T B Aσ TB µσ − Tµ ρσ Tρ A σ − T B ρA Tρ B µ A δe µ S1 := tλµν vλ aµ vν , ◦ G = TG + BG ,  δf |vec = 2fTvec vµ ω ρ Aρ − v B ω µ AB − Tµ AB vB − vµ vA A δe µ L = f(Taxi , Tten , Tvec , B, P1 , P2 ), ◦ could be considered as a subclass of higher order theories. In the teleparallel framework its form becomes  δf 2 |axi = − ϵBC DJ fTaxi a B ED µ TC AJ − EJ µ ω C AD A δe µ 3  +e−1 ∂ν e ϵBA CD fTaxi a B EC ν ED µ , (5.85a) − 2e−1 ∂ν [e fTvec (vµ EA ν − vν EA µ )] , ◦ G = R 2 − 4Rµν Rµν + R αβµν Rαβµν , (5.87) In principle, because of the nature of the torsion tensor and the fact that it depends only on first derivatives of the tetrad (in contrast with the Ricci scalar), one can construct infinite-order where we introduced the following tensors  CABD YB IJ := efTG XB IJ − 2δELKJ ∂µ efTG ED µ KL C E KIA K , 66 (5.92) Rep. Prog. Phys. 86 (2023) 026901 Review and 5.8. Scalar-tensor theories ABCD BACD XA IJ := KJ E B KK FC KFL D δIEKL + KE IB KK FC KFL D δEJKL Details about the variations can be found in the appendix A. Applications of this theory will be discussed in the following sections. Another very well motivated class of theories, capable of describing the accelerating expansion of the Universe at present and early times has been the scalar-tensor theories. The first and probably the simplest scalar-tensor theory introduced, was the Brans–Dicke theory [287] which describes a linear non-minimal coupling of a scalar field to gravity that introduces a new, scalar DoF to the theory and the gravitational constant becomes dynamical. In that spirit, gravity is described both by the dynamics of the metric and the scalar field. In principle there are no restrictions in the number of scalar fields introduced. 5.7. Higher-order derivatives 5.8.1. Scalar-tensor theories with couplings with T and B. K EF L CBAD DBCA + K EC K B KJ D δKFIL + KF ED KEL B KK IC δFLKJ ABCD BACD + 2KK EB KEL F KF CD δIJKL + 2KKE B KJ L F KF CD δKEIL FBCD DBMA + 2KEL F KK IB KA CD δELKJ + 2KFC D KD EB KEL I ηMJ δFCKL ABCD BACD + 2KK EB δIJKL ∂D KEL C + 2KKE B δKEIL ∂D KJ L C . (5.93) In the same spirit, instead of coupling the scalar field to the metric and invariants constructed from the Levi-Civita connection, it is interesting to see what happens in the teleparallel framework, meaning if we couple it to the tetrad and invariants constructed from the teleparallel connection. These theories have been known as scalar-torsion theories and in this section we will present interesting and well-studied models, their field equations as well as how they transform under conformal transformations. Inspired mostly by the curvature case where theories with higher order derivative terms were studied, such ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ as f(R, Rµν Rµν , R αβµν Rαβµν , □R, (∇R) 2 , . . .) and more, together with the fact that such terms arise naturally in dimensional reductions of higher-dimensional theories, such as Kaluza–Klein, and in general in quantum corrections [285], the higher-order teleparallel theory described by the action ˆ ◦ ◦ 1 SHOT := (5.94) d4 x e f(T, (∇T) 2 , □T) + Sm , 2 2κ Before we study specific scalar-torsion models, we will start from a more general action that includes an arbitrary function of the torsion scalar T, the boundary term B, a scalar field ϕ and its kinetic term 5.8.1.1. Action and field equations. where ◦ ◦ ◦ ◦ ◦ (∇T) 2 ≡ gµν (∇µ T)(∇ν T) = η AB EA µ EB ν (∇µ T)(∇ν T), (5.95) ◦ X ≡ −(1/2) gµν ∂µ ϕ∂ν ϕ = −(1/2)(∇ϕ) 2 . Its action reads and ◦ ◦ ◦ ◦ ˆ ◦ µν AB µ ν □T ≡ g ∇µ ∇ν T = η EA EB ∇µ ∇ν T, Sf(T,B,ϕ,X) := (5.96) was proposed by Otalora and Saridakis [286]. The field equations in the Weitzenböck gauge for this theory become  1 d xe f(T, B, ϕ, X) + Lm . 2κ2 (5.98) Variations of equation (5.98) wrt the tetrad yield the field equations h i ◦ ◦ ◦ 1 δβµ □fB − ∇µ ∇β fB + BfB δβµ − (∂ν fB ) + (∂ν fT ) Sβ µν 2 1 1 A µν − fT e β ∂ν (eSA ) + fT TB νβ SB νµ − fT ω B βν SB νµ − fδβµ e 2 1 µ 2 µ (5.99) − fX ∂ ϕ∂β ϕ = κ Θβ , 2 1 1 − fT ∂ν (eSA µν ) − SA µν (∂ν fT ) + fT TB ν A SB νµ − EA µ f e 2     2 1X ∂Xi 1 ∂Xi − fX i A − ∂α efXi 2 ∂e µ e ∂(∂α e A µ ) i=1   ∂Xi − ∂α ∂ν efXi ∂(∂α ∂ν e A µ )   1 ∂X2 + ∂σ ∂α ∂ν efX2 2e ∂(∂σ ∂α ∂ν e A µ )   1 ∂X2 − ∂β ∂σ ∂α ∂ν efX2 = κ2 Θ A µ , 2e ∂(∂β ∂σ ∂α ∂ν e A µ ) (5.97) ◦  4 while the field equations of the spin connection will be the antisymmetric part of equation (5.99), h i W[µν] = 2 (∂ρ fB ) + (∂ρ fT ) S[µ ρ ν] = 3Tρ [µν ∂ρ] (fT + fB ). (5.100) Additionally, the field equations for the scalar field read   ∂µ efX gµν ∂ν ϕ + efϕ = 0. (5.101) ◦ where Xi = ((∇T) 2 , □T) and then fXi = ∂f/∂Xi . In principle, such theories that contain higher order derivatives in the field equations are vulnerable to Ostrogradsky ghosts. However, as stated in [286] this is not necessarily the case here since the theory is formulated on different foundations it may be an Details about the variations can be found in the appendix A. An important theory that can be constructed from the above action is the following ◦ indication of extra DoF rather than ghosts, as in f(R) gravity. 67 Rep. Prog. Phys. 86 (2023) 026901 Review  ˆ 1 d xe (f(T) + F1 (ϕ)T + F2 (ϕ)B) 2κ2  ◦ ◦ 1 µ − ω(ϕ)∇µ ϕ∇ ϕ + V(ϕ) + Sm , 2 S= STTST := ◦ ◦ ◦ whereas the modified Klein–Gordon equation (5.101) is reduced to ◦ × (F1′ (ϕ)T + F2′ (ϕ)B) = 0. " dV 1 + dϕ 2κ2 f(T, B, ϕ, X) = −T + 2κ ˆ 2 (5.104) s # 1 2X 1 f(ϕ)T + g(ϕ)B + V(ϕ) 1 − . 2 2 V(ϕ) (5.108) As already mentioned, the action (5.98) is a very general one and incorporates a lot of interesting scalar-torsion models. One of them is the ‘Teleparallel dark energy’ model [288]. Motivated mainly by the quintessence model in GR [289– 291], its action reads STel. DE := (5.107) where V(ϕ) is the potential. As one can immediately see, considering f(ϕ) = −g(ϕ), a non-minimal coupling between the scalar field and the Ricci scalar is recovered because of the first relation in equation (5.49); further, the boundary term does not play any role when g(ϕ) = constant and it can be shown that such theories allow the crossing of the phantom divide line [298, 299]. The coupling of the scalar field with the boundary term emanates from the fact that without fine-tuning, the system will evolve to a late-time accelerating attractor solution [295, 296]. This theory is a sub-class of equation (5.98), and thus the equations of motion will be exactly the same with (5.99) and (5.101), if we set − (F1 (ϕ) + fT )Gµβ + δβµ □F2 (ϕ) − ∇µ ∇β F2 (ϕ)   1 − ∂ν fT + F1 (ϕ) + F2 (ϕ) Sβ µν − δβµ (f − TfT ) 2 − fT e A β ω B Aν SB νµ − κ2 ω(ϕ)(∂ µ ϕ)(∂β ϕ) 1 + κ2 ω(ϕ)(∂ν ϕ)(∂ ν ϕ)δβµ − κ2 V(ϕ) = κ2 Θµβ , (5.103) 2 ω(ϕ)□ϕ + ω ′ (ϕ)(∂α ϕ)(∂ α ϕ) + T 1 1 + f(ϕ)T + g(ϕ)B 2 2κ 2 2 s # 2X +V(ϕ) 1 − + Lm , V(ϕ) d xe − 4 (5.102) where there are non-minimal couplings between the torsion scalar T and the boundary term B. For this specific theory, the field equation (5.99) can be rewritten as ◦  ˆ 4 A dynamical systems analysis of this theory is presented in [297] and it shown that the tachyonic field might be a good candidate for the solution of the cosmic coincidence and the dark energy problems. In a series of papers [127, 132, 300, 301] the most general teleparallel scalar-torsion theories of gravity in their covariant formulation are discussed. The only two necessary restrictions are invariance of the theory under diffeomorphisms and local Lorentz invariance as well as no coupling between the matter fields and the spin connection. The general formalism and the relation between different classes of theories with conformal transformations are discussed in [300]; a discussion on multiple scalar fields is also presented, as is also done in [132, 302]. In the second paper of the series [301] a Lorentz invariant theory is presented that consists of the torsion scalar, a scalar field, its kinetic term and a derivative coupling between the torsion and the scalar field and its Lagrangian reads     1 1 2 T µ d4 xe − + ξϕ − ∂ ϕ∂ ϕ + V (ϕ) + Sm , µ κ2 2 2 (5.105) where we have set ω(ϕ) = 1, f(T) = −T and F1 = κ2 ξϕ, F2 = 0 in equation (5.102). Some Teleparallel dark energy models were found to be invariant under the Gasperini–Veneziano duality transformation [292, 293]. Obviously, when the coupling of the scalar field is minimal, the theory coincides with the known quintessence models; but once the non-minimal coupling is switched on, the theory presents a much richer phenomenology. In particular, one can obtain a dark energy sector that behaves not only as quintessence, but also as phantom or phantom-divide crossing during evolution. This means that phantom behavior appears naturally without the need of ghost fields which come with ambiguous quantum behavior. The theory (5.105) was further extended to include other irreducible parts of the torsion scalar as well [294]. A specific class of theories of equation (5.102) with f(T) = −T was considered in [295, 296]    ˆ 1 T 1 S = d4 xe − + F(ϕ) − G(ϕ)B 2 κ 2 2  ω(ϕ) − ∂µ ϕ∂ µ ϕ + V(ϕ) + Sm . (5.106) 2 Sf(T,Y,ϕ,X) := 1 2κ2 ˆ d4 x e f(T, Y, ϕ, X) + Sm , (5.109) where X is the kinetic term of the scalar field ϕ and Y is the coupling between the torsion and the derivative of the scalar field Y = gµν Tρ ρµ ∂ν ϕ. (5.110) Part of this action, without the derivative coupling has been studied in [303]. It is interesting to note that linear terms of Y in the action, either minimally or non-minimally coupled, would lead to a B−like contribution after an integration by parts; however, non-linear terms that could be possibly be part of the theory lead to more interesting phenomenology. Variations of the action (5.109) wrt to the tetrad yield the field equations Another interesting class of theories are the so-called tachyonic teleparallel scalar-torsion theories, where a generalized non-minimal coupling of the scalar field with the torsion scalar and the boundary term is considered [297]. Their action reads 68 Rep. Prog. Phys. 86 (2023) 026901 Review  1 µ◦ρ 1 1 ◦µ ∇ fY ∂β ϕ − δβ ∇ ( fY ∂ρ ϕ) + fY 2 2 2   1 ρ µ νµ ρ × g T(βν) ∂ρ ϕ + T β ∂ρ ϕ + Tρ ρβ ∂ µ ϕ − (∂ν fT )Sβ µν 2 1 A − fT e β ∂ν (eSA µν ) + fT TB νβ SB νµ − fT ω B βν SB νµ e 1 µ 1 µ (5.111) − fδβ − fX ∂ ϕ∂β ϕ = κ2 Θµ β. 2 2 with U(Ω) = Ω−4 V(ϕ). We can rewrite the second term in the above action (5.117) introducing a function G(Ω) such that Ω−3 A(Ω)∂µ Ω = ∂µ G(Ω). Integrating equation (5.117) by parts we get ˆ   1 STDE = 2 d4 xẽ −A(Ω)Ω−2 T̃ − 2G(Ω)B̃ + U(Ω) . 2κ (5.119) Further, variations of the action wrt to the spin connection coincide with the antisymmetric part of the tetrad equation (5.111), meaning that only those pairs of tetrads and spin connection that solve both field equations are considered solutions of the theory. The spin connection equations read 3 3∂[ρ fT Tρ µν] + ∂[µ fY ∂ν] ϕ − fY Tρ [µν ∂ρ] ϕ = 0. 2 We can see now that there is no kinetic term for the scalar field Ω, meaning that it is just an auxiliary field. Varying the action (5.119) and solving the respective equations for Ω, we obtain a function of the form Ω = Ω(T̃, B̃). This means that the integrand in equation (5.119) is just a function f(T̃, B̃), thus making the teleparallel dark energy model (5.114) conformally equivalent to f(T, B) gravity. A more general class of disformally invariant theories can be found in [83]. (5.112) Finally, variations wrt the scalar field yields the Klein–Gordon equation ◦ gµν ∇µ (fY Tρ ρν − fX ∂ν ϕ) − fϕ = κ2 Θ, The motivation for Horndeski’s theory was to write down the most general Lagrangian with a single scalar field, that leads to second order field equations. Almost 30 years after its publication, it was realized that this theory is completely equivalent to generalized covariant galileons [304–306]. However, part of it was severely constrained after the GW 170817 and GRB 170817A events, which was the motivation to study if the constraints persisted in the teleparallel framework as well. Bahamonde–Dialektopoulos–Levi Said (BDLS) theory of gravity is nothing but the teleparallel analogue of the known Horndeski gravity, i.e. the most general scalar-tensor theory with a single scalar field in four dimensions leading to second order field equations, satisfying some necessary conditions to retain a finite class of theories. The theory itself was formulated in [307]. The necessary criteria for the theory to be formulated are the following: 5.8.2. Teleparallel Horndeski gravity—BDLS theory. (5.113) where Θ is the trace of the energy-momentum tensor. Conformal transformations as well as relation of scalar-torsion to scalar-tensor theories are discussed in [127, 301]. In addition, disformal transformations in scalar-torsion gravity are discussed in [83] as well as in section 3.7. 5.8.1.2. Conformal transformations teleparallel dark energy. Considering a teleparallel dark energy model of the form   ˆ 1 1 4 µ d xe −A(ϕ)T − ∂ ϕ∂ ϕ + V(ϕ) + Sm , STDE := µ 2κ 2 2 (5.114) and applying a conformal transformation of the form (5.58) in an attempt to remove the kinetic term of the scalar field, we end up with STDE = ˆ 1   d xẽ − A(ϕ) Ω 2 T̃ − 4ΩT̃µ ∂µ Ω − 6g̃µν (∂µ Ω)  1 ×(∂ν Ω)) − Ω 2 g̃µν ∂µ ϕ∂ν ϕ + V(ϕ) , 2 • The field equations both for the tetrad and for the scalar field must be of second order, • the scalar invariants should not be parity-violating, • and contractions of the torsion tensor must be at most quadratic. 4 2κ 2 (5.115) The first condition appears in order to avoid ghost instabilities while the second one refers to contractions of the scalar field (through its derivatives) with the irreducible parts of the torsion tensor. Those scalars should be invariant under parity transformations. The last one is a particular feature of TG. In GR and its extensions, thus in Horndeski gravity as well, Lovelock’s theorem prohibits other terms to be part of these Lagrangians. In TG though, this is not the case; the torsion tensor consists of first derivatives in the tetrad, in contrast with the Riemann tensor that consists of second derivatives in the metric. This means that, in principle, infinite contractions of the torsion tensor could be considered and still the field equations would be second order. However, it is unclear how physical where we considered only the gravitational part of the action (5.114). In order for the kinetic term of the scalar field to vanish, we have to set ! ˆ 1 p Ω = exp dϕ , (5.116) 2 3A(ϕ) and expressing it in its inverse form, meaning ϕ = ϕ(Ω) we can substitute it in equation (5.115) and get 1 STDE = 2 2κ ˆ (5.118) h i d4 xẽ −A(Ω)Ω−2 T̃ + 4Ω−3 A(Ω)T̃µ ∂µ Ω + U(Ω) , (5.117) 69 Rep. Prog. Phys. 86 (2023) 026901 Review L2 := G2 (ϕ, X), such higher order contributions will be. For this reason, in [307], it was demanded that, the contributing scalar invariants of the theory be at most quadratic contractions of the torsion tensor. In order to form the most general Lagrangian that satisfies the above criteria, we should formulate first the one in the tangent space and then raise it to the general manifold through the coupling prescription. The theory considers also the following conditions for the scalar field [308]: ◦ L3 := G3 (ϕ, X)□ϕ, (5.122c) ◦ 1 L5 := G5 (ϕ, X)Gµν ϕ;µν − G5,X (ϕ, X) 6  3  ◦  ◦ ν × □ϕ + 2ϕ;µ ϕ;ν α ϕ;αµ − 3ϕ;µν ϕ;µν □ϕ , (5.122d) ◦ ci L i , ◦ (5.120) where ci are arbitrary constants and the subscript refers to the number of appearances of the scalar field in each component of the Lagrangians which are (5.121a) L2 = X, (5.121b) ¯ L3 = X□ϕ, (5.121c) I1 := tµνσ ϕ;µ ϕ;ν ϕ;σ , ,µν ¯ 2 + (□ϕ)ϕ, ¯ L4 = −X(□ϕ) µ ϕ,ν ϕ + Xϕ,µν ϕ,µν −ϕ,µ ϕ,µν ϕ,νρ ϕ,ρ , ◦ G5 (ϕ, X)Gµν ϕ;µν , where Gµν is the Einstein tensor formulated in teleparallel geometry, i.e. equation (4.168). That is because the covariant derivatives do not commute, in order to compensate for the higher order terms that would appear in the field equations, we have to add these extra terms. However, the theory is not complete yet. In the teleparallel geometry there exist more terms that satisfy the aforementioned criteria. Specifically, if we consider up to quadratic contractions of the torsion tensor, criterion (c), the most general Lagrangian satisfying the conditions explained above of TG (without a scalar field) is f(Taxi , Tvec , Tten ) that was discussed in section 5.5, which means that (at least) these invariants, i.e. the irreducible decomposition of the torsion tensor should be part of our theory. Indeed, in TG there exists a set of scalars that should be included in a new component of the Lagrangian, LTele , which consist of the irreducible parts of the torsion tensor, contracted with derivatives of the scalar field. Linear contractions of the torsion tensor are i=1 L1 = ϕ, ◦ where Gi,X = ∂Gi /∂X and ϕ;µν = ∇µ ∇ν ϕ. In the L4 and L5 Lagrangians there are two new terms, G4 (ϕ, X)(−T + B) and The Lagrangian for a scalar field ϕ with a kinetic term X = − 21 ∂µ ϕ∂ µ ϕ, that is invariant under the Galilean transformation ϕ → ϕ + bµ xµ + c, in the tangent space has the form 5 X (5.122b)   2  ◦ ;µν L4 := G4 (ϕ, X) (−T + B) + G4,X (ϕ, X) □ϕ − ϕ;µν ϕ , (a) The Lagrangian contains up to second order derivatives of the scalar field, (b) the Lagrangian is a polynomial in second order derivatives of the scalar field, (c) the corresponding field equations are at most second order in derivatives of the scalar field (which coincides with the first criterion above). L= (5.122a) I2 := vµ ϕ;µ , I3 := aµ ϕ;µ . (5.123) This is the full set of scalar that can be constructed considering linear contractions of the torsion tensor because of the symmetry of tµνσ in its first two indices and the fact that tσµ σ = tσ σ µ = tµσ σ = 0. However, due to the the fact that t(µνρ) = 0, it can be easily shown that I 1 vanishes. Moreover, conforming with the second criterion, I 3 cannot be considered, together with any scalar terms that contains an odd number of the axial part of the torsion tensor, because it is not parity invariant. Contracting the tensorial part of the torsion tensor with second order derivatives of the scalar field would result higher order derivatives in the field equations. Following the third criterion mentioned above, the theory includes also quadratic contractions of the torsion tensor with derivatives of the scalar field. Those scalars that are not parityviolating are (5.121d) ,µν ¯ 3 − 3(□ϕ) ¯ 2 ϕ,µ ϕ,ν ϕ,µν + 6X□ϕϕ, ¯ L5 = −2X(□ϕ) µν ϕ ,ρ ,µν ¯ + 6□ϕϕ, ϕ,νρ −4Xϕ,µ ν ϕ,ν ρ ϕ,ρ µ µϕ ϕ + 3ϕ,µν ϕ,µν ϕ,ρ ϕ,λ ϕ,λρ − 6ϕ,µ ϕ,µν ϕ,νρ ϕ,λρ ϕ,λ . (5.121e) ¯ = ∂µ ∂ µ which changes The d’Alembertian is given by □ in any gravitational theory since it is a derivative operator, comma ‘,’ denotes partial derivative and barred quantities refer to Minkowski spacetime. Covariantizing the action (5.120), i.e considering it in a general non-flat manifold, we first need to replace the arbitrary constants ci to general functions of the form Gi (ϕ, X). In this way, the Lagrangians (5.121a)–(5.121e) take the form J1 := aµ aν ϕ;µ ϕ;ν , J2 := vµ vν ϕ;µ ϕ;ν , J3 := vσ tσµν ϕ;µ ϕ;ν , J4 := vµ tσµν ϕ;σ ϕ;ν , (5.124a) 70 Rep. Prog. Phys. 86 (2023) 026901 Review J5 := tσµν tσ µ̄ ν ϕ;µ ϕ;µ̄ , J6 := tσµν tσ µ̄ν̄ ϕ;µ ϕ;ν ϕ;µ̄ ϕ;ν̄ , J7 := tσµν tσ̄µ̄ σ ϕ;µ ϕ;ν ϕ;σ̄ ϕ;µ̄ , (5.124b) important theories that are part of the BDLS theory along with some important references. This results in richer phenomenology because it contains new terms through the extra Lagrangian LTele and because of this extra terms it revives several Horndeski models that were severely constrained in the curvature formulation after the GW 170817 and GRB 170817A events. More details can be found in [307, 309] and also in section 7 were cosmological perturbations are discussed. Recently, it was found that the most general BDLS theory propagates up to 7 DoF. In cosmology, it was also found that this theory can generate a late-time, low energy de-Sitter vacuum state by using the well-tempered method [310, 311]. Keeping with recent developments in cosmology, in [312] Noether’s theorem was used to classify the many models that can be produced in this new class of theories. We conclude this section by mentioning a similar theory introduced in [313] which considers a teleparallel vector-tensor theory constructed as an analogue of the so-called generalized Proca theory. This theory assumes a vector Aµ field coupled to gravity which only propagates up to 3 DoF in order to avoid Ostrogradsky ghosts and comply also with the representation of the massive spin 1 Lorentz group. This theory contains up to second order derivatives in the field equations (for the tetrads and the vector) and only parity preserving scalars were considered. The field equations of the theory are of second order for both the tetrad and the vector field and only parity preserving scalars were considered. One important difference wrt teleparallel Horndeski is the fact that the scalars constructed from the torsion tensor were considered to be only linear. J8 := tσµν tσµ ν̄ ϕ;ν ϕ;ν̄ , J9 := tσµν tσ̄µ̄ν̄ ϕ;σ ϕ;µ ϕ;ν ϕ;σ̄ ϕ;µ̄ ϕ;ν̄ , (5.124c) J10 := ϵµ νρσ aν t αρσ ϕ;µ ϕ;α . Nevertheless, these scalars are not all independent with each other. In particular, one can notice that J2 = I22 , J3 = J4 and J7 = −2J6 . In addition, as with I 1 , the vanishing of the totally symmetric part of the tensor torsion makes J 9 vanish. In addition, contractions of quadratic order torsion tensor with second order derivatives of the scalar field, lead to higher order field equations. This means that the most general theory that satisfies the above criteria ends up having seven independent scalars in total and apart from the sum of (5.122a)–(5.122d) it includes also an extra Lagrangian that has the form LTele := GTele (ϕ, X, T, Taxi , Tvec , I2 , J1 , J3 , J5 , J6 , J8 , J10 ), (5.125) where, instead of choosing its tensorial decomposition, we used the torsion scalar itself, which is possible due to the relationship in equation (5.4). Notice that instead of the three irreducible parts of the torsion tensor, we considered the torsion scalar with its axial and its vectorial parts. What is also interesting to notice is that the boundary term B does not contribute in equation (5.125). Even though B itself is second order in the Lagrangian, it produces the fourth order elements of the theory, thus not satisfying the first criterion15 . This is similar to ◦ the standard Horndeski theory case where f(R) does not appear in the action since that term contains higher order derivatives. Thus, the Lagrangian of BDLS theory, i.e. the teleparallel analogue of Horndeski gravity is given by SBDLS := 1 2κ2 ˆ 1 X 2κ2 5 d4 x eLTele + 5.9. Theories non-minimally coupled with matter Another route that became popular in modified gravity is to construct theories with non-minimally couplings between the matter sector and gravity. The most popular of them is the ˆ d4 x eLi + Sm . ◦ so-called f(R, Θ) where Θ := Θµ µ is the trace of the energy- i=2 ◦ (5.126) momentum tensor and R the Ricci scalar [195]. Due to the nature of these theories, the continuity equation may also be modified leading to further generalizations and richer phenomenology. In TG, a similar theory was introduced in [322, 323] by considering the following action Once one sets LTele = 0, the action takes the standard Horndeski form, formulated in the teleparallel geometry. Variations of the scalars and the arbitrary functions, together with the equations of motion of the tetrad and the scalar field can be found in the appendix A. Before we move on, we would like to remark the motivation for BDLS theory. Apart from a competitive exercise and an analytically interesting theory, that is why Horndeski formulated his theory in the first place, it is an interesting terrain to test the possible supremacy of teleparallel model over their curvature-based counterparts. BDLS not only contains the standard Horndeski theory, as well as most of the extended teleparallel theories of gravity (e.g. f (T), f(Taxi , Tvec , Tten ), teleparallel dark-energy models, Gauss–Bonnet theory in the conformal frame and more) but it extends it just by changing the background geometry. The figure 11 shows some 15 These contributions make f( 1 Sf(T,Θ) := 2 2κ ˆ d4 x e f(T, Θ) + Sm , (5.127) where T is the torsion scalar instead of the Ricci scalar. By taking variations wrt the tetrads, we find the following field equations 1 − (∂ν fT )Sβ µν − fT eA β ∂ν (eSA µν ) + fT TB νβ SB νµ e 1 B νµ − fT ω βν SB − fδβµ 2  δΘαβ  2 µ = κ Θβ − fΘ Θµβ + e A ν g αβ A , (5.128) δe µ ◦ R) theory have fourth order field equations. 71 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 11. Relationship between the Teleparallel Horndeski analogue and various theories. 72 Rep. Prog. Phys. 86 (2023) 026901 Review ◦ ◦ ◦ 1 δβµ □fB − ∇µ ∇β fB + BfB δβµ − [(∂ν fB ) + (∂ν fT )] Sβ µν 2 1 − fT eA β ∂ν (eSA µν ) + fT TB νβ SB νµ − fT ω B βν SB νµ e 1 µ 1 1 − fδβ − fL Lm δβµ = fL Θµβ , (5.133) 2 2 2 where fΘ = ∂f/∂Θ and the variation appearing on the RHS of the equation can be expanded as e A λ gµν δΘµν ∂ 2 Lm ρ ρ νρ αβ = −2Θ + δ L − 2g g . λ λ m δe A ρ ∂gλν ∂g αβ (5.129) Here Lm is the density matter Lagrangian. It is common to consider that the matter Lagrangian does not depend on the derivatives of the metric (or equivalently derivatives of the tetrads), therefore, the last term in the above equation is usually neglected. If one assumes that the matter is described by a perfect fluid with energy density ρ and pressure p, then Θµν = (ρ + p)uµ uν − pgµν with uµ being the 4-velocity of the fluid which is normalized as uµ uµ = +1. For this particular case, we can take Lm = −p and the above term becomes e A µν δΘµν = −2Θλ ρ − pδλ ρ . λg δe A ρ where fL = ∂f/∂Lm . The TEGR case is then recovered when f = −T + Lm and further modifications can be made by other forms of f. For this theory, one also finds that the energymomentum tensor is non-conserved. The process of induced particle production was recently studied in [332] without including the boundary term. 5.10. Non-local theories One of the most well known proposals for a non-local modification in gravity, and specifically of the Einstein–Hilbert action, was suggested in [197], where their action reads ˆ ◦ ◦  √ ◦ 1 SNL := 2 d4 x −gR 1 + f(□−1 R) + Sm , (5.134) 2κ (5.130) This particular choice was assumed in [323] too in order to analyze the cosmology of this theory filled with a standard perfect fluid. It can be shown than in this theory, the standard energy-momentum tensor is non conserved [324]  ◦ ◦ fΘ µν (Θµν + gµν p) ∇µ log fΘ ∇µ Θ = 2 κ − fΘ   ◦ 1 µν +g ∇µ p + Θ , (5.131) 2 where f is an arbitrary function that depends on the retarded Green function evaluated at the Ricci scalar, i.e. ˆ ◦ −1 4 ′ ′ ′ ′ (5.135) □ F(x) ≡ d x e(x )F(x )G(x, x ). Such corrections arise naturally when one considers quantum loop effects, while they are considered as a possible solution to the black hole information paradox [333, 334]. Since then, non-locality theories have been studied in different contexts and have given some interesting results in open problems of cosmology, see more details in [335] and references therein. Some interesting non-singular and ghost-free higher order theories that include torsion are studied in [336, 337]. For the same reasons, the non-local TG (TNL) was proposed in [338]. Its action has the form ˆ ˆ ◦  1 1 STNL := − 2 d4 x e T + 2 d4 x e Tf □−1 T + Sm , 2κ 2κ (5.136) which can be interpreted as an additional force which is usually labelled as the fifth force [195]. One can further include the boundary term B in the action. Further, for that extended theory, the process of quantum fluctuations induced by mattergravity couplings were studied [325]. Further, it was found that these fluctuations are compatible with cosmological observations [326]. Another route for constructing theories with couplings between matter and the gravitational sector is by considering actions with couplings between the Lagrangian density and the teleparallel scalars. The first theory considering this was formulated in [327] where the authors considered an action based on f1 (T) + (1 + λf2 (T))Lm (without Sm ) where λ is a constant. This theory was further extended including the boundary term in [328] with the action ˆ 1 Sf(T,B,Lm ) := 2 d4 x e f(T, B, Lm ). (5.132) 2κ where the non-local operator, i.e. the retarded Green function is evaluated at the torsion scalar T. The authors show that the theory is consistent with cosmological data from SNeIa + BAO + CC + H 0 observations. Varying this action wrt the tetrad we get the field equations of the theory that read   1 1 − SA βµ ∂µ + ∂µ (eSA βµ ) − T σ µA Sσ µβ − TEA β e 2 h ◦ i ◦ ◦ × f(□−1 T) + □−1 (Tf ′ (□−1 T)) i ◦ ◦ 1 h◦ + GA β − ∂ρ □−1 (Tf ′ (□−1 T))  2  × ∂σ T gσρ EA β − 2gβ(ρ EA σ) = κ2 ΘA β . (5.137) The above theory also contains the theory f(T, Lm ) which was firstly presented in [329]. These theories are the teleparallel versions of other non-minimally matter-gravity coupled theories constructed from a Riemannian case with the Ricci ◦ scalar R being the quantity which is coupled with Lm [330, ◦ 331]. Furthermore, the f(R, Lm ) case can be derived from equation (5.132) by considering the case f(−T + B, Lm ). The variations of the above action wrt the tetrads yields 73 Rep. Prog. Phys. 86 (2023) 026901 Review ◦ 1 (1 − ξ f(ϕ, φ) + θ)EA β Gβ µ + (∂ λ θ)(∂λ ϕ)EA β 2 1 β 1 1 β  α − ∂ θ (∂α ϕ)EA − (∂ ϕ)(∂α θ)EA α + (∂ λ ζ)(∂λ ψ)EA β 2 2 2 1 β 1 β α α − (∂ ζ)(∂α ψ)EA − (∂ ψ)(∂α ζ)EA + EA ρ Sρ µβ ∂µ 2 2 h i  ◦ ◦ ◦ × (ξ + χ)f(ϕ, ψ) − θ − ζ + EA µ ∇µ ∇β − EA β □ A generalization of equation (5.136) was proposed in [335] where the authors introduced also the effect of the d’Alembertian operator on the boundary term B. Specifically, the action of the generalized teleparallel non-local theory (GTNL) has the form ˆ ˆ 1 1 4 d x e T + d4 x e (ξT + χB) 2κ 2 2κ 2 ◦  ◦ × f □−1 T, □−1 B + Sm . (5.138) SGTNL := − × (ζ − χf(ϕ, φ)) = κ2 ΘA β . (5.142) It would be interesting to study the behavior of the theory at lower scales, e.g. in astrophysical systems, since cosmologically it seems very promising. In the second term ξ and χ are two constant coupling terms. It can be easily seen that by choosing ξ = −χ = −1 one can ◦ obtain the Ricci scalar, since R = −T + B. From the definition of the inverse d’Alembertian operator (5.135) we directly see that 5.11. Other approaches There are other teleparallel theories that are less popular in the literature but it is worthwhile to mention them. For example, −1 −1 −1 (5.139) there are some approaches to the so-called Kaluza–Klein the□ R = −□ T + □ B. ory in the context of TG. The standard version of Kaluza– ◦ ◦ Effectively, this means that in the case when f(□−1 T, □−1 B) = Klein theory formulates a five-dimensional theory which uni◦ ◦ fies electromagnetism with gravity with the Lagrangrian being f(−□−1 T + □−1 B) and simultaneously ξ = −χ = −1 we composed by a five-dimensional Einstein–Hilbert action. In recover the action (5.134). With the equations of motion being the teleparallel equivalent of Kaluza–Klein theory [342, 343], too complicated to study the theory, it is less cumbersome only the internal (or tangent) space is five-dimensional but to study a localized version of equation (5.138), introducing the spacetime remains four-dimensional. This formulation scalar fields in the form of Lagrange multipliers. This method is achieved by considering a five-dimensional translational was also used for the theory (5.134) in [339]. gauge theory which turns out to unify gravity with electroRewriting the action (5.138) in a more suitable way using magnetism. Later, it was found that the Kaluza–Klein theory four scalar fields ϕ, ψ, θ, ζ, it takes the following form reduction of TEGR at low energies in the absence of elecˆ tromagnetism introduces a non-minimally coupling between h ◦ 1 SGTNL = 2 d4 x e −T + (ξT + χB)f(ϕ, φ) + θ(□ϕ − T) a scalar field and the torsion tensor [344]. The Lagrangian of 2κ this reduced theory becomes L = ϕT + 2Tµ ∂µ ϕ, which can be i ◦ +ζ(□φ − B) + Sm (5.140a) integrated out to find that the scalar field is non-minimally coupled to both the torsion scalar and the boundary term B. This reduced theory was then generalized to the case of f (T) ˆ h 1 gravity in the Kaluza–Klein reduced formulation. 4 µ = 2 d x e − T + (ξT + χB)f(ϕ, φ)θT − (∂µ ζ)(∂ φ) Another approach that has been explored in TG is the so2κ i called mimetic gravity. The standard version of mimetic gravµ − (∂µ θ)(∂ ϕ) − −ζB + Sm , (5.140b) ity is a modified version of GR which respects conformal symmetries as an internal DoF. This is achieved by rewriting the where the last two terms in the first line act as Lagrange mul- metric as gµν = −g̃µν g̃ αβ ∂α ∂β ϕ where g̃µν is an auxiliary tipliers. The mixed kinetic terms of the auxiliary fields intro- metric and ϕ is a scalar field [345, 346]. This parametrizaduce ghost-like terms that cannot be gauge transformed away tion is invariant under Weyl transformations. To be consist[340]. This would in principle lead to non-viability of the the- ent, one must impose that gµν ∂µ ∂ν ϕ = −1. These types of ory, however, one can choose the mass of these modes to be theories are usually formulated using Lagrange multipliers. larger than the cut-off of the theory. A ghost-free non-local For example, the simplest modification is constructed with ◦ theory has also been presented in [341]. Varying this action wrt the Ricci scalar as L = R + λ(gµν ∂µ ∂ν ϕ + 1). These theor◦ ◦ −1 −1 θ and ζ we get ϕ = □ T and φ = □ B respectively, while ies have the interesting feature that they can unify dark matter varying the action wrt ϕ and φ we get with dark energy. In [347], the authors formulated a teleparallel version of mimetic gravity by considering the Lagrangian ◦ ◦ ∂f(ϕ, φ) ∂f(ϕ, φ) L = f(T) + λ(gµν ∂µ ∂ν ϕ + 1). This is achieved by considering , □ζ = −(ξT + χB) , □θ = −(ξT + χB) ∂ϕ ∂φ that the tetrads are unchanged and the conformal transforma(5.141) tions are performed in the Minkowski metric. Last but not least, an interesting theory has been proposed which are the field equations for the dynamical scalar fields. in [348], which is a generalization of [349], where the authors The corresponding field equations for the tetrad in the study a general teleparallel theory that is a linear combinaWeitzenböck gauge read tion of the five scalar invariants, quadratic in the torsion tensor. ◦ ◦ ◦ ◦ 74 Rep. Prog. Phys. 86 (2023) 026901 Review tetrad in f (T) gravity (5.53), we end up with a nonvanishing antisymmetric field equation being equal to Two of these scalars have odd parity and thus they allow coupling to pseudo-scalar fields, in order for the Lagrangian to be even under parity. Then, one can formulate a TG theory which is similar to the well-known Nieh–Yan parity-violating theory [350]. In the same spirit as in gauge theories, these fields are called teleparallel axions and they present some interesting phenomenology at cosmological scales, especially in the early Universe. Finally, a local Lorentz covariant version of [349] was considered in [351] and non-flat universes were also taken into account. It was recently reported that some TG parity-violating models suffer from ghost instabilities but the so-called modified TG Nieh–Yan is free of them [352]. fTT In general, finding a tetrad spin connection pair which satisfy the antisymmetric field equations is a difficult task. In this section, we will give an overview of different good tetrad-spin connection pairs found in the literature for different spacetimes satisfying different symmetries. As discussed above, a good tetrad-spin connection pair depends on the theory, therefore, some results provided below will be also theory dependent. As a first glimpse to understanding how important is to choose the correct tetradspin connection pair, we will perform a very simple calculation to show that one needs to be very careful when one is considering teleparallel theories of gravity. Let us start by considering a static spherically symmetric metric 5.12.1. Static spherically symmetric case. W[tr] ∝ C3 C5 (fT′ + fB′ ) = 0, ◦ but this case is only true for either f(R) gravity (which trivially satisfies the antisymmetric field equations for any spacetime) or T, B constants (which is equivalent to TEGR plus a cosmological constant). Since we are interested in the nontrivial cases, we can assume (fT + fB ) ′ 6= 0. Then, to solve the system we notice that there are two possibles branches, (i) C3 = 0; (ii) C3 6= 0. For the first branch, one uniquely solves the system for C3 = C6 = 0, and then, due to the remaining metric DoF, without loosing generality, we can eliminate the cross p setting C2 = 0 and also setting pterms by further A(r) = C1 (r), B(r) = C4 (r) and C5 (r) = ξM(r), where ξ = ±1. Note that the sign of C5 does not affect the metric but it affects the form of the tetrad. By taking all of these forms of Ci = Ci (r) in equation (3.133), we find that the first branch that solves the antisymmetric field equations in f(T, B, ϕ, X) gravity in the Weitzenböck gauge reads and choose a tetrad which is diagonal in the Weitzenböck gauge: A(r), p B(r), M(r), M(r) sin ϑ). (5.144) Clearly, the above tetrad reproduces the metric (5.143) but the teleparallel connection does not respect spherical symmetry (if the spin connection vanishes) since (LXξ Γ)µ νρ 6= 0 (see section 3.6 for more details). In principle, there is no a problem if the connection does not have the same symmetries as the tetrad. However, in this case, is it correct to choose this tetrad-spin connection pair? The answer is simple: the choice of equation (5.144) is incompatible with a vanishing spin connection. This can be easily seen in an example. If we use this  p  e Aµ =   W[ϑφ] ∝ C1 C6 (fT′ + fB′ ) = 0, (5.146) where primes are derivatives wrt the radial coordinate, and the Ci functionals are first shown in equation (3.133). The easiest way to solve the system is by taking (fT + fB ) ′ = 0, ds 2 = A(r)dt 2 − B(r)dr 2 − M(r) 2 (dθ 2 + sin θ 2 dφ 2 ), (5.143) p (5.145) which means that either fTT = 0 (which is f(T) = −T + Λ) or T = const., which again has the same dynamics as TEGR plus a cosmological constant. This small example shows that one needs to be careful on how to choose the correct tetrad-spin connection pair, otherwise the theory will be trivially GR. This important result was totally omitted in many papers and due to this, many incorrect results can be found in the literature. The way to circumvent this issue is to always consider a good tetrad-spin connection, which is related to the symmetries of the problem and also to solve the antisymmetric field equations of the studied theory. We will now derive the correct pair for spherical symmetry. Let us consider the most general tetrad satisfying spherical symmetry in the Weitzenböck gauge where the spin connection is zero, which is given by equation (3.133). In this section, we will further assume stationarity. As a first step, we will just consider the case in f(T, B, ϕ, X) gravity but then, one can conclude that the result can be easily extended this to other teleparallel theories. For this theory, the antisymmetric field equation are given by equation (5.100). For the static spherically symmetric tetrad, there are only two nonvanishing equations 5.12. Good tetrad spin connection examples e A µ = diag( d T = 0, dr  A(r) p 0 0 0  0 pB(r) sin ϑ cos φ ξM(r) cos ϑ cos φ −ξM(r) sin ϑ sin φ  , B(r) sin ϑ sin φ ξM(r) cos ϑ sin φ ξM(r) sin ϑ cos φ  0 p 0 B(r) cos ϑ −ξM(r) sin ϑ 0 75 ξ = ±1, (5.147) Rep. Prog. Phys. 86 (2023) 026901 Review  M ′ 4BMA ′ − 2A MB ′ + 2ξB 3/2 B= AB 2 M 2  ′2 BMA + A −2BMA ′ ′ + MA ′ B ′ + 4ξB 3/2 A ′ − 2A 2 B 2 M ′′ ′2 4M 4M + + . (5.152b) BM BM 2 which reproduces the metric in the standard form ds 2 = A(r)dt 2 − B(r)dr 2 − M(r) 2 (dϑ 2 + sin 2 φ dφ 2 ). (5.148) We can switch on the spin connection by performing the following Lorentz transformation e ′A µ = Λ A B e B µ , with Λ A B being,  Since these quantities are related to gravitational effects, one should expect that they should vanish in the Minkowski limit. One notices that in the limit A, B → 1, M → r, T = B = 0 are zero for √ ξ = 1 but √ non-zero for ξ = −1. However, after redefining A = Ã, B = B̃, one can have the possibility of obtaining the Minkowski metric for Ã, B̃ → −1, M → r and giving us the opposite result, which is that T = B = 0 for ξ = −1 and they are non-zero for ξ = 1. It is still not clear in the literature what is the physical interpretation of having these scalars different to zero in Minkowski. Some authors argue that the teleparallel connection is not satisfying the Minkowski symmetries and therefore, the torsion scalar is non-zero even though one assumes the metric being the Minkowski one. To avoid this issue, several studies [353–356] assumed ξ = 1. However, there are other papers ignoring this issue [357–360] and studied the ξ = −1 where both T, B do not vanish in the Minkowski limit unless we take the limit with the negative functions as we just discussed. Interestingly, the sign of ξ drastically changes the form of the field equations of f(T, B, ϕ, X) gravity in spherical symmetry. This will be explored further in section 9. Let us now explore the second non-trivial branch appearing in equation (5.146) which is the one when C3 6= 0. For this case, the system (5.146) is solved uniquely if C1 (r) = C5 (r) = 0. For this branch, one needs to be careful since the metric change its signature unless we assume that the functions C2 (r) = i B(r), C3 (r) = i A(r) are imaginary. Furthermore, one still has some freedom to choose C4 = 0 to eliminate the cross term in the metric. After doing this  1 0 0 0  0 sin ϑ cos φ sin ϑ sin φ cos ϑ  Λ AB =  0 cos ϑ cos φ cos ϑ sin φ − sin ϑ , (5.149) 0 − sin φ cos φ 0 which gives us another good tetrad-spin connection pair which is equivalent to equation (5.147) with ω A Bµ = 0 having a diagonal tetrad p p e ′A µ = diag( A(r), B(r), ξM(r), ξM(r) sin ϑ), ξ = ±1, (5.150) but now having a non-zero spin connection ω A Bµ with the following non-zero components ω ′1 2ϑ = −ω ′2 1ϑ = −1, ω ′1 3φ = −ω ′3 1φ = − sin ϑ, ω ′2 3φ = −ω ′3 2φ = − cos ϑ. (5.151) One important remark is that the Lorentz transformation must be carried out for both the tetrads and the spin connection. It is interesting to note that the scalar-torsion T and the boundary term B become T=−   √ √  2 ξ B − M ′ MA ′ + A(M ′ − ξ B) ABM 2 , (5.152a)   0 iB(r) 0 0  iA(r) sin ϑ cos φ 0 −M(r) sin φ −M(r) sin ϑ cos ϑ cos φ  , eA µ =   iA(r) sin ϑ sin φ 0 M(r) cos φ −M(r) sin ϑ cos ϑ sin φ  iA(r) cos ϑ 0 0 M(r) sin 2 ϑ where we have set C6 = M(r). It should be noted that this tetrad is complex but the metric is real and its signature remains unchanged, i.e. quantity is complex, both the torsion scalar and the boundary term T= ds 2 = A(r) 2 dt 2 − B(r) 2 dr 2 − M(r) 2 dΩ 2 . (5.153) (5.154) B= For this tetrad the sign of M(r) does not play any role. Thus, the complex tetrad in (5.153) also satisfy the antisymmetric field equations. It is worth noticing that even though this 76 4A ′ M ′ 2M ′ 2 2 + + , 2 2 2 AB M B M M2 (5.155a) M ′ (8BA ′ − 4AB ′ ) 2 (BA ′ ′ − A ′ B ′ ) 4M ′ ′ + + 2 AB 3 M AB 3 B M 4M ′ 2 + 2 2, (5.155b) B M Rep. Prog. Phys. 86 (2023) 026901 Review to solve equation (5.159a). We have further assumed C5 6= 0 since that branch is very restrictive and for FLRW, this case only becomes the flat FLRW case. Moreover, due to the remaining gauge freedom in the metric we can further set C3 = 0 without loosing generality. Let us emphasise here that T and B were assumed to depend only on t but after assuming the above assumptions, this condition is not true yet. To continue, we will focus on the non-flat FLRW case, where the functions are q C1 (t, r) = ξ1 N(t), C5 (t, r) = ξ2 r 2 a(t) 2 − C6 (t, r) 2 , (5.161a) are real, so that, the action will be well-defined under this tetrad. It is still unclear if other important physical quantities can be complex for this tetrad. In the Minkowski limit, T, B 6= 0 unless we take the Minkowski limit for A, B = −1, M = r. Similarly as we did above, one can perform the following local Lorentz transformation   0 −i sin ϑ cos φ −i sin ϑ sin φ −i cos ϑ  −i  0 0 0 , Λ AB =   0  − sin φ cos φ 0 0 − cos ϑ cos φ − cos ϑ sin φ sin ϑ (5.156) √ C6 (t, r) = ξ3 kr 2 a(t), simultaneously for the tetrad and the spin connection, to gives us a spin connection with the following non-zero components ω ′ 0 2φ = ω ′ 2 0φ = i sin ϑ, ω ′2 3φ = −ω ′3 ω ′ 0 3ϑ = ω ′ 3 0ϑ = −i, 2φ = − cos ϑ, Then, for FLRW cosmology one can write down the following good tetrad in the Weitzenböck gauge in f(T, B, ϕ, X) gravity (5.157) and a diagonal tetrad in the form e µ = diag(A(r), B(r), M(r), M(r) sin ϑ). A e0 = ξ1 N(t) dt, a(t) cos φ sin ϑ dr + ra(t) e1 = p 1 − kr 2 ξ2   p √ × −ξ3 kr sin φ + ξ2 1 − kr 2 cos ϑ cos φ dϑ  √  p − ra(t) sin ϑ ξ3 kr cos ϑ cos φ + ξ2 1 − kr 2 sin φ dφ, (5.162b) e2 = In this section, we will perform a similar analysis as in the previous section for f(T, B, ϕ, X) gravity but we will assume that ϕ = ϕ(t), T = T(t), B = B(t). This means that we have drd (fT + fB ) = 0 but dtd (fT + fB ) 6= 0. This case would be important for cosmological scenarios, for example. Consider the FLRW cosmological scenario, where the cosmological principle will impose that these quantities must be time-dependent only. Using the most general time-dependent spherically symmetric tetrad (3.133), the antisymmetric field equation (5.100) for this case is reduced to the following system 5.12.2. Time-dependent spherically symmetric case. W[tr] ∝ (ḟT + ḟB )(C4 C5 − C5′ C5 − C6 C6′ ) = 0, (5.159a) W[ϑφ] ∝ (ḟT + ḟB )C2 C6 = 0, (5.159b) a(t) sin φ sin ϑ p dr + ra(t) 1 − kr 2  √  p × ξ3 kr cos φ + ξ2 1 − kr 2 cos ϑ sin φ dϑ   p √ + ra(t) sin ϑ −ξ3 kr cos ϑ sin φ + ξ2 1 − kr 2 cos φ dφ, (5.162c) p a(t) cos ϑ dr − ξ2 a(t)r 1 − kr 2 sin ϑ dϑ e3 = √ 1 − kr 2 √ + ξ3 kr 2 a(t) sin 2 ϑ dφ, (5.162d) where ξ1 = ±1, ξ2 = ±1 and ξ3 = ±1. It should be noted that this result is compatible with the one found in section 3.6.3 where it was shown that the same above good tetrad for nonflat FLRW applies for any generic modified teleparallel theory. Independently of the sign chosen for the tetrad, the scalars T and B become where dots and primes denote derivatives wrt time and the radial coordinate, respectively. We have two possible branches: (i) C6 6= 0; (ii) C6 = 0. Let us first explore the first branch. For this case we need to impose (fB 6= −fT )  1  C5 C5′ + C6 C6′ , C5 (5.162a) (5.158) The tetrad in (5.153) in the Weitzenböck gauge is equivalent as the pair (5.157) and (5.158). It is straightforward to show that all the good tetrad-spin connection presented in this section for f(T, B, ϕ, X) gravity, also satisfy the antisymmetric field equations of more general teleparallel theories such as the generalized NGR theory (see section 5.2), and also for TEGB extensions such as the ones discussed in section 5.6. C4 (t, r) = ξ1 = ±1, ξ2 = ±1, ξ3 = ±1. (5.161b) T = −6H 2 + 6k , a2 Ḣ B = −18H 2 − 6 , N (5.163) where we have introduced the Hubble parameter H(t) = ȧ/(Ha). We can again perform a Lorentz transformation for the tetrad-spin connection pair (5.162) and find that under the Lorentz transformation, (5.160) 77 Rep. Prog. Phys. 86 (2023) 026901  −1 ξ1  0 A Λ B=  0 0 Review  0 0 0 1  ξ2−√1 cos φ sin ϑ √ ξ2− ξ2−1 cos ϑ . √ sin φ sin ϑ √ √ 2 2 2 − ξ3 kr sin φ ξ2 cos ϑ √ sin φ 1 − kr √ + ξ3 kr cos φ −ξ2 √ 1 − kr sin ϑ ξ2 cos ϑ cos √ φ 1 − kr √ −ξ2 sin φ 1 − kr 2 − ξ3 kr cos ϑ cos φ ξ2 cos φ 1 − kr 2 − ξ3 kr cos ϑ sin φ ξ3 kr sin ϑ (5.164) the tetrad (5.162) (independently of the sign) becomes diagonal   a(t) e ′A µ = diag N(t), √ , a(t)r, a(t)r sin ϑ , (5.165) 1 − kr 2 Thus, another good tetrad-spin connection for FLRW cosmology is given by √ p ξ1 ξ2 −kra(t) e0 = N(t) 1 − kr 2 dt + ξ3 √ dr, (5.168a) ξ3 1 − kr 2 and the spin connection ω ′ A Bµ have the following non-zero components, p ω ′ 1 2ϑ = −ω ′ 2 1ϑ = − 1 − kr 2 , p ω ′ 1 3φ = −ω ′ 3 1φ = 1 − kr 2 sin ϑ, (5.166a) √ e1 = ξ1 N(t) −kr sin ϑ cos φ dt + ξ2 a(t) sin ϑ cos φ dr ω ′1 3ϑ = − ω ′3 1ϑ = ξ3 √ kr, ξ2 ω ′2 3r = −ω ′3 2r = ξ2 ξ3 p √ + ξ2 ra(t) cos ϑ cos φ dϑ − ξ2 ra(t) sin ϑ sin φ dφ, (5.168b) √ e 2 = ξ1 N(t) −kr sin ϑ sin φ dt + ξ2 a(t) sin ϑ sin φ dr + ξ2 ra(t) cos ϑ sin φ dϑ + ξ2 ra(t) sin ϑ cos φ dφ, (5.168c) √ ′2 ′1 ω± 1φ = −ω 2φ = ξ2 ξ3 kr sin ϑ, (5.166b) k 1 − kr 2 , √ e3 = ξ1 N(t) −kr cos ϑ dt + ξ2 a(t) cos ϑ dr − ξ2 ra(t) sin ϑ dϑ, (5.168d) ω ′2 3φ = −ω ′3 2φ = − cos ϑ. (5.166c) with a zero spin connection, which has the scalars given by √ −kH 6k 2 T = −6H + 12ξ1 ξ2 + , a √ a2 6Ḣ 12ξ1 −kH B = −18H 2 − + . (5.169) N ξ2 a Note that we have used ξi = 1/ξi since ξi = ±1. As we have remarked before, the pair (5.162) with zero spin connection is identical to the pair (5.165) with non-zero spin connection (5.166a), so both pairs are good tetrad-spin connection pair which solve the antisymmetric field equations for f(T, B, ϕ, X) gravity while respecting the symmetries. Let us now explore the second branch (ii) where C6 = 0. For this case, the first antisymmetric equation (5.159a) will be zero if C4 = C5′ . To further express the metric in its diagonal form, we must take C1 = C3 C5′ /C2 with C2 6= 0 (the case C2 = 0 is also more restrictive). Furthermore, by taking the FLRW example, we need to set √ √ ξ3 kr a(t) C2 (t, r) = √ , C3 (t, r) = ξ1 −k rN(t), kr 2 − 1 C5 = ξ2 r a(t), ξ1 = ±, ξ2 = ±1, ξ3 = ±1. (5.167)  ξ1 ξ2 √ 1 − kr 2 ξ3 √  −ξ3 −k r Λ AB =   0 0 This tetrad reproduces the standard non-flat FLRW metric and is real only for k = −1 or k = 0. For the k = 1 case, the scalars become complex (as explained in 3.6.3.3). It is easy to see that both tetrads (5.162) and (5.168) are equivalent for the flat case k = 0. There is one important difference in this tetrad which is that depending on the sign chosen of the tetrad (where ξ1 = ±1 and ξ2 = ±1), one could have different T and B. Finally, for the above tetrad in FLRW cosmology (5.168), one can perform the following Lorentz transformation √ √ √  −ξ1 −k r sin ϑ cos φ −ξ1 −k r sin ϑ sin φ −ξ1 −k r cos ϑ √ √ √ ξ2 1 − kr 2 sin ϑ cos φ ξ2 1 − kr 2 sin ϑ sin φ ξ2 1 − kr 2 cos ϑ ,  ξ2 cos ϑ cos φ ξ2 cos ϑ sin φ −ξ2 sin ϑ −ξ2 sin φ ξ2 cos φ 0 78 (5.170) Rep. Prog. Phys. 86 (2023) 026901 Review end). As was discussed in section 3.6.1, there are two different branches solving the symmetry condition which respects axial symmetry, which we labelled as the regular branch and the solely axially symmetric branch. In [362], the authors studied the Kerr case in f (T) gravity and they found a good tetrad-spin connection pair such that for this spacetime T = 0. This tetrad choice is also consistent with the fact that the Ricci scalar ◦ is R = 0 for the Kerr geometry, and then solves the antisymmetric and the symmetric field equations in f (T) gravity. This tetrad is part of the solely axially symmetric branch, so that, it cannot respect spherical symmetry in the teleparallel point of view in any limit. This means that the metric can become spherically symmetric (for the Schwarzschild limit) but the teleparallel connection cannot become spherically symmetric, meaning that LXζ Γ σ µν 6= 0. The failure of this condition for this tetrad is manifest in the nonvanishing component of the torsion tensor Tφ ϑφ = cot ϑ. Moreover, the tetrad found in [362] is time-dependent whereas the Kerr geometry is known to be static. Another point to remark about this work is that T = 0 but the three scalars Tvec , Taxi or Tten are not vanishing. Then, only in f(T, B) gravity, this tetrad will solve the antisymmetric field equations and this occurs in a trivial way since T = B = 0 generates a theory which is equivalent to GR plus a cosmological constant. The first non-trivial known good tetrad-spin connection pair for a more general axial symmetric case was derived in [363] for f(T, ϕ) gravity, but it can be proved that this pair is also a good pair for many TG theories (similarly as the pair found before in spherical symmetry). This solution does not contain any Plebański–Demiański spacetime as special cases (except from Schwarzschild). Later, in [364] the authors analyzed in detail the possibility of finding good tetrad-spin connection pairs in axial symmetry for a very general theory known as f(T, B, ϕ, X) gravity which includes the boundary term and a scalar field with its kinetic term. They focused mainly in the regular branch in order to obtain a smooth limit from axial symmetry to spherical symmetry. First, they started with the tetrad (3.126) in the Weitzenböck gauge satisfying axial symmetry that can be also written in the following form to get the same diagonal FLRW tetrad (5.165) with the following non-zero components for the spin connection √ ξ1 √ −k ′0 ′1 ω± 1r = ω± 0r = ξ1 ξ2 √ , ω±′0 2ϑ = ω±′2 0ϑ = −kr, 2 ξ2 1 − kr ξ1 √ ω±′0 3φ = ω±′3 0φ = −kr sin ϑ, (5.171a) ξ2 p ω±′1 2ϑ = −ω±′2 1ϑ = − 1 − kr 2 , p ω±′1 3φ = −ω±′3 1φ = − 1 − kr 2 sin ϑ, ω±′2 3φ = −ω±′3 2φ = − cos ϑ. (5.171b) Therefore, the above spin connection with the diagonal tetrad components (5.165) are also a good tetrad-spin connection pair. Let us remark here that the scalars T and B (or any other constructed from the torsion tensor) will have the same value after making the Lorentz transformation since both the spin connection and the tetrad compensates the transformation. It is also important to mention that as we have pointed out before, the good tetrad-spin connection pair is field equation dependent. However, for the FLRW cosmological pair found in this section, it has been proved in [74] that both (5.162) and (5.168) with zero spin connection are good tetrad-spin connection pairs valid for any modified teleparallel theory of gravity. This is not the same for the static spherically symmetric good tetrad-spin connection pair (5.147) which may not be a good pair for some specific theories. Depending on the teleparallel theory, one must check if this pair is good or not. 5.12.3. Good tetrad-spin connection pair: axial symmetry. The situation is very complicated in axial symmetry since it is hard to find a good tetrad-spin connection pair in modified TG due to the impact of the time dependence. This pair is really essential in performing calculations in axial symmetry in modified TG. There are some works in TEGR such as in [361], but they are trivial since in this case any choice would be a good tetrad-spin connection pair (all would be identical to GR in the   H00 H01 −H02 H03  H10 cos φ − H20 sin φ H11 cos φ − H21 sin φ H22 sin φ + H12 cos φ H13 cos φ − H23 sin φ   e Aµ =   H10 sin φ + H20 cos φ H11 sin φ + H21 cos φ H12 sin φ − H22 cos φ H13 sin φ + H23 cos φ  , H30 H31 −H32 H33 where C A µ where replaced by H ij and they denote 16 different functions which depend only on t, r, ϑ. The metric reproduced by the above tetrad contains all the possible cross terms (dt, dr, dt, dφ, dt, dϑ, dr, dφ, dr, dϑ, dϑ, dφ). Then, there is some gauge freedom related to the metric. For example, one can choose a gauge such that only the dt dφ cross term differs from zero. This choice is the standard one chosen for the Kerr (5.172) metric in Boyer-Lindquist coordinates. The authors in [364] studied the stationary case and set H01 = H02 = H20 = H33 = H30 = H10 = H21 = H22 = H13 = 0, H H H31 = 11 12 , (5.173) H32 79 Rep. Prog. Phys. 86 (2023) 026901 Review giving the following metric   2 H12 2 2 2 2 2 2 ds = H00 dt − H11 + 1 dr 2 − (H12 + H32 )dϑ 2 2 H32 2 2 − (H23 − H03 )dφ 2 + 2H00 H03 dt dφ. W[rϑ] = where commas denote differentiation and (5.174) Qϑ = This choice was inspired by the spherically symmetric good tetrad in the Weitzenböck gauge (5.147) since the the tetrad (5.172) with (5.173) gives us   H00 0 0 H03  0 H11 cos φ H12 cos φ −H23 sin φ  , e Aµ =   0 H11 sin φ H12 sin φ H23 cos φ  0 H11 H12 /H32 −H32 0 (5.175) which reproduces (5.147) in the limit p p H00 = A(r), H11 = B(r) sin ϑ, H23 = H32 = p C(r) sin ϑ, H12 = p (5.176b) H00  0   e µ=  0  0 A   0 H00,r H23 + H23,r H00   Qr = H12 − H23,ϑ H00,ϑ − . H23 H00 (5.178) (a) Case (Qr = Qϑ = 0): A good tetrad that does not contain any Plebański–Demiański spacetime which is valid for any form of f(T, B, ϕ, X), and also generalizes the tetrad found in [363], which reads If we assume stationarity and replace (5.175) in the f(T, B, ϕ, X) field equations, one notices that there is only one nonvanishing antisymmetric equation, that reads  H00,r H11 H23,r − + , H00 H23 H23 The tetrad (5.175) contains one additional free function compared to the metric (5.174). This extra free function can be found by solving the remaining antisymmetric field equation (5.177). Thus, in order to find a good tetrad in the Weitzenböck gauge, one would need to solve equation (5.177) for one of the six functions H ij . However, doing this in general is a very complicated task since T and B become very cumbersome in axial symmetry. In order to tackle this problem, the authors in [364] studied four different particular cases where they were able to find good tetrads. The most important good tetrads in the Weitzenböck gauge found are: C(r) cos ϑ, (5.176a) H03 = 0.  1 (fT,ϑ + fB,ϑ )Qϑ + ( fT,r + fB,r )Qr = 0, (5.177) 2  cos φ H00,r H23 + H23,r sin φ H00    H00,r H23 −1 H00,ϑ H23 H32 + H23,ϑ + H23,r H00 H00  0 H03  H00,ϑ H23 + H23,ϑ cos φ H00  H00,ϑ H23 + H sin φ 23 ,ϑ H00 −H32  −H23 sin φ    . H23 cos φ   0 (5.179) (b) Case (Qr = 0 and fT,ϑ + fB,ϑ = 0): A good tetrad behaving as a family of Taub-NUT-like spacetimes which needs ϕ = ϕ(r)  p  e Aµ =   A(r) p 0 0 p 0 B(r) sin ϑ cos φ C(r) cos ϑ cos φ p p B(r) sin ϑ sin φ C(r) cos ϑ sin φ 0 p p 0 B(r) cos ϑ − C(r) sin ϑ p  A(r) p (C2 + C1 cos ϑ) −p C(r) sin ϑ sin φ  ,  C(r) sin ϑ cos φ (5.180) 0 where C1 and C2 are constants. The corresponding metric is then so that, the Taub-NUT spacetime is recovered by setting ds 2 = A(r)dt 2 − B(r)dr 2 − C(r)dϑ 2 h i 2 − C(r) sin 2 ϑ − A(r) (C1 cos ϑ + C2 ) dφ 2 D(r, ϑ) = 2b cos ϑ, (5.182) √ where r± = M ± M 2 + b 2 and b is the Taub-NUT parameter and M the mass. It should be noted that when C2 = + 2A(r) (C1 cos ϑ + C2 ) dt, A(r) = 1/B(r) = (5.181) 80 (r − r+ )(r − r− ) , r2 + b2 C(r) = r 2 + b 2 , Rep. Prog. Phys. 86 (2023) 026901 Review (c) Case (Qϑ = 0 and fT,r + fB,r = 0): A good tetrad having a metric which is always axially symmetric (unless a trivial case is considered) needing that ϕ = ϕ(ϑ) C1 = 0, this tetrad exactly matches the spherically symmetric good tetrad (5.147).   B(ϑ) 0 0 B(ϑ)  0 D1 (ϑ)D2′ (r) cos φ 0 −D1 (ϑ)D2 (r) sin φ  . e Aµ =  ′  0 D1 (ϑ)D2 (r) sin φ 0 D1 (ϑ)D2 (r) cos φ  0 0 −C(ϑ) 0 (5.183) (d) Case ((fT,r + fB,r )Qr ̸= 0 and (fT,ϑ + fB,ϑ )Qϑ ̸= 0): A good tetrad in f(T, B) gravity for a slowly rotating Kerr spacetime which is given by the tetrad (5.175) with r H00 = 1− 2Mr H32 , H11 = √ , H12 = Σ ∆ v u u H23 = tsin 2 ϑ A(r, ϑ) = q 2a2 Mr sin2 ϑ + a2 + r2 Σ 2 , Σ − H32 (5.184a) ! 2 , + H03 (5.184b) 2aMr sin 2 ϑ H03 = − p , Σ(Σ − 2Mr) (5.184c) H32 (r, ϑ) = r sin ϑ + a 2 A(r, ϑ) + O(a4 ), (5.184d)  √ sin ϑ cos 2 ϑ 4µ5 + 6µ 2 r3/2 + r5/2 + 4µr 2 + µ4 r − 16µ3 r √ , 2µ 2 r3/2 (−µ 2 − 4µ r + r) where Σ = r 2 + a 2 cos 2 ϑ, ∆ = r 2 − 2Mr + a 2 , µ = √ r − 2M and a is the angular momentum (per unit mass) which is assumed to be small (a  1) . Note that H00 , H11 , H23 and H 03 need to be expanded up to second order in a. (5.184e) general form of the tetrad respecting these symmetries. In [365], the authors found that for cylindrical symmetries (stationary case), it is convenient to work with Weyl-type coordinates xµ = (t, r, φ, z) such that the metric can be written as ds 2 = e2u(r) dt 2 − e2(k(r)−u(r)) dr 2 − w(r) 2 e−2u(r) dφ 2 − e2(k(r)−u(r)) dz 2 , All the above tetrads were computed in the Weitzenböck gauge but one can also perform a local Lorentz transformation and then one could have a non-zero spin connection. The three first cases solve all the antisymmetric field equations for f(T, B, ϕ, X) gravity whereas the last case is only valid for f(T, B) gravity. These results can be used to further analyze the symmetric field equations and explore axial symmetry in modified TG. For example, as was done in [354] for Schwarzschild, the slowly rotating Kerr good tetrad can be used to find perturbative teleparallel modifications of slowly rotating axially symmetric solutions around Kerr. (5.185) where u(r), k(r) and w(r) are arbitrary functions depending only on the radial coordinate. It can be shown that a good tetrad-spin connection pair for f(T, B) gravity is the one in which the spin connection vanishes and has the following diagonal tetrad   e A µ = diag eu(r) , ek(r)−u(r) , w(r)e−u(r) , ek(r)−u(r) , (5.186) which gives the following scalar-torsion and boundary term T= There are few works in TG dealing with other symmetries such as cylindrical or planar ones. There are no works dealing with the most 5.12.4. Cylindrical and other symmetries. B= 81  e2u−2k 2k ′ w ′ − 2wu ′2 , w (5.187a) 2e2u−2k (w (k ′ ′ − u ′ ′ ) + w ′ (k ′ − u ′ ) + w ′ ′ ) . (5.187b) w Rep. Prog. Phys. 86 (2023) 026901 Review Here primes denotes derivatives wrt r. It can be shown that the tetrad (5.186) satisfied the antisymmetric equations for f(T, B) gravity and also for other more general theories. For the GR case in vacuum (f = −T), one finds the following solutions Cosmology is dominated by the two assumptions of homogeneity and isotropy which are supported to a high degree by unprecedented observational evidence [371, 372] resulting in the widely accepted cosmological principle [373]. This has led to the well known FLRW metric which is a universal solution in cosmology. Considering its form in cosmic time t gives it the four-dimensional representation   dr 2 2 2 2 2 2 2 2 2 ds = N(t) dt − a(t) + r (dϑ + sin ϑdφ ) , 1 − kr 2 (6.1) c2 w(r) = c0 + c1 r, k(r) = log(c0 + c1 r), c1 r c2 log(c0 + c1 r), u(r) = ± (5.188) c1 which after making some coordinate transformations, it can be recast as the well known Levi-Civita cylindrical solution [366]. In [365], the authors found perturbed solutions for torsion squared gravity behaving similarly as the above metric. In [367], the authors used the Noether’s symmetry approach to find solutions for f (T) gravity but they are just written implicitly and not in an exact form for the metric coefficients. In [368, 369], the authors studied the collapse for cylindrical stars but they unfortunately, used a dynamical cylindrical tetrad that is not compatible with the choice of setting the spin connection to zero, and then, the antisymmetric equations presented there are not satisfied. Other more exotic symmetries such as planar ones have also been explored in the literature. In [370], these geometries was studied in f (T) gravity. It can be shown that for the coordinates xµ = (t, r, x, y), a diagonal tetrad of the form e A µ = diag(A(r), B(r), C(r), C(r)) which reproduces a planar metric ds 2 = A(r) 2 dt 2 − B(r) 2 dr 2 − C(r) 2 (dx 2 + dy 2 ), is compatible with the choice of a zero spin connection, leading to the conclusion that [370] correctly chose the tetrad ensuring that the antisymmetric field equations in f (T) are satisfied. In this paper, in the end, they only studied the case of GR since they always chose either T = 0 or T = const. This is the reason why they found the same exact solutions known in GR for these geometries which are the Møller, Kattler-Wittaker and planar de-Sitter solutions. where N(t) and a(t) represent the lapse function and scale factor respectively. In many cases, the lapse function is found to not be dynamical, to determine this feature the metric 6.1 has to be substituted into the dynamical equations to determine is the lapse function plays a role in the equations of motion of the theory. In those cases where the lapse function is found to be nondynamical, it can be absorbed into the definition of cosmic time. Another important property to point out is that this is being defined in spherical polar coordinates in which the Lorentz frame is not represented by the Minkowski metric which directly leads to a nonvanishing appearance of the spin connection components, as explained in section 5.12. Finally, the curvature parameter k corresponds to open (k = −1), flat (k = 0), or closed (k = +1) cosmologies. As we can notice from equation (6.1), the scale factor a(t) is a function of time t, where a(t0 ) = a0 at the present time t0 . It is standard to use the redshift z ≡ (a0 /a) − 1 as a proxy for the age or scale factor. The redshift can be measured for distant sources; it is the fractional amount by which the wavelength of a photon has been stretched by the expansion between the time the photon is emitted and the time it is received. The expansion rate H ≡ ȧ/(Na) is a function of time, with the value H 0 as the Hubble–Lemaı̂tre constant and where the dot denotes a derivative wrt t. The cosmological deceleration parameter is then q ≡ −(ä/a)/H 2 = (1 + z)Ḣ/H − 1, where dots refer to derivatives with cosmic time. Once this is considered, a GR equation of motion (Friedmann and acceleration equations) for a(t) for a flat Universe filled with fluids i (e.g. nonrelativistic matter, radiation, and dark energy) of energy densities ρi can be written as 6. Cosmology in TG The relatively recent observational discovery of the accelerating expansion of the Universe [19, 20] has led to an intense few decades of theoretical and observational work to determine the origin and properties of the so-called dark energy (DE). In terms of numerical analysis, the most consistent model that fulfil these requirements is ΛCDM where CDM acts as a gravitational well in galaxies while a cosmological constant Λ dominates on cosmological scales, and GR acts at all scales as the fundamental theory of gravitation. However, theoretically this has led to numerous problems [23] which may ultimately be addressed by revisiting this approach to gravity. By investigating possible cosmology beyond the concordance model, we may design a better formulation of gravitation that requires less physics beyond the standard model of particle physics in the dark matter sector, as well as having a more nature emergence of late-time acceleration in the Universe, and possibly resolving some issues of the inflationary epoch and its origins.  2 8πG ȧ = ρ, a 3 (6.2a) ä = −4πG(ρ + 3p). a (6.2b) H2 = In these equations ρ and p are the total density and pressure and dots represent differentiation wrt cosmic time. Hence, they can be written as sums of the individual P of the contributions P components as ρ ≡ i ρi and p ≡ i pi . If the fluids have pressures pi , then the change d(ρa3 ) in the total energy (ρ = P volume is equal to the work −pd(a3 ), i ρi ) per comoving P where p = i pi , done by the fluid. If we define equationof-state parameters wi ≡ pi /ρi (e.g. wm = 0 for matter and wr = 1/3 for radiation), then the second form of the Friedmann equation can be written as q0 = (1 + 3wTot )/2, where wTot ≡ 82 Rep. Prog. Phys. 86 (2023) 026901 Review P i pi /ρi is the total equation-of-state parameter. Thus, if GR is correct, the observations require that the Universe has wTot < −1/3. Thus, some DE, a negative-pressure fluid, is postulated to account for cosmic acceleration. Another way to represent this, and other, contributions is through the density parameter which we define in terms of the critical density within the ΛCDM Friedmann equations, namely ρc = 3H 2 /8πG (where we retain SI units for convenience). We can then define density parameters for each contribution ρi as Ωi := ρi /ρc . An analogous description can be prescribed for certain TG theories of gravity, those which FLRW equations can be recast in a GR-like format, such as in f (T) and f(T, B) gravity. When this formulation is permitted, the gravitational source acts as an effective fluid (with energy density ρeff and pressure peff ) with EoS weff . Consequently, the Universe’s acceleration is p+peff achieved when the total EoS wTot = ρ+ρ < −1/3. eff Moreover, the overwhelming historical evidence has pointed to a flat Universe [374–377] with further support from inflationary theories [14, 15]. However, some recent observations have prompted a reexamination of the closed Universe model [372, 378–380], which provides a substantial improvement of the fit of early-time data, but is in disagreement with external additional data sets as acoustic oscillations seen in cosmic microwave background anisotropy data. Saying that, a non-flat cosmology would require a radical overhaul of our standard approaches to cosmology and the interpretation of data and so should be taken with caution. In the context of TG, the vast majority of works have been based on a flat cosmological background while some works have investigated the possibility of a non-flat cosmology, as will be explored in this section. Considering again the general metric in equation (6.1) which describes a Universe that observes an isotropic and homogeneous cosmology. The tetrads feature a branching effect in which the positive curvature parameter (k = +1) can be described by the good tetrad in equation (5.162) while the negative curvature parameter (k = −1) branch are described by equation (5.168) (otherwise one would have a complex tetrad). For the case of a flat FLRW cosmology, both good tetrads given in equations (5.162) and (5.168) coincide with each other. It is important to say again that this is not the only choice that describes the FLRW background but it is the only once in which the spin connection is allowed to vanish while the teleparallel connection respects the cosmological symmetries. Another important tetrad to highlight is used particularly for flat cosmologies (k = 0), where the metric can be written as  ds 2 = N(t) 2 dt 2 − a 2 (t) dx 2 + dy 2 + dz 2 , (6.3) One can also relax the condition that the teleparallel connection enjoys the symmetries of cosmology, but then, the corresponding cosmological equations would not respect the symmetries of cosmology. For the diagonal tetrad (6.4) with a vanishing spin connection, one finds that by using the definitions of the torsion scalar (1.6d) and the boundary term (4.156), they become T = −6H 2 , 6Ḣ . N (6.5) Here, we again used that H = ȧ/(aN). One naturally reproduces the standard Ricci scalar in this scenario ◦ Ḣ R = −T + B = −12H 2 − 6 . N (6.6) In the following sections, the different manifestations of this cosmological setup are explored in the dynamics of the various extensions beyond ΛCDM, which are investigated together with their cosmological evolution, and properties. In this section, we explore the predominant cosmological constructions in TG. Given the enormity of the task, we first describe the background on the different tools that have been employed to explore the background cosmology of the different TG theories, and then delve into the specific theories themselves. In the supplementary annexes (supplementary 1) we provide some basic cosmological methods that will be used in the following sections such as the reconstruction method, the Noether symmetry approach and dynamical systems in cosmology (see supplementary annexes (supplementary 1)). Additionally, some definitions regarding bouncing solutions are also explained in the supplementary annexes (supplementary 1). Those definitions and methods will be applied for several TG theories in the next sections for the case of FLRW cosmology. A more advanced topic regarding anisotropic cosmology is also presented in the supplementary annexes (supplementary 1) as a complementary study for the interested reader. 6.1. f(T) cosmology In this section we will analyze f (T) cosmology in flat FLRW using the techniques/methods explained in the previous sections. If we use the diagonal tetrad (6.4) in Cartesian coordinates and replace it in the f (T) field equations (5.53), we arrive at the modified FLRW equations described by 1 −6H 2 fT − f = κ2 ρ, 2 1 −2fT (3H 2 + Ḣ) − 2HḟT − f = −κ2 p, 2 and the diagonal tetrad results in the tetrad eAµ = diag (N(t), a(t), a(t), a(t)) , B = −18H 2 − (6.4) (6.7a) (6.7b) where dots are derivatives wrt the time, so that ḟT = fTT Ṫ. It is important to mention that the most general teleparallel theory constructed from of up to quadratic contractions of torsion, f(Taxi , Tten , Tvec , P1 , P2 ) gravity (see section 5.5), is dynamically equivalent to f (T) gravity in flat FLRW since Tten = Taxi = P1 = P2 = 0 and Tvec = 9H 2 , meaning that only one scalar is which turns out to be in the Weitzenböck gauge for the extensions to TEGR explored here. An important remark here is that the above tetrad (with vanishing spin connection) is the only one that has the property that both the tetrad and the teleparallel connection obey cosmological symmetries for flat FLRW. 83 Rep. Prog. Phys. 86 (2023) 026901 Review where ΩΛ = 1 − Ωm0 , T 0 represents the torsion scalar evaluated at current times, and nonvanishing and due to (5.4), the theory can be always represented as a f (T) gravity theory. This conclusion is only valid for flat FLRW at the background level. These equations will be analyzed in detail in the following sections. ∞ 2 F1 (a, b; c; z) ≡ Γ(c) X Γ(a + k)Γ(b + k) k z Γ(a)Γ(b) Γ(c + k) k! (6.13) k =0 In the trivial extension of f (T) gravity, the reconstruction procedure has been explored extensively in the literature. One common approach involves describing the matter sector as a manifestation of torsion usually by taking the ansatz f(T) = −T + F(T). The latter’s F(T) role acts as the source for the matter sector as can be seen from equation (6.7a) 6.1.1. Reconstruction method. 3H 2 = κ2 ρ + F(T) − TFT , 2 represents Gauss’s hypergeometric function which is convergent for |z| < 1 [395]. Naturally, the standard result is recovered for w = 0. If the fluid is instead assumed to be composed of a spinor field [391], the Lagrangian becomes a rescaled version ΛCDM, i.e. TEGR with the addition of a constant. In the case of HDE and QCD DE [396, 397], the reconstructed solution does not satisfy the Friedmann equations and hence are not reported. An alternative proposal is to reconstruct through a specification of the deceleration parameter q(z). During periods where the matter content is mostly composed of baryonic matter, the reconstructed function is obtained from the relation [393] (6.8) where we recall that T = −6H 2 . A torsional fluid ρeff can therefore be defined which can be used to reconstruct the F(T) function [381] κ2 ρeff = √ F(T) − TFT =⇒ F(T) = κ2 −T 2 ˆ f(z) = −6Ωm0 H0 e 2 ρeff ´ z 1+q(z̄) 0 1+z̄ dz̄ ˆ z (1 + z̄) 2 (1 + q(z̄)) ´ z 1+q(z̄) e0 0 3 dT. 1+z̄ dz̄ dz̄. (6.14) (−T) 2 (6.9) This is derived from equation (6.11) by expressing it in terms of redshift z and the deceleration parameter q(z) via the rela´ z 1+q(z̄) 1 +z̄ dz̄ and T = −6H 2 . In particular, the 0 tions H(z) = H0 e parametrization q(z) = q0 + q1 X(z) [393] for constants q0 , q1 and for two functional choices of X(z) which determines the cosmological evolution was investigated. The first two models X(z) = z1(1++z z2) [398] and X(z) = e+z) ln(N e e 1+z − ln N, N > 1 [399] do not yield a viable cosmology. On the other hand, the final model explores the effective fluid EoS weff = − 1+α(11+z)n for some constants α, n [400], or equivalently in terms of q(z), This approach has been applied for various fluids including Ricci DE, HDE and (m, n) HDE, Tsallis HDE including its power-law and logarithmic corrected variants, PlDE and (m, n) PlDE [382–388]. However, in the aforementioned works, a further assumption on the cosmological expansion behavior leads to a disagreement with the Friedmann equations. Meanwhile, in [389], the reconstruction procedure can be applied in the absence of fluids ρ = p = 0 through an effective EoS weff = −1 − 16TḢfTT , f (6.10) q(z) = −1 + provided the evolution of H(t) is set. Meanwhile, in [390], a correspondence between HDE and f (T) gravity is produced via a suitable identification of the IR cut-off length. An alternative proposal is to make use of the matter component in order to reconstruct the Lagrangian as opposed to an a priori desired form for ρeff [391]. This leads to an alternative relation ˆ √ ρm f(T) = −κ2 −T (6.11) 3 dT. (−T) 2 3α(1 + z)n . 2 [1 + α(1 + z)n ] (6.15) The model describes an early matter domination and late de Sitter phases. In particular, n = 3 yields the standard ΛCDM cosmology, hence n is expected to deviate slightly from this value. In this case, the reconstructed function is 3 f(z) = −9Ωm0 H0 2 [1 + α(1 + z)n] 2n ˆ z 0 (1 + z̄)n+2 3 1+ 2n [1 + α(1 + z̄)n ] dz̄. (6.16) For the ΛCDM cosmology in the presence of dust matter, the Lagrangian reduces to the standard ΛCDM Lagrangian [392, 393]. However, it can be generalized for arbitrary perfect fluids with EoS w to give [394] n Here, Ωm ⩽ 1 provided α > αmin = Ωm0 3 3n . For α slightly lar1−Ωm0 ger than this minimum value leads to the following torsional fluid behaviors:  1+w    ΩΛ 1 T 1 f(T) = T0 Ωw0 − , −w; ; F − 2 1 Ωm0 2 2 T0 ΩΛ   T 1 3 T (6.12) + , −w; ; , 2 F1 T0 ΩΛ 2 2 T0 ΩΛ (a) n ≲ 3: Quintessence with weff ≳ −1 at present; (b) n ≳ 3: Quintom (quintessence → phantom transition at z ∼ 4) with weff ≲ −1 at present; (c) n = 3: Quintessence with weff ∼ −1 at present. 84 Rep. Prog. Phys. 86 (2023) 026901 Review 3. w = 1: quintom behavior; weff ∼ 1 at early-times, crosses the phantom-divide line and approaches weff → −1 at late-times. (c) j = 1 − ϵf(a) with slowly varying function f(a) = 89ϵ + ϵa0 , 4 ϵ, ϵ0  1: A decelerating Universe a(t) ∼ t 5 results and thus not explored in further detail. As a final remark, for all the aforementioned jerk parameter considerations, the models are asymptotically stable against homogeneous perturbations. The reconstruction procedure can also be applied to generate the spatially flat FLRW cosmology through non-flat FLRW spacetimes [403]. Accounting for the spatial curvature k, the Friedmann equations are obtained following the first branch solution described in section 3.6.3.2 with tetrad equation (3.154), leading to f κ2 ρ = − − 6H 2 fT , 2 −κ2 p = −2fT (3H 2 + Ḣ) + Figure 12. The torsional fluid EoS behavior for the model Here, Ωm0 = 0.297 and δα = 10−3 . Reprinted figure with permission from [393], Copyright (2019) by the American Physical Society. For all cases, weff → 0 at early-times and asymptotically approaches DE in the future (see figure 12). Overall, this produces a viable cosmological model. A similar reconstruction procedure uses the combination of the equations (6.7a) and (6.7b) [394, 401], 4Ḣ [fT + 2TfTT ] = 0, 1+w (6.18b) with T = −6H 2 + a6k2 as given in equation (3.159). In the avenue that the second branch solution (section 3.6.3.3) is used, the form of T changes to equation (3.167) which is complex for k > 0. In the following works, the former description has been employed. Following [404, 405], an almost flat Universe is constructed by eliminating the explicit spatial curvature dependence in equation (6.18b) which leads to the reconstructed f (T) function to be weff = − 1+α(11+z)n for different choices of n as presented in [393]. f − 2TfT + 2k f fT − 2HḟT − , 2 a 2 (6.18a) ˆ t f(t) ∼ exp "ˆ # ... t̃ 2 k + (3äa − 5ȧ2 )k + 2ä2 a2 + 4ȧ4 − 7ȧ2 äa + ȧ a a2 ′ × dt dt̃. ȧa (äa − ȧ2 + k) (6.17) (6.19) and the specification of the perfect fluid’s EoS. Due to the presence of the Ḣ terms, solutions are obtained only in special limited scenarios such as specification of the evolutionary behavior and setting w to take constant values. For instance, for 3n a power law cosmology a(t) ∝ tn and w = 0, f(T) ∝ (−T) 2 which, however, does not host GR as a limit unless n = 23 . In [401], the evolutionary behavior is specified through the ... jerk parameter j(t) ≡ a /(aH3 ) where three models were considered (see the supplementary annexes (supplementary 1)): Here, k reappears by virtue of the torsion scalar. The dependence of k can be further suppressed by setting 3äa − 5ȧ 2 = 0,  − 3 i.e. a(t) ∝ 1 − 2H3 0 (t − t0 ) 2 with t0 representing the cutoff time and H0 := H(t0 ). Thus, the evolutionary behavior is independent of k and therefore flat-like. In addition, it has been shown to be consistent with early inflation. The associated reconstructed function P for this flat-like scenario takes on n ∞ the series expansion f(T) = n=0 αn (−T)− 2 for some coefficients αn . Use of spatial curvature also appears in [406, 407] to generate an accelerating de Sitter cosmology. In the limit when the matter content behavior approaches that of DE yields16 [406] (a) j = 1: A de Sitter accelerating behavior can be produced. At early-times, the torsional fluid varies in nature (−1 < weff < 1) whereas it always approaches phantom behavior at late-times. 2 (b) j = Hs 2 : late-time acceleration when m  n provided s > 0.   T a 2 (äa − ȧ 2 − k) f(T) ∼ exp , 12 ȧ 2 (äa − ȧ 2 + k) p) 2 µ := s(1s−(1p− ) 2 +4mn A best-fit value µ ∼ 0.39 where was obtained using SNIa Union 2.1 luminosity distance data [402]. Here, the nature of the matter fluid influences the behavior of weff : 1. w = 0: weff > 0 at all times with weff → 0 at late-times; 2. w = ± 12 : weff → ± 12 at late-times; (6.20) 16 The reconstructed function obtained in [406] is not a solution. Substituting equations (3.5) and (3.6) in (3.21) does not lead to equation (3.22); instead, the equation identically reduces to zero. 85 Rep. Prog. Phys. 86 (2023) 026901 Review −ȧ −k) provided that aȧ 2 (äa is assumed, (äa−ȧ 2 +k) is constant. If a vacuum   T+6H0 2 f (T) then takes on the form f(T) ∝ exp − 12H0 2 [407]. Contrary to the standard spatially flat cases, f (T) no longer reduces to a cosmological constant, instead it becomes dynamical which asymptotically approaches the latter behavior. 2 2 6.1.3. Dynamical system approach. The dynamical systems approach has been extensively investigated in f (T) gravity [420–426]. Given the simple nature of the field equations FLRW equations (6.7a) and (6.7b) as they can be solely expressed in terms of the Hubble parameter and its time derivative, allows for the straightforward construction of the one dimensional autonomous system 1 + w f − 2TfT f(H) − HfH = 3(1 + w) 4 fT + 2TfTT fHH 3 2 = − (1 + wTot ) H ≡ F(H), (6.21) 2 Ḣ = − In the first trivial extension of teleparallelism, f (T) gravity, Noether symmetry has been studied in numerous works. Generally, the configuration space is taken to be Q = {t, a, T} from which the Noether condition is then solved (see the supplementary annexes (supplementary 1)). Due to the simplicity of the Lagrangian, irrespective of whether the time symmetry generator is included in the analysis, the Noether condition imposes the Lagrangian constraint f(T) ∝ (−T)n for some constant n [408–416]. This 3 generates a power-law expansion behavior, a(t) ∼ t 2n , and henceforth includes a number of cosmological solutions. However, according to [410], n can only take a singular value of n = 32 . This restriction arises once the Euler–Lagrange field equations are solved in conjuction with the resulting conserved charge arising from the Noether symmetry. Together, they restrict n = 32 leading to a coasting cosmology. An analogous 6.1.2. Noether’s symmetry approach. where w represents the perfect fluid EoS and wTot is the total EoS, and where f H and f HH represent first and second derivatives of f (H) wrt H. Clearly, the critical points always correspond to Minkowski or de Sitter points which stability is identified by FH . This approach has been considered in [420, 423–426] for various choices of the f (T) Lagrangian. √ In [420, 426], the model f(T) = αT + β −T + γ, where α, β and γ are constants, was considered. However, this is simply rescaled TEGR with a cosmological constant hence not leading to other new dynamics beyond the ΛCDM framework. Meanwhile, [424] considers a dust matter fluid for different f (T) model ansatz: ◦ (a) −T + α(−T)b : Semi-stable Minkowski and accelerating (decelerating) stable (unstable) de Sitter critical points. For b  1, the sequence Big Bang singularity → matter domination → de Sitter acceleration is recovered. Similar conclusions are drawn√ in [425]; −p T/T0 (b) −T + αT0 (1 − e ): Semi-stable Minkowski and stable or unstable de Sitter critical points. A cosmological turnaround can be realized while a correct matter era is recovered for p > 0;  (c) −T + αT0 1 − e−β T/T0 : Similar critical points to model (ii). Cosmological evolution changes depending on p, but a deceleration → acceleration sequence can be realized for p = 1/2. Larger values of p yield better agreements with observations; (d) Teβ T0 /T : The number of critical points depend on the parameter β. 1. β < 0: Only a semi-stable Minkowski solution arises. A future Big Crunch singularity (H < 0) or deceleration → acceleration which ends at the Minkowski point (H > 0) can be realized; 2. β > 0: Semi-stable Minkowski and accelerating (decelerating) stable (unstable) de Sitter point. The cosmological evolutionary behavior√depends on the magnitude of H. In particular, H > 2β|H0 | appears to realize a viable cosmological scenario. restriction appears in f(R) theories where the model reduces to ◦ ◦3 f(R) ∝ R 2 [417, 418], a result which also holds true in the Palatini formulation [419]. Despite this similarity, the resulting cosmological behaviors are distinctively different from one another due to the fundamental differences between the two theories. In [413], the Noether symmetry approach was further extended to include velocities in the Noether symmetry vector. In the absence of these velocities, the model reduces to the f(T) ∝ (−T)n result. Otherwise, further solutions are √ obtained, including f(T) = ± 2c1 T + 2c2 for integration constants c1,2 . This leads to an entirely different cosmological behavior where it now describes a cosmological turnaround and acceleration/deceleration phases during CDM domination. Furthermore, the total EoS parameter does not cross the phantom-divide line and is always non-phantom. Nevertheless, these models do not have a TEGR background, making them less popular. Meanwhile, in [414], the Noether condition is not generally solved but only after specific f (T) ansatz are considered. Provided a symmetry exists for the considered ansatz, the model is then constrained against SN+H(z)+BAO data. In summary, only the f(T) = αT + βT n ansatz leads to meaningful results as it can match with observations with n = −1 being P the best-fitting parameter. The f(T) ∝ (−T)n , enT or f(T) ∼ n (−T)n ansatz are insufficient to match with observations. Finally, [415, 416] investigate the correspondence with the Wheeler–de Witt equation where the resulting classical Universe trajectories lead back to the power-law scale factor solu3 tion a ∼ t 2n . For arbitrary f (T) functions, the complete dynamical nature of the theory was obtained in [421] in the presence of dust and radiation fluids. To account for the two-fluid system, a further r was defined, leading to a twophase-space variable X = ρmρ+ρ r dimensional system. This also modifies equation (6.21) to 86 Rep. Prog. Phys. 86 (2023) 026901 Review Table 1. Summary of the critical points (H = H⋆ , X = X⋆ ) for an arbitrary f (T) model in the presence of dust and radiation fluids as found in [421]. Here, W⋆ and W⋆H represent the function W(H) and its derivative evaluated at H = H⋆ . (H⋆ , X⋆ ) Existence Stability Properties P1 (H⋆ ̸= 0, 0) W⋆ = 0 de Sitter vacuum P2 P3 (H⋆ ̸= 0, 1) (0, X⋆ ) W⋆ = 0 W⋆ > 0, W⋆H ̸= 0 P4 P5 (0, X⋆ ) (0, X⋆ ) W⋆ = 0 W⋆ > 0, W⋆H → ±∞ Attractor H⋆ > 0 Repeller for H⋆ < 0 Saddle Attractor W⋆H > 0 Repeller W⋆H < 0 Not determined Not determined Non-vacuum static universe Not determined Type IV finite singularity P6 (0, X⋆ ) W⋆ > 0, W⋆H = 0, H/WH → 0 Same behavior as P1 Non-empty Static Minkowski geometry Minkowski vacuum solution Figure 13. The dynamical behavior for the f(T) = −T + α(−T)b power-law model17 for different choices of b. Here, the gray (white) region represents decelerated (accelerated) expansion with the dotted line representing the transition. Circles represent the critical points with empty (filled) being repeller (attractor) and grey being semi-stable. Focusing on the H > 0 plane, a matter domination → late-time de Sitter is achieved for b = − 13 . Reproduced from [424]. © 2018 IOP Publishing Ltd and Sissa Medialab srl. Ḣ = −(X + 3)H W , WH Therefore, a late-time acceleration can be recovered through P1 . Furthermore, for any f (T) model, no cyclic and oscillating solutions are possible. Moreover, the sequence where the Universe starts with an initial accelerating phase (inflationary), become decelerating and finally evolves to an accelerating de Sitter phase cannot be realized. In the particular case when f(T) = −T + α(−T)b , the critical points as found in [424, 425] are recovered with the extra critical point P2 which cannot appear in the aforementioned works due to the different dynamical phase-space analysis. A simple dynamical representation for different choices of b are highlighted in figure 13. (6.22) where W(H) = f(H) − HfH . The critical points are listed in table 1. Evidently, the existence conditions depend on the behavior of W and W H . Meanwhile, attractor solutions are P1 (vacuum de Sitter) and P3 (non-empty static Minkowski geometry). 17 To match with the review convention, the TEGR term’s signature is reversed compared to [424]. 87 Rep. Prog. Phys. 86 (2023) 026901 Review Table 2. Summary of the existence of matter and radiation domination epochs, the matter to DE domination and finally the existence of a stable de Sitter cosmology, for various f (T) models as carried out in [428]. Here, m := TffTTT and r := − TffT . mr > −1, r = −1 m = − 12 , r = − 12 F(T) m = 0, r = −1 Ωm /Ωr Domination Ωm → DE Domination Stable de Sitter (−T)a (−T)a ebT b (−T)a e T (−T)a [ln(γT)]b Always a=1 a=1 √ a = 11±136 3 a<1 a > −2 a>0 a > −2 a = 12 a = 18 a = 18 √ a = 11±136 3 F(T) Model For the remaining works [427–432], instead of a onedimensional system, a larger autonomous system is constructed which allows for a more general investigation of non-de Sitter critical points. Of note, two notable approaches were employed. The first follows [427–431] which consider dust and radiation fluids with density Ωm and Ωr , respectively, and the phase-space variables18 f(T) + T , ỹ := −2(fT + 1), T 2 κ ρm κ2 ρr Ωm := , Ω := . r 3H 2 3H 2 x̃ := z̃ := 2TfTT , (6.23) Following [428, 429], for arbitrary choices of the gravitational Lagrangian, a stable de Sitter attractor, a stable or saddle scaling solution with matter and a possibly unstable scaling radiation solution were found whereas an inflationary solution was not observed. For particular f (T) models, the cosmological sequence starting with radiation domination → matter domination → de Sitter acceleration was investigated. As summarized in table 2, a viable model which realizes the sequence is obtained either in the power-law model T a or in b the logarithmic model Ta [ln(γT)] while the exponential models are able to realize the matter to DE domination sequence leading to a stable de Sitter phase but do not host dust and radiation domination epochs. Of further note, in the powerlaw model, a stable de Sitter solution is realized for a = 12 √ which corresponds to the case where a −T term appears in the Lagrangian. As this term does not contribute to the field equations, the stability of the solution is to be treated lightly. In [430], a DE dominated de Sitter sink is found which suggests the resulting late-time behavior of the theory. In addition, the critical points (x̃⋆ , Ωm ⋆ ) = (0, 0) and (0, 1) are found which correspond to radiation and matter dominated epochs respectively, are also obtained for the models f(T) = −T + α(−T)n and f(T) = −T + α(−T)β ln T, β = { 12 , 1}. In the power-law and the β = 12 cases, the domination phases correspond to their expected decelerating expansion rate. However, for the logarithmic β = 1 case, the rate of deceleration (or possibly acceleration) depends on the magnitude of α, hereby Figure 14. The√dynamical behavior for the logarithmic f(T) = −T + α −T lnT model with the critical points marked with circles (blue—radiation, violet—matter, red—de Sitter). The orange line (x = 2) represents a singularity present in the dynamical system. In fact, the evolutionary behavior is distinct between x < 2 and x > 2. To match our conventions, x → x̃, y → Ωm and f(T) → −T + f(T). This figure first appeared in [430]. Reproduced with permission from [430]. © World Scientific Publishing. in agreement to the bifurcation observed in the dynamical analysis. Overall, a suitable radiation → matter → de Sitter evolutionary process can be realized for suitable parameter choices (see figure 14 for the β = 12 logarithmic model phase-space representation). Following [427] and the improved results in [431], critical points of the form (x̃⋆ , Ωr ⋆ ) were obtained for arbitrary choices of f (T), as summarized in table 3. Overall, the sequence starting from an inflationary period all the way to a late-time acceleration can be realized provided the following conditions are met: (i) x̃0 and x̃1 exist, (ii) −1 < x̃0 ⩽ 0 and (iii) ỹx̃ (x̃0 ) > −2. These conditions were examined for a selected number of f(T) = −T + F(T) models with F(T) being equal to: 18 With the exception of [430] where the combination of phase-space variables X̃ = x + y is considered. 88 Rep. Prog. Phys. 86 (2023) 026901 Review Table 3. The critical points along with their stability and cosmological description for an arbitrary f (T) model based on  the analysis carried out in [431]. Here, x̃0 is a solution to the equation ỹ(x̃) + 2x̃ = 0, x̃1 for ỹ(x̃) + x̃ = 1 and x2 when 1 − 12 ỹ + ỹx̃ x̃ + 12 ỹ diverges. P1 P2 P3 P4 (x̃⋆ , Ωr ⋆ ) Stability Cosmological description (x̃0 , 1 + x̃0 ) (x̃0 , 0) (x̃1 , 0) (x̃2 , 0) Saddle or unstable Stable or saddle Stable Stable or saddle Decelerating radiation or DE dominated solution Decelerating matter or dark energy dominated solution Dark energy dominated de Sitter accelerating phase Inflationary phase (without graceful exit) Figure 15. The dynamical behavior for the exponential f(T) = −T + T0 1 − e −p TT  0 model for (a) p = 1 and (b) p = 1.5. For this model, x0 = 0 meaning the critical points P1 (0, 1) (radiation) and P2 (0, 0) (matter) appear in both plots. The critical point P3 (de Sitter) also appears with (a) x1 = −4.59 while in (b) x1 = 1 and 5.04. Thus, for both p choices, the sequence radiation → matter → de Sitter can be achieved. Note the identification x → x̃ for this figure. Reproduced from [431]. © 2017 IOP Publishing Ltd and Sissa Medialab srl. (a) Tn s   T qT0 ln qT0 T   T0 (c) T n tanh T  q  −p TT 0 (d) T0 1 − e   −p T (e) T0 1 − e T0 (b) n < 1, Finally, in the second consideration examined in [432], the phase-space variables q > 0, x̃ = − n < 1, p ⩾ 0.5, f̈ , Hḟ ỹ = f , 4H 2 ḟ z̃ = 3H 2 + Ḣ , H2 (6.24) were chosen and two scenarios were considered: (i) absence of matter fluids and (ii) presence of dust and radiation components. Furthermore, the case when the parameter m = − HḦ3 takes on constant values m = 0 (quasi-de Sitter evolution) and m = − 92 (matter dominated evolution) was explored. Starting with the vacuum case, an asymptotic stable de Sitter late-time cosmology can be reached while the matter domination is necessarily an unstable point. On the other hand, in the presence of matter, dust dominated unstable critical points and an asymptotic DE dominated de Sitter critical point are obtained. However, it is unclear whether a radiation domination epoch can be generated. 0.203 ⩽ p ⩽ 1.977, p ⩾ 6.110, p 6= 1.256, which highlight the possibility of realizing the desired sequence of cosmological behaviors (see, for instance, the exponential model dynamical behavior in figure 15). However, this appears to be in disagreement with the results obtained in [421] where it is claimed that inflation → deceleration → accelerating de Sitter phase cannot be realized for arbitrary f (T) models. 89 Rep. Prog. Phys. 86 (2023) 026901 Review Another interesting work which uses dynamical system is [433] where the authors studied the cosmology of Ddimensional f (T) gravity. They focused on eleven dimensions (with seven dimensions being compactified) and found that in this case it is possible to find an early inflationary epoch without introducing any extra matter source. In another work [434], the compactification of extra dimensions in flat FLRW cosmology for f (T) was further studied obtaining a correct tetrad in six and seven dimensions which can be used as the starting point to compactify the extra dimensions. This means that for this branch the Friedman equations become 1 3HḟB − 3H 2 (3fB + 2fT ) − 3fB Ḣ − f(T, B) = κ 2 ρ, (6.30a) 2 −(3H 2 + Ḣ)(2fT + 3fB ) − 2HḟT + f̈B + (6.30b) Recently, this branch was studied using dynamical systems finding that Milne, Milne-like and de-Sitter solutions exist [437]. It is worth mentioning that the second branch (3.161) can only be used for k ⩽ 0 (since T, B become complex) has not been used in the literature yet. Following the same idea as we did in the previous section, we will analyze these equations further in the next sections. 6.2. f(T, B) cosmology If we use the FLRW diagonal tetrad in equation (6.4) in the f(T, B) gravity equations (5.80), we arrive at the following modified Friedmann equations [257, 260, 261] 1 3HḟB − 3H 2 (3fB + 2fT ) − 3fB Ḣ − f(T, B) = κ 2 ρ, (6.25a) 2 One notices from the f(T, B) cosmological equations (6.25a) and (6.25b) that reconstructing for a general f(T, B) function is not generally possible; thus, the ansatz choices f(T, B) = f1 (T) + f2 (B) and its sub-case with f1 (T) = −T were considered, for the (i) power-law a(t) ∝ th , h > 0, (ii) de Sitter, (iii) ΛCDM, and (iv) phantom dominated, cosmological behaviors [257, 260, 261]. The respective solutions are summarized in the supplementary annexes (supplementary 1). A noticeable feature appears for ΛCDM where the sub-case scenario offers a much simpler Lagrangian compared to the separable ansatz as it negates the existence of Gauss’s hypergeometric function. Beyond the reconstruction procedure, the solutions were then tested against the validity of the generalized second law of thermodynamics as well as their stability against homogeneous perturbations. In the former, each model is able to satisfy the constraint. In the latter, only the power-law and de Sitter behaviors were explored where both are able to exhibit stability. More importantly, however, is the fact that the power-law cosmology is only stable for h > 1, i.e. an expanding behavior. 6.2.1. Reconstruction method. 1 −(3H 2 + Ḣ)(2fT + 3fB ) − 2HḟT + f̈B − f(T, B) = −κ 2 p. 2 (6.25b) The equations can be recast into a TEGR-like form when the TEGR contribution is maintained, i.e. f(T, B) = −T + F(T, B), where the gravitational component F(T, B) acts as a gravitational fluid with associated energy density ρeff and pressure peff . In other words, equations (6.25a) and (6.25b) are rewritten as 3H 2 = κ2 (ρ + ρeff ) , (6.26a) 3H 2 + 2Ḣ = −κ2 (p + peff ) , (6.26b) with the gravitational fluid defined as κ2 ρeff = 3H 2 (3FB + 2FT ) − 3HḞB + 3ḢFB + κ2 peff = − F(T, B) , 2 (6.27a) F(T, B) − (3H 2 + Ḣ) (3FB + 2FT ) − 2HḞT + F̈B . 2 (6.27b) Following [268, 438], various f(T, B) ansatz have been considered to examine the Noether symmetries, namely 6.2.2. Noether’s symmetry approach. Consequently, the effective EoS is [263, 435, 436] weff := (i) b0 Bk + t0 (−T)m , (iv) Bn , F̈B − 3HḞB − 2ḢFT − 2HḞT peff = −1+ 2 . ρeff 3H (3FB + 2FT ) − 3HḞB + 3ḢFB + 12 F (6.28) (ii) Bl (−T)n , (iii) − T + F(B), where b0 , t0 , k, m, l and n are constants and F(B) is some arbitrary function not linear in B. Of note, the first model ansatz only satisfies the Noether condition provided k = 1 reducing the model to the power-law f (T) model investigated in section 6.1.2. For the remaining models, further restrictions and cosmological implications are summarized in table 4. It is noted that for the −T + F(B) ansatz, a similar F(B) functional form appears in [257, 261]. Under the assumption that the matter fluids do not interact with the gravitational fluid, the latter satisfies the standard fluid conservation equation ρ̇eff + 3H(ρeff + peff ) = 0. 2kfT 1 − f(T, B) = −κ 2 p. 2 a2 (6.29) ◦ For the f(R, T) gravity sub-case, the Noether symmetry has been explored in [439]. As the Noether condition yields an intractable system of equations to solve generally, two As discussed in section 3.6.3, for the non-flat FLRW case there are two branches. Interestingly, in the first branch (3.154), the boundary term does not depend on k (see equation (3.159)). 90 Rep. Prog. Phys. 86 (2023) 026901 Review Table 4. A summary of the existence of symmetries and the cosmological implications of the f(T, TB ) ansatz considered in [268, 438]. Here, α0 , . . . , α5 and β are integration constants. f(T, TB ) Symmetry and cosmological implications Bl (−T)n 1  βt 3 1+l √ l 6 +α 3 . Depending on the symmetry, power-law a(t) ∝ t Condition enforces m = 1− or exponential a(t) = α e 0 1 2 −T + F(B) Bn behaviors can be recovered. F(B) constrained to be F(B) = − B3 ln B.  1/3 When dust is considered, a(t) = α2 e α3 t + α4 t + α5 . In the case of dark energy, the solution is identical except α4 = 0. Symmetries exist for dust and w = 2n1−1 . However, no cosmology has been reported. ◦ ◦ general description of the resulting cosmology. In order to conform with the signature notation used in the Review, the auxiliary scalar field ϕ relations are modified to ϕ = −FB and V(ϕ) = F(B) − BFB . Thus, the chosen phase-space variables are f(R, T) ansatz were considered which reduce to: (i) f(R, T) = ◦ ◦ ◦ n 1−n g0 R + h0 T and (ii) f(R, T) = f0 R (−T) where f0 , g0 , h0 and n are arbitrary integration constants. As the first model is a TEGR rescaling, the second model offers a richer cosmo◦ logy provided that n 6= 1 and is absent in both f(R) and f (T) formulations. x̃ := The dynamics of the specific class f(T, B) = −T + F(B), for some unknown function F(B), has been investigated via its scalar-tensor equivalent form by introducing an auxiliary scalar field ϕ = FB with an associated potential V(ϕ) = BFB − F(B) [257]. This scalar field (or equivalently the F(B) contribution) is assumed to act as the source for DE. In particular, the case when DE is absent during certain phases of evolution was studied, which leads to matter dominated phases. Here, the matter fluids are assumed to have an EoS wm ∈ [0, 1). One avenue is to set the DE energy density and pressure 2−γ to be vanishing which leads to V(ϕ) ∝ ϕ 2+γ (or equivalently, 2−γ F(B) ∝ B− 2γ ). However, according to dynamical analysis, the cosmology is unable to evolve past the matter domination phase. Instead, DE was then considered with non-vanishing energy density and pressure while still being magnitudes smaller than that of matter. In this case, the scalar field evolution equation becomes19 6.2.3. Dynamical system approach. ϕ̈ +  V (1 − γ) + 4 − 3γ 2 κ2 ρm0 = 0, 2 3γ 2 t 2 ϕ̇ , H ỹ := V(ϕ) , 6H 2 λ := − Vϕ . V(ϕ) (6.32) Here, the choice of λ determines the nature of the potential (and hence, the choice of F(B)). In particular, two scenarios were explored: (a) λ = constant leading to V(ϕ) ∝ e−λϕ , and (b) λ 6= 0, i.e. a general potential. As the critical points of (a) are contained in the general case (b), we highlight the latter results. For the above phase-space variables, five critical points are obtained which nature and existence are summarized in table 5. Evidently, the model hosts stiff fluid behaviors (P1 , P5 ) and two possible stable de Sitter behaviors (P2 , P4 ). Meanwhile, the matter domination epoch is achieved via P2 . In the instance when λ = constant, only the first critical points P1 –P3 remain. For constant λ, matter domination → late-time acceleration can be achieved through P2 → P3 whereas for the general model, a similar behavior is achieved through P2 → P3 /P4 . Recently, the scalar-tensor approach has also been explored for the f(T, B) = F1 (T) + F2 (B) model ansatz where F1 (T) = B −T + f0 (−T)n and F2 (B) = B ln where f0 , λ and n are λ constants [441]. In particular, the F2 (B) function leads to a scalar potential V(ϕ) ∝ e−λϕ . Thus, this model serves as an extension of the previous λ = constant scenario where a torsion power-law contribution has now been included. For this choice, in the absence of matter fluids, the chosen phase-space variables are (6.31) where γ = 1 + ωm is the barotropic index. Naturally, during matter domination, the bracketed term is vanishing reducing the system to the previous scenario. In this case, a solution exists for the power-law potential V(ϕ) = V0 ϕµ with µ 6= 1 without the necessary constraint that µ = 22−γ +γ . Additionally, pϕ ≃ 0 for this potential meaning that the scalar fluid behaves as dust at leading order. The dynamical systems analysis reveals the existence of three critical points for µ 6= −1 with deviation from matter domination occurring only for µ > 1. An extension of the above approach is presented in [440] where the dynamical system procedure was applied with a different set of phase-space variables which permit a more x̃ := p ϕ̇ 1 + H2 , ỹ := V(ϕ) , 6H 2 z̃ := (−T)n , 1 + H2 η := p H 1 + H2 . (6.33) Since ỹ is constrained by the Friedmann constraint (6.25a) and z̃ = z̃(η) by virtue of T = −6H 2 , the system’s dimensionality reduces to a two-dimensional one. This leads to six finite critical points (x̃, η) being: (a) two unstable critical points where the effective fluid behaves as stiff matter. Only the kinetic term of the auxiliary field contributes to the fluid’s behavior; 19 The matter contribution equation (74) in [257] was not expressed correctly. 91 Rep. Prog. Phys. 86 (2023) 026901 Review Table 5. The critical points for f(T, B) = −T + F(B) together with their existence conditions viewed from the introduction of an auxiliary 3 3 scalar field ϕ = −FB [440]. Here, λ0 is an arbitrary value, Γ̄(λ) ≡ Vϕϕ2 − 1, λ± 0 ≡ 2λ0 (1 ± w) and λ̃ ≡ 2 (3 + w). V ϕ ⋆ ⋆ ⋆ (x̃ ,ỹ , λ ) Existence Stability Properties P1 (1, 0, λ0 ) Γ̄(λ0 ) = 0 Unstable P2 − (2λ+ 0 , λ0 , λ0 ) λ0 ⩾ λ̃, Γ̄(λ0 ) = 0 λ0 > λ̃ Matter fluid absent; effective fluid behaves as a stiff fluid (weff = 1) Effective fluid behaves as dust (weff = 0) P3 (− λ0 + 2, λ0 − 1, λ0 ) Γ̄(λ0 ) = 0 λ0 < λ̃, (−3 + λ0 )Γ̄λ (λ0 ) > 0 P4 (0, 1, 3) Always ℜ(Γ̄(3)) > 0 P5 (1, 0, 0) Always Unstable 6 6 which each i denotes the component density parameter Ωi = κ 2 ρi /(3H 2 ). The prime (0) denotes derivatives wrt Nf = ln a, with a chain rule given by d/dt = H(d/dNf ). These set of equations impose a condition over the form of the derivative f ′ (T, B). Taking the derivative wrt Nf in equation (6.34a) and making use of the FLRW equations, the relation  ′  ′  ′2  H H H H′′ F′ 6 + = 0, FB + 2 FT + F + B H H H2 H 6H 2 (6.35) (b) two critical points where the effective fluid behaves either as a perfect fluid with weff = 23λ − 3 for λ 6= 3 or as a cosmological constant for λ = 3. These points are attractors for λ < 3 and n < 1, otherwise, they act as sources/saddle points; (c) a de Sitter critical point which can serve as a late-time attractor; (d) a Minkowskian spacetime in which stability can be determined through center manifold theory. Nonetheless, for the considered numerical analysis, it is observed to be unstable. is obtained which can be used to impose a condition, at least at a background level, over the form of the derivative F ′ (T, B). Naturally, further constraints can be obtained once cosmological perturbations are considered (see section 7). Following equations (6.34a) and (6.34b), a set of conveniently specified dimensionless variables are chosen to construct the autonomous system [431, 443]. First, the parameter [444] Meanwhile, the critical points lying at the infinite boundary cannot be attractors. Thus, a viable cosmology can be realized for suitable parameter choices. However, to confirm the existence of matter and radiation dominated phases, the contribution from matter fluids still needs to be incorporated in the analysis. Scenarios where the contributions of DE without a scalar field are assumed to be the source has not been studied widely yet. In [266] a dynamical analysis was presented where, contrary to the previous approach, the cosmic acceleration effect is analyzed without the introduction of an auxiliary scalar field. In the following, an f(T, B) DE which is fluid-like is used to obtain a richer population of stability points that can be constrained by observational surveys [442]. The models can also be tested against thermodynamics and cosmological stability [261, 262] as well as energy conditions [260]. Taking the ansatz f(T, B) = −T + F(T, B), in order to construct the dynamical system, the FLRW equations (6.25a) and (6.25b) are first rewritten as Ω + Ωeff = 1,  H′ 3+2 H  λ=  ′ H F + 9FB + 6FT + = (3FB + 2FT − FB′ ) 2H 2 H x̃ := FB , 2 Ωeff ≡ 3FB + 2FT − FB′ + F + 6H 2 (6.34b)  (6.36) ỹ := FB′ , z̃ := H′ Ḣ = 2, H H g := F 6H 2 (6.37) are introduced. The energy constriction, equation (6.34a), in terms of the above variables becomes  ′ H H Ḧ H ′2 H ′ ′ = 2 + , 3 H H H is introduced. For cases where λ = constant, some cosmological solutions can be recovered including de Sitter/quaside Sitter (λ = 0) and matter domination (λ = 9/2). This ansatz, which shall be assumed in the following, shows cosmological viable scenarios as analogous to models with barotropic fluids with EoS w. However, this choice restricts the number of critical points which in turn leaves an incomplete picture of the whole cosmological evolution for the chosen F(T, B) model. To remove this imposition, either a different dynamical system approach or suitable choice of the F(T, B) function must be adopted. In fact, the latter case is discussed in [445] for the mixed power-law model F(T, B) ∝ Bk T m for constants k, m. Following this, the dynamical phase-space variables (6.34a) κ p + 2FT′ − FB′ ′ − 2 , H Matter fluid absent; effective fluid EoS weff = 2λ03−9 (accelerating for λ0 < 92 and de Sitter for λ = 3) Matter fluid absent; de Sitter behavior (effective fluid behaves as DE) Same as P1 FB , (6.34c) Ω + 3x̃ + 2FT − ỹ + g + z̃x̃ = 1, 92 (6.38) Rep. Prog. Phys. 86 (2023) 026901 Review where Ω = κ2 ρ/(3H 2 ) depends on the other dynamical variables. We can write the autonomous system for this theory using (6.34a) as [266, 445] ′ x̃ = ỹ, ỹ ′ consider T0 = B0 = 0. Labeling the constant coefficients as Ai , the Lagrangian can be rewritten as F(T, B) ≃ A0 + A1 T + A2 T 2 + A3 B 2 + A4 TB, (6.39a) where the linear boundary term does not contribute so we have omitted it. We notice from this specific form that the first term can be seen as A0 ≈ Λ, therefore we are dealing with a cosmic acceleration as a consequence of the F(T, B). Thus, the form of this model can be written in terms of the dynamical variables as κ2 p = 3g + (9 + 3z̃)x̃ + FT (6 + 2z̃) + 2FT′ − z̃ỹ − 3 − 2z̃ − 2 , H (6.39b) z̃ ′ = λ − 2z̃ 2 , (6.39c) g ′ = −6z̃x̃ − 2z̃FT − λx̃ − 2z̃g̃. (6.39d) FT = −(3 + z̃)x̃ − 2g − A1 , To follow the constraint of the system in equations (6.38)– (6.39d), we need to write F T either as a dynamical variable or expressed in terms of the described variables. The latter can be done by considering a specific form of F(T, B). Finally, we can rewrite the EoS (6.28) as weff = H 2 z̃(2z̃ + 3) − κ 2 pz̃ . 3H 2 [g ′ + z̃(g − x̃z̃ + 3x̃ + ỹ) + λx] z̃ = ± λ , 2 ỹ = 0. (6.44) at linear order in torsion and with A0 = 0, i.e. we are switching-off the cosmological constant. This can be done since an explicitly time-dependent factor appears and then a different approach has to be taken. The critical points for this model are (6.40) g = −A1 , From equations (6.39a) and (6.39c), we notice that there is not an explicit dependency of F(T, B); therefore, for the critical points, we require that r (6.43) x̃ = A1 − 1 q , 3 ± λ2 (6.45) q where the case 3 ± λ2 6= 0 was not explored. The constriction evolution equation (6.38) sets Ω = 0 hereby denoting contributions only related to F(T, B) gravity which imply that the constriction evolution equation in equation (6.38) is now Ω = 0 which denotes only contributions related to F(T, B) gravity. According to these points, the following eigenvalues result r r λ λ ω1 = −3 ∓ , ω2 = −3 ∓ 2 , 2 2 r r λ λ (6.46) ω3 = ∓4 , ω4 = ±2 . 2 2 (6.41) As z̃ = Ḣ/H 2 , this variable becomes an important determining factor for the stability and cosmological nature (power-law z̃ 6= 0 or de Sitter z̃ = 0) of the critical points. To achieve this, we are going to consider three cosmologically viable models, determine the critical points (by determining the remaining critical values x̃ and g), then consider the eigenvalues of the stability matrix and the constriction equation equation (6.38). In the following, the results reported from [266] are for the case when the perfect fluid is sourced by dust w = 0. Considering values λ 6= 0, we get that Re(ω3 ) = −Re(ω4 ) 6= 0 implying that, for this system, all the critical points are saddle-like. The solutions for this case are in agreement with the cosmological constraints found in [263]. According to these results, the critical points behave as Ai < Ai+1 (with A0 = 0, A1 = 1), show a decelerating behavior with 0 ⩽ weff < 1 when B dominates at redshift z ≈ 1 followed by a ΛCDM behavior. (b) Power Law model—Considering a Lagrangian of separated power law style models for the torsion and boundary scalars, we can write a model like [268] (a) General Taylor Expansion model—The form for this model was presented in [446], given by F(T, B) = F(T0 , B0 ) + FT (T0 , B0 )(T − T0 ) 1 + FB (T0 , B0 )(B − B0 )+ FTT (T0 , B0 )(T−T0 ) 2 2! 1 + FBB (T0 , B0 )(B − B0 ) 2 2! + FTB (T0 , B0 )(T − T0 )(B − B0 ) + O(T3 , B3 ), (6.42) F(T, B) = b0 Bk + t0 (−T)m , which gives the general Taylor expansion of the f(T, B) Lagrangian about its Minkowski values for the torsion scalar T and boundary term B. The Minkowski space is chosen as the background for this model since all models should contain this spacetime in some limit. We notice from here that we need to take into account beyond linear approximations since B is a boundary term at linear order. Where locally spacetime appears to be Minkowski, we can (6.47) for constants b0 and t0 . This is an interesting model since it was already been shown in [389] that for m < 0, the Friedmann equations will be effected mostly in the accelerating late-time Universe while for m > 0, this impact will take place for the early Universe, assuming no input from the boundary contribution. By incorporating the boundary term, this analysis will reveal an effect of B on the combined evolution within f(T, B) cosmology. The form for 93 Rep. Prog. Phys. 86 (2023) 026901 Review this model can be written in terms of the dynamical variables as m FT = −mg − (3 + z̃)x̃. (6.48) k The critical points for this scenario are   k−m k m−1 1 q , g= , x̃ = − m(1 − k) m k−1 3± λ However, the more general approach in [445], which includes both radiation ρr and a perfect fluid with EoS w, ρw , leads to a different conclusion. Most notably, λ is no longer constrained to be constant as it is expressed in terms of the dynamical variables. This alters the former matter dominated critical point to a late-time de Sitter one with Ωr = Ωw = 0, Ωeff = 1 which can be stable under suitable parameter conditions. Investigation of the cosmological evolution close to the critical point was then used to best-fit H 0 using H(z) observations. The obtained values can match with both early (Ωr 6= 0) and late-time (Ωr = 0) observations. Such H 0 values can be derived for three cosmological scenarios which links their behavior together with the error propagation on the free variables of the system. A full description of this matter can be seen in figure 3 in [445]. (6.49) 2 q where, once more, the case 3 ± λ2 6= 0 was not explored. In this scenario, we find attractors or saddle points for the positive branch whereas repellent or saddle points for the negative branch. Additionally, for the former case, keeping b0 , t0 > 0, ΛCDM and late cosmic acceleration can be recovered. (c) Mixed Power Law model—In order to reproduce several important power law scale factors relevant for several cosmological epochs, in [268] a form of F(T, B) given by F(T, B) = f0 Bk (−T)m , The above results can be linked in a more straightforward manner if we consider directly the form for the EoS equation (6.28). At the critical point, for the first two cases (Taylor and Power law model), r 2 λ weff = −1 ∓ . (6.55) 3 2 (6.50) was presented, where the second and fourth order contributions will now be mixed, and f0 , k, m are arbitrary constants. We can recover ΛCDM when the index powers vanish, i.e. when k = m = 0. For this case, the model can be written in terms of the dynamical variables through FT = −mg. Meanwhile, for the Mixed Power Law model r 2 λ [266] λ = constant: weff = −1 ∓ (k + m) , 3 2 [445]non-constant λ : weff = −1. (6.56) (6.51) In comparison to the latter F(T, B) scenarios, this case implies x̃ = FB = kf0 Bk−1 (−T)m = F=− Notice that we recover ΛCDM in the GR limit. In both EoS scenarios, we recover a ΛCDM model when λ vanishes as z̃ = 0. Notice that we can rewrite the z̃-variable from equation (6.37) using the definition of the second cosmographic parameter, the deceleration parameter q, as z̃ = −(q + 1). Thus, λ = 0 leads to q = −1. We can also rewrite the ansatz for λ given in equation (6.36) in terms q of the k k F=− B 6(3H 2 + Ḣ) gk F k =− , 6H 2 3 + Ḣ2 3 + z̃ (6.52) H third cosmographic parameter, the jerk, as j = λ ∓ 3 λ2 + 1. Notice that when λ = 0, we recover the standard value j = 1. Notice that an important feature of this analysis (as summarized in figure 16) is the role of couplings between the torsion tensor T and the boundary term B. These represent the second order and fourth order contributions to the field equations respectively, and in a particular combination, form- from which we can notice that x̃ is not an independent variable of the dynamical system. In the same way, as ỹ = x̃ ′ , we directly obtain that ỹ = ỹ(g,z̃). With these conditions, the autonomous system can be reduced to a twodimensional dynamical phase-space   6z̃(k+m−1) + λk+2z̃ 2 (m − 1) z̃ ′ = λ − 2z̃ 2 , g ′ = g , 3+z̃ (6.53) ◦ ◦ ing f(R) = f(−T + B) gravity. In f(R) gravity, these couplings do appear but in a very prescribed format. with critical points r z̃ = ± 6.3. Teleparallel Gauss–Bonnet cosmology λ , 2 and g = 0. (6.54) In the case of f(T, TG ) gravity, since the tensorial field equations are very involved, it is easier to use the minisuperspace for flat FLRW cosmology. To do this, we require the torsion scalar and the teleparallel Gauss–Bonnet scalar in the minisuperspace, which are given by Under these values, the constriction of the system is given by Ω = 1. Once more, z̃ turns out to be the determining factor in the behavior of the dynamics of the system. On the other hand, the system turns out to be relatively straightforward to analyze with clear cut results which tally with the general results of the power law model. TG = 94 24H 2 Ḣ + 24H4 , N T = −6H 2 , (6.57) Rep. Prog. Phys. 86 (2023) 026901 Review Figure 16. Dynamical behavior for f(T, B) = −T + F(B) models and their results. and then the Friedmann equations for f(T, TG ) gravity can be directly found by using the minisuperspace analysis and then fixing the gauge such that N = 1, giving us Of note, models (iii) and (iv) are distinct despite their apparent similarity. Nonetheless, the resulting reconstructed solutions are related via:  1 −6fT H 2 − 12H3 ḟTG + 12fTG H 2 Ḣ + H 2 − f(T, TG ) = κ2 ρ, 2 (6.58a) Model (iv) f(T, TG ) = −T + − 2HḟT − 2fT Ḣ + 3H  2 + 12fTG H 2 Ḣ + H 3TG + Solution of model (iii). 4T (6.59) As such, only the results of model (iii) are summarized. Although a majority of the ansatz and cosmological behaviors considered do yield an analytical solution (see supplementary annexes (supplementary 1)), some scenarios do not host a solution. Note that for the power-law cosmology a(t) ∼ tα reconstructed solutions, the coasting cosmology (α = 1) has been investigated separately as TG = 0 in this case. Certain reconstructed solutions constitute relatively more complicated forms including a lack of a closed form or are expressed in terms of Gauss’s hypergeometric function. Despite this apparent complexity, the nature of the perfect fluid may yield simpler results. For instance, the ΛCDM cosmology with w = 0 reduces to the standard ΛCDM Lagrangian for model (i). However, non-trivial models are also obtained, such as  2  1 − 8HḟTG Ḣ + H 2 − 4H 2 f̈TG − f(T, TG ) = −κ2 p, 2 (6.58b) where fTG = ∂f/∂TG . It is worth mentioning that BG = 0 in flat FLRW and then the standard Gauss–Bonnet term is ◦ just G = TG . It is worth mentioning that the scalars TG1 = TG2 = 0 and TG3 = 24H4 , TG4 = 24H 2 Ḣ/N in flat FLRW (see equations (3.201a)–(3.201d) for their definitions), then, TG = P4 T is reproduced as expected. i=1 Gi The TEGB extension has been investigated in [86, 90] for various late-time behaviors. Due to the introduction of the TEGB scalar, a number of model ansatz were considered 6.3.1. Reconstruction method. f(T, TG ) = 3TG (ΩΛ T0 − 2T) , 8T 2 (6.60) by virtue of model (iii). This could be resourceful in other sectors such as cosmological perturbations as a different choice of Lagrangian may lead to different conclusions even though the background evolution is still that of ΛCDM. The reconstructed models were then examined and checked if they are able to host vacuum solutions i.e. whether f(0, 0) = 0. The reconstructed ΛCDM. model behavior cannot host vacuum solutions (see Supplementary annexes (Supplementary 1)). As shown in Table 6, the reconstructed ΛCDM model (i) f(T, TG ) = g(T) + h(TG ), (ii) f(T, TG ) = Tg(TG ), (iii) f(T, TG ) = TG g(T). (iv) f(T, TG ) = −T + TG g(T), (v) f(T, TG ) = −T + µ(−T)β TG γ . 95 Rep. Prog. Phys. 86 (2023) 026901 Review Table 6. The conditions for the reconstructed models found in Table 7. Summary of conditions and cosmological behaviors [86, 90] to obey the vacuum condition f(0, 0) = 0 for the ΛCDM and power-law cosmology subject that all individual contributions are maintained. In the latter cosmology, the distinction between α = 1 and α ̸= 1 is necessary due to branching solutions. obtained for the model f(T, TG ) ansatz considered and analyzed in [281, 416, 438]. Here, a0,1,2 are integration constants. Ansatz ΛCDM Power-law (i) 8 α ̸= 1: w < −1, α < 0 or w > −1, 0 < α < 1 α = 1: w > −1 or vacuum, and depending on h(TG ) α ̸= 1: 3α(1 + w) > 1, −3 < α < 7 α = 1: g(0) is finite 4 + 3α(1 + w) > 0 Vacuum or 4 + 3α(1 + w) > 0 Always (ii) Unknown (iii) (iv) (v) 8 8 8 The cosmological dynamics of f(T, TG ) gravity was investigated in [449]. In p particular, the model f(T, TG ) = −T + α1 T 2 + α2 TG , where α1,2 6= 0 are constants, was studied. In this case, the presence of a perfect dust fluid was assumed and the following dimensionless phase-space parameters were defined s x̃ = f0 (−T)m + f1 Bn + f2 TG q , (ii) 2α2 1+ 3   Ḣ 1+ 2 , H Ωm = κ2 ρm . 3H 2 (6.61) As the phase-space (x̃, Ωm ) is noncompact, the boundary can be studied with Poincaré’s projection method. Two new phasespace variables, (r, θ), having domain r ∈ [0, 1) and θ ∈ [0, π2 ], are introduced to determine any critical points lying on the boundary. These are related to the original parameters via (iii) TG n , where g0 , t0 , k, m, l and n are constants. Model (i) imposes the constraint k = 1 reducing the Lagrangian to the f (T) power-law case. For the remaining cases, symmetries can be found along with their associated cosmological behaviors, a summary of which is provided in table 7. The TG l (−T)m ansatz leads to interesting phenomenology as besides the standard power-law cosmology, trigonometric and hypergeometric behaviors can also be recovered. Additionally, the power-law behavior is recovered as a classical trajectory following the Wheeler–de Witt formulation [416]. Meanwhile, there is no clear evolutionary behavior for model (iii) although a power-law behavior is observed if a1,2 = 0. For the extended generalized teleparallel class f(T, B, TG , BG ) gravity [438], since for the FLRW cosmology BG = 0, the theory reduces to that of f(T, B, TG ) gravity. Despite this simplification, given its general complexity, specific ansatz choices were considered for the Noether symmetry approach (i) l Noether condition constrains m = 1− 2 . Depending on the symmetry considered, power-law, trigonometric a(t) ∼ tan t or hypergeometric a(t) ∼ tanh t behaviors are achieved. Symmetries exist for dust and w = 4n1−1 . A partial analytical solution is obtained for n ̸= 1, 14 , being  n−1 n−1 ´  (4n − 5)a0 a4 − a1 4n−5 [4(n − 1)] 4n−5 da = t + a2 . 6.3.3. Dynamical system approach. Following [281, 416, 438], the following f(T, TG ) model ansatz were chosen to simplify the Noether condition system of equations, being TG l (−T)m , TG l (−T)m where f0,1,2 , m, n and q are constants. The Noether condition reveals that the first model imposes the constraint n = q = 1 reducing it to the f (T) power-law model (section 6.1.2). Meanwhile, the second model contains two Noether symmetries but its general cosmological behavior was not explored. 6.3.2. Noether’s symmetry approach. (ii) Symmetry and cosmological implications TG n behavior cannot host vacuum solutions. On the contrary, the power-law can satisfy the vacuum constraint subject to conditions. While the de Sitter scenario is not listed, the vacuum condition can be satisfied by suitable Lagrangian choices. On a different framework, the reconstruction procedure has been applied to generate models yielding PlDE with the IR cutoff length taken as the future event horizon in [447] and Hubble horizon in [448]. While analytical solutions for the power-law and de Sitter cosmologies were obtained along with numerical solutions for bouncing cosmologies and the unification of matter-accelerated cosmological phases, no solution behaves as PlDE throughout the cosmic phase. Primarily, only the power-law cosmology leads to the desired phantom behavior at late times. (i) g0 TG k + t0 (−T)m , f(T, TG ) x̃ := r cos θ, 1−r Ωm := r sin θ. 1−r (6.62) In this way, the boundary is reached by taking r → 1− . The critical points can therefore be determined, and are summarized in table 8. Immediately, one notes that the critical points Q2 , Q3 are unlikely to be physically viable as Ωm > 1. Focusing on the remaining points, a realistic cosmology can only be obtained if a transition from P1 (akin to a matter dominated solution) to P2 /P3 is realized provided that the latter points describe an accelerating cosmology. Meanwhile, the critical points P4 and Q1 are saddle or unstable points which could represent intermediary phases. Additionally, [449] also investigate the behavior for certain parameter choices which are illustrated in √ figure 17. For α1 = − 33 and α2 = 4, a late-time de Sitter attractor (P2 ) without formation of any singularities results. On the other hand, late-time phantom (P2 ) for α1 = −α2 = 1 2 and quintessence (P3 ) for α1 = 2α2 = 3 behaviors can be recovered. f0 (−T)m Bn TG q , 96 Rep. Prog. Phys. 86 (2023) 026901 Review Table 8. The critical points for the f(T, TG ) = −T + α1 √ T 2 + α2 TG , α1 , α2 ̸= 0, model as√ obtained in [449], together with their stability √ α2 ) 9−3α2 nature and cosmological description. The variables Ωm1 := 1 − α1 (6−65(α , x± := 2 −3) P1 P2 (x̃⋆ , Ω⋆m )  p 1 − α32 , Ωm1 (x− , 0) P3 (x+ , 0) P4 (0, 0) Q1 (∞, 0) Stable/saddle Stable node, unstable node or saddle Stable node, unstable node or saddle Unstable node or saddle Saddle Q2 Q3 (∞, ∞) (0, ∞) Unstable or stable Undetermined 3± 3α1 2 (4α2 −3)+9 have been defined for simplicity. 3α1 Stability Cosmological description Decelerating scaling solution a(t) ∝ t2/3 Dark energy dominated universe, fluid can behave as phantom, cosmological constant, quintessence or non-phantom depending on parameter constraints Same behavior as the critical point P2 Dark energy dominated universe, decelerating for α2 > 0 and accelerating for α2 < 0. Fluid can be phantom, a cosmological constant or quintessence Super-accelerated phantom solution (α2 > 0) or an an asymptotically decelerating universe (α2 < 0) Future singularity for α2 < 0, past singularity for α2 > 0 Future, past or intermediate singularity √ T 2 + α2 TG model for different choices of α1 and α2 . Depending on the latter’s magnitudes, certain critical points appear. Observe that in (b) (and (c)), the Universe can evolve toward a future singularity Q2 (Q3 ) as opposed to the phantom P2 (quintessence P3 ) behavior. Reproduced from [449]. © IOP Publishing Ltd. All rights reserved. Figure 17. Pioncarè phase-space representations of the f(T, TG ) = −T + α1 On the other hand, [450] focuses attention on the stability of the Minkowski solution by examining the de Sitter space behavior (similar approaches appear in [451, 452]). Here, the phase-space variables are taken to be (H, Ḣ) which results in an autonomous system yielding critical points of the form (H⋆ , 0), namely de Sitter behavior. Thus, the matter density component approaches a constant value, Λ. In this way, a Minkowski solution can be investigated in the limit when H⋆ , Λ → 0. Additionally, however, a true Minkowski vacuum exists provided the conditions f(0, 0) = 0, fT (0, 0) < 0 arepalso satisfied. For the previous model f(T, TG ) = −T + α1 T 2 + α2 TG , a stable Minkowski solution is obtained when20 α1 < 0, 0 < α2 < i p 3 h 2 2 . (6.63) 1 + α + 3 1 + 3α 1 1 α1 2 Combined with the results in [449], the existence of a Minkowski solution together with a viable cosmological sequence can be achieved. For instance, the choice α1 = √ − 33 and α2 = 4 hosts a stable Minkowski solution and a saddle matter domination behavior which later evolves toward a stable de Sitter behavior. 6.4. Teleparallel cosmology with non-minimal couplings to matter Different theories with non-minimally couplings between the matter sector and gravity have been considered (see section 5.9). One popular one is described by the trace of the energy-momentum tensor Θ := Θµ µ . For this theory, considering a perfect fluid, and that the density Lagrangian does not depend on derivatives of the tetrads, the 20 The stability conditions are significantly different from [450]. This is due to evaluation of the derivatives FH , FD of the function F(H, D) (equation (12)) at the critical point, where the expressions equations (17) and (18) have not been recovered. 97 Rep. Prog. Phys. 86 (2023) 026901 Review (b) λ 6= 0: Here, a reconstructed solution is obtained provided the cosmological behavior is specified. The authors explore the unification of matter and DE domination eras and matter-DE transition phase which reconstructed solutions are listed in the supplementary annexes (supplementary 1). field equations (5.128) in flat FLRW cosmology are reduced to 1 −6fT H 2 − (ρ + p)fΘ − f(T, Θ) = κ2 ρ, 2 (6.64a) 1 −2HḟT − 2fT (Ḣ + 3H 2 ) − f(T, Θ) = −κ2 p, 2 (6.64b) 6.4.2. Dynamical system approach. In [328, 458], the dynamics of the general torsion-matter coupling model f(T, B, Lm ) (see section 5.9)) was investigated. Given the complexity of the field equations, particular functional ansatz were chosen for the dynamical analysis, namely: where fΘ = ∂f/∂Θ. Along with this theory, f(T, B, Lm ) gravity has been studied in cosmology. Using the field equation (5.133) for flat FLRW and a perfect fluid, we arrive at 1 3HḟB − 3H 2 (3fB + 2fT ) − 3fB Ḣ − f(T, B, Lm ) = 0, (6.65a) 2 ◦ (a) f(T, B, Lm ) = f(R, Lm ) which is the curvature-matter coupling scenario previously explored in various works [459– 462]; (b) f(T, B, Lm ) = f(T, Lm ); 1. f(T, Lm ) = f1 (T) + 2κ2 f2 (T)Lm : [458] investigates six model ansatz of which three (i) f1 = −T + α1 T 2 + 2Λ, f2 = −1 + β1 T, (ii) f1 = −T + 2Λ, f2 = 2 −1 + β2 T + β3 T 2 , (iii) f1 ∝ e α2 T , f2 ∝ e−T host a repeller or saddle power-law behavior (such as matter domination) critical points which evolves toward a DE dominated de Sitter attractor. Thus, an appropriate late-time cosmology  can be realized;  1 2. f(T, Lm ) = −Λ exp − (T + Lm ) , Λ>0 and Λ 1+ϵ f(T, Lm ) = M−ϵ (T + Lm ) , ϵ, M are constants. Here, contrary to their curvature analogues investigated in [331, 462], require no dynamical system analysis since the scale factor can be directly solved; (c) f(T, B, Lm ) = c1 T + c2 Bs + (c3 + c4 B r ) Lm with c4 6= 0, r and s are constants. Here, the critical points cannot be determined for arbitrary choices of the parameters. To this end, the following three scenarios were considered [328]: (a) r = 1, (b) c2 = 0 and (c) r = 1 − s, c2 6= 0, s = {−2, −1, 2, 3, 4, 5}. While all sub-cases yield stable critical points (for sub-case (c), it is only for s = −2) and is able to describe an accelerating Universe (with sub-case (a) describing a de Sitter cosmology), it is unclear whether they are able to host the desired sequence of cosmological epochs. 1 − (3H 2 + Ḣ)(2fT + 3fB ) − 2HḟT + f̈B − f(T, B, Lm ) 2 = −fL (ρ + p), (6.65b) where fL = ∂f/∂Lm . In the following, we will mainly concentrate on analyzing these two set of FLRW equations. 6.4.1. Reconstruction method. In [453], the authors reconstruct the f(T, Θ) Lagrangian for a diverse number of scenarios (see the supplementary annexes (supplementary 1)). Overall, a reconstructed solution is obtained with some cases containing exceptions and further extensions. In the case of HDE and PlDE [454, 455], the reconstructed f(T, Θ) solutions do not satisfy the Friedmann equations and thus are not reported. It is remarked that a ΛCDM reconstructed solution appears in [456] although the result is incorrect as the spatial components of ∇µ Θµν are identically zero as opposed to constituting a condition for f(T, Θ). In [457], the reconstruction procedure in the presence of a perfect fluid and the ansatz choice f(T, Θ) = f1 (T) + f2 (Θ) was explored via the equivalent scalar-tensor formulation by introducing an auxiliary scalar field Φ and proper functions P1,2 , Q satisfying f(T, Θ) = P1 (Φ)T + P2 (Φ)Θ + Q(Φ), f1 (T) = P1 T + Q1 (Φ), Q = Q1 + Q2 . (6.66a) f2 (Θ) = P2 Θ + Q2 (Φ), (6.66b) 6.5. Teleparallel scalar-tensor cosmology This section will be devoted to studying cosmology for different scalar-tensor teleparallel theories (see section 5.8 for more details). One particular class of theory that has been studied in cosmology is the one described by the action (5.98) which is constructed from the torsion scalar and the boundary term. For this theory, the field equation (5.99) in flat FLRW becomes The crucial relation for this approach is 8HṖ1 + 4ḢP1 = −ρ(1 + w)(1 + 2P2 ). (6.67) For simplicity, [457] apply the method of separation of variables by setting the above equal to some constant λ whose magnitude alters the form of the solutions: 1 1 − f(T, B) + 3HḟB − 3H 2 (3fB + 2fT ) − 3fB Ḣ − fX ϕ̇ 2 = κ 2 ρ, 2 2 (6.68a) 3 wΘ (a) λ = 0: This leads to f(T, Θ) = 2ek(−T) 4 − 1− 3w for some integration constant ek, a solution which contrasts with that reported in [457]. By virtue of the Friedmann equations, 1 the power-law cosmology a(t) ∼ t 2−2w results. 1 −(3H 2 + Ḣ)(2fT + 3fB ) − 2HḟT + f̈B − f(T, B) = −κ 2 p, 2 (6.68b) 98 Rep. Prog. Phys. 86 (2023) 026901 Review whereas the modified Klein–Gordon equation (5.101) yields ḟX ϕ̇ + 3fX Hϕ̇ + fX ϕ̈ + fϕ = 0, χ(σ̈ + 3Hσ̇) + (6.69) where fX = ∂f/∂X and fϕ = ∂f/∂ϕ. A particular case which has been extensively used in the literature for cosmology is the one described by (5.102) where the kinetic term appears linearly and the scalar field is non-minimally coupled to both T and B. For this theory, the flat FLRW are 3H 2 ′ 3 f2 (ϕ) + 2 g2′ (ϕ)(3H 2 + Ḣ) + Vσ = 0, 2 κ κ (6.74b) where Vϕ = ∂V/∂ϕ and Vσ = ∂V/∂σ. In the following, we will mainly concentrate on these kind of theories. In [463], the reconstruction procedure was applied for the minimal coupling scenario F1 (ϕ) = F2 (ϕ) = 0 where the matter component is solely described by the scalar field with kinetic coupling ω(ϕ) = 1 n+1 and potential V(ϕ) = ϕn+1 , where n 6= −3, ±1, 0 is a constant. In particular, the authors explore the case when the Klein– Gordon equation, generally expressed in Euler’s form [464] 6.5.1. Reconstruction method. 1 − 6fT H 2 − f(T) − 3H 2 F1 (ϕ) + 3HḞ2 (ϕ) 2   1 = κ2 ρ + ω(ϕ)ϕ̇ 2 + V(ϕ) , (6.70a) 2 1 − 2HḟT − 2fT (Ḣ + 3H 2 ) − f(T) − F1 (ϕ)(3H 2 + 2Ḣ) 2   1 2 − 2HḞ1 (ϕ) + F̈2 (ϕ) = −κ p + ω(ϕ)ϕ̇ 2 − V(ϕ) , 2 (6.70b) ϕ̈ + f1 (t)ϕ̇ + f2 (t)ϕ + f3 (t)ϕn = 0, (6.75) for time-dependent functions f1,2,3 (t) satisfy the relation [465] whereas the modified Klein–Gordon equation becomes −  3 H 2 (F1′ (ϕ) + 3F2′ (ϕ)) + ḢF2′ (ϕ) κ2 1 + 3Hω(ϕ)ϕ̇ + ω(ϕ)ϕ̈ + ϕ̇ 2 ω ′ (ϕ) + V ′ (ϕ) = 0, (6.71) 2 1 f̈3 n+4 − n + 3 f3 (n + 3) 2 + where primes denote differentiation wrt the scalar field. On the other hand, a further extension of the f(T, B, X, ϕ) action is the quintom model investigated in [302] where two canonical scalar fields ϕ and σ are introduced, both being non-minimally coupled with the torsion T and boundary B scalars. The gravitational Lagrangian is thus generalized to !2 + n − 1 f1 ḟ3 (n + 3) 2 f3 2ḟ1 2(n + 1)f1 2 + = f2 . n+3 (n + 3) 2 (6.76) This condition guarantees an analytical solution for ϕ(t). As f1 = 3H, f2 = 0 and f3 = 1 for this particular case, the above 2(1+3n) n +3 relation leads to a(t) ∝ t 3(n+1) leading to ϕ(t) ∝ t 1−n 2 . Thus, the reconstructed form of f (T) is obtained via the Friedmann equation (6.70a) to be 1 [T + ( f1 (ϕ) + f2 (σ)) T + (g1 (ϕ) + g2 (σ)) B] 2κ2 ξ χ − ∂µ ϕ∂ µ ϕ − ∂µ σ∂ µ σ + V(ϕ, σ), (6.72) 2 2 L=− 2(1+3n) f(T) ∼ (−T)n −1 + (−T) n−1 . 2 (6.77) Other scalar potentials were also investigated, namely power law (mixed and Higgs), exponential and logarithmic. Except for the exponential case, which leads to a rescaled TEGR solution, no solutions were obtained for the remaining cases. In [254], the conformal f (T) theory was studied having non2 2 minimal couplings F1 (ϕ) = Cϕ2 , F2 (ϕ) = Dϕ2 , kinetic coup- where ξ and χ are constants and V(ϕ, σ) represents the combined potential of both scalar fields. For this model, the flat FLRW equations are  χ ξ 3H 2 (1 + f1 (ϕ) + f2 (σ)) = κ2 ρ + V(ϕ, σ) + ϕ̇ 2 + σ̇ 2 2 2  +3H(ġ1 (ϕ) + ġ2 (σ)) , (6.73a) m+1 2 0ϕ ling ω(ϕ) = 2κ2 and potential V(ϕ) = 2κ mV+ , where C, D, 1 V 0 and m are constants. To ensure conformal invariance, C = 1 1 6 , D = 3 and m = 3 which is assumed in the reported results. For simplicity, only simple cosmological scenarios were considered for reconstruction. Starting with a power-law cosmology a(t) ∝ tα for some constant α, in the absence of the potential V(ϕ), f (T) takes the form  1−5α y−2α c1 4 + 2(α − 1) 2 c2 2 c1 3 c2 y 2 f(T) = − − 4(4α − 1) 2 − 5α (3H 2 + 2Ḣ)(1 + f1 (ϕ) + f2 (σ))  χ ξ = −κ2 p − V(ϕ, σ) + ϕ̇ 2 + σ̇ 2 + 2H(ḟ1 (ϕ) + ḟ2 (σ)) 2 2  −g̈1 (ϕ) − g̈2 (σ) , (6.73b) whereas the modified Klein–Gordon equations for both scalar fields are ξ(ϕ̈ + 3Hϕ̇) + ḟ3 f3 3−7α 3H 2 ′ 3 f1 (ϕ) + 2 g1′ (ϕ)(3H 2 + Ḣ) + Vϕ = 0, 2 κ κ (6.74a) c1 2 c2 2 y1−3α c1 c2 3 y 2 + 2α − 1 7α − 4 − + 2κ2 ρ0 y− 2 α(1+w) , 3α(1 + w) − 1 6α 2 , T 3 99 c2 4 y2−4α 4(5 − 8α) + y≡− (6.78) Rep. Prog. Phys. 86 (2023) 026901 Review 4 f(T) = c1 (−T)3/4 + c3 , 3 where c1,2 are integration constants originating from the modified Klein–Gordon equation. It is remarked that the reconstructed solution is invalid for certain choices of α and w, such as radiation domination (α = 21 ). For a de Sitter cosmology, f − 2TfT becomes constant in time (see section 6.1.1) making the scalar field act as the source for the accelerated behavior. The latter’s evolution, obtained from the Klein–Gordon equation (6.71), is dependent on the choice of potential. For the qpower-law potential, ϕ approaches a constant value21 ϕ = ± (a) Quintessence—The scalar field grows with time while its EoS becomes close to the phantom-divide line. The number of e-foldings increases exponentially with time; (b) Phantom—The scalar field decays in an oscillatory fashion with its EoS rapidly decreasing into the phantom regime. The number of e-foldings becomes asymptotically flat. −2H 2 V0 whereas in its absence (V0 = −Ht Finally, the minimal coupling case consisting of TEGR, a scalar field with ω(ϕ) = 1 coupled with an electromagnetic tensor scalar field coupling was explored in [474]. Namely, this introduces the contribution 41 f(ϕ) 2 Fµν Fµν to the gravitational action, i.e. L=− 1 T 1 − ϕ;µ ϕ;µ + V(ϕ) + f(ϕ) 2 Fµν Fµν , 2 2κ 2 4 (6.81) where f(ϕ) represents the coupling strength, to the gravitational action which role is to explain inflationary phases [475] as well as to generate late-time acceleration [476]. Since TEGR is considered, the field equations are identical to its curvature analogue meaning the Noether symmetries match with those obtained, for instance, in [476]. However, there appear to be differences in the Noether equations. For instance, neglecting time symmetries for simplicity, the condition 2αa = 0 appearing in [474] should be α + 2aαa = 0 as correctly reported in [476]22 . In the following, the Noether symmetry approach has also been explored within teleparallel scalar-tensor theories. Generally, the symmetries also constrain the form of the scalar-tensor coupling as well as the scalar potentials. For simplicity, the minimally and nonminimally coupled cases shall be discussed separately. 6.5.2. Noether’s symmetry approach. In [466], the authors investigate the teleparallel analogue of Saez–Ballester theory [467], which has been extensively explored in curvature-based models [468–472]. Following the action given in equation (5.102), the theory is recovered when F1 (ϕ) = F2 (ϕ) = 0. For this case, the theory exhibits one Noether symmetry together with the functional constraints 6.5.2.1. Minimally coupled scalar fields. ω(ϕ) ∝ ϕ−4 , (6.80) where c1 , . . . , c5 are integration constants. For c1 = 34 , c2 = c4 = 0 and c5 = 1, the following distinct evolutionary behaviors were obtained: 0), the scalar field decays with time as ϕ(t) ∝ e . Finally, for a ΛCDM cosmology in the absence of the scalar potential, ´ f (T) reduces to the standard ΛCDM Lagrangian with ϕ = e− H dt making the evolution independent of the scalar field’s behavior. In view of the above observations, the scalar field can be made to act only as a source for the late-time acceleration by constructing a potential which is predominantly absent during earlier phases (such as matter domination) but becomes dominant as the Universe transitions to accelerated phases (such as de Sitter). f(T) = t0 T 2 + C, 2 V(ϕ) = c4 + c5 (c1 ϕ + c2 ) , Non-minimal scalar field couplings with the torsion and boundary term scalars have been explored in [416, 477, 478] with f(T) = 0, ω(ϕ) = 123 . Starting with [477, 478], in order to simplify the Noether symmetry analysis, the following models were considered: 6.5.2.2. Non-minimally coupled scalar fields. V(ϕ) = V0 ϕ−4 + C, (6.79) 6 Model I [477] F2 (ϕ) = 0, V(ϕ) = λϕ 2n+3 where n 6= −3/2, where C, t0 and V 0 are integration constants. Aslam et al [466] then fixes the latter parameters to investigate the resulting cosmology. In particular, a ΛCDM-like behavior is observed with a scalar field which is phantom in nature and approaches DE at late-times. Despite this, it has been noted that the C term disappears from the Lagrangian and hence has no physical consequences. However, its contribution still appeared in the cosmological equations which may lead to a different cosmological behavior. A further extension is explored in [473] which includes the contribution of dust matter for the particular choice ω(ϕ) = ϵ, where ϵ = +1, −1 correspond to a quintessence or phantom field respectively. In this case, the Noether condition imposes the following functional constraints Model II [478] F2 (ϕ) 6= 0, V(ϕ) = λϕ 2 , with λ ⩾ 0. The first model describes the absence of the boundary term coupling while a general power-law potential is considered whereas the second model maintains both couplings while the potential is taken to be quadratic. Through Noether symmetry, both F1 (ϕ) and F2 (ϕ) coupling functions turn out to be constrained, as summarized in table 9. Evidently, both coupling terms generally take on a power-law form and each consist of a quadratic coupling which is reminiscent of teleparallel DE [288]. 22 The typo in [474] did not effect the main results which are in a different convention to other works. 23 The Lagrangians between [477, 478] differ by a factor of 2. The necessary factors have been included for sake of consistency. 21 An extra factor of 4 appears in [254]. 100 Rep. Prog. Phys. 86 (2023) 026901 Review 6 (2n + 3) 2 2 ϕ , V(ϕ) ∝ ϕ 2n+3 , 24 (II) F1 (ϕ) ∝ ϕ 2 , V(ϕ) ∝ ϕ 2 , Table 9. The resulting functional constraints imposed by Noether (I) F1 (ϕ) = − symmetry for the two models considered in [477, 478]. The integration constants c1,2 and k̃ arise from the Noether constraint. Observe that in Model II, when k̃ = 0 and c1 = − 21 , F2 (ϕ) = 0, which matches the Model I result when n = 0 as expected. (III) F1 (ϕ) = constant, V(ϕ) ∝ e αϕ , (IV) F1 (ϕ) ∝ c1 + c2 ϕ + c3 ϕ 2 , Model F1 (ϕ) and F2 (ϕ) I II +3) F1 (ϕ) = − (2n24 ϕ 2 , F2 (ϕ) = 0 (a) k̃ ̸= −3 F1 (ϕ) = 38 ϕ 2 + 3c21 ϕ V(ϕ) ∝ (c1 + c2 ϕ + c3 ϕ 2 ) 2 , 2 2(k̃+3) 3 2 (b) k̃ = −3 1 ϕ F2 (ϕ) = − ϕ4 − 2(3c k̃+3) for constants n and α. Observe that the first two models are precisely the cases described previously in [477, 478] whereas the third and fourth model are new symmetries which could not be derived in the previous analysis. While these scenarios were not explored in further detail, the quantum cosmological description using the Wheeler–de Witt equation and Hartle’s criterion was applied for the second model. In this case, the Universe’s wave function satisfies Hartle’s criterion and thus the classical trajectories reduce to the cosmological behaviors as obtained in [477, 478]. In the case of generalized couplings, the results obtained in [479, 480] are reviewed. Tajahmad [479] considers the f(T, ϕ, X) sub-class , 2(k̃+3) 3 2 + 3c21 , F1 (ϕ) = 38 ϕ 2 − 3c ϕ2 2 F2 (ϕ) = − ϕ4 − c1 ln ϕ Prior to examining the resulting cosmological behaviors, the coordinates were first transformed to a new set of coordinates (u, z) with z being a cyclic coordinate24 . Thus, this yields a constant of motion I 0 which influences the nature of the cosmology as shown in table 10. Starting with I0 = 0, the potential becomes zero in both models. Depending on the choice of integration constants, both models can contain static (z1 6= 0, z2 = 0) and powerlaw (z1 = 0, z2 , n 6= 0) behaviors while Model I can also host (anti)-de Sitter (z2 6= 0, n = 0) behavior. Transition periods may be realized for more general parameter considerations. Moving on to the I0 6= 0 case, each model exhibits distinct behaviors. In Model I, n 6= 0, −3, −6 leads to a power-law behavior which can result into an ever-accelerating phantom or quintessence cosmology. For n = 0, the scale factor yields an evolution which can be associated with a multi-fluid scenario. Furthermore, a late-time de Sitter phase is achieved since the scalar field component approaches DE behavior. Lastly, for n = −3, the scalar field component can behave as a nonphantom fluid which approaches a DE state at late-times. Similar to the n = 0 case, this Universe becomes accelerating and approaches a de Sitter-like state. For Model II, for k̃ 6= −3, −6, various cosmological behaviors can be obtained. For instance, this Universe may experience accelerating and decelerating phases whereas the scalar field fluid can behave as either a phantom or a non-phantom fluid. Therefore, this solution can be used to explain transitioning periods. In particular, a late-time de Sitter-like behavior can be achieved for small k̃ values. Finally, k̃ = −3 can host a de Sitter solution which is expanding provided c1Iα0 1 < 0 and leads to a growing potential which serves as the main driving force for the expansion. In the absence of the boundary term coupling (F2 (ϕ) = 0), four Noether symmetries are obtained which constrain F1 (ϕ) and V(ϕ) to be [416]: L = f(ϕ)T − U(ϕ, X)g(T) − 2k̃+3 (6.82) where U(ϕ, X) is a coupling function chosen be U = h(ϕ)ϕ̇. Applying the Noether condition leads to the following functional constraints √ f(ϕ) ∝ ϕ 2 , g(T) ∝ −T, h(ϕ) ∝ ϕ, V(ϕ) = V0 ϕ 2 , ω(ϕ)is constant, (6.83) where V 0 is an integration constant. The corresponding cosmological behavior was investigated for the particular choices 3 2 f(ϕ) = 32 ϕ and ω(ϕ) = 1, leading to the scale factor  a(t) = (c1 t + c2 ) 2/3    8 16 V0 c2 2 exp V 0 t + c3 − t + c4 , 3 3 c1 (6.84) with c1 , . . . , c4 representing integration constants. However, as both f(ϕ), ω(ϕ) > 0, this may lead to a non-canonical scalar field, depending on the form of the second term in the Lagrangian (6.82). On the other hand, [480–482] consider a fermionic coupling in the form L = h(u)f(T) −   i µ ψ̄Γ Dµ ψ − D̄µ ψ̄ Γµ ψ + V(u), (6.85) 2 where ψ is the fermion function, Γµ = EAµ γ A with γ a representing the Dirac matrices, and u = ψ ψ̄. Overbars represent the respective adjoint operators. The TEGR coupling case f(T) = T has been explored in 3 + 1 [481] and 2 + 1 [482] spacetimes. Starting with the former, the Noether condition imposes the 2n+1 constraint on the coupling functions h(u) ∝ u 3 and V(u) ∝ u, where n is some constant which greatly influences the nature of the cosmology. This can be summarized as follows: 24 The Model II point-like Lagrangian in terms of these new coordinates is 2 ω(ϕ) ϕ,µ ϕ,µ + V(ϕ), 2 2 2 L = − u̇2 + 2c1 α1 u 3 u̇ż + λu 2 + 4c2 αu 21 ż with the c2 contribution being absent for k̃ ̸= −3, and α1 is an integration constant. This is different from that reported in [478], hereby affecting the nature of the cosmological solutions. As c2 only appears for the case when k̃ = −3, the presented solutions shall hold true only for c2 = 0. 101 Rep. Prog. Phys. 86 (2023) 026901 Review Table 10. The corresponding cosmological behaviors for each Noether symmetry depending on the constant of motion I 0 as obtained in n+3) 3) [477, 478]. Here, b := αI00 ((2n , ℓ := − I03(αn+ and u1 , . . . , u4 , z1 , . . . , z6 are integration constants. It is remarked that the cases n = −6 and +3) 0 c1 k̃ = −6 were not studied. Model Scale Factor Case I: I0 = 0 Model I Model II ( a∝ (z1 + z2 t)− 2n , for n ̸= 0 , for n = 0 1 2α − 3 0 (z1 +z2 t) e 1 a ∝ (z1 + z2 t)− k̃ Case II: I0 ̸= 0 Model I 3 − n(n+ 3) (a) n ̸= 0, −3, −6 a ∝ (bt + c1 ) (b) n = 0 0 3 I0 a = (α0 u1 + I0 t) 6 e 3 1   I t 6 3 − I0 t − 0 a ∝ e 6α0 c2 + 2αI00λ2u2 e α0 (c) n = −3 1 Model II  2  −α z + 2α0 u1 λt 3λα1 c1 (u3 +ℓt) 2 − 8α3c k̃ I0 2 (k̃+6) 1 1 (a) k̃ ̸= −3, −6 a∝ (b) k̃ = −3 h i 13 I t − 0 a ∝ z5 u4 2 e c 1 α 1 + z6 2 (a) n = − 12 yields a(t) ∼ t 3 with the fermionic field behaving as dust; (b) n = 1 gives de Sitter and hence can describe inflationary epochs. Here, the fermionic field behaves as a cosmological constant; 1 (c) n 6= 1, − 12 leads to a(t) ∼ t 1−n . For 0 < n < 1, the fermionic field has phantom behavior while n > 1 leads to quintessence behavior.  +λt 2  − (u3 + ℓt) k̃ k̃+3 − 1 + z4 k̃ Similar to the Noether symmetry approach, the dynamical system approach is explored for a number of teleparallel scalar-tensor theories, categorised under minimally and non-minimal couplings. These shall be discussed separately with a comparison made in cases where the non-minimal dynamics reduce to the minimal ones. 6.5.3. Dynamical system approach. In [423, 483], the f (T) gravity minimal coupling model having F1 (ϕ) = F2 (ϕ) = 0 and ω(ϕ) = 1 was studied. In [483], an interaction between a dust fluid and the DE √ scalar fluid was considered for the model f(T) = −T + α −T. As this reduces to standard minimally coupled TEGR with a scalar field, the results are not explored further. Instead, the deviation from TEGR is explored in [423] in the absence of matter fluids with the scalar (inflaton) field obeying the constant-roll approximation ϕ̈ = βHϕ̇, where β represents the constant-roll parameter. This, in turn, leads to the assumption that the Hubble parameter takes on the form r ! β H(ϕ) = M cos κ ϕ , (6.86) 2 6.5.3.1. Minimally coupled scalar fields. In 2 + 1 spacetime, the functional constraints take on various forms depending on whether the gauge term g(t, qi ) in the Noether symmetry condition is maintained (see the supplementary annexes (supplementary 1)). When neglected (g = 0), the resulting cosmology becomes identical to the 3+1 scen2 2k ario, else we find h(u) ∝ u k+1 and V(u) ∝ u k+1 for some constant k, which magnitude alters the cosmological behavior, namely: k +1 (a) k 6= 1: a(t) ∼ t 2(1−k) which can be decelerating (−1 < k < 1 1 3 ), coasting (k = 3 ) or otherwise accelerating. Meanwhile, the fermionic field can behave as a quintessence field ( 12 < k < 1) or a phantom field (k < −1or k > 1); (b) k = 1: de Sitter with fermionic field behaving as a cosmological constant. with M ⩽ κ−1 . For this model, the f (T) dynamical system approach discussed in section 6.1.3 can be directly applied by using equation (6.21) with the inflaton’s EoS Finally, for the general case [480], the Noether condition m−1 imposes the constraints f(T) ∝ T n , V(u) ∝ u and h(u) ∝ um , 3n leading to a power-law cosmology a ∼ t 2 . Interestingly, the parameter m does not affect the evolution. As only a powerlaw behavior is obtained, the model appears to be insufficient to describe the whole Universe’s history but it may be viable for specific epochs. 1 wϕ = 21 ϕ̇ 2 − V(ϕ) 2 ϕ̇ 2 + V(ϕ) , (6.87) in conjunction with the FLRW equations (6.70a) and (6.70b). This leads to the one-dimensional autonomous relation 102 Rep. Prog. Phys. 86 (2023) 026901 Ḣ = Review  6  f − HfH + 2κ2 V(ϕ) . fHH Furthermore, the critical points (B–E) only appear if the scalar field is minimally coupled (ξ = 0). Meanwhile, the boundary critical points Q1,2 are unstable and leave an arbitrary DE fluid behavior meaning that these points could describe transition, inflationary or de Sitter phases. For minimally coupled scenarios, a viable cosmology can be realized following the trajectory A → D while for non-minimally coupled cases A → F/G/J provided that the relevant existence and stability criteria are satisfied. A finite phase-space representation of the latter for both minimally and non-minimally coupled cases is illustrated in figure 18. An extension of the previous model, explored in [485], takes into account the interaction between the scalar field and the matter content via an interaction term Q for a number of interactions and potentials, namely (i) Q = 0, (ii) Q ∝ ρm ϕ̇, (iii) Q ∝ Hρ, (iv) Q ∝ Hρm , and (A) V(ϕ) ∝ e−λϕ , (B) V(ϕ) ∝ ϕ n . Here, λ and n are constants. Interestingly, the interaction models (iii) and (iv) yield no critical points. Meanwhile, only a few selected models can realize a viable cosmology. A(i) can realize a matter → de Sitter sequence whereas only the latetime de Sitter phase is apparent in B(i). On the other hand, A(ii) and B(ii) yield a scaling solution to de Sitter. Meanwhile, the special non-interacting case Q = 0 in the presence of a dust fluid was investigated in [489] but for a linear V(ϕ) ∝ ϕ and exponential V(ϕ) ∝ e−ϕ potentials. The critical points therefore reduce to the ones given in [485] by setting n = λ = 1. In this case, the exponential model is still able to generate a viable cosmology from matter domination → de Sitter. In [486], the T-coupling is investigated for the generalized power-law form F1 (ϕ) = ξϕÑ for some Ñ > 0 and power-law potential V(ϕ) = V0 ϕn with V0 > 0 and n is some constant. Here, the dynamics have been explored in vacuum using a new set of dimensionless phase-space variables (6.88) The f (T) function can be reconstructed by combining the FLRW equations (6.70a) and (6.70b) whereas the potential is obtained from the modified Klein–Gordon equation (6.71). The resulting functions take on the forms f(Φ) = 3M 2 cos(2Φ) + c1 [Φ cos(Φ) − sin(Φ)] + c2 , (6.89a) M 2 [(β + 3) cos(2Φ) − β + 6] (β + 3)c1 sin(Φ) + 2κ 2 6κ 2 c2 βc1 2 − , (6.89b) − 144κ 2 M 2 2κ 2 q where Φ := κ β2 ϕ, c1 and c2 being integration constants. It is remarked that the f (T) solution can be expressed in terms of the torsion scalar by inverting the relation T = −6M 2 cos 2 (Φ). Substituting back into the autonomous relation equation (6.88) leads to V(Φ) =  βc1 p Ḣ = β H 2 − M 2 + M 2 − H 2, 12M (6.90) which is different than the one given in [423]. This leads to a different set of critical points and hence, a possible different dynamical behavior. Nonetheless, we highlight the reported results. For c1 < 0, the Universe interpolates between a Type IV initial singularity (which replaces the initial Big Bang singularity) and a late-time de Sitter cosmology. Throughout the cosmic evolution, the Universe experiences a decelerated expansion and a late-time behavior compatible with ΛCDM especially for small c1 values. Meanwhile, for c1 > 0, a cyclic Universe is obtained, one which interpolates between two Type II singularities. m := The dynamical system analysis for the non-minimally coupled scalar field case has been extensively studied in numerous works. Here, we review [295, 302, 484–489]. Starting with the T-coupling cases (F2 (ϕ) = 0) with f(T) = −T, in [487], the coupling F1 (ϕ) = ξϕ 2 with kinetic coupling ω(ϕ) = 1 and potential V(ϕ) ∝ e−λϕ was explored using the dimensionless phase-space variables 6.5.3.2. Non-minimally coupled scalar fields. x̃ 2 := κ2 ϕ̇ 2 , 6H 2 ỹ 2 := κ2 V(ϕ) , 3H 2 z̃ := p |ξ|ϕ, Ωm := ϕ̇ , Hϕ A := ÑϕÑ . 1 − ξϕÑ (6.92) Depending on whether Ñ = 2 or Ñ 6= 2, different critical points are obtained. In the latter, only a de Sitter behavior appears. On the other hand, for Ñ = 2, the choice of ξ determines the cosmological behavior: (a) ξ > 0: no stable points and therefore unable to describe any late-time behavior; (b) ξ < 0: this Universe can evolve from de Sitter → powerlaw provided n 6= 2 and n is odd. To maintain a late-time accelerating behavior, (n 2 − 4)ξ < 2. κ2 ρm 3H 2 (6.91) √ Finally, [488] considers f(T) = αT + c1 −T + c2 , F1 (ϕ) ∝ ϕ 2 , ω(ϕ) = ±1 and potential V(ϕ) ∝ eβϕ where c1,2 , α and β > 0 are constants. For the phase-space variables equation (6.91) leads to two unstable critical points which their physical description was not explored. Nonetheless, through numerical analysis, a late-time phantom behavior was found. Moving on to the case which explores the B-coupling, [295] is reviewed which investigates the model F1 (ϕ) = 0, F2 (ϕ) = in the presence of dust. Solving the autonomous system and use of Poincaré’s projection method yields a number of critical points as summarized in table 11. The critical points D, E, F, J can be dynamically stable, corresponding to a quintessence Universe (D), decelerating matter dominated Universe with DE behaving as matter (E), and DE dominated de Sitter accelerating universes (F, J). 103 Rep. Prog. Phys. 86 (2023) 026901 Review Table 11. Summary of the critical points investigated in [487] for the non-minimal torsion scalar coupling model. A–J represent the finite boundary critical points whereas Qi are the critical points found at the boundary derived using Poincaré’s projection method. Here, the p Poincarè variables are x̃r = ρ cos θ sin ψ, ỹr = ρ cos ψ, z̃r = ρ sin θ sin ψ where ρ = √ r 2 with r = x̃ 2 + ỹ 2 + z̃ 2 , θ ∈ [0, 2π] and 1−r ψ ∈ [− π2 , π2 ]. (x̃⋆ ,ỹ⋆ ,z̃⋆ ) Existence Stability Properties A (0, 0, 0) Always Saddle B(+) /C(−) (±1, 0, 0) ξ=0 √ Unstable (λ(+) < 6, (−) λ > 6), saddle otherwise Matter dominated decelerating universe DE dominated dec. universe behaving as a stiff fluid ξ = 0, λ 2 ⩽ 6 Stable node (λ 2 < 3), saddle otherwise DE (quintessence) dominated universe 2 with a(t) ∝ t λ 2 ξ = 0, λ 2 ⩾ 3 √  q 2(ξ± ξ(ξ−λ 2 )) 0, , λ2 √ √  2 Stable node (3 < λ 2 < 24 7 ), stable spiral (λ 2 > 24 ) 7 Decelerating matter dominated phase with DE behaving as dust 0 < λ 2 ⩽ ξ, or ξ < 0 (G) F: Stable node (λ 2 < ξ); G: Saddle Accelerating de Sitter DE domination J (0, 1, 0) ξ ̸= 0, λ = 0 Stable spiral ( 38 < ξ), stable node (0 < ξ < 38 ), saddle otherwise Accelerating de Sitter DE domination Existence Stability Properties Q1,± (x̃⋆r ,ỹ⋆r ,z̃⋆r )   ∓ √12 , 0, ± √12 ξ>0 Unstable (0 < ξ < 38 ), saddle otherwise Accelerating (ξ > 16 ) with phantom behavior (ξ > 38 ). Arbitrary DE EoS ξ>0 Saddle Non-accelerating universe. Arbitrary DE EoS  D q  2 1 − λ6 , 0 6 λ √ , q E F (−) /G (+) (ξ−  Q2,± 3 1 2 λ, q  3 1 2 λ,0 ξ(ξ−λ )) λξ ± √12 , 0, ± √12 |ξ|  Figure 18. Projection of the finite phase-space on the x-y plane for the T-coupling (a) minimal and (b) non-minimal scalar field cases. In the former, the evolution approaches the late-time quintessence dominated behavior (D). Once the non-minimal coupling is introduced, the behavior changes considerably. While the critical points B–E no longer exist, the cosmology can remain for a long finite time near E. Then, the system evolves to the matter dominated state A followed by the de Sitter critical point G. However, as this point is saddle, it does not represent the final evolutionary state. Here, x → x̃ and y → ỹ to match our conventions. Reproduced from [487]. © IOP Publishing Ltd. All rights reserved. 104 Rep. Prog. Phys. 86 (2023) 026901 Review Table 12. Summary of the critical points for the boundary-scalar coupling model considered in [295]. The stability and cosmological implications only reflect the dust matter case. (x̃⋆ ,ỹ⋆ ,z̃⋆ ) Existence Stability Properties O (0, 0, 0) Always Saddle A± (±1, 0, 0) χ=0 Matter dominated decelerating universe DE dominated dec. universe behaving as a stiff fluid q B  3 1+w 2 λ , q √ 3 2 λ  q λ2 1 − , 0 6 6 λ √ , C 1−w 2  ,0 χ = 0, λ 2 ⩾ 3(1 + w) Stable node (3 < λ 2 < 24 7 ), stable spiral (λ 2 > 24 ) 7 Scaling solution χ = 0, λ 2 < 6 Stable node (λ 2 < 3), saddle otherwisea Quintessence DE domination with 2 a(t) ∝ t λ 2 Stable spiral (48χ > λ 2 + 6), stable node (0 < 48χ < λ 2 + 6), saddle otherwise Stable spiral ( 18 < χ), stable node (0 < χ < 18 ), saddle otherwise Accelerating de Sitter DE domination D  q  0, 1, 23 λ Always E (0, 1, 0) χ ̸= 0, λ = 0 a √ P2,+ : Unstable (λ < 6), saddle otherwise; √ P2,− : Unstable (λ > 6), saddle otherwise Accelerating de Sitter DE domination A parameter β appears in [295], however no source for its significance was found. Nonetheless, the point corresponds to the same behavior found in [487]. χϕ 2 , ω(ϕ) = 1 and V(ϕ) ∝ e−λϕ . In this case, the phase-space variables are x̃ 2 := √ κ2 ϕ̇ 2 κ2 V(ϕ) κ2 ρm 2 , ỹ := , z̃ := 2 6χϕ, Ω := . m 6H 2 3H 2 3H 2 (6.93) To maintain physical solutions, the following additional conditions were imposed: x̃ 2 + ỹ 2 + x̃z̃ ⩽ 1 (ensuring that the matter density is positive) and y > 0 (the potential is assumed to be positive). Similar to [487], Poincaré’s projection method is necessary to complete the analysis. However, as the boundary critical points turn out to be non-physical, only the finite critical points shall be discussed which are summarized in table 12. It is remarked that the stability and physical behavior has been carried out only for the dust matter case (w = 0) but can be easily extended for more general fluids. In addition, the special case χ = 0 leads to the identical critical points obtained in [487] for the ξ = 0 model as they lead to to the same gravitational Lagrangian, the only difference being the generalization to an arbitrary fluid constant EoS. When χ 6= 0, only the critical points O (decelerating matter domination) and D, E (accelerating de Sitter) remain which can constitute a viable cosmological evolution (see figure 19). Lastly, the generalized quintom case model which considers both T and B-couplings is discussed in [302]. In particular, the following non-minimal couplings and potentials have been considered Figure 19. Dynamical behavior of the B-non-minimal coupling scenario described via the projection onto the x-y phase-space plane. Clearly, the evolution evolves toward a late-time de Sitter attractor (D). Note the identification x → x̃ and y → ỹ to match the review convention. Reprinted figure with permission from [295], Copyright (2015) by the American Physical Society. where c1 , . . . , c4 , V1 , V2 and λ1 , λ2 are constants. To build the dynamical autonomous system, the following dimensionless variables were defined Ωm = ũ 2 = f1 (ϕ) = c1 ϕ 2 , f2 (σ) = c2 σ 2 , g1 (ϕ) = c3 ϕ 2 , g2 (σ) = c4 σ 2 , (6.94a) V(ϕ, σ) = Ṽ(ϕ) + V̄(σ) = V1 e−λ1 ϕ + V2 e−λ2 σ , κ2 ρm , 3H 2 κ2 σ̇ 2 , 6H 2 x̃ 2 = κ2 ϕ̇ 2 , 6H 2 ṽ 2 = ỹ 2 = κ2 V2 (σ) , 3H 2 κ2 V1 (ϕ) , 3H 2 √ z̃ = 2 6ξϕ, √ w̃ = 2 6χσ, (6.95a) (6.95b) where the dynamical variable s can be determined via the Friedmann equation (6.73a) hereby reducing the dimensionality of the autonomous system. Here, the matter content is (6.94b) 105 Rep. Prog. Phys. 86 (2023) 026901 Review assumed to behave as a perfect fluid with EoS w ∈ [−1, 1] whereas the constraints ỹ,ṽ ⩾ 0 (to ensure non-negative potentials as in standard quintessence theories) and Ωm > 0 (to ensure a positive matter energy density) were considered. Solving the autonomous systems yields 13 critical points characterizing a matter dominated phase and DE dominated phases, the latter of which consists of: Table 13. A summary of the distinct Noether symmetries Si together with the associated constrained f(ϕ, ψ) function derived in [335, 490]. Here, c1 , . . . , c11 are integration constants, c12 ≡ c5 c6 − c4 c7 + c5 c9 , while F(x), G(x), H(x) and I(x) are arbitrary functions. S1 (a) one of the scalar fields being absent with the other frozen in time with either zero or non-zero potential energy; (b) both scalar fields frozen in time while the Universe accelerates in a de Sitter phase. S2 S3 S4 Based on their dynamical stability, a matter → DE evolutionary sequence can be achieved under suitable parameter choices. 6.6. Other theories In the following, we explore other theories investigated in the context of cosmology. These approaches have not yet been extensively explored as the previous TG models especially under the approaches considered in this section. For this reason, a brief summary of the results is provided together with associated FLRW equations where applicable. + c11cc3 12 exp h i c3 c12 (c5 φ − c7 ϕ) c11 + F(−c7 ϕ + c5 φ) c3 −c10 + c11 exp ξc3 c7,10 = 0, c3,5 = ̸ 0, c5 ̸= −c6 c3,7,10 = 0, c5 = ̸ 0, c5 = −c6 = c9  c3 c6 c9 φ  c11 + G(φ)   c9 c3 ϕ/c4 − c6 + c4 ϕ + φ e S5 c5,7,10 = 0, c3,4 ̸= 0 c3 −c10 +H ξc3 S6 c3,4,5,7 = 0, c6 ̸= −c7 c10 ψ + I(ϕ) (c6 +c9 )ξ S7 c3,4,5,7,10 = 0, c6 = −c7 = −c9 Arbitrary  (6.98) with c1 , . . . , c10 being integration constants and ψ ≡ φ − ϕ.25 Each symmetry constraints the form of f(ϕ, ψ) (and ultimately the non-local theory) as summarized in table 13 and, correspondingly, the cosmological behavior. The obtained res- 1 1 κ2 ρ = 3H 2 (θ + 1 − ξf(ϕ, φ)) + 3χHḟ − ζ̇(3H + φ̇) − θ̇ϕ̇, 2 2 (6.96a) ◦ ults match with the non-local f (T) (χ = φ = ζ = 0) and f(R) (ξ = −χ = −1, θ = −ζ and f(ϕ, φ) = f(ψ)) gravity limiting cases [335, 490]. Interestingly, some forms of f(ϕ, ψ) host exponential contributions which are known to make the theory renormalizable 1 −κ2 p = (2Ḣ + 3H 2 )(1 + θ − ξf(ϕ, φ)) + ϕ̇(θ̇ − 4ξHfϕ ) 2 1 − 2ξHφ̇fφ + 2Hθ̇ − ζ̈ + ζ̇ φ̇ + χf̈, (6.96b) 2 ◦ (at least in the non-local f(R) cases) [491–493]. In fact, a similar observation is found in the study of Noether symmetries while the scalar field equations are ◦ 0 = −6χḢfϕ − 6H 2 (ξ + 3χ)fϕ + 3Hθ̇ + θ̈, (6.97a) 0 = −6χḢfφ − 6H 2 (ξ + 3χ)fφ + 3Hζ̇ + ζ̈, (6.97b) in non-local f(R) gravity [494]. Only the cosmological expansion behavior of S1 has been investigated in detail where it hosts both de Sitter and powerlaw behaviors [335]. Nonetheless, it is claimed that these behaviors also appear in the remaining symmetries. For the nonlocal f (T) gravity sub-case, the de Sitter behavior appears as a solution to the field equations [490]. As the general analytic scale factor behavior is not obtained, use of statefinder parameters showed the cosmological sequence consisting of deceleration → acceleration → asymptotically de Sitter with periods where this Universe lies in a phantom domination epoch which may suggest quintom behavior. However, the deceleration phase may persist for a sufficiently long time leading to a fine tuning problem of the initial conditions [490]. 0 = Ḣ + 3Hφ̇ + 18H 2 + φ̈. (6.97c) ◦ 1 ξ + c7 (6 ln a + φ) + c9 , c10 + c3 θ, (c3 − c7 )ζ − c5 θ + c8 , by (5.138) are 0 = 3Hϕ̇ + 6H 2 + ϕ̈, c10 = 0, c3,7 ̸= 0, c4 c7 ̸= c5 (c6 + c9 ) c3,10 = 0, c7 ̸= 0, c4 c7 = c5 (c6 + c9 )  1 X = c1 t + c2 , (c1 − c3 )a, c4 + c5 (6 ln a + ψ) , c6 3 The FLRW for non-local given 6.6.1. Non-local theories. f(ϕ, ψ) Symmetry ◦ ◦ In the above, ϕ = □−1 T, φ = □−1 B, □θ = −(ξT + ◦ χB) ∂ f(ϕ,φ) and □ζ = −(ξT + χB) ∂ f(ϕ,φ) ∂ϕ ∂φ . The study of Noether symmetries in GTNL gravity has been investigated in [335, 490] using its scalar-tensor equivalent form equation (5.140b). Taking the configuration space Q = (t, a, ϕ, φ, θ, ζ) yields 7 distinct symmetries which stem from the generalized Noether vector 6.6.1.1. Noether’s symmetry approach. 25 Following [335], some integration constants do not appear for particular symmetries (for example, c10 in S1). However, as these constants do not appear in both the Noether vector and the function f(ϕ, ψ), they are taken to be zero. 106 Rep. Prog. Phys. 86 (2023) 026901 Review 6.6.1.2. Dynamical system approach. ◦ 1 κ2 ρ = − f(T, X1 , X2 ) − 6H 2 f̈X2 + 6HḟX2 (Ḣ − 3H 2 ) 2 − 6fX2 (HḦ + 3H 2 Ḣ + Ḣ 2 ) − 144H3 ḟX1 Ḣ The cosmological dynamics of the non-local f(T) = T(A□−1 T + B − 1) gravity model, where A, B are constants, has been investigated in [495]. Here, the presence of interacting dust (matter), radiation, DE and scalar field fluids are considered. This interaction is described through the interaction coefficients Γa X Γa = 0, Γa = 3H3 a X αab Ωb , − 144fX1 H3 (Ḧ + 3HḢ) − 6fT H 2 , (6.101a) ... 1 −κ2 p = − f(T, X1 , X2 ) − 2 f X2 H − 2f̈X2 (Ḣ + 6H 2 ) 2 − 18HḟX2 (Ḣ + H 2 ) − 6fX2 (HḦ + 3H 2 Ḣ + Ḣ 2 ) (6.99) − 48H 2 f̈X1 Ḣ − 48HḟX1 (2HḦ + 6H 2 Ḣ + 3Ḣ 2 )   ... − 48fX1 6H3 Ḧ + 9H4 Ḣ + Ḣ3 + H 2 (H + 12Ḣ 2 ) + 4HḢḦ b where − 31 αmΛ = αΛΛ = αrΛ = αϕΛ = 2β for some constant β. The phase-space variables are taken to be the density parameters Ωa , leading to a four-dimensional autonomous system. However, careful re-examination illustrates the following key observation. P The Friedmann equations yield the constraint a Ωa = 1 − 2B which, in principle, allows for an elimination of one of the dynamical phase-space variables reducing the dimensionality of the system. Solving the four-dimensional system instead fixes the value of B. More importantly, a realistic critical point is obtained provided all the density parameters are non-negative. This ultimately leaves the critical point (Ω⋆m , Ω⋆Λ , Ω⋆r , Ω⋆ϕ ) = (0, 0, 0, 0) (Point C) as the only viable point as the remaining critical points do not appear for the following reasons: − 2HḟT − 2fT (Ḣ + 3H 2 ), ◦ (6.101b) ◦ where fX1 = ∂f/∂(∇T) 2 and fX2 = ∂f/∂ □T. It should be noted that the second FLRW (6.101b) is not the same as the one reported in [286]. This means that the study performed in that reference could contain some mistakes. 6.6.2.1. Noether’s symmetry approach. ◦ In [416], the Noether symmetry approach for the f(T, □T) sub-class has ◦ been investigated for the configuration space Q = (a, T, □T). However, the derived FLRW equations do not match with the ones given previously. It appears that the issue stems from the determination of the Lagrange multipliers λ1 , λ2 . Considering the minisuperspace higher-order theory Lagrangian ◦  ◦ L = a3 f(T, □T) − λ1 (T + 6H 2 ) − λ2 □T − T̈ − 3HṪ (a) Point A becomes non-physical as the matter density becomes negative; (b) Points B, D and E reduce to the critical point C. (6.102) This critical point represents an empty static Universe which is either stable or saddle. Thus, it is not suited to describe the late-time behavior. Finally, the stability of a de Sitter solution for the special case when A = −1 and B = 0 was investigated. In the absence of matter fluids, the de Sitter cosmology turns out to be unstable, while inclusion of matter fluids yields an asymptotically stable static Universe critical point. Thus, the model does not appear to be a viable alternative in describing a latetime, de Sitter behavior. and taking variations wrt T yields     ∂L d ∂L d 2 ∂L − + 2 =0 ∂T dt ∂ Ṫ dt ∂ T̈ =⇒ a3 fT − λ1 + 3Ḣλ2 + 3Hλ̇2 + λ̈2 = 0. Contrary to that reported, it does not constrain λ1 directly due to the terms emerging from the Ṫ and T̈ contributions. Re-expressing the minisuperspace Lagrangian in terms of the scale factor a(t) instead would resolve this issue leading to the correct set of field equations. Thus, the Noether symmetry results may no longer hold and thus are not reported. 6.6.2. Higher-order derivative TG. The higher-order derivative theory described in section 5.7 adds two scalars that contain depend on the torsion scalar. These scalars for flat FLRW become ◦ (∇T) 2 = ◦ □T = − 144H 2 2 Ḣ , N2  12Ḣ 2 12HḦ 12HḢ − − 3HN 2 − Ṅ , N2 N2 N3 (6.103) For the general higher-order theory theory, the dynamical system approach has been explored in [286]. Generally, an autonomous system can be easily constructed following the definition of the dynamical variables 6.6.2.2. Dynamical system approach. (6.100a) l Ḣ H Z1 := H, Z2 := 2 , . . . , Zl+1 := l+1 , H H (6.100b) and then, by performing variations wrt the lapse and the scale factor (and then setting N = 1), we arrive at the modified FLRW given by (6.104) where overhead indices represent the number of time derivatives. Any critical point of the system would then be classified under two distinct cases: 107 Rep. Prog. Phys. 86 (2023) 026901 Review (a) Z1 ⋆ > 0, Z2, ..., l+2 ⋆ = 0: de Sitter cosmology; (b) Z1 ⋆ =, Zl+2 ⋆ = l!(l + 2)Z2 ⋆(l+1) : power law evolution − 1 a(t) ∝ t Z2 ⋆ . is achieved. However, the inflationary phase cannot be realized. The background cosmological dynamics of teleparallel axion theory has been investigated in [348, 349, 351]. In particular, the particular sub-case where the vector, axial and tensor quadratic contributions give rise to the TEGR contribution is considered, namely the gravitational action  ω(ϕ) 1  L = − 2 T + bϕP1 + b̃ϕP2 − gµν ∂µ ϕ∂ν ϕ + V(ϕ), 2κ 2κ2 (6.107) 6.6.4. Teleparallel axions. Although the general dynamical behavior of the theory was not explored, two ansatz forms were considered: ◦ (i) − T − ◦ δ(∇T) 2 α1 (∇T) 2 − α2 e T 4 , 2 T ◦ ◦ ◦ σ ∇T β1 □T β2 (□T) 2 T3 , (ii) − T − − − β e 3 T T3 (6.105) β1 where α1,2 , β1,2,3 with β2 = 734 , δ and σ are constants. For simplicity, the dynamics have been investigated in the presence of a perfect matter fluid having constant EoS. However, as equation (6.101b) does not match with the one reported in [286], the resulting dynamical behaviors may be different from those reported. Here, we summarize the results for completeness. The first model contains quintessence dominated and de Sitter phases as late-time attractors. Furthermore, the cosmological sequence radiation → matter → DE can be realized for suitable choices of α1 , α2 and δ. In the second model, only a stable de Sitter critical point appears while intermediary phases are not well realized making the model not viable. where P1 = uµ aµ and P2 = ϵµνρσ tλµν tλ ρσ are the odd-parity quadratic torsion contributions (see section 5.5), ϕ is a pseudoscalar field, b and b̃ are constants. For the FLRW geometry in the presence of spatial curvature k, there are two distinct tetrad branches which lead to the same metric as discussed in section 5.12.2. Due to the presence of the odd-parity terms, the chosen branch affects the forms of the FLRW equations:   ω(ϕ) 2 3k 2 2 ϕ̇ + V(ϕ) , 3H + 2 = κ ρ + (6.108a) a 2κ2 2Ḣ + 3H 2 + In the unimodular formulation [250, 251], the f (T) Lagrangian has been reconstructed for the power law cosmology a(t) ∝ tq in the presence of a perfect fluid with constant EoS, leading to26  − 3q2((3q1+−w1)) (3q − 1)κ2 ρm0 T f(T) = Λ + , 1 − 3q(2 + w) T0  − 3q2((3q1+−w1)) T 2 λ = Λ − 2κ ρm0 , (6.106) T0 ω(ϕ) 2 k = −κ2 p − ϕ̇ + κ2 V(ϕ) a2 2 ( 6.6.3. Mimetic and unimodular teleparallel cosmology. + buϕ̇ a , 0, for the first (axial) branch, for the second (vector) branch, (6.108b) √ where u = ± k. Observe that in both cases, the contribution from P2 is absent since tλµν = 0 for both branches. Meanwhile, the P1 contribution is absent in the vector branch as aµ = 0 reducing the model to a minimally coupled scenario whereas in the axial branch, the contribution is retained leading to a non-minimal coupling. Furthermore, the axion coupling does not influence the late-time dynamics due to the u/a dependence but it will affect the pre-inflation cosmology. It is also convenient to define the effective axion field energy density ρϕ and pressure pϕ : where λ represents the Lagrange multiplier which imposes the unimodularity constraint and Λ is an integration constant. For the case of HDE, the Lagrangian reduces to rescaled TEGR [252]. Meanwhile, the cosmological dynamical behavior of mimetic f (T) gravity has been investigated in [347] in the presence of dust, radiation and dark matter fluids. Generally, the model hosts both decelerating and accelerating DE dominated phases, radiation and matter dominated periods, and an early inflationary epoch. In particular, for the power-law model f(T) = −T + α(−T)b with b < 1, the cosmological sequence ρϕ = ω(ϕ) 2 ϕ̇ + V(ϕ), 2 2κ ( − buϕ̇ κ2 a , 0, pϕ = ω(ϕ) 2 ϕ̇ − V(ϕ) 2κ2 for the axial branch, for the vector branch. (6.109) In view of the above, [348] focus their attention to explore the cosmological dynamics of the axial scenario in vacuum (p = ρ = 0) by introducing the phase-space variables (α, β) satisfying the relations s s 2κ2 V(ϕ) α κ2 V(ϕ) √ cos β, ϕ̇ = , H= ω(ϕ) 3(1 − α 2 ) 1 − α2 s !−1 κ2 V(ϕ) a=u sin β , (6.110) 3(1 − α 2 ) radiation → matter → (decelerating) dark matter → (accelerating) dark energy 26 The torsion scalar equation (40) in [250] is incorrect. Correcting yields the solution found in [251]. 108 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 20. Dynamical behavior for the teleparallel axion model in vacuum for two choices of b illustrated in the α-β phase-space. Here, filled (open) circles denote stable, expanding (unstable, contracting) de Sitter critical points whereas filled (empty) diamonds denote Big Crunch (Big Bang) critical points. Stars denote the saddle transition points. Shades from dark to white indicate phantom behavior weff < −1, acceleration −1 < weff < −1/3 and deceleration p weff > −1/3, where weff = pϕ /ρϕ , respectively. It can be noted that no phantom behavior arises for b = 0 but appears for b = 8/3. In both cases, the late-time attractor is either de Sitter or Big Crunch. Reproduced from [348]. CC BY 4.0. which are constrained by −1 < α < 1 and sign u = sign sin β. As pre-inflationary epochs were considered, the constant roll approximation was assumed with ω(ϕ) = 1 and constant potential V(ϕ). In turn, critical points describing a Big Bang/Big Crunch singularity (α = ±1), an infinitely expanding or contracting de Sitter Universe (α = 0, β/π ∈ Z) and a saddle transition phase between acceleration and deceleration √ 2 (α± = b±2√b6+8 , πβ − 12 ∈ Z) were obtained. A representation of the dynamical behavior is given in figure 20. by some exotic particle physics beyond the standard model in the matter sector. One of the interesting properties of gravity beyond GR is that it allows for designer models to naturally induce an inflationary mechanism in the very early Universe which may give a more natural source to such a scenario. In this section, we explore some of the possible teleparallel inflationary scenarios proposed in the literature. While interesting, there are still many possible realizations of cosmic inflation that are yet to be explored, as well as more precise predictions that these models may produce. On the other hand, there are already interesting possible scenarios in which inflation is either induced naturally or in the same way of as in ΛCDM. 6.7. Inflation Cosmic inflation was proposed to solve the horizon, flatness and cosmic magnetic-monopole problems [14, 15] which had plagued theories of early Universe cosmology for a long time. The number of inflationary scenario proposals [26, 496, 497] has since drastically increased where various slow- and fastroll possible decay mechanisms for the inflationary scalar field(s) have been put forward. Despite all this, direct observational evidence of an inflationary era in the cosmic history of the Universe continues to evade measurement keeping the possibility of other exotic solutions to the early Universe puzzles as a possibility. ΛCDM contains no natural mechanism to produce an inflationary scenario on its own. This may of course be resolved The pioneering work by Born and Infeld [498, 499] assumes a finiteness for physical quantities at all times in order to avoid the divergences that were emerging in physics at the time. Their initial work concerned the self-energy of point-like particles [500], but has since been expanded as the root inspiration into several different branches. In gravity theory, there have been numerous manifestations of Born–Infeld inspire gravity theories starting with [501] where the first proposal turned out to be plagued by ghosts [502]. Since this time, Born–Infeld theories have been incorporated into metric-affine formalisms [503], 6.7.1. Born–Infeld inspired inflation. 109 Rep. Prog. Phys. 86 (2023) 026901 Review where αi are dimensionless constants. The only meaningful condition on Fµν (and thus its constants) is that it must limit to GR in the low energy limit (F  λBI ), which in this setting means Tr(Fµ ν ) = T. For the case where α1 = 0 = α2 , the Lagrangian in equation (6.114) becomes again an f (T) model, but is distinct to f (T) in all other cases. In a flat FLRW setting and a perfect fluid source term ρ, [508] finds the Friedmann equation Eddington-inspired Born–Infeld gravity [504], and many others. These have been used to eliminate a range of singularities from black holes to the big bang singularity. Given the status of black hole physics in TG (as explained further in section 9), the core work in the TG literature has been in the realm of early Universe cosmology. In this background, [38, 274, 505] explore a TG model à la Born–Infeld within the context of an f (T) gravity Lagrangian in the spirit of the action in equation (5.48) 27 in which  s 2 (T + 2Λ) − 1 , LBI−1 = λBI  1 + (6.111) λBI p √ where A1 = 6 (α2 + 2α3 ) /λBI and A2 = 2 (2α1 + α2 + 6α3 ) /λBI . The study also finds that the fluid conservation equation is observed at all scales. The authors study an intriguing form of the Ai parameter, namely the case where A2 = 0 which implies that A1 = 12/λBI . For the flat FLRW case they find again that in this limit # "r λBI t , a(t) ∼ exp (6.117) 12 where λBI is a scaling parameter which recovered ΛCDM in the limit where λBI → ∞. For a flat FLRW Universe, for very early-times (a → 0), the scale factor and Hubble parameter are given, respectively, as "s   # λBI 4Λ a(t) ∼ exp 1+ t , 12 λBI s   λBI 4Λ H(t) ∼ (6.112) 1+ , 12 λBI p for the very early Universe, giving H ∼ λBI /12 which is a maximum in this regime and which drives the early Universe to a de Sitter phase. The work also goes on to the study the evolution in nonflat scenarios but the tetrad that is adopted is not compatible with the Weitzenböck gauge tetrad we derive in equations (5.162) and (5.168) (for the k = +1 and k = −1 choices, respectively). While the results in this case are interesting, the antisymmetric field equations (see section 5.1) will render an incompatibility with this gauge choice. Finally, in [510] another more exotic branch of these parameter settings is explored, namely where A1 = A2 which reduces the Friedmann equation to which regularizes the divergence of TEGR at very earlytimes. This naturally induces an inflationary era from geometry without the need of an inflaton field. The redshift relation z = 1/a(t) − 1 implies that z tends to a constant for the early Universe which produces a divergence in the particle horizon ˆ 1 da η= → ∞, (6.113) 0 aȧ giving a Universe that is entirely causally connected, in agreement with the isotropic part of the cosmological principle and the CMB. It is important to point out that despite these properties at early-times in the Universe, this model tends to regular ΛCDM for late-times (provided that λBI  Λ) which leads to a cosmic history that is extremely similar to that of ΛCDM up to some very fine differences. In this way, the scaling parameter λBI acts takes on the role of an effective initial vacuum energy driver for the inflationary stage. In [506], very preliminary results are shown where the λBI is constrained using SN data from [507]. Here, it is found that λBI ≲ 10−11 Mpc−2 . In [508, 509], a slightly more generalized scenario is considered in which "s  #  p 2Fµν LBI−2 = λBI det gµν + − det gµν , (6.114) λBI   9H 2 6H 2 1 − = 2κ2 ρ, 2λBI (6.118) which produces two disconnected solutions for H 2 with the positive sign being ignored since it is detached from GR. To retain strong enough resemblance to GR, the λBI must be positive, in which case, the scale factor near the very early Universe gives a4 (t) = 6κ2 λ ρ0 + O(λBI t 2 ), BI (1 ± 4λBI t) (6.119) and a Hubble parameter H(t) = ∓λBI / (1 ± 4λBI t) + O(λBI t 2 ), where ρ0 is the energy density at present time for the perfect fluid. This produces a minimum for the scale factor at t = 0 where where Fµν contains only first derivatives of the tetrad and is chosen to be Fµν := α1 Sµρσ Tνρσ + α2 Sρµσ Tρνσ + α3 gµν T,  1 − A2 H 2 2κ2 ρ , (6.116) 1 + 2A2 H 2 − 3A1 A2 H4 − 1 = λBI 1 − A1 H2 a4Min = (6.115) ρ0 , 6κ2 λBI (6.120) and where HMin = λBI . This is called a brusque bounce, and represents a spacetime in which finite objects do not crash into zero volume. 27 Note that the convention of the torsion scalar in this Review is twice the torsion scalar in the above references. 110 Rep. Prog. Phys. 86 (2023) 026901 Review to a perfectly scalar-invariant spectrum (ns = 1), contradicting CMB observations ns = 0.9649 ± 0.0042 [516]. Additionally, this does not lead to a reheating phase. As such, the scalar potential must provide [517–519]: Loop quantum cosmology within the FLRW framework hosts the matter bounce scenario [511], described by the Friedmann equation   ρ ρ 2 1− , H = (6.121) 3 ρcr 6.7.2. Power spectra in LQC f(T) gravity. (a) Matching with spectral indices and power spectra observations; (b) Account for reheating processes; (c) Stability against small perturbations. where ρ is the matter fluid energy density and ρcr = λ 23γ 2 is a critical q density originating from quantum considerations with √ 3 λ= 2 γ and γ ≈ 0.2375 the Barbero–Immirzi parameter. Assuming a dust perfect isotropic fluid, the scale factor and Hubble parameter take on the forms  a(t) = 3 ρcr t 2 + 1 4  13 , H(t) = 2ρcr t . 4 + 3ρcr t 2 Some works also include a late-time accelerating expansion as a criterion but this is not necessarily the case. Many such scalar fields cannot achieve both goals of an early-time inflationary epoch coupled with late-time accelerating expansion. For instance, a quadratic potential does not meet both this extra criterion. However, it may be modified to include a small cosmological constant provided instant preheating and matter dominated fields are obtained at low and large ϕ values respectively. Similar considerations can be achieved via quintessence type potentials. Another possibility focuses on the quasi-matter contraction scenario which focuses on the matter fluid’s EoS to be ω  1 as opposed√to being identically zero. In this case, the potential V(ϕ) ∝ e− 3(1+ω)ϕ and the scalar spectral index changes to ns = 1 + 12ω allowing for a non-invariant spectrum. A further √ generalization considers V(ϕ) ∝ e 3ϕ(1+f(ϕ)) with | f(ϕ)|  1 and |ϕfϕ |  1. Despite such models do not allow for a transition between the contraction and expansion phases in GR, this does not appear in the f (T) formulation [520]. We now turn our attention to the scalar and tensor power spectra. To investigate the scalar power spectrum, the longitudinal gauge is considered with Bardeen potentials Φ and (6.122) Clearly, the bounce occurs at t = 0. In contrast to curvature-based theories, the teleparallel formulation allows for a direct reformulation of the LQC Lagrangian purely in terms of the torsion scalar by virtue of the direct correspondence between T and H, namely T = −6H 2 . In this way, the LQC Lagrangian can be recast in terms of f (T) gravity [512–514] ! r r 3 T T f(T) = − − arcsin 2λγ − λγ 6 6 " !# r 1 3 T + 2 2 sin 2 arcsin 2λγ − . (6.123) γ λ 2 6 Alternatively, the Lagrangian can be obtained via reconstruction of the Friedmann equations [267, 514]. An extension to the standard LQC formulation is considered in [515] where the critical density is instead parametrised as ρcr = α122 where α is a constant which measures the deviation from TEGR. This modifies the teleparallel Lagrangian to be recast in the following form " !# r r r 12 α 2T T T f(T) = − 2 1 − 1 + − α − arcsin α − , α 6 6 6 0 Ψ. Perturbing the scalar field about a background value ϕ as 0 ϕ = ϕ +δϕ, for the potentials considered, no anisotropic contributions arise. From the perturbed f (T) field equations, the Φ and Ψ modes are equal according to [521] and in follow up works [517–520, 522, 523]. However, this is not the case as a proper account of the perturbed tetrad was not considered. The perturbed tetrad introduces a further DoF which has not been accounted for in the aforementioned works. As this missing term appears in the anisotropic parts of the field equations, the Φ and Ψ potentials are no longer equal despite being of the same order at least within sub-horizon length scales. Even if the latter scales are considered, only one of the antisymmetric f (T) field equations has been accounted for despite the existence of a second antisymmetric field equation. Neglect of this secondary equation in conjunction with having equal potentials leads to having a greater number of field equations than DoF which is only reconciled once the extra scalar DoF is taken into account. On the other hand, the tensor power spectrum is obtained by exploring the tensor perturbations which yields the relation (expressed in Fourier space, details provided in section 7.4) ! ḟT k2 ḧλ + 3H + (6.125) ḣλ + 2 hλ = 0, fT a (6.124) which recovers TEGR f(T) = −T in the limit |α| → 0. Despite this apparent equivalence, there are crucial differences between the LQC and teleparallel formulations. Most notably, the critical density is quantum mechanically derived in LQC but becomes a free parameter in the teleparallel formulation. This has important consequences when matching with spectra observations as the freedom in ρcr may be sufficient to reconcile different data sets. Meanwhile, equivalence only holds at a background level with deviations appearing at perturbative regimes. In this setting, the matter sector is sourced by some scalar field ϕ having an associated potential V(ϕ). The latter’s nature is crucial for the matter bounce cosmology as it affects both spectral indices and power spectra. A simple choice for such scalar 2 field is one which2 effective fluid acts as dust with ρ = ϕ̇2 + V(ϕ) and p = ϕ̇2 − V(ϕ) = 0. However, this leads 111 Rep. Prog. Phys. 86 (2023) 026901 Review which is a subset of the general equation (7.40), and where λ denotes the two polarization states of the tensor perturbations. √ Introducing the Mukhanov-Sasaki variables zt = a fT and v = hzt recasts the equation in the form z′′ v −k v− v = 0, z ′′ 2 where λ and v are constants, and where  K = K0 , F = F0 1 + ζϕ 2 , G = G1 ϕ 2 , where K0 , F0 and G1 are constants. In this work, the tetrad postulate of TG is used to expand the particular terms in the action. This is done so that the problematic term in section 5.3.2 (namely, the vector torsion coupling term) can be isolated and set to zero for one branch of possible function settings. This study is, however, not purely contained to TG, and also includes a broader contribution from non-metricity. Despite this, for the branch where an Einstein frame exists, regular Higgs inflation (ζϕ  1) produces a very high tensor-to-scalar ratio which is well constrained by observations [372]. In the other branch, in which an Einstein frame is not recovered by a conformal (or disformal) transformation, a working inflationary model is not recovered due to linear perturbations not being dominant. This is conducted using an FLRW metric in conformal time. Thus, in [527], Higgs inflation appears problematic. However, the particular prescription being explored is very specific and may yield more realistic results. Moreover, other scalar field coupling terms may also produce more practical inflationary scenarios such as those explored in [530–533]. The question of the naturalness of inflation in TG remains an open question and poorly tackled in the literature. An interesting recent work on the topic is [534] where inflation is explored through a nonminimally coupled scalar field in f (T) gravity [535] (see section 5.8) in the same way that previ- (6.126) where primes denote derivatives wrt conformal time (defined through dη = dt/a [524]). This leads to the computation of the tensor power spectrum k3 v(η) 2π 2 z(η) ˆ ∞ dη , RT := 2 −∞ zt 2 PT := = ρcr 2 2 16ρcr 2  π  RT = Si , ρ 3ρ 2 (6.127) where Si(x) is the sine integral function. Under the assumption that the scalar power spectrum result is correct, the tensor-scalar ratio r ≈ 6.7187, considerably larger than currently observed bounds r < 0.056 [516]. However, the magnitude of this ratio together with the magnitudes of the power spectra changes depending on the choice of potential [517] as well as the magnitude of the background scalar field ϕ̄ at the bounce [517, 522]. Another important point to highlight is that this results depends on a particular solution of the scalar field equation of motion where a vanishing pressure for matter is assumed, which may not be the case. [517] goes on to find numerical solutions which do satisfy current bounds on the tensor-scalar ratio. ◦ ous works have approach the same setting in f(R) gravity [22, 24]. They find the slow-roll conditions for a near scale invariant power spectrum in this general setup and explore the dynamical systems for quadratic and quartic potentials. In both cases, inflation is very difficult or impossible to produce in the scalar-torsion approach. However, there are many other possible options for the scalar field potential which may produce an inflationary epoch that is more consistent with cosmic evolution in the early Universe. Other works also consider the impact of a coupled scalar field in generalized f(T, ϕ) gravity such as [536, 537] where power spectra for scalar and tensor modes are calculated (which have led to determinations of the spectral index and scalar-to-tensor ratio). In [538], a powerlaw model of f (T) gravity was probed for its possible inflationary scenarios where the spectral index and power spectrum is again calculated but where the model was not totally related to possible late-time accelerated expansion, for consistency. Finally, [539] explores possible tachyonic inflationary scenarios which has some interesting explanations for recent observational measurements. In the standard model of particle physics all the matter fields couple minimally with the gravitational section except for the case of a Higgs field where no suppression needs to take place [525]. Thus, the Higgs field is an interesting avenue in which to explore differences between GR, TEGR, and STEGR. In GR, Higgs inflation then emerges as quantum corrections between the Einstein–Hilbert action and the Higgs coupling field [526]. By considering direct couplings with TG quantities like the torsion scalar instead of the Ricci scalar, we can probe possible differences in these different inflationary scenarios. Taking a generic scalar that can then be non-minimally coupled to the scalar torsion through [527–529] ˆ h ◦ 1 SHiggs = − 2 d4 e F(ϕ)T + 2G(ϕ)∇α T α + K(ϕ)g αβ 2κ  i  ◦ × ∂α ϕ∂β ϕ + V(ϕ) + Sm ϕ, Ψ, eAµ , ω ABµ , 6.7.3. Higgs inflation. (6.128) where all other matter fields Ψ are minimally coupled to the torsion scalar, and K and V are kinetic and potential functions, and which is a particular form of the action introduced in equation (5.109). In [527], this model is investigated for Higgs inflation in which the potential is given by V= 2 λ 2 ϕ − v2 , 4 (6.130) 6.8. Essential conditions for viability in cosmological models [Jackson: In this section a number of prominent modified theories within TG have been explored through the works that appear in the literature in the cosmological setting at background level. However, cosmology can also yield theoretical limits on models in the context of their behavior against various kinds of instabilities such as ghosts and other forms (6.129) 112 Rep. Prog. Phys. 86 (2023) 026901 Review of instability. Ghost instabilities represent unstable propagating DoF where a wrong sign in the propagator gives rise to unbounded states in the phase space of the theory [168, 540, 541]. This became a notable feature due to the popularity in fourth-order gravitational theories in recent decades but is also a property that can arise in second-order gravitational theories. The issue can also be expressed through a model exhibiting the correct sign in its kinetic term which can be used to put limits on potential models that can be realized within such classes of theories [542, 543]. There has been recent work on producing stable models for theories that feature a ghost [544] but this has not yet been realized in a TG setting.] [Jackson: Another possible problematic property in modified gravitational theories is gradient or Laplacian instabilities [545, 546] where even small wavelength perturbations grow are potentially exponential rates unless otherwise tamed. The instability arises out of an imaginary propagation velocity in one of the perturbation sectors, and is only avoided by demanding positive propagation speeds at all scales. Similarly, if a massive field is admitted, tachyonic instabilities can appear for regions of negative effective mass [547]. These instabilities can cause problems at background level which are related to instabilities in the vacuum. A special case of this instability is the scenario where a negative effective mass appears but which grows slower than the Hubble rate. This issue appears in ΛCDM [548] and may not gives grounds to discard a particular theory. However, it is something that must be accounted for and studied for potential problematic features.] A healthy theoretical model or class of models must be exorcised of any such instabilities which may produce unnatural expressions of a theory that are incompatible with the real Universe. Another crucial condition on which to assess whether a theoretical framework is healthy or not is the possibility of strongly coupled solutions. This was introduced in sections 3.6.3, 5.2.2 and will be discussed further in the remainder of the Review. In essence, strong coupling occurs for specific solutions where the DoF do not all appear in the linearization of the theory. This means that, for those solutions, that possibly unbounded modes may coupling different order perturbations in an irregular way. Moreover, given that solutions are approximations to real systems such as in cosmology but also for astrophysical solutions, it means that such spacetimes may not approximate well these setups since even minor perturbations away from background assumptions may have a large impact at background level. This will pollute the background solution since perturbations will no longer be contained to perturbative order level. Hence, it is crucial that any theoretical setup not exhibit strong coupling for the most physical solutions such as those that express the cosmological principle, i.e. the FLRW spacetime. A viable theory should first be based on a healthy theoretical framework in which instabilities are adequately addressed and that physically realistic spacetimes are not strongly coupled. at least from a background cosmological viewpoint. Briefly, f (T) gravity is the simplest class of models to reconstruct and can provide solutions which match with observations as well as reproduce non-flat spacetime cosmologies purely from a gravitational standpoint. However, no f (T) Lagrangian can perfectly reproduce the ΛCDM cosmology unless it is ΛCDM Lagrangian. This is not the case for other teleparallel extensions such as f(T, B) and f(T, TG ) gravity. For these theories, the Lagrangian reproduces the background ΛCDM cosmology with the possibility of generating differing behaviors at perturbative regimes. The ΛCDM cosmology can also be produced via a conformal scalar theory by requiring the scalar field to be the driving field to cause late-time acceleration. Introducing the trace of the stress-energy tensor Θ increases the complexity of the solutions which may make them less appealing for broader use. Nonetheless, this formulation provides an interesting alternative to the Einstein static Universe as the closed spacetime geometry requirement is lifted. Additionally, unification and transitional eras can be purely described from this torsional-matter coupling viewpoint. The dynamical system approach serves as an interesting alternative to explore the Universe’s cosmological dynamics prescribed by a gravitational model. Through critical points and their stability, the corresponding cosmological sequence can be determined, although, differing phase-space variables choices influence the ensuing analysis especially in f (T) gravity. Overall, the models under review here together with associated tables and figures of the relevant dynamical analysis are summarized in figures 21 and 22. As the dynamical analysis can provide a general overview of the cosmology, a selection criterion to distinguish between suitable models has been adopted, one which hosts the cosmological sequence described by an unstable inflationary period → saddle radiation/matter domination epochs → (stable) late-time accelerated expansion. In hindsight, the teleparallel description offers a vast selection of models whose evolutionary sequence are summarized in figures 23 and 24. However, only a few choices of TG models appear to host the complete cosmological sequence while the majority of models start from matter domination. Even so, such TG models may indeed host the full cosmological sequence needed to describe the Universe. The absence of such behavior is also imparted by the choice of the dynamical phase-space variables which greatly influence the number of critical points, and hence the evolutionary behavior, of the theory. Of note, a late-time accelerating attractor is present in practically all TG theories. Additionally, other behaviors including bouncing scenarios, phantom and quintessence regimes can be realized. In some cases, such as in f (T) gravity, absence of cosmological sequences or epochs including cyclic or oscillatory solutions can also be determined through this approach. Moreover, only the dynamical behavior of f (T) gravity has been explored for arbitrary functional choices, making the study on other TG extensions to be of further interest. Focusing our attention to the topic of inflation, the different approaches revealed the potential existence of an inflationary period in TG, namely in scalar-tensor formulations and 6.9. Perspectives of teleparallel cosmology The reconstruction procedure is a viable alternative approach as a means to explore the gravitational teleparallel action 113 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 21. Breakdown summary of the reviewed works involving the use of the dynamical system approach in the absence of scalar fields. Figure 22. Breakdown summary where the dynamical system approach has been applied in the presence of scalar fields. from cosmological dynamics. In the latter, a viable cosmological sequence can be realised for f (T) gravity and its mimetic equivalent. Other models mainly the TG equivalents of Born– Infeld, LQC and Higgs inflation have focussed their attention in the realization of this inflationary epoch. For instance, in the Born–Infeld formulation, having λBI ≲ 1 × 10−11 Mpc−2 ensures casual connectedness, and realization of an inflationary epoch with ΛCDM-like behavior at late times. However, the main issue outlined in the reviewed works lies in the cosmological perturbative regime. In the TG LQC and Higgs formulations, they both suffer from large tensor-scalar ratios, marking it an important viability requirement for TG models. In view of the above, it must still be emphasized that generating the background cosmology is insufficient to confirm the viability of the model. For instance, the model could also ensure the generation of the desired cosmological sequence through dynamical systems but unable to conform with local or cosmological observations. Ensuring existence of vacuum solutions is another important criterion, a property which has in fact helped to greatly reduce the number of models in the reconstructed solutions obtained in f(T, B) and f(T, TG ) gravity. Ultimately, all these different approaches help in providing a set of necessary criteria to build a viable model. The next step is the exploration of the perturbative regime which holds 114 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 23. Representation of how f (T) gravity realizes different cosmological sequences based on the investigated model. The choice of Lagrangian also invokes the existence of certain phases and development toward future phases. Some model examples have been included. Figure 24. Representation of how other teleparallel gravitational models (including possible Lagrangian choices) realize different cosmological sequences. another realm of important observations and conditions crucial to ensure the validity of the model. in a more direct way. This is the origin of cosmological perturbation theory and was first introduced by Lifshitz in 1946 [551]. See [552] for a modern exposition of the theory of cosmological perturbations. In this framework, we attribute the inhomogeneous and anisotropic structure to the perturbations of the flat FLRW metric. These perturbations turn out to describe the observed CMB anisotropy and large scale structure very well. Although, one needs to be very careful with the perturbative framework since it is only meaningful in very specific limits [553]. For example, at late-times the matter overdensities become significantly more dominant to higher perturbative levels than the linear one. In the perturbative framework, up to linear order, the metric assumes the expansion [552] 7. Cosmological perturbations in TG The background geometry of the Universe, as also discussed in depth in section 6, has been considered as being spatially flat FLRW geometry. The general class of FLRW geometries is generated by assuming that the Universe is homogeneous and isotropic (cosmic principle [373]). Observations related to the distribution of structure formation at large scales, due to the isotropic nature of CMB render such spatially flat cosmological geometries well motivated and founded [371, 374–377]. This is the case even if a few percent nonzero spatial curvature cannot be ruled out at high precision [372, 378–380]. Dropping the weak field approximation of Minkowski spacetime, and considering full perturbation theory in more general backgrounds (such as the FLRW background solution), these deviations from the cosmic principle can be tackled 0 gµν = gµν + δgµν , 0 (7.1) where gµν is the background value of the metric and δgµν are the small inhomogeneous and anisotropic correction to the background spacetime, which satisfies |δgµν |  1. As the 115 Rep. Prog. Phys. 86 (2023) 026901 Review background evolution is exhaustively reviewed in section 6, our main focus in this section will be to study the perturbation of the metric. In a very similar manner one can formulate the perturbation of the tetrad as 0 eA µ = eA µ + δeA µ , 7.1. 3 + 1 split in cosmology We start by reviewing the geometry of the FLRW spacetime, as presented in [562] and in section 4.2.1, in order to facilitate the introduction of perturbations, and to lay out the notation to be used. Assuming that the spacetime is modeled after a globally hyperbolic manifold M, which is diffeomorphic to the product manifold R × Σ, where Σ is a maximally symmetric, threedimensional manifold [563, 564]. We will denote by i : R × Σ → M the diffeomorphism relating these manifolds and we define the time coordinate t on M as the projection t = pr1 ◦ i−1 : M → R. In addition, we equip Σ with local coordinates (xi ), thus inducing the coordinate system (xµ ) = (t, xi ) for M. For the details regarding the conventions, see section 1.3. The space manifold Σ is taken to be maximally symmetric, which means that it is equipped with a metric (7.2) 0 where eA µ denotes the background value of the tetrad while δeA µ is the first order perturbation of the tetrad obeying |δeA µ |  1. In general, in order to have proper perturbative analysis the background solution of any variable under consideration 0 0 (eA µ or gµν ) must not lead to trivialization of the Hamiltonian constraints. This would mean that some of the fields may assume anomalous propagation or completely lose their kinetic term thus signaling potentially fatal problems in the analysis. The most common issue of this sort is the so called strong coupling issue as discussed also in section 5.2.2, where some constraint in the Hamiltonian analysis is trivialized due to specific (bad) background solution thus stripping away constraints some DoF loses its dynamics. Studies in the direction for the Minkowski background in f (T) theory has been carried out in [231, 554]. Furthermore, in cosmological back- γ = γij dxi ⊗ dxj , (7.3) where ⊗ denotes the usual tensor product, admitting the maximal number of Killing vector fields. Hence the metric can be split as gµν = nµ nν − ḡµν , (7.4) into the hypersurface (co-)normal 0 grounds where a background tetrad solution eA µ was such that fT ≡ fT (t), it was shown in [154] to be linked with a strong coupling issue. Whether or not this is the case, the overall nature of the strong coupling problem is traced back to the non-linearity of the constraints in the velocities which is introduced by modifying the theory T → f(T) [555–557]. Still we need more studies in other TG theories such as −T + F(B) gravity (or other extensions) to understand if this situation also appears in other theories. We also mention, that there is a class of perturbations which are called homogeneous [558–560]. In this class of perturbations, one directly perturbs the background quantities and field equations. For instance one perturbs the scale factor a(t) as nµ ∂µ = N−1 ∂t , nµ dxµ = N dt, (7.5) ḡµν dxµ ⊗ dxν = a2 γij dxi ⊗ dxj , (7.6) and the spatial metric where N = N(t) is the lapse and a = a(t) the scale factor. They satisfy the normalization nµ nµ = 1 and orthogonality nµ ḡµν = 0. The metric is thus given by g = gµν dxµ ⊗ dxν = N2 dt ⊗ dt − a2 γij dxi ⊗ dxj . (7.7) This metric expressed in spherical coordinates was previously introduced in equation (6.1). In case one chooses Cartesian coordinates, and a flat FLRW geometry for the metric (7.7), we obtain the form of equation (6.3). Both of these cases were discussed in detail in section 6. Further, we denote by ϵµνρσ the totally antisymmetric tensor defined by the spacetime metric gµν . Its spatial part is denoted as εµνρ and defined by 0 a(t) = a(t) + δa(t) and then one can find the field equations of δa(t) by perturbing the background Friedman equations of the theory. In this approach one could perturb the matter density and pressure to determine the behavior of the effective EoS at perturbative level. We can also use this approach to find the deceleration parameter and how it behaves under specific choices of the perturbation variables along with particular parameters arising from choosing specific models. Considering the effective EoS and the deceleration parameter, one is able to obtain stability conditions, as it was done for f (T) in [392, 561] and for f(T, B) in [261]. This method of perturbations in completely different than the standard one, for example outlined in [552], since we perturb already background evaluated quantities as indicated with the scale factor. This only affects the background DoF which for a metric in flat FLRW background are just 2 whereas in the general perturbative scheme they are 10. One of the main uses of the homogeneous perturbations is to probe the background stability of a theory in contrast to the general perturbations which probe the stability of the perturbed background. εµνρ = nσ ϵσµνρ , ϵµνρσ = 4ε[µνρ nσ] . (7.8) We note that by construction, nµ εµνρ = 0 and we remark that εµνρ is related to the totally antisymmetric tensor υ ijk of the metric γ ij via εµνρ dxµ ⊗ dxν ⊗ dxρ = a3 υijk dxi ⊗ dxj ⊗ dxk . (7.9) We will make use of these relations in the following sections. We split the arbitrary tensors into spatial and temporal components by first introducing the spatial tensor fields Πcµ ∂c ⊗ dxµ = aδbc ∂c ⊗ dxb , 116 Πµc ∂µ ⊗ dxc = a−1 δcb ∂b ⊗ dxc . (7.10) Rep. Prog. Phys. 86 (2023) 026901 Review It follows that they are related to the unit (co-)normal nµ and induced spatial metric ḡµν by nµ Πµi = 0, In order to relate the introduced decomposition of covariant derivatives on model spacetimes M to those projected on the spatial manifold Σ, we denote by di the Levi-Civita derivative of the metric γ ij , which acts on vector fields ξ i on Σ as ḡµν Πµi Πνj = γij , nµ Πiµ = 0, γij Πiµ Πjν = ḡµν . (7.11) di ξ j = ∂i ξ j + Γ̄j ki ξ k , Consequently, we can write the temporal and spatial projectors in the form δνµ = nµ nν − ḡµν = nµ nν − Πµj Πjν , Πiµ Πµj = δji . with the spatial Christoffel symbols of the spatial metric γ ij (7.12) 1 Γ̄i jk = γ il (∂j γlk + ∂k γjl − ∂l γjk ). 2 For a vector field X = Xµ ∂µ , we then introduce the notation for the temporal and spatial components: X = N−1 X̂0 ∂t + a−1 X̂a ∂a , X̂0 = nµ Xµ = NX 0 , X̂b = −Πbµ Xµ = −aXb . Conversely, for a covector field Ξ = Ξµ dxµ we write µ Ξ = NΞ̂0 dt + aΞ̂b dx , Ξ̂0 = n Ξµ = N −1 Ξ̂b = Πµb Ξµ = a−1 Ξb . Aµ = 0, Ξ0 , X̂a = −γ ab X̂b , (7.15) and analogously for Ξ. Hence, the decomposed tensor fields indices are raised and lowered with the background metric ds2 = g = dt ⊗ dt − γij dxi ⊗ dxj . Note that the components of the latter are independent of the time coordinate t, in contrast to the components of the spacetime metric gµν . We now make use of the tensor decomposition introduced above to decompose the covariant derivative of the Levi-Civita ◦ ν ∇X Y = (X ∇µ Y )∂ν , δgµν  := (7.16) is not necessarily spatial. Its decomposition ◦ ◦ ∇X Y = DX Y + nK(X, Y), (7.17) where n := n ∂α is the unit normal vector, defines the spatial covariant derivative ◦ ◦ ◦ (7.18) as well as the extrinsic curvature ◦ K(X, Y) = Kµν Xµ Yν = −nν Xµ ∇µ Yν , (7.19) which acts as a measure geometric deformation wrt a hypersurface [73], and where equation (7.4) was used. The latter is related to the acceleration [565] vector field Aµ by ◦ Kµν = ∇µ nν + nµ Aν , ◦ Aµ = nν ∇ ν nµ . (7.23)  −2φ a (∂i B + Bi )  . a (∂i B + Bi ) 2a2 −ψδij + ∂i ∂j h + 2∂(i hj) + 21 hij (7.24) The constituent fields that have an index are all divergenceless, i.e, ∂ i hij = 0 and also for any of the vectors ∂i Xi = 0. In addition hij is symmetric and traceless hij δ ij = 0. Since the metric is a symmetric (0,2) tensor in general, it assumes 10 DoF which are distributed as 4 scalars {φ, B, ψ, h} (account for 1 DoF each), 2 vectors {Bi , hj } (account for 2 DoF each) and 1 tensor hij (accounts for 2 DoF). This is the most general form of the perturbation of the metric in a flat FLRW background and it was first introduced by Lifshitz [66, 552, 553, 566, 567]. On the other hand, the tetrad can be represented as a generic 4 by 4 matrix with no symmetries which means that it enjoys the full set of 16 DoF contrary to the metric which can be also represented as symmetric matrix assuming only 10 DoF. Hence we distribute these 16 DoF of the perturbation of the tetrad δeI µ as follows α DX Y = (Xµ Dµ Yν )∂ν = (Xµ ∇µ Yν )ḡρν ∂ρ , 1 ∂a , N a Expanding on the form of the perturbations, the general first order expansion of the perturbed metric of equation (7.1) can be explicitly determined in specific backgrounds. Fixing the background to be spatially flat FLRW, the perturbation of the metric δgµν can be parametrized in a irreducible form as ◦ µ H= 7.2. Irreducible scalar-vector-tensor decomposition connection ∇ (as defined in sections 1.3, 2 and 4.2). By considering X µ and Y µ as spatial, we have that nµ Xµ = nµ Yµ = 0, and then, the covariant derivative ◦ Kµν = Hḡµν , where the dot denotes differentiation wrt time and H is usual Hubble function already introduced in various places as sections 1.3 and 6. This just signifies the geometrical origin of the Hubble function as derived from a fully geometrical point of view. Having laid out explicitly the underlying geometrical framework, we will introduce the cosmological perturbation scheme in detail in the next sections. (7.14) Hence we see that the temporal and spatial components are related by X̂0 = X̂0 , (7.22) In the case of the FLRW metric, the acceleration and extrinsic curvature are given by (7.13) b (7.21) (7.20) 117 Rep. Prog. Phys. 86 (2023) 026901 Review  δeI µ := τµν = 0 τµν := ηAB eA µ δeB ν ,  −φ −a (∂i β + βi )  a(∂i b + bi ) a2 −ψδij + ∂i ∂j h + 2∂(i hj) + 12 hij + ϵijk ∂ k σ + σ k δΘµ := δρ (ρ + p)(vi + ∂ i v) 2 −a (ρ + p)(vi + ∂i v) −δpδji (7.25) (7.26) which calculated explicitly for equation (7.25) reads as ˆ and turns out to be more suitable for calculations since the Lorentz indices can be dropped in the process. Just as the inverse metric perturbation, the inverse tetrad perturbation can be calculated using the relations eA µ EA ν = δµν and gµα g αν = δµν which were introduced in sections 1.3 and 2.2.1. The following convention for the perturbation of the energy-momentum tensor of a perfect fluid will be employed ν , are due to the Lorentz group which the metric is already invariant under. The antisymmetric part, that introduces the terms ∂ k σ and σ k , was elusive at first by some early works like [568] and were later correctly included in [242, 569, 570]. We note also that one can introduce the spacetime indexed tetrad where the capital Latin letters indicate three-dimensional spatial Minkowski indices (see section 1.3). Here, again all the vectors are solinoidal ∂i Xi = 0 and to be compatible with the metric B ≡ b − β, Bi ≡ bi − βi . In this case, the 16 DoF are split as follows: 6 scalars {φ, b, β, ψ, h, σ}, 4 vectors {bi , βi , hi , σi } and 1 tensor hij . The counting is again the same as the in case of the metric. The main differences schematically are the introduction of a pseudo-scalar σ and a pseudo-vector σ i which gives us B → {b, β}, Bi → {bi , βi }. The extra 6 DoF   −φ  −a (∂i β + βi )  δ I i ∂ i b + bi aδ Ii −ψδij + ∂i ∂j h + 2∂(i hj) + 12 hij + ϵijk ∂ k σ + σ k X (t, x) = dω ˆ  , (7.27) d3 k 1/2 3/2 (2π) (2π) h i j −iω t+ikj xj × X(ω, kj )e + X† (ω, kj )eiωt−ikj x , (7.30) where ω is the time component and kj is the spatial wave covector of the Fourier transform, and X† is the conjugate of X. We also raise and lower the indices as X0 = X 0 , Xj = −Xj and use the conventions □ := ∂µ ∂ µ = ∂02 − ∂ 2 , where the spatial Laplacian is defined as ∂ 2 := −η ij ∂j ∂i = δ ij ∂i ∂j . Along these lines, the norm of the wave covector ki is defined as k2 := −η ij ki kj = δ ij ki kj . In the supplementary annexes (supplementary 2), we provide some useful expressions for the SVT decomposition for different important tensors in TG.  , (7.28) where ρ is the matter density, p denotes the pressure, and v, vi denote the scalar and vector parts of the perturbation of the velocity field. In general, the spatial part δΘi j can include an anisotropic stress piece Πi j which can be further split into a scalar-vector-tensor (SVT) decomposition as 7.3. Gauge transformations Πij = ΠSij + ΠVij + ΠTij , (7.29a) We will utilize the active approach of [571, 572] in setting up the gauge transformations of tensor fields. In this approach the perturbation of a tensor field Z along the direction of a vector field Y at any order is given by ΠSij := ∂i ∂j ΠS , (7.29b) e = eLY Z, Z 1 ΠVij := − (∂j ΠVi + ∂i ΠVj ), (7.29c) e is the transformed tensor field and LY is the Lie derivwhere Z ative along Y. The vector field Y that generates the transformation and the tensor field Z can be further expanded up to second order as ΠTij (7.29d) 2 is the tensor part, where ∂ ΠVi ≡ 0 ≡ ∂ i ΠTij , ΠTij ≡ ΠT(ij) and δ ij ΠTij ≡ 0. i We close off by introducing our convention for the Fourier transform of a perturbation X as 118 (7.31) 1 Y = Y1 + Y2 , 2 (7.32a) 1 3 Z = Z1 + Z2 + O((LY ) ), 2 (7.32b) Rep. Prog. Phys. 86 (2023) 026901 Review Table 14. Common gauge choices for the metric and corresponding and the exponential map can be expanded up to second order as eLY = 1 + LY1 + 1 1 2 2 3 (LY1 ) + (LY2 ) + O((LY ) ). 2 2 tetrad. (7.33) Combining the equations above, we finally arrive at e 0 = Z0 , Z (7.34a) e1 = Z1 + LY1 Z0 , Z (7.34b) Gauge Metric Tetrad Flat Unitary Newtonian Synchronous ψ = 0, h = 0 δΦ = 0, h = 0 B = 0, h = 0 φ = 0, B = 0 ψ = 0, h = 0 δΦ = 0, h = 0 b = β, h = 0 φ = 0, b = β where in general, for a scalar field ϕ, a vector field X µ and a (0,2) tensor tµν we have can always set the pseudo-vector σ i to zero but not the pseudoscalar σ since it is gauge invariant. Using the transformations from equation (7.37) and the grouping of the fields in order to gather a few popular gauge fixing choices in table 14. For completeness we introduce a generic scalar field Φ and a commonly used scalar called the shear potential χ := a(β − b + aĖ) which transform as LY ϕ = Yλ ϕ,λ , (7.35a) χ e = χ + Y 0, (7.38a) LY Xµ = Xµ,α Y α + Xλ Yλ ,µ , (7.35b) f = δΦ + Φ̇Y 0 . δΦ (7.38b) e2 = Z2 + LY2 Z0 + (LY1 )2 Z0 + 2LY1 Z1 , Z (7.34c) LY tµν = tµν,λ Yλ + tµλ Yλ ,ν + tλν Yλ ,µ . From all the scalar DoF we can construct the following gauge invariant variables found in the literature but adapted to the perturbed tetrad, (7.35c) Without digressing, let us state that the passive approach, counterpart to the active we just presented is just stating that an infinitesimal change of coordinates along the vector field Y µ is described by exµ = xµ + Yµ1 which is usually further split as et = t + Y10 and exi = xi + Y1i . This is the usual first order approach followed for example in [553, 566, 573], but that method is rather cumbersome if one needs to consider higher orders. Next, we will calculate the gauge transformation of the components of the perturbed tetrad. In what follows we will only determine the first order perturbations so that, we drop the subscripts Yµ1 → Yµ . The transformed perturbed tetrad according to the active approach is e A µ = δeA µ + LY eA µ . δe ΦB = ΨB = −ψ + a(−β + b)H − a2 Hḣ, δΦf = δΦ + ψe = ψ + Hξ 0 ,   e b = b − a ζ̇ξ + ζ ξ˙ ,  (7.39c) (7.39d) Starting with the most important and simple perturbations (the tensor ones), one finds that the most general case of GW propagation equation for this sector yields [574]   k2 ḧij + (3 + αM )Hḣij + (1 + αT ) 2 + m2 hij = ΠTij , (7.40) a (7.37a) e bi = bi + a ζ̇ξi + ζ ξ̇i , 1 σ ei = σ i − ϵi jk ∂ j ξ k , 2 1 e hij = hij . h = h − ζξ, e hi = hi + ζξi , e 2 σ e = σ, (7.39b) 7.4. Tensor perturbations βei = βi ,  (7.39a) where ΦB , ΨB are the usual Bardeen potentials [573], δΦf is called the Mukhanov-Sasaki variable and δΦN is a Bardeen type variable built from a scalar field and the metric/tetrad DoF. Finally, it should be noted that in [562] there is an extensive and quite general perturbative framework for TG that one could follow. This framework was also applied to TEGR using gauge invariant variables. (7.36) 1 βe = β − ξ 0 − ξ 0 , a Φ̇ ψ, H δΦN = δΦ + a(−β + b)Φ̇ − a2 Φ̇ḣ,  We further split Y µ as Yµ = Y 0 , ζ(Y i + δ ij ∂ j Y) where ζ ∈ {1, a1 }, which incorporates both conventions (in terms of how Y µ can be split in the 3 + 1 decomposition), ∂ i Y i = 0 and Y the scalar part coming from splitting Y i φ e = φ − ξ˙0 , d d (a(b − β)) − (a2 ḣ), dt dt (7.37b) where αT := c2T − 1 is the tensor speed excess with cT := cg /c (here revert to SI units for clarity’s sake). This quantity describes the modification in the GW propagation speed. In addition, the Planck-mass run rate αM enters as a friction term and it is related to the cosmological strength of gravity M2∗ d log(M2 ) (the kinetic term of tensor perturbations) by αM := d log a∗ , m denotes the effective graviton mass and ΠTij is the tensor (7.37c) These transformations suggest the grouping {φ, ψ, β}, {b, h} and {bi , hi , σi } for the fields. This grouping is generated by gathering all fields that transform similarly. In this way, we can set to zero only one element from each group due to the similarity in their transformation. For example we 119 Rep. Prog. Phys. 86 (2023) 026901 Review the results only through the time derivative ḟT = fTT Ṫ + fTB Ḃ. Another very famous theory that enjoys the same functional part of the anisotropic stress (7.29d). For the case of GR, equation (7.40) is reproduced by setting αM = αT = m = 0. For the rest of the analysis, only cases where ΠTij = 0 are considered. The detection of the GW event GW170817 and the γray burst GRB 17 0817A place a strong constraint on the speed at which GW propagate, leading to the following constrain [575] −3 × 10−15 < cg − 1 < 7 × 10−16 . c ◦ picture in the sense of f (T), is f(R) gravity [42] since the GWs propagation at the speed of light and the the Planck mass run rate is given as αM = ḟR /(HfR ). These similarities were to be ◦ expected to some degree since both f (T) and f(R) are subclasses of f(T, B). The next major class of theories we consider is the BDLS theory, which encapsulates all possible theories constructed with up to second order derivatives of the tetrad and a scalar field (see section 5.8.2). Following the same procedure, one can find the gravitational propagation equation for the tensor perturbations as equation (7.40) by setting m2 = 0, then (7.41) In addition, observations for the mass of the graviton set an upper bound of mg < 1.2 × 10−22 eV/c2 from [576]. There is also a stronger constraint coming directly from Solar System tests from which we have the stronger upper bound mg < 10−23 eV/c2 [577, 578]. Hence suggesting that our current understanding of the graviton regarding its mass was not really changed from the GW170817 event [579]. In what follows we will present the results of two major classes of TG theories, namely f(T, B) (see section 5.4) and BDLS (Teleparallel analogue of Horndeski) gravity (see section 5.8.2) since these are the only results currently reported in the literature. In order to consider the tensor perturbations we only need the tensor part of the perturbed tetrad in equation (7.25) which is a δeI j = δ Ii hij . 2 αT = 1 ḟT . H fT  M2∗ = 2 G4 − 2XG4,X + XG5,φ − ϕ̇XHG5,X + 2XGTele,J8  1 + XGTele,J5 − GTele,T . (7.46) 2 (7.42) Let us stress that at this point the only nonvanishing contributing background scalars to the GTele term are T = −6H2 , Tvec = −9H2 , and I2 = 3Hϕ̇, while all the other scalars vanish up to first order. The fact that the GTele terms enter equations (7.45) and (7.46) means that in contrast to the standard Horndeski theory there is the potential of revised speed of waves instead of trivialization of most interesting classes of models. Hence one can still find a wide variety of modes that respect the light of speed propagation constraint by solving αT = 0. For the case of GTele = 0 we directly recover the standard results for the GW propagation equation [318]. Both of the above major classes correctly reproduce the most trivial result of TEGR and f (T) theories which corresponds to αM = αT = 0, i.e, (7.43) It should be noted here that ḟT = fTT Ṫ + fTB Ḃ. AS in GR, the propagation of the tensor waves in f(T, B) gravity is also the speed of light [582] c2T = 1, (7.45) and the effective Planck mass is given by Starting with the f(T, B) case, by inserting the above perturbation into the second order action in Fourier space (7.30), one can derive directly the GW propagation equation which is just equation (7.40) by setting αT = 0 and m2 = 0. From this equation, the stability condition fT < 0 is implied28 . The Planck mass run rate, representing a frictional term [574, 580, 581] turns out to be in this case αM = 2X  2G4,X − 2G5,ϕ − G5,X (ϕ̈ − ϕ̇H) M2∗  1 −2GTele,J8 − GTele,J5 , 2 (7.44) and it is in agreement with [575]. In this context, f(T, B) gravity is not that strongly constrained by present observations related to equation (7.41). However a full stability analysis of the perturbations needs to be performed, i.e. also including the scalar sector. Let us point out that exactly the same results hold true for the class of f (T) theories and hence one can deduce that the boundary term only influences the functional form of ḧij + 3Hḣij + k2 hij = 0, a2 (7.47) which is also identical to the result of GR. For other theories like the ones containing the teleparallel Gauss–Bonnet terms (see section 5.6), it is expected that cT 6= 1 as it was found for similar theories as in the case of the modified Gauss–Bonnet one [583]. This study has not been analyzed yet. Recently, two TG parity-violating models have found that some of those theories can produce a GW birefringence and dispersion meaning that the dispersion relation could depend on the wave number and their polarization [584, 585]. 28 The stability condition f T < 0 comes from the fact that the equation (7.40) with αT = 0 and m2 = 0 can[also be generated ] from an action analogous to ´ d3 k k2 3 ij ij the form dt(−a f ) ḣ ḣ + h h . This type of action in order to T ij 3/2 a2 ij (2π) be healthy (−a3 fT ) > 0 ⇒ fT < 0 needs to be imposed. 120 Rep. Prog. Phys. 86 (2023) 026901 Review f(T, Θ) gravity theories are reported since they are quite general including a wide range of theories. One can find more details about the derivations in the supplementary annexes (supplementary 2). 7.5. Vector perturbations In order to study the vector perturbations at linear order, the tetrad perturbation of equation (7.25) reduces to    A  0 aβi . δe µ = (7.48) δ I i bi aδ Ii ϵijk σ k The linearized field equations of the scalar perturbations for f(T, B) are given by 7.6.1. f(T, B) gravity. In the literature, so far, there have not been reported explicitly any theories with dynamical vector perturbations in flat FLRW. It is instructive to go through the result [158] of f(T, B) gravity B δfB − 6H2 δfT 2 2Hk2 fT 1 b + ψ̇(12HfT − 3ḟB ) − fT δT − 2 a 2k2 fT + 2 ψ + 6Hφ(2HfT − ḟB ), (7.53a) a W00 : κ2 δρ = 3Hδ ḟB + ◦ which also includes GR, f (T) and also f(R) as sub-cases. For that calculation the gauge can been fixed by setting hi ≡ 0 (see section 7.3 for details). Perturbing the field equations which we now denote as Wµν and for ḟB + ḟT 6= 0 (which is a theory ◦ different to f(R)) we get that βi = 0 and σi = 0. Hence one eventually arrives at ! ḟT . Wij (i 6= j) : 0 = ḃj + bj 2H + (7.49) fT Wij (i 6= j) : ψ − φ =  Wii : −κ2 δp = δ f̈B + δfB  k2 a2 + 1 (a(f˙T + ḟB )b − δfB ), fT 2k2 B + 3a2 2  − 2(3H2 + Ḣ)δfT 1 2k2 (ḟB + 3HfT + ḟT )b − 2Hδ ḟT − fT δT 3a 2 2k2 fT + ψ + 2fT ψ̈ + 2ψ̇(6HfT + ḟT ) + φ̇(2HfT − ḟB ) 3a2     2k2 fT + φ 4fT − 2 + 3H2 + Ḣ − 2f̈B + 4HḟT , 3a The above equation is just a constraint equation since only first order derivatives on time appear and hence the vector perturbations are not propagating. One can also read off the same stability condition fT < 0 as in the tensor case. It so happens that f (T) gravity [243] is again portrayed by the same functional picture as f(T, B). In the case of ḟB + ḟT ≡ 0 which means fT = −fB = −fR or in − (7.53c) ◦ other words, f(T, B) → f(R), all antisymmetric field equations are trivialized (W[µν] ≡ 0). By further introducing Yi := bi − βi , we end up with ! ḟR Wij (i 6= j) : 0 = Ẏj + Yj 2H + , (7.50) fR W0i : κ2 av(p + ρ) = δ ḟB − 3HδfB + 2fT ψ̇ − 2HδfT + (2fT H − ḟB )φ, + 2(ḟT + ḟB )ψ + (2fT H − ḟB )φ, (7.53e) which is a well known result for f(R) theories [42]. Notice again that this result has the exactly the same functional form as the one for f(T, B) above. Lastly, for the simplest case of TEGR one recovers the constraint equation 0 = ḃj + 2Hbj , (7.53d) Wi0 : κ2 av(p + ρ) = δ ḟB − HδfB + 2fT ψ̇ ◦ Wij (i 6= j) : (7.53b) where δfT = fTT δT + fTB δB and δfB = fBT δT + fBB δB, while the antisymmetric part of the field equations reads as Wi0 − W0i : 0 = H(δfT + δfB ) + ψ(f˙T + ḟB ), (7.54) and the energy-momentum conservation in the case of dust (for the general case, see the supplementary annexes (supplementary 2) is given by (7.51) and thus there are no propagating vectorial DoF which agrees with the literature. ◦ ρ 2 µ ∇µ Θ0 : δ ρ̇ + 3Hδρ = k v + 3ψ̇ρ, a 7.6. Scalar perturbations ◦ µ ∇µ Θi : av̇ + aHv = −φ. Focusing on the scalar sector which is the most involved, the analysis will be confined to the scalar part of equation (7.25) that reads    A  φ a∂i β  δe µ = , δ I i ∂ i b aδ Ii −ψδij + ∂i ∂j h + ϵijk ∂ k σ (7.52) (7.55a) (7.55b) Notice that σ completely drops out from the field equations, just like in f (T) gravity. On top of this, also the antisymetric part of the field equations is not trivialized (7.54), just like in f (T) [242, 243, 569]. In the limit of f(T, B) → f(−T + ◦ B) = f(R) where (fT → −fR , fB → fR ), one can recover after a few trivial substitutions the usual equations reported in [42]. In addition, in this limit the antisymmetric part of the field equations in equation (7.54) is trivialized and the scalar field in which the Newtonian gauge (b = β and h = 0) will be adopted. In the following, the field equations for f(T, B) and 121 Rep. Prog. Phys. 86 (2023) 026901 Review  ∂ 2 ΠS δp + 3 h = (1 + FT ) Hϕ̇ + 3H2 ϕ + 3Hψ̇ b completely drops off from the field equations29 . This is to κ2 2 Wii : ◦ be expected since for f(R) there are no antisymmetric field equations and there should be only 4 scalar fields in the equations [42].   1 +2Ḣφ + ψ̈ − a−2 ∂ 2 (ψ − φ) − FTT 3H2 δT + 2ḢδT 3    + Hδ Ṫ + 12H2 Ḣφ + FTΘ − 3H2 + Ḣ δΘ − Hδ Θ̇   δT +(ρ̇−3ṗ) +Hϕ +12H2 Ḣ (FTTT δT + FTTΘ δΘ) 12H   FΘ 2 −H (ρ̇−3ṗ) [FTTΘ δT + FTΘΘ δΘ]− δΘ− ∂ 2 π S , 4 3 (7.59c) For the case of f(T, Θ) gravity, as also introduced in section 5.9, we use the model ansatz f(T, Θ) = −T − F(T, Θ) for some arbitrary function F(T, Θ) which is formulated as TEGR plus a modification. The analysis follows from [586] where the field equations are first presented in the flat FLRW background (see equations (6.64a) and (6.64b)) 7.6.2. f(T, Θ) gravity. κ2 T + F FΘ ρ = (1 + FT ) 3H2 + + (ρ + p) , 2 4 2 (7.56a) W0i : κ2 − FΘ (p + ρ) = (1 + FT ) Ḣ − 12H2 ḢFTT − 2 + H (ρ̇ − 3ṗ) FTΘ , (7.56b) where we have used Θ = ρ − 3p and ḟT = ḞT = FTT Ṫ + FTΘ Θ̇ = −12HḢFTT + (ρ̇ − 3ṗ)FTΘ . Combining these equations (or using equation (5.131)) yields the modified continuity equation  2 κ − FΘ     a2 (ρ + p) κ2 − FΘ ∂i v = (1 + FT ) ∂i ψ̇ + Hφ 2 − H (FTT ∂i δT + FTΘ ∂i δΘ) , (7.59d) Wi0 : −  a2 (ρ + p) κ2 − FΘ ∂i v 2   = − (1 + FT ) ∂i ψ̇ + Hφ   − ∂i ψ −12HḢFTT + (ρ̇ − 3ṗ) FTΘ ,  F [ρ̇ + 3H (ρ + p)] = Θ ρ̇ − ṗ + (ρ + p) 2   × −12HḢFTΘ + (ρ̇ − 3ṗ) FΘΘ . (7.59e) and the antisymmetric part Wi0 + W0i : H (FTT ∂i δT + FTΘ ∂i δΘ)   = −∂i ψ −12HḢFTT + (ρ̇ − 3ṗ) FTΘ , (7.57) Special care is needed in the presence of the trace of the energy momentum-tensor since it modifies the continuity equation, which leads to the non-conservation of the stressenergy tensor. Nevertheless we impose the conservation as a constraint in order to limit the possible functional form. This constraints reads as     ρ̇ 0 = FΘ (ṗ − ρ̇) + 6(ρ + p) 4HḢFTΘ + ṗ − FΘΘ . 3 (7.58) (7.60) where WAρ denotes the linearized field equations, δT can be found in the supplementary annexes (supplementary 2) and δΘ = δρ − 3 (p + δp) − ∂ 2 ΠS . These equations properly reproduce the results reported in [323, 569]. Furthermore, the scalar perturbation σ does not appear in the field equations, leaving its effect to be negligible [243, 569]. In addition to the field equations, we also need to perturb the conservation laws for the stress-energy tensor and the velocity perturbation. Thus, the perturbed continuity equation is     ∂ 2 πS µ 2 Θ : κ − F δ ρ̇ + 3H δρ + δ p + ∇µ 0 Θ 3  − 3 (ρ + p) ψ̇ + (ρ + p) ∂ 2 v ◦ On the other hand, the perturbed field equations of equation (5.53), in the Newtonian gauge, read a.s. W00 :  h i κ2 δρ = (1 + FT ) a−2 ∂ 2 ψ − 3Hψ̇ − 3H2 φ 2    FΘ  2 + 3H FTT δT + FTΘ δΘ + 3δρ − δp − ∂ 2 ΠS 4   ρ+p FTΘ δT + FΘΘ δΘ , (7.59a) + 2   Wij (i 6= j) : κ2 − FΘ ∂j ∂ i ΠS = a−1 ∂j ∂ i β −12HḢFTT + (ρ̇ − 3ṗ) FTΘ ] =  FΘ  δ ρ̇ − δ ṗ − ∂ 2 π˙S 2 h  + FTΘ −2a−2 ∂ 2 ψ (ρ̇ − 3ṗ) + (ρ + p) 3Hδ T + δ Ṫ   i −12HḢ (δρ+δ p)+ 6H Hφ+ ψ̇ (ρ̇− 3ṗ)+(ρ̇+ ṗ) δ T    1 + FΘΘ (ρ + p) 3HδΘ + Θ̇ + 3δρ − δ p − ∂ 2 π S 2 × (ρ̇ − 3ṗ) + (ρ̇ + ṗ) δΘ] − 12HḢ (ρ + p) (FTTΘ δ T + a−2 (1 + fT ) ∂j ∂ i (ψ − φ) , (7.59b) +FTΘΘ δΘ) + (ρ + p) (ρ̇ − 3ṗ) (FTΘΘ δ T + FΘΘΘ δΘ) . (7.61) and the linearized conservation of momentum equation equation 29 After expanding δT, δB, δf and δf as prescribed in supplementary T B annexes (supplementary 2). 122 Rep. Prog. Phys. 86 (2023) 026901 Review  2  ◦ 1 µ 2 S ∇µ Θi : − FΘ a (ρ̇ − ṗ) ∂i v + ∂i δρ − δp − ∂ π 2    = κ2 − FΘ (ρ + p) a2 ∂i v̇ + 2a2 H∂i v + ∂i φ +a2 ṗ∂i v + ∂i δp + ∂i ∂ 2 π S . critical one Ωm = Ωm0 DE EoS as Geff = GQ, d ln Ωm 1 . 3(1 − Ωm ) d ln(1 + z) (7.66) where Q = 2 + 4Ω2m , 3 + 3Ω2m (7.67) where we can recover δ m from f. Notice that f = f0 (with f 0 constant) if δm ∝ a f0 . In particular, f → 1 for the ΛCDM case and for large z, while for f = 1 we obtain the Einstein-de Sitter Universe. As is standard, we can derive the growth of perturbations by parameterizing f in terms of the DE related quantities as f = Ωm (z)γ(z) . (7.69) We note that in this case it is important to study whether γ can be exactly constant or not. In such cases, the the modified evolution of matter perturbations assumes the form of equation (7.76) which is δ̈m + 2Hδ̇m − 4πρGeff δm = 0, (7.70) where Geff , as the effective gravitational constant which depends on the choice of the theory. As an example, for effectively massless scalar-tensor models Geff varies wrt t. The galaxies and quasars observed in spectroscopic observations are biased tracers of such structure formation, leading to degeneracy between the amplitude of matter fluctuations and biasing parameters. This issue could be a consequence of the complicated calculation of the growth function from the isotropic power spectrum derived from clustering of cosmic sources. However, spectroscopic observations contain extra information about the velocity field arising from gravitational collapse by separately measuring the power spectrum along and perpendicular to the line-of-sight. Measurements of the velocity field can help to differentiate between the effect of DE and modified gravity as the source of the accelerating Universe through measurements of Redshift-Space Distortions (RSD) [588]. These distortions emerge due to the gravitational pull of matter over-densities which causes velocity deviations from the smooth Hubble flow. These peculiar velocities are imprinted in galaxy redshift surveys in which recessional velocity is used as the line-of-sight coordinate for galaxy positions. This in turn leads us to an apparent compression of radial clustering. (7.63) in the matter component in the Universe with DE components (we neglect radiation at the matter and DE dominated stages). Inside the Hubble radius dynamics of δ m is given by (7.64) In the absence of spatial curvature, the evolution of the H as a function of the redshift z = aa0 − 1 at z  zeq reads   ˆ z ′ ′ 1+wDE (z ) h (z) = Ωm0 (1+z) +(1−Ωm0 ) exp 3 dz , 1 + z′ 0 (7.65) 2 , we can write the effective which allows the matter perturbation equation (7.64) to retain it form while absorbing modified terms into this effective parameter. While we can work with δ m , it is convenient to use the growth function f := ddlnlnδam and the DE EoS (7.66) in order to arrive at the non-linear equation   df d ln Ωm 3 1 2 1− f = Ωm , (7.68) +f + d ln a 2 d ln a 2 A viable way to carefully explore the phenomenology of proposed models is to compare with observations. To achieve this goal, we need to find tools that can help efficiently in discriminating between different classes of models. The growth index γ (see equation (7.69)), which describes a way of parameterizing the growth of density perturbations in a nonrelativistic matter component is a good example of such a tool. In recent years, the parameter γ has been used to discriminate spatially open from spatially flat universes. As an extension, γ has been used in the context of DE in order to study deviations from standard GR models. The reason why this parameter is important is due the fact that it has a clear quasi-constant signature at very low redshifts. At this point, it is important to study (theoretically) in which models the growth index γ can be exactly constant, γ ≈ 3(w − 1)/(6w − 5) which is GR plus DE fluid or γ ≈ 6/11 in the case of ΛCDM. Also some indirect values for modified gravity are reported in [587]. Hence, the value of γ offers a way to distinguish between modified gravity models. Let us start by considering the evolution of linear scalar perturbations δ m as δ̈m + 2Hδ̇m − 4πGρδm = 0. −1 In this representation, the effective Newton’s parameter Geff can be parametrized as 7.7. Growth index and the fσ8 parameter δρ + 3HV, ρ a 2 h a30 wDE = − (7.62) The scalar field σ completely drops out from the field equations and in addition a non-trivial antisymmetric part of the field equations is obtained (7.60). This is also case in f(T, B) gravity (7.6.1). In this instance the scalar perturbations are coupled with the perturbations of the energy-momentum components and so this is not enough information to determine the impact of these cosmological perturbations on observational parameters. δm :=  3 3 with h(z) := HH0 and wDE (z) := pDE (z)/ρDE (z). This equation will hold for all FLRW models inside GR. Now, if we consider that the relative density of matter component in terms of the 123 Rep. Prog. Phys. 86 (2023) 026901 Review The resulting anisotropy in the clustering of galaxies is correlated with the speed at which structure grows; deviations from GR which cause slower or faster growth result in smaller or larger anisotropic distortions in the observed RS clustering. The rate of change of the amplitude of the clustering is given by fσ8 (a) = dσ8 (a)/d ln a, where a = 1/(1 + z) and f ≡ f(a) has been defined in equation (7.69). Because RSD measurements are sensitive to the product of the growth rate and the amplitude of matter fluctuations, a wide range in redshift coverage is important to constrain the evolution in clustering amplitude and directly probe GR. On the other hand, if one were to assume a ΛCDM model where GR properly explains gravitational collapse, the growth rate can be predicted to high precision and RSD results can be used to constrain σ8 (a), thereby provide insight into other fundamental physics such as neutrino masses. redshifts distortions reported by Planck 2015 Collaboration in a sample detailed in [593]. Using machine learning techniques, as will be further introduced in section 10.4.2 and detailed in [594] the role of the growth data was studied by reconstructing observations from RSD fσ8 data together with the Hubble data that comes from cosmic chronometer and SN type Ia data. The advantage of this reconstruction, is that it acquires more simulated data points in order to constrain modified gravity models with better precision. In addition, in [595, 596] a f (T) power law model was constrained using the CMB temperature power spectrum from early-time estimations that includes baryonic acoustic oscillations and also, local Hubble constant measurements. By including this analysis, it was possible to set a playground where the evolution of tensor modes in f (T) gravity could influence the power spectrum and the B-modes polarization. In [595, 596], the power model was further constrained by considering GW multi-messenger astronomy, especially on the effects of primordial GW on the CMB anisotropies and the BB spectrum. It was shown that only an amplitude modification could be compared with the standard ΛCDM cosmology, i.e the more the parameter n in the model deviates from GR, the larger the GW amplitude decay is compared to the canonical ΛCDM. Additionally, one can constrain the growth factor by including detections from neutron stars mergers that include an EM counterpart. In [568], an analysis of the scalar cosmological perturbations in f (T) gravity was presented. In particular, they chose to investigate a power-law model scenario where it is possible to find constraints on the growth matter factor in sub-horizon scales around O(100 Mpc). This is illustrated by the field equation for the perturbation of the additional scalar field of the theory, δϕ read as at first order perturbations when f (T) is fixed to a constant. In [381], in comparison to [568], a reconstruction of f (T) was considered as a quintessence model, where it was noticed that strong deviation of the matter over density evolution at small scales and weak otherwise. In this direction, it was shown that at linear perturbative order it is possible to break the degeneracy between a family of f (T) models. As an extension of the aforementioned work on f (T), this quintessence analysis showed some deviations for large scales. According to equation (7.69), it is possible to analyze cases where the growth index γ can be parameterized. In such cases, as in [589], a parameterization was given in the form γ(z) = γ0 + γ1 1+z z in order to study if a power law model can be constrained using observations. Including a new scalar field, αm in the linear equations, as in [590], could contribute further to the anisotropic effects between the gravitational potentials ψ and φ. This new scalar degree of freedom vanishes in the sub-horizon scales of f (T) gravity and it is constrained with observations related to the flux power spectrum of the Lyman-alpha forest in quasar absorption spectra. According to this analysis it was found that there is a faster growth rate if the tracker solution is followed by the usual quintessence-like evolution. While proposing f (T) power law models can help avoiding degeneracies at linear order, in [591], it was found that an asymptotic form of this particular f (T) model can be given by γ ≈ 11−6 6b . This served a generalization to the described analyses in the regime where γ can vary with redshift. This allows for an accurate determination of b and also the ability of testing it in the range of validity of the observations. Furthermore, in [592], the evolution of the weighted growth rate, denoted by fσ8 (a), was investigated for several different values of the parameter n in the power law model. The results were based in observations derived from 7.7.1. f(T) gravity. 7.7.2. f(T, B) gravity. Considering dust for the perfect fluid, and following [42, 597], the variable V := av is introduced and along with the gauge invariant variable dubbed density contrast δ m as in equation (7.63). In order to determine the time derivative of this parameter, we need to utilize the continuity equation to obtain the density parameter time derivative, which is δ ρ̇ + 3Hδρ = k2 ρV + 3ρψ̇. a2 (7.71) The time derivative of the density contrast parameter can then be written as ◦ d ∇2 V δ̇m = − 2 + 3ψ̇ + 3 (HV), a dt (7.72a) V̇ = −φ, (7.72b) where the time derivative of V is also presented. By combining both derivatives, we obtain ◦ δ̈m + 2Hδ̇m = d2 d ∇2 φ + 3ψ̈ + 3 2 (HV) + 6Hψ̇ + 6H (HV). 2 a dt dt (7.73) In the sub-horizon approximation k  aH, k being well inside the Hubble radius, the dominant terms are k and δρ. Now that we have all the prerequisites we need to proceed, let us first summarize the dominant terms in this limit 124 Rep. Prog. Phys. 86 (2023) 026901 Review Table 15. Summarizing the cases of all the subclasses of f(T, B) in the sub-horizon approximation (recalling that definitions for A2 , A6 , ∆4 , ∆10 are in appendix B). For more details see [158]. Σ Class Conditions Models Geff 1 1a {Π ̸= 0, Υ ̸= 0} Ξ ̸= 0 General f(T, B) Υ −G 36H2 ( f fTT4+ 2Ξ)+3ΥfT − ΥfT +12H2 (Υf fTT +2Ξ) 1b Ξ=0 Includes f (T) G AA26 ∆4 = − AA26 ∆10 2 2a {Π ̸= 0, Υ ≡ 0} Ξ ̸= 0 Less general f(T, B) 2b Ξ=0 Only f(R) 4 −G 3( f +12H 2f ) TB  T 1 4 G 3fR + k2 BB ◦ BB  3(−fR +3 2 fRR ) 1 − fT +12H 2f TB Σ ∼ f1R a  2  k k2 k2 k2 k2 |φ|, |ψ|, |β|, |δf |, |δf | T B a2 a2 a2 a2 a2  2  H |φ|, H2 |ψ|, H2 |β|, H2 |δfT |, H2 |δfB | , 2.a {Π 6= const, Υ = 0, Ξ 6= 0} ◦ 2.b {Π = const, Υ = 0, Ξ = 0} the unique f(R) case (7.74) These branches may also be indicators of variable DoF, ◦ and ˙ ≲ |HX| |X| n since we know for sure that f(R) has 3 DoF and f (T) has either 3 or 5 DoF [145, 154, 246, 556] as also discussed in section 5.3. The classification induced by these branches is illustrated in table 15 and the relevant coefficients are included in the appendix B. o where X ∈ φ, ψ, β, δfT , δfB , φ̇, ψ̇, β̇, δ f˙T , δ f˙B . (7.75) Thus, it follows directly that in Fourier space (7.30) of the subhorizon limit of equation (7.73) that it reduces to (7.70) as κ2 k2 φ δ̈m + 2Hδ̇m ≃ − 2 = 4πρGeff δm = ρGeff δm , a 2 Investigating the growth formation for epochs well within the matter domination periods, we set the pressure components p = δp = 0 along with the absence of anisotropic stress. In addition, models that obey the conservation of the stress-energy tensor will be assumed. Even when the models do not obey the conservation law, the result reported in [586] is still obtained. Also, the conclusions regarding the resulting growth evolution remain the same. Following the procedure of section 7.7.2, and combining equations (7.59e) and (7.59d), the relation   FTT 12Hψ̇+12H2 ϕ−12Ḣψ−4a−2 k2 ξ = −FTΘ (δρ−3ρψ) , 7.7.3. f(T, Θ) gravity. (7.76) from which it follows that the only contributing scalar is φ. Along a similar vein, Σdef is a parameter sensitive to weak lensing which appears when we write the lensing potential − (φ + ψ) in terms of the matter density contrast δ m , so Σdef plays a similar role to Geff but between the lensing potential and δ m specifically. This parameter is defined as   1 Geff ψ Σ := 1+ , (7.77) 2 G φ (7.79) which is calculated in along with Geff in the sub-horizon approximation. Due to the nature of the f(T, B) there is branching in various sectors by solving the field equations (7.53a)– (7.54). By introducing the useful notation Π := fB + fT , Υ := fBB + 2fTB + fTT = ΠT + ΠB , is obtained where we define ξ := aHβ. Here, we have eliminated φ using equation (7.59b), where we have obtained φ=ψ− (7.78a)  3ξ 4ḢFTT + ρFTΘ , 1 + FT (7.80) which, then combined with equation (7.79) results in (7.78b) 4ξFTT Ξ := f2TB − fTT fBB = −ΠT ΠB + fTB Υ, h  i ρFTΘ δm − 3Ha2 v − 3ψ + FTT 12Hψ̇ + 12ψ H2 − Ḣ , =  k2 9H2 a2 + 1+FT 4ḢFTT + ρFTΘ (7.78c) induced branches are classified as: (7.81) 1. {Π 6= const, Υ 6= 0} Which can further be classified using Ξ = −ΠT ΠB + fTB Υ 1.a {Π 6= const, Υ 6= 0, Ξ 6= 0} most general case of f(T, B) 1.b {Π 6= const, Υ 6= 0, Ξ = 0} includes f (T) 2. {Π 6= const, Υ ≡ 0} Which can further be classified using Υ ≡ 0 ⇒ ΠB ≡ −ΠT into equation (7.78c) as Ξ = Π2T = Π2B which in the in the sub-horizon approximation assumes the form  a2 n 4ξFTT ≈ 2 ρFTΘ δm − 3Ha2 v − 3ψ k h io +FTT 12Hψ̇ + 12ψ H2 − Ḣ . (7.82) 125 Rep. Prog. Phys. 86 (2023) 026901 Review 2 It is straightforward to check that the limits A = κ2 and F(T, Θ) = 0 are recovered as expected. Usually, at this stage, one can read off some Geff but this is rather not the case since we have a modified continuity equation as can be seen from equations (7.61) and (7.62). Regarding its solubility, this system of equations would yield two different solutions for δ m dubbed the growing and decaying modes. However, in most cases the interest lies in the formation of structure and thus only the growing solution is probed. A very similar model from the curvature-based gravity framework with very similar Approximating by leading order, we obtain ξ∼  a2 H 2 δm − a2 Hv + ψ , 2 k (7.83) which can be further reduced using equation equation (7.62) as  FΘ φ v̇ + 2Hv + 2 (κ2 − FΘ ) = − 2 δm , (7.84) a 2a ◦ and from equation (7.80) ξ∼ a2 H 2 (φ + ψ)  ψ. k2 behavior is f(R, Θ) [598]. Henceforth, for the above reasons, extra caution while using first order perturbations is needed. The use of the correct perturbed tetrad is quite straightforward at this stage. However, in general one should carry out a full Hamiltonian analysis to fully understand if strong coupling appears in a theory. Then try to probe a specific background in order to extract information about linear perturbations and their dynamics. (7.85) Hence, we find that φ ≃ ψ. We note that in this analysis we implicitly imposed FΘ , FTΘ , FTT 6= 0 but nevertheless the conclusion φ ≃ ψ remains the same for the two gravitational potentials. This is important since these potentials play a dominant role in the calculation of the deflection parameter (7.77). Using the sub-horizon limit approximation, one can then proceed to derive the evolution equation for the gaugeinvariant fractional overdensity δ m . Using equations (7.59a), (7.59d) and the definition of δ m in equation (7.63) As closing remarks we would like to point out that in most of these works, and in general, one needs to be very careful considering (the scalar) perturbations. There are actually two three major reasons for that, namely: 7.7.4. Closing remarks.   2 κ 3FΘ ρFΘΘ − − δm 2 4 2 h   i k2 ψ 1 n = (1 + FT ) 2 + FTΘ 12H ψ̇ + Hφ − 4a−1 Hk2 w a ρ 2 o 3F Θ 2 a Hv, +3Ha2 vρFΘΘ + (7.86) 4 (a) A correct choice of perturbed tetrad—In most of the references discussed in section 7.7 a wrong perturbed tetrad was used that only assumed 13 DoF. The missing DoF is traced back to the antisymmetric part of the perturbation of the tetrad which is described by the scalar field field σ and the vector field σ i . The tetrad is neither symmetric, like the metric tensor, nor antisymmetric and thus it must assume 16 DoF in general, thus not including the antisymmetric contributions one is left with 3 DoF less. This is illustrated in a more clear way via the full spacetime indexed perturbation of the tetrad dubbed as τµν defined in equation (7.27). Having said that, it was proven in [158, 242], that the pseudo-scalar field σ is not dynamical for neither f (T) nor f(T, B) in a flat FLRW background. Once one extends these theories, by adding more scalar fields or more contributions to the actions, then a full analysis using the correct perturbed tetrad equation (7.25) must be performed. (b) Overfixing the gauge—One needs to be very careful first to use the correct perturbed tetrad assuming the correct number of DoF as discussed in (a) and then fix the gauge properly as discussed in section 7.3. In some instances but not properly understanding the gauge transformation of the fields one can be mislead in overfixing the gauge thus artificially removing DoF which lead to a false result in principle. (c) The strong coupling issue—The next and maybe the most important reason is related to the strong coupling issue in modified theories like f (T) seem to be haunted from, as also discussed in section 5.3.1. This phenomenon is directly linked with the existence of some (scalar) DoF which in general is propagating but in specific (highly symmetric) backgrounds becomes non-dynamical, at least it seems 2 which in conjunction with ψ ≃ φ and ak2 Hψ2  ψ reduces to Aδm = (1 + FT ) 3FΘ 2 k2 ψ 3Ha2 v + ρFΘΘ + a Hv, a2 ρ 2 4 (7.87) 2 where A := κ2 − 34 FΘ − 12 FΘΘ ρ. In addition, From equation (7.59e), in the sub-horizon limit, we obtain the relation Hφ ∼ ρa2 v which can be further used to simplify (7.87) as Aδm = (1 + FT ) k2 ψ , a2 ρ (7.88) from which one generates the generalized matter density equation as  1 aAH ′ + 2A ′ aH + 3AH aAH   3AH 4H ′ aHFTΘ + ρFΘΘ δm′ (a) − 2 κ − FΘ  1 3A ′ aH2 + 2 2 a2 H2 A ′ ′ + 3A ′ aH2 + A ′ a2 HH ′ − 2 Aa H κ − FΘ  2 2  κ − FΘ k FΘ Aρ − δm (a) = 0. × 4H ′ aHFTΘ + ρFΘΘ − 2(1 + fT ) 4a2 δm′ ′ (a) + (7.89) 126 Rep. Prog. Phys. 86 (2023) 026901 Review that way in first order perturbations. If one goes beyond first order perturbations then eventually one will find these missing DoF in higher perturbative orders. Hence if there are hidden DoF in higher perturbative orders, the linear order perturbations might not be dominant anymore and a full analysis of all orders until the missing DoF is located will be needed. This is directly linked to the fact that linear perturbations are not sensitive to non-linear velocity constraints in general. Finally, we stress that, specifically for f (T) this is a problem of choosing a proper background solution, not a problem of the actual theory [154]. Further, we do not know if this problem will also appear in other modified TG different to f (T) gravity. For a GW incident toward the Earth, the polarization content of the GW determines the relative amplitudes that are measured at different detector locations on the Earth’s surface, modulated also by each detectors’ antenna pattern [174]. The current ground-based advanced interferometer network— comprising LIGO Hanford, LIGO Livingston and Virgo—is therefore not yet sufficient to fully reconstruct all the GW polarization of detected signals [602]. However, the existing three-detector network can already be used to compare the evidence for some specific subsets of all the possible polarization combinations [604]. In [605–607], the LIGO and Virgo collaborations carried out a Bayesian model comparison between some extreme alternative hypotheses—specifically, comparing full tensor with full vector or full scalar—for the polarization content of the GW signals detected from three events in the second LIGO-Virgo Observing Run: GW170814, GW170817 and GW170818. The first and third events were binary black hole mergers while GW170817 was a binary neutron star merger with an explicitly identified electromagnetic counterpart in its host galaxy NGC4993. Consequently the constraints obtained from these analyses were strongest for GW170817, which favored the full tensor hypothesis over both full vector and full scalar—in each case with a log Bayes factor greater than 20. For the two binary black hole mergers also, however, the full tensor hypothesis was also modestly favored. To further clarify, therefore, this means GWs are highly more like to have only tensor polarization modes compared to only vector or only scalar ones. However, we do not yet have any information about a mixture of polarizations such as tensor and scalar modes or vector and scalar modes etc. Hence we cannot yet exclude theories that predict such mixtures of polarization pairs. More recently, the LIGO and Virgo collaborations have carried out a further Bayesian analysis that again compares full tensor with full vector and full scalar hypotheses, but without assuming a specific waveform model. (The analyses carried out in [605–607] had adopted GR template waveforms). This analysis was applied in [607] to a subset of the GW detections reported in the GWTC-2 catalog, which included events observed during the first part of the 3rd LIGO-Virgo Observing Run, denoted O3a [608]. Again, none of the mergers studied showed significant evidence favoring a non-GR hypothesis. Notwithstanding the limitations of the current three detector global network, the near-future prospects for conducting high-precision polarization tests of general metric theories of gravity are very good. By the late 2020s, the Advanced LIGO detectors will have undergone significant further improvements, enhancing their sensitivity to the socalled ‘A+’ configuration: a mid-scale, incremental upgrade to Advanced LIGO that is scheduled to happen between 2026– 2028 [609]. This should increase their reach by a factor of several. Similar enhancements to the Advanced Virgo detector [610] are anticipated on the same timescale, and more importantly by that time the KAGRA [611] and LIGO India [612] detectors will also have joined the global network, with LIGO India planned to begin operations at the ‘A+’ configuration. Hence we can expect that by the end of this decade an extended Henceforth, for the above reasons, extra caution while using first order perturbations is needed. The use of the correct perturbed tetrad and avoid over-fixing the gauge is quite straightforward at this stage. Regarding the strong coupling issue, it would be wise to first make sure that a proper Hamiltonian analysis of the action has been carried out and the number of DoF is known. Then try to probe a specific background solution and its perturbations. Although it is expected that f (T) is strongly coupled for the available background solutions used so far, it should be stressed that this is purely due to the nonlinearity f (T) introduces. In the case of −T + Λ for example there is no such issue since the dependence on T is linear. 8. Polarization of GWs in TG In this section, the polarization status of some major classes of TG theories will be reviewed. The NGR will be the starting point for calculating the polarizations by utilizing the Newman–Penrose formalism. In this case it will be evident that the polarizations depend heavily on the parameters of NGR. Subsequently, the case of f(T, B) gravity will be illustrated where the polarizations are calculated with a more traditional method using the geodesic deviation equation. This case differs significantly from NGR since there is a massive scalar on top of the tensor modes. Finally, the polarization modes of the BDLS theory, the most general scalar-tensor theory built on TG, will be exhibited. For this case there is yet another distinct method, of calculating the polarization modes, used by employing the SVT decomposition. First the number of DoF are calculated and then the various results are applied for the calculation of polarizations. 8.1. Polarization modes in metric theories of gravity In metric theories of gravity, due to the nature of the Riemann tensor, only six GW polarizations are allowed [6, 599, 600]. These six polarizations can be also classified in terms of their helicity content as two tensor (helicity ±2) modes plus (+) and cross (×), two vector (helicity ±1) modes called x and y and two scalar (helicity 0) modes named breathing and longitudinal modes. This classification is illustrated in figure 25. In general, we can explore the polarizations of the GWs by measuring their amplitudes with GW detectors [174, 601–603]. 127 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 25. All possible polarizations of GW travelling in z-direction, starting with the scalar modes breathing, longitudinal, the vector modes x, y, and the tensor modes +, ×. The GW deforms a sphere of freely falling test particles. The symbols → and ⊙ denote propagation parallel and perpendicular to the paper plane accordingly. ‘Big Five’ global network of advanced interferometers will be fully operational, with significantly higher event rates and improved sky localization capability [613]. Data from these five detectors will be able to fully characterize the polarization content of GW sources, significantly reducing or eliminating degeneracies, and hence will provide an excellent means for testing the predictions of teleparallel theories of gravity. The six aforementioned polarizations are directly linked with the symmetries of the electric components of the Newman–Penrose variables can also be classified wrt their helicity states through ◦ ◦ (8.1)  R0i0j =  ◦ 2 (<Ψ4 + Φ22 ) 1 2 =Ψ4 −2<Ψ3 1 2 =Ψ4 − 12 (<Ψ4 − Φ22 ) 2=Ψ3 −2<Ψ3 Φ22 : s = 0, Ψ3 : s = −1, Ψ3 : s = 1, Ψ4 : s = −2, Ψ4 : s = 2, l = ∂0 + ∂3 , 1 n = (∂0 − ∂3 ), 2 (8.3) 1 m = √ (∂1 + i∂2 ), 2 1 m̄ = √ (∂1 − i∂2 ), 2 where dots represent coordinate time derivatives, (t, x, y, z) = (0, 1, 2, 3), i = {1, 2, 3} and xj = (x, y, z). Since in TG we mostly assume that matter is coupled to the metric (and the Levi-Civita connection), the force-like equation which describes the motion of particles in TG coincides with the geodesic equation (see section 3.2). This means that the information related to the polarizations would be only encoded in the Riemann tensor (as in GR). The tetrad contains more DoF than the metric, but the extra ones cannot be directly measured due to the matter coupling prescription adopted. On the other hand, the extra information that could carry the tetrad could be found by indirect GW observations. Restricting to massless GWs there is the E(2) classification [174, 599, 600], that utilizes the Newman–Penrose variables in order to facilitate the calculation of the polarizations in a given gravitational theory. Within this framework the six components of the electric Riemann tensor can be represented as  1 : s = 0, where the overbar denotes complex conjugation. From equation (8.3) we can deduce that Φ22 , Ψ2 are related to scalar DoF, Ψ3 is related to vectorial DoF and Ψ4 is related to tensorial DoF. Finally, we can visualize this parametrization of equation (8.2) in figure 25. We will use the notation of [614] and denote the basis vectors by lµ , nµ , mµ , m̄µ as Riemann tensor denoted as Ri0j0 , which assumes strictly only 6 independent components due to its symmetry. These components are responsible for the evolution of the geodesic deviation equation [66] ẍi = −Ri0j0 xj , Ψ2 (8.4) and we consider a plane wave propagating in the positive x3 direction, which corresponds to a single Fourier mode (7.30). The wave covector then takes the form kµ = −ωlµ , where ω characterizes the norm of the wave covector, and any generic perturbation pµν will be expanded as pµν = Pµν eiωu , (8.5) where we introduced the retarded time u = x0 − x3 and the wave amplitude is denoted Pµν . As shown in [600], the Riemann tensor of a null plane wave is determined completely by the six so-called electric components. For the wave in equation (8.5), these can be written in the Newman–Penrose basis as 1 1◦ Ψ2 = − Rnlnl = ḧll , 6 12 1◦ 1◦ 1 1 Ψ3 = − Rnlnm̄ = − Rnlnm = ḧlm̄ = ḧlm , 2 2 4 4   2=Ψ3  . −6Ψ2 (8.2) ◦ ◦ 1 1 Ψ4 = −Rnm̄nm̄ = −Rnmnm = ḧm̄m̄ = ḧmm , 2 2 ◦ 1 Φ22 = −Rnmnm̄ = ḧmm̄ , 2 where Φ22 , Ψ2 , Ψ3 , Ψ4 are Newman–Penrose variables, < represents the real part and = the imaginary one. These 128 (8.6a) (8.6b) Rep. Prog. Phys. 86 (2023) 026901 Review Table 16. E(2) classes as labeled by the Petrov type of its nonvanishing Weyl tensor and the maximum number of nonvanishing amplitudes as seen by any observer. These classes are independent of observer [614]. Class Conditions Description II6 Ψ2 ̸= 0 III5 Ψ2 = 0, Ψ3 ̸= 0 N3 N2 O1 O0 Ψ2 = 0, Ψ3 = 0, Φ22 ̸= 0, Ψ4 ̸= 0 Ψ2 = 0, Ψ3 = 0, Φ22 = 0, Ψ4 ̸= 0 Ψ2 = 0, Ψ3 = 0, Φ22 ̸= 0, Ψ4 = 0 Ψ2 = Ψ3 = Φ22 = Ψ4 = 0 All standard observers measure the same value for Ψ2 , but disagree on the presence or absence of all other modes. All standard observers agree on the absence of Ψ2 and on the presence of Ψ3 , but disagree on the presence or absence of Ψ4 and Φ22 . Presence or absence of all modes is observer-independent. Independent of observer. Independent of observer. Independent of observer: no wave. Λj := −bj + 2ḣj . where dots denote derivatives wrt the retarded time u. This representation of Newman–Penrose variables serves as the standard way of calculating the polarization content of massless GW. However, note the amplitudes in equation (8.3) are not observer-independent quantities. For example, if in one frame Ψ4 6= 0 but Ψ2 6= 0, then there exists a frame in which Ψ4 = 0. Hence, the absence or presence of the components of various helicities depends upon the frame chosen. Nonetheless, there are still some frame invariant statements for the amplitudes which include a set of quasi-Lorentz invariant classes of GW. Each class is labeled by the Petrov type of its nonvanishing Weyl tensor and the maximum number of nonvanishing amplitudes as seen by any observer. These labels observer independent quantities. These classes are illustrated in table 16. This is exactly the so called E(2) classification for GWs and each of these classes of a particular metric theory is defined to be the class of its most general wave. More details also including almost null waves can be found in [614]. We present yet another way to probe the polarizations via the use of the SVT decomposition as introduced in section 7.2. The idea is to use the SVT decomposition of the tetrad (and consequently the metric) on Minkowski spacetime and then calculate the electric components of the Riemann tensor [615], (8.8e) The representation of the electric components of the Riemann tensor in equation (8.7) actually holds for any case, irrespectively if the GW is massless or massive and hence is more flexible. 8.2. Polarization modes in NGR We review the analysis of polarizations of GW in NGR, introduced in section 5.2, as developed in [616, 617]. The Lagrangian of NGR [34] reads as  L = e c1 Tρ µν Tρ µν + c2 Tρ µν Tνµ ρ + c3 Tρ µρ T σµ σ = eGαβ µνρσ T α µν Tβ ρσ , (8.9) where the three real parameters c1 , c2 and c3 define different NGR theories and the supermetric [144] or constitutive tensor [125, 130] [µ σ] Gαβ µνρσ = c1 gαβ gρ[µ gν]σ − c2 δβ gν][ρ δασ] − c3 δα[µ gν][ρ δβ , (8.10) which serves in compactifying the calculations later on. ◦ R0i0j which reads as We start by linearizing the field equations by introducing the perturbations of the tetrad, inverse tetrad and the Lorentz transformation defining the spin connection 8.2.1. Linearized regime and field equations. ◦ R0i0j   = ψ̈ − 12 ḧ+ − 12 ḧ× − 12 ḧ× ψ̈ + 12 ḧ+ − 12 ik(β̇1 + Λ̇1 ) − 12 ik(β̇2 + Λ̇2 ) − 21 ik(β̇1 + Λ̇1 )   − 21 ik(β̇2 + Λ̇2 )  , ψ̈ − k2 (χ̇ + Φ) (8.7) where we introduced the gauge invariant variables χ := b − ḣ, (8.8a) Φ := φ − β̇, (8.8b) Σj := hj k2 + iεjlp kl σ p , (8.8c) Ξj := ik2 bj − 2εjlp kl σ̇ p , (8.8d) e A µ = δµA + ϵ u A µ , (8.11a) EA µ = δAµ + ϵ vA µ , (8.11b) Λ A B = δBA + ϵ w A B , (8.11c) where ϵ  1 acts as the perturbation parameter. One can easily show, at first order in ε, from eA µ EA ν = δµν that vA µ δµB = −u B ν δAν and from ηAB ΛA C ΛB D = ηCD implies that wAB = −wBA . We consider a tetrad ê A µ that is related to the original tetrad e A µ , by a local Lorentz transformation Λ̃ A B , i.e. ê A µ = Λ̃ A B e B µ . Consequently, the torsion tensors of the respective 129 Rep. Prog. Phys. 86 (2023) 026901 Review tetrads will be T̂ A = Λ̃ A B T B , and the connections are given in terms of two further Lorentz transformations Λ̂ and Λ by We can further split xβσ as xβσ = sβσ + aβσ (sβσ being the symmetric part and aβσ being the anti-symmetric part) and thus replacing it back into equation (8.17) we get ω̂ A B = Λ̂ A C d(Λ̂−1 )C B , ω A B = (Λ̃−1 ) A C Λ̂C D d(Λ̃E B (Λ̂−1 )D E ) = Λ A D d(Λ−1 )D B . W τ κ = ∂ρ [(2c1 − c2 )∂ ρ a τ κ − (2c1 − c2 )∂ κ a τ ρ + (2c2 + c3 ) × ∂ τ aρκ ] + ∂ρ [(2c1 + c2 )∂ ρ s τ κ − (2c1 + c2 )∂ κ s τ ρ  +c3 η τ κ (∂ ρ sβ β − ∂β sρβ ) − η τ ρ (∂ κ sβ β − ∂β sκβ ) . (8.18) (8.12) In the case when Λ̃ = Λ̂ the spin connection of the tetrad vanishes, ω A B = 0, meaning we choose the Weitzenböck gauge. Do not confuse the ˆ symbol used here to denote the transformed tetrad, with the b one used in sections 2–5 and denotes the quantities calculated with the symmetric teleparallel connection. The gauge transformations regarding the perturbations in this framework are êA = Λ̃A B eB ⇒ ûA µ = w̃A µ + uA µ , (8.13a) ΛA D = (Λ̃−1 )A C Λ̂ C D ⇒ ŵA B = w̃A B + wA B . (8.13b) In the same manner we split also the field equations W τ κ into symmetric and antisymmetric parts thus arriving at W(τ κ) and W[τ κ] as   W(τ κ) = ∂ρ − (2c1 + c2 + c3 )∂ (τ aκ)ρ + ∂ρ [(2c1 + c2 )∂ ρ s τ κ −(2c1 + c2 + c3 )∂ (τ sκ)ρ + c3 η τ κ (∂ ρ sβ β − ∂λ sρλ ) i −η ρ(τ ∂ κ) sβ β , (8.19a)   W[τ κ] = ∂ρ (2c1 − c2 )∂ ρ a τ κ + (2c1 − 3c2 − c3 )∂ [τ aκ]ρ   + ∂ρ (2c1 + c2 + c3 )∂ [τ sκ]ρ ) . (8.19b) We change indices using δµA and δAµ while raising/lowering any kind of index is done with the Minkowski metrics η AB and ηµν or its inverse. The torsion tensor up to first order reads T A µν = 2∂[µ e A ν] + 2ω A B[µ e B ν]  = 2ϵ ∂[µ u A ν] − ∂[µ w A ν] + O(ϵ 2 ). A few remarks on the structure of these fields equations: (a) It is possible to fully decouple W(τ κ) and W[τ κ] by demanding (2c1 + c2 + c3 ) = 0. (b) If one demands that equation (8.19b) vanishes identically, in addition to the decoupling conditions (2c1 + c2 + c3 ) = 0, (2c1 − c2 ) = 0 also (2c1 − 3c2 − c3 ) = 0 has to be satisfied, which implies c1 = − 14 c3 and c2 = − 21 c3 . It turns out that the family of Lagrangians satisfying these conditions are multiples of TEGR. In TEGR the antisymmetric part of the field equations is satisfied trivially because the spin connection enters only as a boundary term in the action (see section 5 for an extended discussion) up to a boundary term [222]. (c) The weak field limit of TEGR has already been studied in [618] and the fully general case, albeit in a different representation, in [227, 617]. (8.14) This perturbative expansion of the torsion tensor does not have a zeroth-order contribution since we assumed that the background is described by the Minkowski metric with the diagonal background tetrad δµA which indeed ensures that the teleparallel connection and the metric respect maximally symmetric spacetimes (see section 3.6.4). For further convenience we will transform all the indices of the torsion tensor into Greek indices, up to first order, as Tλ µν = EA λ TA µν . We expand the NGR Lagrangian (8.9) up to the first non-trivial order  L(2) = ϵ 2 Gαβ µνρσ ∂µ uα ν − ∂µ wα ν  (8.15) × ∂ρ uβ σ − ∂ρ wβ σ + O(ϵ3 ), In the next step, the principal polynomial will be used in order to investigate the propagation properties of the GW. After that the polarizations of the GW will be obtained. and note that the determinant e and Gαβ µνρσ contribute only their zeroth order since the quadratic torsion terms are already second order. From this Lagrangian we can derive the field equations wrt the variables u and w as 0 = ∂λ ∂L(2) ∂∂λ u τ κ 8.2.2. Principal symbol approach. The principal polynomial is a tool that helps us understand the propagation properties of a field from its equation of motion. More specifically the principal polynomial is defined as the determinant of the principal symbol [619, 620]. The vanishing of the principal polynomial defines the wave covectors kµ of the propagating DoF of the theory. If these covectors are known then one can calculate the speed of propagation and the mass (if not massless) of the fields under examination. The principal symbol is defined as the highest derivative operator of the partial differential equation in Fourier space equation (7.30). In order to derive the principal polynomial we need to rewrite the field equation (8.17) as  ⇔ 0 = Gτ β λκρσ ∂λ ∂ρ uβ σ − ∂ρ wβ σ , (8.16a) 0 = ∂λ ∂L(2) ⇔ 0 = (Gτ β λκρσ − ηγτ η ξκ Gξβ λγρσ ) ∂∂λ w τ κ  × ∂λ ∂ρ uβ σ − ∂ρ wβ σ . (8.16b) It is clear that u and w are not independent variables of the theory thus we introduce the new gauge invariant with equation (8.13b) variable xβσ = uβσ − wβσ gives us W τ κ := G τ βλκρσ ∂λ ∂ρ xβσ = 0. Ŵ τ κ = G τ βλκρσ kλ kρ x̂βσ = P τ βκσ (k)x̂βσ , (8.17) 130 (8.20) Rep. Prog. Phys. 86 (2023) 026901 Review where x̂βσ is the Fourier transformation of our original field variable xβσ (and not the hat accent that was used in the previous section) and Ŵ τ κ kτ = (2c1 + c2 + c3 )K2 Wκ = 0, (8.25c) c1 τ β c2 β κ στ η (Kη κσ − kκ kσ ) − k k η − kβ k τ η κσ 2 4  c3 τ κ σβ β τ κσ k k η −k k η +kσ k τ η βκ −Kη βκ η στ − 4 +kσ kβ η τ κ − Kη τ κ η σβ ), (8.21) Ŵ τ τ = (2c1 + c2 + 3c3 )KQ τ τ = 0, (8.25d) P τ βκσ (k) = where K = η(k, k) = η µν kµ kν is the norm of the wave covector. The principal polynomial P(k) is given by the determinant of the principal symbol, which is interpreted as a metric on the space of fields y τ κ = P τ βκσ (k)x̂βσ . From the anti-symmetry of the field equations in the indices λκ and ρσ, it is immediately clear that the principal symbol is degenerate, since fields of the form x̂βσ = kσ Vβ (k) solve the field equations trivially. The principal symbol being degenerate indicates that there are gauge DoF in the theory. Thus we need to derive a non-degenerate version of the principal symbol by restricting the field equations to the subspace of fields, on which they are non-degenerate30 . In order to do that we just perform the following split x̂βσ = kβ kσ U + Vβ kσ + kβ Xσ + Qβσ , Ŵ kα V = 0, α kα X = 0, α kα Q β = 0,       η τκ 1 − k τ kκ K  (8.25f )  0 0 0 0 K2 (2c1 + c2 + c3 )I3 0 0 0 0 K(2c1 − c2 )I2 0 0 0 0 K(2c1 + c2 )I2 Qτ τ   Xκ  ×   τ κ  Ŝ τ κ   0      0     , =   0     0        (8.26) where we condensed the matrices corresponding to X κ (3 DoF),  τ κ (2 DoF) and Ŝ τ κ (2 DoF) by writing them in terms of the identity matrix I n where n stands for the dimension. For example   1 0 K(2c1 − c2 )I2 = K(2c1 − c2 ) 0 1   K(2c1 − c2 ) 0 = . (8.27) 0 K(2c1 − c2 ) The 4 DoF from U and V α cannot be dynamical since they identically solve the field equations and so do not propagate. The remaining 12 DoF are divided with 4 − 1 = 3 in X α and 16 − 7 = 9 in Qαβ , which span the subspace V. Expanding Q τ κ further into its symmetric traceless and antisymmetric part as well as its trace by writing Q τ κ = S τ κ + A τ κ + τ κ 1 τκ − k Kk )Q σ σ , and using equation (8.10), the Fourier 3 (η space (7.30) field equations become Ŵ τ κ = (2c1 + c2 + c3 )Kk τ Xκ + (2c1 + c2 )KS τ κ  1 + (2c1 − c2 )KA τ κ + Qσ σ K 2c1 + c2 + 3c3 3 1 τ κ κτ × η − k k . (8.24) K Due to their simple nature the principal polynomial is now easily obtained as determinant of the above matrix P(k) = (2c1 + c2 + c3 )3 (2c1 + c2 + 3c3 )(2c1 − c2 )3 × (2c1 + c2 )5 K15 . To analyze them further we observe that they decompose into their contractions with k, their trace, their symmetric traceless and antisymmetric part as follows (8.25b)  kσ − 13 K(2c1 + c2 + 3c3 )  kα Qβ = 0. (8.23) Ŵ τ κ kκ = 0, |σ|κ) − k Ŵ (τ  α (8.25a) (τ κ) (8.25e) The first two equations are satisfied trivially for any choice of parameters c1 , c2 and c3 . The remaining four non-trivial field equations can be represented by a block diagonal matrix acting on a field space vector which is an element of V (8.22) Ŵ τ κ kτ kκ = 0, 1 [τ |σ|κ] k Ŵ kσ = (2c1 − c2 )KA τ κ = 0, K Ŵσ σ = (2c1 + c2 )KS τ κ = 0. where the scalar U, the 1-form components V α and X α and the (0, 2)-tensor Qβσ satisfy the constraints α Ŵ[τ κ] − (8.28) In order to solve the field equation (8.28) in a non-trivial way, a dispersion relation needs to be found from P(k) = 0 such that P(k) is non-degenerate. This solution is K = 0 which means that the fields Xκ , Q τ τ , S τ κ and A τ κ travel at the speed of light and they are massless. Hence we find that for NGR theories of gravity, perturbations propagate with the speed of light. On the other hand, for K 6= 0 the only solution of the field equations is that the fields themselves vanish identically, that is, no combination of coefficients can consistently solve each field equation expressed in equations (8.25a)–(8.25f ) simultaneously. Note that For the W κ mode a double pole in its propagator was found, which is consistent with [226, 227]. 30 This feature is common in field theories with gauge DoF and appears also in general premetric theories of electrodynamics [621] for example. 131 Rep. Prog. Phys. 86 (2023) 026901 Review Table 17. Summarizing the cases appearing in figure 26. Conditions Class 2c1 + c2 = c3 = 0 2c1 (c2 + c3 ) + c22 = 0 and 2c1 + c2 + c3 ̸= 0 2c1 (c2 + c3 ) + c22 ̸= 0 and 2c1 + c2 + c3 ̸= 0 2c1 + c2 + c3 = 0 and c3 ̸= 0 II6 (6 polarizations) III5 (5 polarizations) N3 (3 polarizations) N2 (2 polarizations) We can now study the polarizations of NGR. The whole preparation we need in the previous sections was performed to show whether the norm of the wave covector K is zero or not. As a matter of fact, K = 0 for our system and thus GWs are massless and travel at the speed of light independently of the parameters of the theory c1 , c2 , c3 . This directly allows us to make use of the well known Newman–Penrose formalism which was introduced in section 8.1 in order to link directly the linearized field equations to particular polarizations. Afterwards we will see which polarizations are acceptable by using the classification scheme detailed in [599, 600]. Recall that we consider minimal coupling between gravity and matter, i.e. coupling only through the metric seen as function of the tetrad, but not through the flat spin connection. As also discussed in section 4.1.2, this is the most straightforward way of making the energy-momentum tensor local Lorentz covariant. This is the usual coupling prescription for non-spinning matter, which we will henceforth assume. It follows from this choice of the matter coupling that test particles follow the geodesics of the metric, and hence the autoparallel curves of its Levi-Civita connection. The effect of a GW on an ensemble of test particles, or any other type of GW detector, such as the mirrors of an interferometer, is therefore described by the corresponding geodesic deviation equation. The observed GW signal hence depends only on the Riemann tensor derived from the Levi-Civita connection. The metric perturbation components δgµν are derived from the perturbation ansatz in equation (8.11) which takes the form 8.2.3. Newman–Penrose formalism and polarizations. Figure 26 0 = Wµν nµ mν = Wnm = Wnm̄ = −c3 s̈ml + (2c2 + c3 )äml , (8.30c) 0 = Wµν mµ m̄ν = Wmm̄ = Wm̄m = −c3 s̈ll , (8.30d) 0 = Wµν lµ nν = Wln = (2c1 + c2 )s̈ll , (8.30e) where the dots denote derivatives wrt retarded time (8.5). In order for the system of equations (8.30) to be satisfied the conditions in table 17 must be met. Specifically, by setting c1 = C sin θ cos ϕ c2 = C sin θ sin ϕ c3 = C cos θ, (8.31) with C = (c12 + c22 + c32 )1/2 being non-zero, we can plot the above classes in polar coordinates as in figure 26. The zenith angle θ (in polar coordinates) is shown by the radial axis, while the azimuth angle ϕ is shown by the polar axis. This concludes our discussion of GW polarizations in NGR. More details regarding the construction details of the figure 26 can be found in [616, 617]. We have seen that depending on the parameters c1 , c2 , c3 we obtain the E2 class II6 , III5 , N3 or N2 , with N3 filling most of the parameter space. We have also seen that there exists a family of theories besides TEGR which is of class N2 and thus exhibits the same two tensor modes as in GR. Theories in this class therefore cannot be distinguished from GR by observing the polarizations of GW alone. Finally let us state that, the exact same results hold true for the case of generalized NGR described by the Lagrangian L = f(T1 , T2 , T3 ) as introduced in equation (5.82). The only difference is that now the constants ci are generated from the function f as c1 → f1 := (∂f/∂T1 )|T1 =0 , c2 → f2 := (∂f/∂T2 )|T2 =0 and c3 → f3 := (∂f/∂T3 )|T3 =0 . Thus at the level of polarizations in a Minkowski background NGR and generalized NGR cannot be distinguished. 0 gµν = gµν + δgµν = ηµν + ϵ(ηµρ uρ ν + ηνρ uρ µ ) = ηµν + 2ϵsµν . (8.29) Note that they depend only on the symmetric perturbation of the tetrad, so that these are the only components whose presence or absence we must determine. We now examine which of the components (8.6b) may occur for GW satisfying the linearized field equations (8.18). Inserting the wave ansatz equation (8.5) and projecting the field equation tensor Wµν in the Newman–Penrose basis equation (8.4), we find the only non trivially satisfied equations are 8.3. Polarization modes in f(T, B) gravity The polarization content of f(T, B) gravity closely resembles ◦ that of f(R) gravity [41, 622–625]. This is due to the fact that both theories in first order perturbations in Minkowski, predict one massive dynamical scalar field on top of the usual tensor perturbations. The exposition closely follows [446] in what follows. Since the theory does predict a massive GW the E(2) classification cannot really be used as it is only valid for small masses. On the other hand, some promising recent progress has been made in using the Newman–Penrose approach 0 = Wµν nµ nν = Wnn = (2c1 + c2 + c3 )s̈nl + 2c3 s̈mm̄ + (2c1 + c2 + c3 )änl , (8.30a) 0 = Wµν mµ nν = Wmn = Wm̄n = (2c1 + c2 )s̈ml +(2c1 − c2 )äml , (8.30b) 132 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 26. Visualization of the parameter space. Reproduced with permission from [616]. Reprinted figure with permission from [617], Copyright (2018) by the American Physical Society. for massive particles [626]. Thus, this can be properly studied by using the traditional approach where the geodesic deviation equation (8.1) is directly used, which holds for any case. A similar calculation was carried out in [627]. The Weitzenböck gauge is employed both in the background and all perturbative orders. The first order perturbation of the torsion scalar is found to be T (1) = 0. Inserting all the above into the field equations of f(T, B) gravity, namely equation (5.80), order by order we obtain 0 = ηµν f(0, 0),  ◦ ◦ ◦ 1) 0 = −fT (0, 0)G(µν + fBB (0, 0) ηµν □ − ∂µ ∂ν R(1) , (8.35b) (8.32) The torsion scalar assumes no first order perturbation which as a consequence has ◦ ◦ R(1) = B(1) = η µν ∂ρ ∂ν hρµ − □h, (8.35a) ◦ where we have used the fact that R(1) = B(1) , and that f(0, 0) = 0 from the zeroth order condition. The latter condition is another statement for the fact that the arbitrary Lagrangian function does not include a cosmological constant. We proceed following [628] and we define an effective mass by considering the trace of the first-order equation. This (8.33) ◦ where h := hµµ and □ := ∂µ ∂ µ . We expand the f(T, B) function in Taylor series assuming it is expandable around the background values values T 0 and B0 , namely ◦ is also similar to the f(R) gravity case. However, our effective mass is different to that of [628]. Indeed, by taking the trace f(T, B) = f(T0 , B0 ) + fT (T0 , B0 )(T − T0 ) + fB (T0 , B0 )(B − B0 ) 1 1 + fTT (T0 , B0 )(T − T0 )2 + fBB (T0 , B0 )(B − B0 )2 2! 2! + fTB (T0 , B0 )(T − T0 )(B − B0 ) + O(T3 , B3 ). (8.34) ◦ ◦ ◦ fT (0, 0)R(1) + 3fBB (0, 0)□R(1) = 0, 133 (8.36) Rep. Prog. Phys. 86 (2023) 026901 Review we identify the effective mass m by bringing the equation in ◦ ◦ the form (□ − m )δ R 2 (1) = 0, which turns out to be [446] from which the solution of the Ricci scalar (8.42) simplifies to 31 ◦ fT (0, 0) m2 := − . 3fBB (0, 0) R (8.37) ◦ case where fBB (0, 0) 6= 0 we can follow the procedure of f(R) ◦ gravity [41, 622–625] since f(R) acts as a particular subclass of f(T, B) gravity, namely f(−T + B) gravity. Firstly, we introduce the tensor h̄µν to be  where h̄ represents the trace of h̄µν . Similarly to the previous section we consider the non-trivial case of fT (0, 0) 6= 0 (otherwise GR cannot be obtained at any limit). This simplifies equation (8.35b) to ◦  1 2 1 2 ◦ (1) 1 k0 h̄+ + p x + k20 h̄× y, R 0 2 2 6m 2 (8.47a)  1 1 2 1 2 ◦ (1) y + k20 h̄× x, ÿ = − k0 h̄+ + p R 2 6m2 0 2 (8.47b)  (8.39) z̈ = As shown in [622], it is possible to consider the Lorenz gauge condition ∂ µ h̄µν = 0, which simplifies the wave equation to ◦ □h̄µν = 0, (8.40) (8.41) where kρ is the four-wavevector, Aµν are constant coefficients, kρ kρ = 0, kµ Aµν = 0 and Aµµ = 0. The last conditions are the Lorenz gauge and traceless conditions respectively. On the other hand, the solution for equation (8.36) is ◦ R(1) = F exp (ipµ xµ ) , 8.4. Polarization modes in the teleparallel analogue of Horndeski gravity We thoroughly review the calculation of the polarization states in BDLS theory as presented in [615]. This is the most complex and inclusive case since it contains the most general scalar-tensor form on top of any teleparallel theory that can be build from scalars quadratic to torsion coupled to a scalar field that leads to second order field equations (see section 5.8.2). The layout will contain two parts, in the first one we will find the DoF and how the overall theory branches. In the second part we will adapt the polarization analysis using the first as a guide. The starting point is considering the linearized field equations for the BDLS theory, which were calculated by expanding the action in equations (5.122a)–(5.122d) fBB (0, 0) ηµν F exp (ipµ xµ ) . (8.43) fT (0, 0) Note that from equations (8.35b) and (8.36), the Ricci tensor is found to be ◦ ◦ ◦ 1 fBB (0, 0) ∂µ ∂ν R(1) , R(1) µν = ηµν R(1) − 6 fT (0, 0) (8.47c) (8.42) where F is a constant and pµ is another four-wavevector such that pµ pµ = −m2 . Hence, the full solution for hµν is constructed as hµν = Aµν exp (ikρ xρ ) + ◦ 1◦ 1 p20 − p23 R(1) z = − R(1) z, 2 6m 6 where in the last equation we have used that pµ pµ = −m2 . Additionally, since the wave propagates in the z-direction, we have used and defined h̄11 = −h̄22 ≡ h̄+ and h̄12 = h̄21 ≡ h̄× , which represent the massless + and × polarizations. As we observe, in the TEGR and f (T) limit, namely at |m2 | → ∞, the remaining modes are the + and × polarizations as expected. However, in the case |m2 | < ∞ we find the presence of the longitudinal and breathing modes in the geodesic deviation equations. The fact that we find a combination of scalar polarizations is due to the fact the scalar component is massive field. Thus f(T, B) gravity includes more polarization modes than f (T) gravity. The reason for this behavior is directly attributed to the boundary term that is not trivialised at first order B(1) 6= 0. as well as the traceless condition h̄ = 0. This allows for the solution h̄µν = Aµν exp (ikρ xρ ) , (8.45) Therefore, the geodesic deviation becomes (8.38) ẍ = ∂ ρ ∂ν h̄ρµ + ∂ ρ ∂µ h̄νρ − ηµν ∂ ρ ∂ α h̄ρα − □h̄µν = 0. µν =  ◦ 1 1 ηµν − pµ pν R(1) . 6 3m2 Hence, it is trivial to verify that taking the trace part of the above equation, it yields a consistent relation for the Ricci scalar, as expected. We proceed by analyzing the polarization states of the GW by using the geodesic deviation equation (8.1). Assuming propagation in the z direction we find the linearized components of the Riemann tensor as i ◦ ◦ 1 1 h 2 ◦ (1) (1) η p + p p . (8.46) Ri0j0 = k20 h̄ij − R R ij 0 i j 2 6m2 We remark that in the |m2 | → ∞ limit (for instance when fBB (0, 0) = 0 and fT (0, 0) 6= 0), the equation reduces to that of GR. Since it is known that f (T) gravity yields no further GW modes [629], this special condition leads to a broader class of theories in which at first order yield the GW solutions. In the ◦ 1 fBB (0, 0) hµν = h̄µν − h̄ηµν + ηµν R(1) , 2 fT (0, 0)  (1) (8.44) 31 This is slightly different to the effective mass that appears in [628] where a typo appears which was corrected in [446]. 134 Rep. Prog. Phys. 86 (2023) 026901 Review b = (GTele,X + G2,X − 2G3,ϕ )δ ϕ̈ W   + δϕ GTele,ϕϕ + G2,ϕϕ + (GTele,X + G2,X − 2G3,ϕ )k2  + (GTele,I2 − 2G4,ϕ ) −k2 (χ̇ + Φ − 2ψ) + 3ψ̈ . (8.50e) and (5.125) up to second order in the perturbation of the tetrad δeA µ and the perturbation of scalar field δϕ. We then perform a change of variables for the tetrad as indicated in equation (7.26) in order to simplify the notation and facilitate the calculation process. Thus our new perturbative variables are τµν , δϕ for which we vary the second order action to obtain their field equations as   Wµν = −2GTele,Tvec ∂ λ ∂µ τλν − ∂µ ∂ν τ − ∂λ ∂σ τ λσ ηµν + □τ ηµν  + (−2GTele,T + 2G4 ) □τ(µν) − ∂ λ ∂µ τ(νλ) − ∂ λ ∂ν τ(µλ)  +∂µ ∂ν τ + ∂λ ∂σ τ λσ ηµν − □τ ηµν   4 + GTele,Tax □τ[µν] − ∂ λ ∂ν τ[µλ] + ∂ λ ∂µ τ[νλ] 9 + (−GTele,I2 + 2G4,ϕ ) (∂µ ∂ν δϕ − ηµν □δϕ) , (8.48a) On the other hand, the vector part consists of three linearly independent equations for the gauge invariant variables (β i , Σi , Λi ) W0j =  h 1n 18GTele,Tvec β̈ j + k2 2GTele,Tax − 9(G4 − GTele,T ) Λj 9  i − 9(G4 − GTele,T ) + 2GTele,Tax β j  Wj0 =  1 n 2 h k 2GTele,Tax − 9(G4 − GTele,T ) β j 9   i  1 n h kl Wlj = − i k2 9(G4 − GTele,T ) + 2(9GTele,Tvec + GTele,Tax ) β̇ j 9   i + 9(G4 − GTele,T ) − 2GTele,Tax Λ̇j − 18GTele,Tvec Σj o (8.51c) + 4GTele,Tax Σ̈j , (8.48b) 0 = GTele,ϕ + G2,ϕ . o (8.51b) where we have already applied the background field equations 0 = GTele + G2 , (8.51a) − 9(G4 − GTele,T ) + 2GTele,Tax Λj + 4GTele,Tax Σ̇j ,   b = (GTele,I2 − 2G4,ϕ ) □τ − ∂λ ∂σ τ λσ + (GTele,ϕϕ + G2,ϕϕ ) δϕ W + (GTele,X + G2,X − 2G3,ϕ ) □δϕ,  o − 2 9GTele,Tvec + 2GTele,Tax Σ̇j , (8.49) while the tensor field is described by one equation In the next step we perform a SVT decomposition of the field equations (8.48a) and (8.48b) by inserting equation (7.27) and splitting them into 3 + 1, transforming into Fourier space equation (7.30) and using the gauge invariants variables equations (8.8a)–(8.8e) we arrive at Wij = (G4 − GTele,T )(ḧij + k2 hij ). (8.52) In order to study the system as a whole, we can combine all the SVT sectors into one master matrix as done in [615]. The determinant of this master matrix, which is also the principal polynomial P(k) is the most important quantity and it reads as W00 = k2 ((GTele,I2 − 2G4,ϕ )δϕ + 2GTele,Tvec Φ −4(G4 − GTele,T + GTele,Tvec )ψ) , (8.50a) 8 P(k) = −214 3−5 (G4 − GTele,T )5 GTele,Tvec 3 GTele,Tax 3 k12 ω 2 − k2    × c̃1 + c̃2 ω 2 − k2 , (8.53)  kj Wj0 = ik2 (GTele,I2 − 2G4,ϕ )δ ϕ̇ + 2GTele,Tvec χk2  −2(2(G4 − GTele,T ) + 3GTele,Tvec )ψ̇ , (8.50b) where c̃1 = 2(GTele,ϕϕ + G2,ϕϕ )(2(G4 − GTele,T ) + 3GTele,Tvec ), (8.54a) ηjl W jl = 3(GTele,I2 − 2G4,ϕ )δ ϕ̈ c̃2 = −3(GTele,I2 − 2G4,ϕ )2 − 2(GTele,X + G2,X − 2G3,ϕ ) (8.54b) × (2(G4 − GTele,T ) + 3GTele,Tvec ).  + 2k2 (2(G4 − GTele,T ) + 3GTele,Tvec )χ̇ + (GTele,I2 − 2G4,ϕ )δϕ + 2(G4 − GTele,T + GTele,Tvec )Φ − 2(G4 − GTele,T + 2GTele,Tvec )ψ   c̃3 = −GTele,ϕϕ − G2,ϕϕ , (8.54c) c̃4 = GTele,X + G2,X − 2G3,ϕ , (8.54d) (G4 − GTele,T )G2,ϕϕ , 2 −3G4,ϕ + (G4 − GTele,T )(2G3,ϕ − G2,X ) (8.54e)  − 6 2(G4 − GTele,T ) + 3GTele,Tvec ψ̈, (8.50c)  8 kl ϵljk Wjk = − iGTele,Tax k2 σ̈ + k2 σ , 3 (8.50d) Z1 = − 135 Rep. Prog. Phys. 86 (2023) 026901 Review Table 18. All branches of the BDLS theory represented according to their PDoF (propagating degree of freedom). To each of the scalar, vector and tensor components correspond 1,2 and 2 DoF respectively. The quantities c̃i are defined in (8.54a)–(8.54d) while Z 1 and Z 2 are defined in equations (8.54e) and (8.54f )) respectively. Sectors Massless ω 2 = k2  Massive ω 2 − k2 = m2  Scalar Vector Tensor Scalar m2 PDoF G4 − GTele,T ̸= 0 GTele,Tvec = 0, GTele,Tax = 0, GTele,I2 = 0, GTele,X = 0, GTele,ϕϕ = 0 — — 1 — — 2 0.I G2,ϕϕ ̸= 0 and −3G4,ϕ 2 + (G4 − GTele,T ) (2G3,ϕ − G2,X ) ̸= 0 — — 1 1 Z1 3 0.II G2,ϕϕ = 0 and −3G4,ϕ 2 + (G4 − GTele,T ) (2G3,ϕ − G2,X ) = 0 — — 1 — — 2 1 GTele,Tvec ̸= 0, GTele,Tax ̸= 0 c̃1 ̸= 0,c̃2 ̸= 0 2 1 1 1 −c̃1 /c̃2 7 2 GTele,Tvec ̸= 0, GTele,Tax ̸= 0,c̃1 = 0,c̃2 = 0 2.I GTele,Tax ̸= 0 2.I.a c̃3 ̸= 0,c̃4 ̸= 0 1 1 1 1 −c̃3 /c̃4 6 2.I.b c̃3 = 0,c̃4 = 0 1 1 1 — — 5 2.II G3,ϕ = Z2 , G2,ϕϕ = −GTele,ϕϕ 2.II.a GTele,I2 − 2G4,ϕ ̸= 0 2 1 1 — — 6 2.II.b GTele,I2 − 2G4,ϕ = 0 1 1 1 — — 5 3 GTele,Tvec ̸= 0, GTele,Tax = 0,c̃1 ̸= 0,c̃2 = ̸ 0 1 — 1 1 −c̃1 /c̃2 4 4 GTele,Tvec ̸= 0, GTele,Tax = 0,c̃1 = 0,c̃2 = 0 4.I GTele,Tvec = − 32 (G4 − GTele,T ), G4,ϕ = 21 GTele,I2 4.I.a c̃3 ̸= 0,c̃4 ̸= 0 — — 1 1 −c̃3 /c̃4 3 4.I.b c̃3 = 0,c̃4 = 0 — — 1 — — 2 4.II G3,ϕ = Z2 , G2,ϕϕ = −GTele,ϕϕ 4.II.a GTele,I2 − 2G4,ϕ ̸= 0 1 — 1 — — 3 4.II.b GTele,I2 − 2G4,ϕ = 0 — — 1 — — 2 5 GTele,Tvec = 0, GTele,Tax ̸= 0,c̃1 ̸= 0,c̃2 = ̸ 0 1 — 1 1 −c̃1 /c̃2 4 6 GTele,Tvec = 0, GTele,Tax ̸= 0,c̃1 = 0,c̃2 = 0 1 — 1 — — 3 7 GTele,Tvec = 0, GTele,Tax = 0, c̃1 ̸= 0,c̃2 ̸= 0 — — 1 1 −c̃1 /c̃2 3 8 GTele,Tvec = 0, GTele,Tax = 0,c̃1 = 0,c̃2 = 0 — — 1 — — 2 Cases Conditions — 0  3(GTele,I2 − 2G4,ϕ )2 + 2(2(G4 − GTele,T ) + 3(GTele,Tvec ))(GTele,X + G2,X ) Z2 = . 4(2(G4 − GTele,T ) + 3(GTele,Tvec )) (8.54f ) tensorial ones. For the Case 1, which is the most general version of BDLS, 7 propagating DoF where found which can be though of as just two extra massless scalars and one massless vector on top of the standard Horndeski DoF. The rest of the cases are simpler versions of Case 1 and include various combinations from just tensor modes to combinations of scalar,vector and tensor modes. In table 19 there is a link between well known theories of the scalar-tensor form In order to properly solve the system we need to find all the cases where P(k) in equation (8.53) is non-degenerate. This was performed in a exhaustive manner in [615] and all the results are gathered in the table 18. In the Case 0, the standard Horndeski gravity was explicitly studied which also matches the results reported in [630, 631] but a new branch that only entails a massless sector was further discovered. In this sector the only propagating DoF are the 136 Rep. Prog. Phys. 86 (2023) 026901 Review Table 19. Common models in the literature (see section 5) shown against the analysis presented in table 18. Theory Case PDoF GR (or TEGR) or f (T) (section 5.3) Horndeski (section 5.8.2) GTele (section 5.8.2) Extented NGR (section 5.5) Generalized teleparallel dark energy (section 5.8.1.2) Generalized teleparallel scalar-tensor (section 5.8.1) Tachyonic teleparallel gravity (section 5.8.1) — 0.I 1 2.I.b 7 7 7 2 3 7 5 3 3 3 and the full analysis included in table 18. Hence, in general there is a plethora of sub branches including most combinations of SVT propagating DoF. Using the results regarding the DoF detailed in table 18 and using the representation of the electric components of the Riemann tensor (8.7), the polarizations of the respective cases were exhaustively studied. The resulting cases [615] of GR and f (T) gravity are first shown as the straightforward instances of the analysis. Next, there is the standard Horndeski Case 0.I which assumes tensor polarizations for the massless sector and and both the scalar polarizations for the massive sector. For Case 0.II, the new branch of Horndeski gravity, only tensor polarizations where found as expected, since only tensorial DoF were found as shown in table 18. In Case 1 (the full BDLS theory), there were found the breathing mode along tensor modes for the massless sector and both scalar modes for the massive sector. One could compare this to standard Horndeski (Case 0.I) as having an extra breathing mode in the massless sector. This exact polarization imprint is, also, only shared with Case 3 despite the fact that the DoF in these two cases are different. Regarding Cases 2.I.a, 2.I.b, 2.II.b, 5 and 6 in general, although there is a massless scalar DoF, it does not leave a polarization imprint. This also happens for Cases 2.I.a and 4.I.a where instead, there is a massive scalar. There is also a similar behavior vectorial DoF for the Cases 1, 2.I.a, 2.I.b, 2.II.a and 2.II.b where although there is a vectorial DoF it is invisible to the detectors. This phenomenon is directly linked to the fact that these DoF are not purely metrical or that they do not couple strong enough with the metric. Hence we see that even in the full BDLS theory Case 1, there are all sorts of scalar-vectorial-tensorial DoF found, only the scalar and tensorial leave a polarization imprint in end. Obviously there is a plethora of possible combinations described by the various subclasses of how this happens but nevertheless this is a very interesting illustration of how metrical and non-metrical DoF behave within this system. Lagrangian DensityLi Si = 2κ1 2 ´ d4 x e Li  ◦ R or f (T) Equations (5.122a)–(5.122d) GTele (ϕ, X, T, Tax , Tvec , I2 , J1 , J3 , J5 , J6 , J8 , J10 ) f(T, Tax , Tvec ) −A(ϕ)T − 12 ∂µ ϕ∂ µ ϕ − V(ϕ) F(ϕ)T + P(ϕ, X) − G3 (ϕ, X)□ϕ f(T, X, ϕ) many important issues that have not been addressed yet. Some of them will be listed at the end of this section. 9.1. Spherical symmetry and solutions There are several studies aiming to find spherically symmetric exact or perturbed solutions in different TG. There are no good tetrad-spin connection pairs in TG for the most general timedependent spherically symmetric metric ds 2 = A(r, t)dt 2 − B(r, t)dr 2 − M(r, t)dΩ 2 reported in the literature. At a first ingredient at this, one needs to consider a good tetrad-spin connection pair in spherical symmetry (see section 5.12) and since this pair depends on the theory, one should be careful in checking if equation (5.147) is still a good tetrad-spin connection pair in the studied theory at hand. The first paper studying spherical symmetry in modified TG was [632] where the authors found an exact solution in f (T) gravity which behaves as the standard Schwarzschild solution. The problem of this solution is that the torsion scalar becomes zero (T = 0), therefore, it is trivial since f (T) becomes TEGR. There are other vacuum exact solutions in f (T) in the literature having a similar property with T = const. [633] or also with T = 0 [634], which again are trivial since those solutions become TEGR plus a cosmological constant (or just TEGR). In [635], the authors found different solutions with T 6= const., but the analysis was not complete since they did not consider one of the field equations. It is also important to mention that there are other several incorrect works in the literature that also did not consider a good tetrad-spin connection pair, and then, they ignored the appearance of a nonvanishing antisymmetric field equation [636–640]. As discussed in section 5.12, one cannot use any arbitrary spin connection-tetrad pair since they must solve the antisymmetric field equations of the theory. In several earlier works, different authors found exact solutions in f (T) gravity but they used the incorrect tetrad in the Weitzenböck gauge (not a good tetrad) which in the end is manifested in a non-zero antisymmetric field equation of the form of fTT T ′ = 0 which is only satisfied for the GR case. The first non-trivial vacuum exact spherically symmetric solutions in TG (different to just GR + Λ) was derived in [355]. In this paper, several solutions were found in f(T, B) gravity using the Noether’s symmetry approach. The problem of these solutions is that all of them have a metric with a power-like form A(r) ∝ r m with B(r) = const. Hence, all of these solutions might not be so physically interesting since 9. Astrophysical systems This section will be devoted to studying the most important results and problems in astrophysics in TG. We will focus on Solar System tests and review the most important results regarding compact objects such as stars, black holes and wormholes. Since this is an active field of research, there are 137 Rep. Prog. Phys. 86 (2023) 026901 Review This means that even though the metric is Minkowski, the connection (teleparallel one) is non-trivial. This idea was then further explored for ξ = 1 in [643], where the authors found perturbed solutions around Schwarzschild in −T − ϵ(α/2)T p with p = 2. This study was then generalized in [355] for different p > 1 with p ∈ Z. Later, in [354, 641] the authors found more perturbed solutions in f(T, B) gravity for the two cases and also for the complex tetrad. We will now provide some of the most important results for the case ξ = 1 since this study does not have the issue of T and B being nonvanishing in the Minkowski limit. The f(T, B) gravitational field equations (5.80) in spherical symmetry with the good tetrad-spin connection pair given by the tetrad (5.147) with ξ = 1 and M(r) = r, which gives the metric they cannot describe black holes nor standard astrophysical systems. Two recent papers found the first non-trivial exact black hole solutions in f(T, B) gravity and also in teleparallel scalar-tensor theories [641, 642]. These solutions are based on the complex tetrad in equation (5.153). Still, these solutions have not been analyzed in detail to know if they might provide realistic black hole solutions. From these works, one can then state that for a generic f(T, B) theory, the Birkhoff’s theorem is not satisfied. Since it is hard to find exact solutions, it is also helpful to use perturbation techniques to find perturbed solutions in TG. The main idea behind this is to use GR in the background and add small perturbations related to the modification of TEGR. Regarding perturbed spherically symmetric solutions, the first study in TG was done in [357]. The authors used the tetrad in equation (5.147) with ξ = −1 and found perturbed solutions around Minkowski. It is important to recall here that for this tetrad, T does not become zero for Minkowski unless one assumes a non-trivial limit in the metric functions. ds 2 = A(r)dt 2 − B(r)dr 2 − r 2 (dϑ 2 + sin 2 φ dφ 2 ), (9.1) and with a zero spin connection become  √ rB ′ fB′ − 4B 3/2 fB′ + fT′ + 4BfT′ 1 rB( B − 1)A ′ + A(rB ′ + 2B 3/2 − 2B) κ ρ = − f− fT + 2 r 2 AB 2 2rB 2 h i r 2 BA ′2 + rA rA ′ B ′ + 4B 3/2 A ′ − 2B(rA ′ ′ + 4A ′ ) + 4A 2 (rB ′ + 2B 3/2 − 2B) fB′ ′ f − − , B 4r 2 A 2 B 2 B √ √ rA ′ + 4A ′ r( B − 2)A ′ + 2A( B − 1) 1 2 f − fT −κ pr = − f − 2 2rAB B r 2 AB   −r 2 BA ′2 + rA −rA ′ B ′ − 4B 3/2 A ′ + 2B(rA ′ ′ + 4A ′ ) − 4A 2 rB ′ + 2B 3/2 − 2B + fB , 4r 2 A 2 B 2 √ 1 rA ′ − 2A( B − 1) ′ rB ′ − 2B 3/2 ′ fB′ ′ −κ 2 pl = − f + fB − fT + 2 2rAB 2rB 2 B   −r 2 BA ′2 + rA −rA ′ B ′ − 4B 3/2 A ′ + 2B(rA ′ ′ + 3A ′ ) + A 2 (−2rB ′ − 8B 3/2 + 4B 2 + 4B) fT + 4r 2 A 2 B 2   −r 2 BA ′2 + rA −rA ′ B ′ − 4B 3/2 A ′ + 2B(rA ′ ′ + 4A ′ ) − 4A 2 (rB ′ + 2B 3/2 − 2B) + fB , 4r 2 A 2 B 2 2 (9.2b) (9.2c) coordinate, i.e. fT′ = fTT T ′ + fTB B ′ and the scalar-torsion and boundary term yield √   √ 2 B − 1 rA ′ − A B + A T=− , (9.3a) r 2 AB where we have assumed an anisotropic fluid with energy density ρ and lateral and radial pressures given by pl and pr respectively. Here, primes denote differentiation wrt the radial B= (9.2a)   −r 2 BA ′2 + rA −rA ′ B ′ − 4B 3/2 A ′ + 2B (rA ′ ′ + 4A ′ ) − 4A 2 rB ′ + 2B 3/2 − 2B . 2r 2 A 2 B 2 138 (9.3b) Rep. Prog. Phys. 86 (2023) 026901 Review  2M −1 B(r) = 1 − + ϵ b(r), r These two quantities are zero for the Minkowski case A = B = 1. As discussed above, there are non-trivial vacuum exact solutions but all of them have B = const. One example is the following vacuum exact solution [355]  A(r) 2 = r r0 where M is the Schwarzschild mass that can be arbitrary large and ϵ  1 is a small tracking parameter. Up to first order in ε, there are not solutions around Minkowski (M = 0). Moreover, to get solutions around Minkowski, one would need to go beyond first order in ε. It can be proved that for power-law f (T) gravity, there is only contributions in a(r) and b(r) if one makes perturbations up to fourth order in ε. On the contrary, when one uses the other choice of the tetrad (5.147) (with ξ = −1) and also the complex tetrad (5.153) which have nonvanishing torsion scalars at Minkowski, one finds non-trivial perturbed solutions around Minkowski (see section 9.5 to see the real case for ξ = −1). For the M 6= 0 one can get perturbed solutions around Schwarzschild by only considering perturbations up to first order in ε for the following form of f(T, B) as  4n(n−2 1)(2n−3) 4n −8n+5 B(r) 2 = const = , (2n − 1)(4n − 5) , 4n 2 − 8n + 5 f(T, B) = f0 T n . (9.4) It is important to mention that the standard conservation equation for the fluid holds. This can be seen by differentiating equation (9.2b) wrt r and then using equations (9.2a)– (9.2c) accordingly, finding the standard conservation equation pr′ = −(ρ + pr ) A′ 2 + (pl − pr ), 2A r (9.5) which is the standard conservation equation obtained in GR. This equation follows from the the fact that the theory is formulated within the standard teleparallel description of gravity which is diffeomorphism invariant (see section 4.1.5). Thus, one finds that the energy-momentum tensor satisfies the stand- 1 f(T, B) = −T − ϵ (αTq + βBm + γBs Tw + ζ(ξT + χB)u ) . 2 (9.8) ◦ Different solutions were found in [354] depending on the parameters chosen. As an example, for the case q = m = 2, ζ = 0 and w = −s + 1, one gets the following perturbed solution ard conservation equation ∇µ Θµν = 0. Let us emphasise here that the system (9.2a)–(9.2c) only has two independent equations for the vacuum case. In [644, 645], the authors noticed that for fT 6= 0 and r(B − 2)A ′ + A(B − 1) 6= 0, these equations can be rewritten in the case of vacuum f (T) gravity as  1 1 3βµ7 − (α + 13β)µ 6 r 2 (µ 2 − 1) 2 2 1 − 4βµ5 + µ4 (15α + 43β + 2C1 r) 2 2 3 1 − µ (32α + 35β) − µ 2 (31α + 51β + 4C1 r) 3 2 1 + 4βµ + (17α + 21β + 2C1 r) 2    β 2 (9.9a) − + 2(α + β) 3µ − 1 log µ , µ  A(r) = µ 2 + ϵ C2 −  0 = −r 2 A ′2 − r 2 A A ′ B ′ − (B − 1)A ′ ′ + A 2 (B − 1) 2 (B + 1), √ √ 1 r( B − 2)A ′ + 2A( B − 1) 0 = − f− fT , 2 r 2 AB (9.6a) (9.6b) where the first equation does not depend on the form of f. This can be achieved by solving equations (9.2a)–(9.2b) for fT′ and f T and then by replacing them in equation (9.2c). By doing this, one can eliminate all the f dependence and then one finds that for any generic form of f, the metric functions must satisfy (9.6a). This equation is not easily solvable for A or B unless one assumes something for those functions. One interesting result that one finds is that when we assume the ansatz B = 1/A, which is the form of many well known black hole solutions (like Schwarzschild), one gets that the unique solution of equation (9.6a) is a Schwarzschild de-Sitter metric A = 1 − 2M/r + Λr 2 . Plugging this result into the second equation (9.6b), we find that f(T) = c1 T − 6Λ. This means that the unique solution in f (T) gravity which behaves as B = 1/A is a trivial one when the theory becomes TEGR plus a cosmological constant. This statement was shown for the two possible tetrads (the real and complex ones) in [641]. In other words, this means that any solution in f (T) gravity beyond GR must have B 6= 1/A. Let us now find perturbed solutions by assuming a Schwarzschild background A(r) = 1 − 2M + ϵ a(r), r (9.7b) " B(r) = µ −2  C2 µ 2 − 1 C1 1 +ϵ + + 2 2 4 4 rµ µ r (µ − 1)  1 × (25α + 37β) − 4(α + 2β)µ 2 2(16α + 13β) 2(α + 3β) 4(α + β) − − + 3µ µ2 µ3  −21α − 25β 2β 2(α + β) + + + log µ , 2µ4 µ5 µ4 (9.9b) where µ 2 = 1 − 2M/r and Ci are integration constants. One important remark is that Ci can be set by imposing different physical situations. One can see this by expanding the above solutions up to 1/r, yielding     16(α + β) 16(α + β) 1 a(r) ∼ C2 + − C + 1 3M 2 M r   1 +O 2 , (9.10a) r (9.7a) 139 Rep. Prog. Phys. 86 (2023) 026901 Review    16(α + β) 1 1 b(r) ∼ C1 − 2C2 M + M +O 2 . 3M r r (9.10b)  symmetric solution around Schwarzschild was found [353] in a theory concerning the five scalars that one can construct from contractions of torsion, which is an extension of new GR (see section 5.5). Let us here finish this section by mentioning two recent papers where exact solutions were found for the complex tetrad (5.153). The first interesting solution presented in [641] reads In [354], they chose the constants as C1 = − 16α + β , M C2 = − 16α + β 3M 2 (9.11)  to recover the Schwarzschild metric at infinity. Since these constants are related to the boundary conditions of the system, one can also choose them in other ways. The above solution is also a generalization of the solution found in [643] where a power-law f (T) gravity was considered with β = 0. For more details about the specific form of a(r) and b(r) for other solutions, see [354]. Taking a similar approach, another spherically  2M Q ds = 1 − + 2 r r 2   2 dt − 2Mr − Q − r 2 2Q − r 2 −1 (9.12) with  Q 2 T 2 − 2QT + 4 f(T) = 4f0  . q p p QT + 2 ± Q 2 T 2 − 2QT + 4 8 − 2QT ± 4 Q 2 T 2 − 2QT + 4 2± p This exact solution is similar to Reissner–Nordström one but with grr 6= −1/gtt and hence, it describes a black hole with two event horizons. A second interesting black hole solution  for the Born–Infeld f (T) gravity q was found f(T) = λ (9.13) f(T, B) gravity has exact spherically symmetric solutions that differs from Schwarzschild (see for example equation (9.4)), then we know that the generalized Birkhoff’s theorem will not hold for all forms of f. It is obvious to mention that if T, B are constants, then the dynamics of f(T, B) will be just TEGR plus a cosmological constant. Then, for any f, the unique vacuum spherically symmetric solution is Schwarzschild de-Sitter with Λ being related to the constants T 0 and B0 . Then, when T0 = B0 = 0, we will recover the standard Schwarzschild solution and the generalized Birkhoff’s theorem will hold. This statement will also be valid for more general teleparallel theories having other scalars. However, this case is not so interesting since it is just GR plus Λ. For example, in f (T) gravity, if T = T0 (constant), the field equation (5.55b) contracted with two tetrads becomes 1 + 2T λ − 1 and is given by " √ !# a12 √ λr −1 ds = λ(a0 λ + r) − 2 tan dt 2 r 2 " √ !#−1 √ λr λ5/2 r5 −1 λ(a0 λ + r) − 2 tan − 2 2 (4 + r λ) 2 2 × dr 2 − r 2 dΩ 2 , dr 2 − r 2 dΩ 2 , (9.14) where a0 , a1 are integration constants. By fixing the constants accordingly, this spacetime is asymptotically flat and it describes a black hole with one horizon. Another recent paper [642] found that for the teleparallel scalar-tensor theory presented in section 5.102, it is possible to find exact scalarized black hole solutions. One example of them are the so-called Bocharova–Bronnikov–Melnikov–Bekenstein (BBMB) provided by a theory with a coupling between the scalar field and the torsion scalar. For more details about other solutions, we refer the reader to see [642]. Further new analysis are needed to understand the nature of these solutions as long as if they are healthy or not since they were constructed from the complex tetrad. ◦ Gαβ − 11 1 (f − T0 fT )gαβ = κ2 Θαβ , 2 fT fT (9.15) where we have used ∂ν fT = 0 and we have divided by f T . Now, since T = T0 , all the functions evaluated at T 0 will be just constants, meaning that one define f1T (f − T0 fT ) = const. := Λeff . Thus, the f (T) field equations with T constant yields ◦ 1 1 Gαβ − Λeff gαβ = κ2 Θαβ =: κ̃ 2 Θαβ , 2 fT (9.16) which clearly is the Einstein field equations with an extra effective cosmological constant. In the following, we will ask the following question: can one have time-dependent vacuum spherical symmetry in a general modified teleparallel theory? To explore this, let us first take the most general time-dependent spherically symmetric tetrad given by equation (3.133) and consider 9.2. Generalized Birkhoff’s theorem The Birkhoff’s theorem states that in GR (in vacuum), the spherically symmetric spacetime must be static and its solution is described by the Schwarzschild metric. Since we know that 140 Rep. Prog. Phys. 86 (2023) 026901 Review ds 2 = (C12 − Ċ52 )dt 2 + 2(C1 C2 − C5′ Ċ5 )dt dr f(T, B, ϕ, X) gravity. The antisymmetric field equation of this theory reads − (C5′ 2 − C22 )dr 2 − C5 dΩ 2 . 0 = ∂[µ (fT + fB )Tµ λν]  1 Tλ ∂ν (fT + fB ) − Tν ∂λ (fT + fB ) + Tµ λν ∂µ (fT + fB ) . = 3 (9.17) Without loss of generality, due to the remaining gauge freedom in the metric, we can choose C5 = C5 (r) = r 2 and also set C2 = 0 to convert our metric into its diagonal form. After doing all of this, one finds that the above metric is highly constrained since grr = −1. Then, for this particular case, the Birkhoff theorem does not hold but the spacetime is static. Let us finish this discussion by emphasizing that the above computation is only valid in the theory f(T, B, ϕ, X) and in the specific case when one does not know the specific form of f and assumes that all the scalars depend on the two coordinates t, r, which means that the only way to solve the antisymmetric equation (9.17) is by imposing that the terms multiplying ḟT + ḟB and fT′ + fB′ vanish which is achieved if equation (9.18) is true. Still the question about the generalized Birkhoff’s theorem can be thought off as being open the context of modified TG. In general, this equation depends on the form of f and also on the form of the tetrad that can give a dependency on r and t in T, B and ϕ. Let us study this equation without assuming an explicit form of the function f. This strategy of course does not consider all the possible models constructed from f since it could be that for some models, the equation is solved by just replacing the form of f and also T and B in the above antisymmetric field equation. Further, let us assume that fT + fB depends on both t and r, meaning that the torsion scalar, the boundary term and the scalar field T = T(r, ϑ) B = B(r, ϑ), ϕ = ϕ(r, ϑ) (or at least one of them depend on the two variables), then the only way to solve the antisymmetric field equation without choosing f would be to constrain some components of the torsion tensor and also the torsion vector. For the most general spherically symmetric tetrad satisfying spherical symmetry in teleparallel geometries in the Weitzenböck gauge (3.133), one finds out that for this case, we must have Tt = −T r tr , Tr = −T t rt , T t ϕθ = Tr ϕθ = 0 9.3. Solar system tests Many Solar System observations have been performed, and they have confirmed that GR can explain these effects in a good agreement with observations [174]. Thus, any meaningful modified gravity must pass the so-called Solar System tests. In this section we will explore two main routes to constrain theories of gravity: either using solutions of the studied theory and then analyze them by performing the standard particle phenomenology of geodesics using observables (i.e. the classical tests); or to use the so-called parametrized post-Newtonian formalism, that basically uses postNewtonian expansions and express the possible deviations from GR in different parameters. (9.18) to solve the antisymmetric equation (9.17). These equations 0 = 2C3 C5 − 2Ċ5 C5 − 2C6 Ċ6 , (9.19a) 0 = 2C4 C5 − 2C5′ C5 − 2C6 C6′ , (9.19b) 0 = C2 C6 , 0 = C1 C6 , 9.3.1. Solar system tests using particle motion observables. (9.19c) One interesting route for testing teleparallel theories in astrophysical systems is to find exact or perturbed solutions and then constrain them using particle phenomenology, by which we mean both classical astrophysical tests of gravity as well as more recent ones such as gravitomagnetic effects. The first paper doing this for Solar System in f (T) gravity was [359, 360], which was performed using a perturbed solution around Minkowski found in [357]. These results are correct but they assumed a tetrad having a nonvanishing torsion scalar in the absence of gravity (see equation (5.147) with ξ = −1). Later in [646], the authors analyzed the tetrad (5.147) with ξ = 1 (which has the correct limit in the scalars) and they found the corrections to the perihelion shift predicted by a powerlaw f(T) = −T − (α/2)T n gravity with n = 2. This work was extended for n = 2, 3, . . . , 10 in [356] and also finding the photon sphere and perihelion shift of the perturbed solutions around Schwarzschild to then constrain the models accordingly with the perihelion of Mercury. Then, in [354] studied f(T, B) gravity, finding new perturbed solutions around where primes and dots are derivatives wrt r and t, respectively. One notices that the only way to solve these equations is by imposing C6 = 0, C3 = Ċ5 , C4 = C5′ . (9.21) (9.20) The first condition is always needed to solve equation (9.19c) since the inverse of the tetrad diverges when C2 = C1 = 0. This solution observes that T = 0 (which can be seen by calculating the torsion scalar using the tetrad in (3.133) and the torsion components in (3.143)) but the boundary term B = B(r, ϑ), ◦ meaning that the Ricci scalar R = B and then f(T, B, ϕ, X) grav◦ ity becomes f(R, ϕ, X) gravity. One can then conclude that if fT + fB depends on both t and r (and we do not choose or replace f ), the only solution of the antisymmetric field equations is the one which matches with the curvature-based ◦ theory f(R, ϕ, X) gravity by having the metric 141 Rep. Prog. Phys. 86 (2023) 026901 Review Particles must obey the normalization condition so that by setting if 2L = σ, then massive (massless) particles would be described by σ = 1 (σ = 0). σ is a constant that characterizes the massive or massless behavior of the type of particle being investigated. By using this normalization in equation (9.22), we arrive at the following expression  2  h2 2 −1 k ṙ = B − −σ , (9.24) A r2 Figure 27. The Sun causes gravitational effects to be exhibited by a that can be further rewritten in term of an effective potential giving us test mass in orbit. The solar orbit is characterized by a closest point perihelion and a furthest point aphelion. ṙ 2 = −2V(r), where we have defined the potential as   1 −1 k 2 h 2 V(r) = − B − −σ . 2 A r2 Schwarzschild (equations (9.9a) and (9.9b) is one solution) and then computing different observables such as the perihelion shift, deflection of light and the Shapiro delay. After finding them, they also constrained the parameters of the models using Solar System observations. Using a similar approach, in [264] the authors also studied f(T, B) gravity in a weak field limit and found the geodetic and Lense-Thirring effects predicted by this theory, noticing that the modifications play an important role and then, these effects may be also use to constrain models. Let us here compute the most important observables that are usually used to constrain models for the real tetrad and ξ = +1. In general, the observables that we will derive are also valid for any physical situation when we have a test body orbiting a massive compact object. This happens in the Solar System (where the Sun is the massive compact object) as depicted in figure 27, but also can be used in other astrophysical scenarios such as modeling the movements of the stars around the supermassive black hole Sgr A* in our galaxy. When the masses of the system have the same order of magnitude, then, the geodesic equation cannot be used and one needs to use the machinery of the two-body problem. It is interesting to mention that it turns out that for some observables, the final result coincides with the one that it will be presented here for the geodesic motion by just replacing the mass of the central massive compact to the sum of the two bodies (M → M1 + M2 ). Let us here consider the case of geodesics of test particles orbiting a certain compact object. The worldline q(τ ) = (t(τ ), r(τ ), θ(τ ), ϕ(τ )) of a test particle, with τ being an affine parameter, can be expressed as Now, we have all the ingredients to compute some interesting observables. For their more detailed derivation and schematic representations, see [73]. For the perihelion shift, which measures how much the perihelion is shifted after an orbit, we need to assume massive objects and a nearly circular orbit such that perihelion shift: σ = 1, and V(r) = V ′ (r) = 0, (9.28) where now r = r(φ) = rc + rφ (φ), giving us that the potential equation (9.25) can be rewritten in the following form  where dots are derivatives wrt τ . Due to spherical symmetry, we can consider the motion of the test particle at the equatorial plane ϑ = π/2, without loss of generality. The energy k and angular momentum h are the conserved quantities that are given by ∂L h=− = r 2 φ̇. ∂ φ̇ (9.26) Up to this point, we have not assumed any form of the metric. Then, equation (9.25) is valid for any spacetime and any test particle. As discussed in section 9.1, there are not many exact spherically symmetric solutions in modified teleparallel theories of gravity. However, there are some perturbed solutions around Schwarzschild, for example, the ones expressed in equation (9.9) for f(T, B) gravity. If we then assume that our metric functions can be expressed as in equations (9.7a) and (9.7b), the potential can be expanded up to first order of ε, yielding    2 1 h 1 2M + σ V(r) = − k 2 + 1− 2 2 r r2 "  ! ϵ 2 a(r) 2M + k + b(r) 1 − 2 r 1 − 2M r #     2 h2 2M −b(r) σ + 2 1− . (9.27) r r 2L = gµν q̇µ q̇ν = A ṫ 2 − B ṙ 2 − r 2 ϑ̇ 2 − r 2 sin 2 ϑ φ̇ 2 , (9.22) ∂L k= = Aṫ, ∂ ṫ (9.25) drφ dφ 2 = −2 (rc + rφ )4 V(rc + rφ ). h2 (9.29) Now, if we assume a nearly circular orbit, we can expand rφ /rc up to second order to approximate the above equation as  (9.23) 142 drφ dφ 2 3 r4 r = − c2 V ′ ′ (rc )rφ2 + O rφ3 , 0 h (9.30) Rep. Prog. Phys. 86 (2023) 026901 Review where V(rc ) = 0 and V ′ (rc ) = 0 were used. The above equationq for rϕ is in the form of an oscillator with a wave num- gravitating central mass). This quantity can be easily found by taking equation (9.25) and rewriting it as r4 ber K = hc2 V ′ ′ (rc ), giving us that the final expression for the perihelion shift becomes ∆φ = 2π 1 K dφ φ̇ h = =p , dr ṙ −2V(r)r 2 !  − 1 = 2π rc2 p h V ′ ′ (rc ) −1 . then integrating this equation from r0 (minimal distance) to infinity. It should be noted that the deflection of light for nonasymptotically flat spacetimes (such as in Schwarzschild deSitter) cannot be obtained by using the above formula since it goes beyond the event horizon. Instead, one needs to use the method used in [648, 649]. Besides these cases, by integrating the above equation and using ṙ(r0 ) = 0 (V(r0 ) = 0), we get (9.31) For any perturbed solution around Schwarzschild, we find that the perihelion shift is ∆φ = ∆φGR + ϵ ∆φϵ = 6πq + 27πq 2 + 135πq3 + O(q4 ) + ϵ ∆φϵ , (9.32a) ∆φ = ±2 (9.32b) ∆φϵ,max Mercury ≈ 0.18 /cen. dr̄ B(r̄)1/2 r̄ 2  1 A(r0 ) − 2 2 r0 A(r̄) r̄ −1/2 − π, (9.36)  2 where r02 = hk A(r0 ) was used and one can choose the + sign when further evaluating. Similarly as we did before, one can find the deflection of light predicted by GR (the Schwarzschild contribution) plus a new contribution coming from the perturbed solution. After considering the approximation r  M, one finds   4M M 2 15π ∆φ ≈ φGR + ϵ φϵ = + 2 −4 r0 4 r0   3 M 244 − 45π + 3 + ϵ φϵ . (9.37) 6 r0 (9.33) This observable was explicitly found for different theories in [353, 354, 356]. As an example, for the theory (9.33), the first order correction of the deflection of light is φϵ = 64αM3 /(45r50 ). One can use observations from the Very Long Baseline Interferometry (VLBI), which uses radio-telescopes on Earth [650], to constrain the teleparallel corrections. In this case, GR predicts that the deflection of light (see the above equation) produced by the Sun is ϑGR ≈ 1.756 ′ ′ . Then, the observed value over the GR predicted value is [174, 651] the first order correction to the perihelion shift is ∆φϵ = 8παq 2 /rc2 , respectively [356, 646]. It is important to emphasize again here that these corrections are different depending on how one sets the integration constants Ci in (9.9a) and (9.9b) (with β = 0). In the Solar System, one can use the observations for the perihelion shift of Mercury which is 42, 98 ± 0.040 ′ ′ /cen [174]. This means that any deviations in the perihelion shift predicted by the modifications of TEGR should have the following maximum value for Mercury ′′ ˆ ∞ r0 where q = M/rc and ∆φϵ denotes the first order correction from the perturbed solution (a(r) and b(r) in equations (9.7a) and (9.7b)). In [353, 354, 356, 646], the authors derived the first order correction of different observables such as the photon sphere and the perihelion shift for different perturbed solutions found in f (T), f(T, B) and f(Tax , Tvec , Tten , P1 , P2 ) gravity, respectively. Then, they used Solar System observations to constrain the corresponding free parameters of their model. For example, for a squared-power law f (T) model 1 f(T) = −T − ϵ αT 2 , 2 (9.35) ϑobs ≈ 1.0001 ± 0.0001, ϑGR (9.34) (9.38) which tells us that for the model (9.33), the maximum value that α can take is αmax ∼ 1019 km 2 . Another interesting experiment that can be used to put bounds in modified theories is the Cassini–Huygens experiment which is an experiment composed by a system of Earthspacecraft-Earth. In this experiment, the fractional frequency shift y of a ray of light after passing this system is measured [652], and its definition is For the model (9.33), the maximum value that the constant should take is αmax ∼ 1020 km 2 to match the Mercury observations. Furthermore, there are other astrophysical systems where perihelion shift has been measured. For instance, the perihelion shift produced in the system composed by the supermassive black hole in the center of our galaxy, namely Sagittarius A* (Sgr A*), and the S stars which are orbiting it [647]. One can also use these observations to constrain teleparallel theories, but this has not been reported in the literature yet. Another interesting observable is the deflection of light ∆φ which is the angle characterizing the difference between the light trajectory at infinity (with or without the presence of a y=2 vCassini–Huygens lEarth + vEarth lCassini–Huygens ∆φ, lEarth + lCassini–Huygens (9.39) where ∆φ is the deflection of light (equation (9.37)), vEarth and vCassini–Huygens are the transverse velocities of the Earth and the Cassini–Huygens spacecraft, respectively, and lEarth 143 Rep. Prog. Phys. 86 (2023) 026901 Review By assuming a small central object r  M, we arrive at the following expression for the Shapiro delay ∆tShapiro (r, r, r0 ) = ∆tShapiro,GR (r, r, r0 ) + ϵ ∆tShapiro,ϵ (r, r, r0 ) (9.43a) q ≈ M r 2 − r02 r + r0 q + 2M log  + ϵ ∆tShapiro,ϵ (r, r, r0 ). calculation that is performed. Here, the central mass is shown as the Sun, the mirror is shown as the planet and the emission point is taken to be from Earth. (9.40) For the power-law model (9.33), the first order fractional frequency shift was found to be equal to yϵ = 128M3 α/(15r50 ) [354]. In this experiment, it has been measured that yobs ∼ 10−10 ± 10−14 [652] and since the GR predicted value (see above equation) is around yGR ≈ 1.690 × 10−9 , the parameter α in (9.33) cannot exceed αmax ∼ 1023 km 2 . Another quantity that can be used to constrain models is the Shapiro delay or the retardation of light. By integrating equation (9.25), we find that the time required for a signal to be emitted from r0 to r is (σ = 0)  ˆr t(r, r0 ) = dr̄ r0 r 2 A(r̄) 1 − 20 r̄ A(r0 )  A(r̄) B(r̄) (9.43b) where ν i are the frequencies measured at ri . Then, the fractional frequency between these two points becomes       ν2 ν2 ν2 ≈ +ϵ ν1 ν1 GR ν1 ϵ   ν2 1 −1 = 1 + M(r− − r ) + ϵ , (9.46) 1 2 ν1 ϵ −1/2 ,  One example of an experiment that has measured the Shapiro delay is the Viking mission on Mars [653], which essentially observed the time travel of a signal measured on the Earth after performing the trajectory Earth-Mars-Sun-Earth. It has tobs been found that ∆ ∆tGR = 1.000 ± 0.001 where ∆tGR ≈ 2.664 × −4 10 s. Thus, for the power-law form in equation (9.33), one finds that the maximum constant αmax ∼ 1020 km 2 . There are other measurements regarding the Shapiro delay outside the Solar System that can be used to constrain the corrections of GR. For instance, there is an important effort in the field of pulsars to measure these effects (see for example [654, 655]). These constrains have not been used in the context of TG yet. The gravitational redshift can be also used to constrain models, for example by using experiments with a hydrogenmaser clock on a rocket. For a light ray propagating from altitudes r1 and r2 (with r1 < r2 ), the gravitational redshift can be expressed as s ν2 A(r2 ) z≡ −1 = − 1, (9.45) ν1 A(r1 ) and lCassini–Huygens are the distances from the Earth to the Sun and from the Cassini–Huygens spacecraft to the Sun, respectively. After assuming that the distance Cassini–Huygens–Sun is much larger than the distance Earth–Sun (lCassini–Huygens lEarth ), we get 8M vEarth + ϵyϵ . r0 r0  For the theory (9.33), the first order correction becomes q 4αM3 r 2 − r02 ∆tShapiro,ϵ (r, r, r0 ) = 30(r + r0 )   7 7 17 20 × 3 + 2 2 + 3 + 4 . (9.44) r r0 r r0 rr0 r0 Figure 28. The Shapiro delay setup is shown in the context of the y = yGR + ϵ yϵ ≈ r 2 − r02 + r (9.41)  2 where r02 = hk A(r0 ) was used. The Shapiro delay is defined as the time required for a signal to propagate from a radius re to a point of closest encounter to the central mass r0 , and then from there to a mirror at radius rm where it gets reflected and returns on the same path to the emitter, as shown in figure 28 for a specific example case. This quantity is then defined as where we have assumed that ri  M. The leading order contribution (9.33) is found to be equal   for the power-law case  −5 −5 ν2 2 3 to ν1 = 5 M α r1 − r2 . There are several experiments ϵ performed on the Earth which measures this effect. The most common one is the one which uses hydrogen-maser clock on a rocket launched to an altitude of about 107 m [656], that found obs that ∆ν ∆νGR = 1.000 ± 0.0002, which effectively constrain the parameter α in (9.33) to have the maximum value of the order of αmax ∼ 1022 km 2 . ∆tShapiro (re , rm , r0 )   q q 1 = t(re , r0 ) + t(rm , r0 ) − re2 − r02 − rm2 − r02 . 2 (9.42) 144 Rep. Prog. Phys. 86 (2023) 026901 Review Another important effect that has been measured is the socalled geodetic effect, which appears when a central body (a source) is rotating and one has a gyroscope which is orbiting it. In this situation, the gyroscope starts to precess. To obtain this effect, one can follow the procedure as it is explained in [657]. By starting from a spherically symmetric metric, we can consider that the system is rotating with an angular frequency ω, such that the angular coordinate φ → φ̃ − ω t. By sitting the gyroscope in the circular polar point ϑ = π/2, one finds that the metric becomes the one predicted by GR. There is only a Newtonian rescaling for this theory, therefore, one cannot constrain this theory by using weak field approximations. It would be interesting to extend this study by not assuming the weak field limit and study some specific models of f(T, B) with its respective perturbed solutions. An important comment that should be taken into account is that the value of all of these observables depend on the choice of the integration constants in the perturbed solutions. In [658], the authors used other boundary conditions (to obtain a black hole) as the ones considered in the results provided above, and then they found that the maximum value that the parameters of the theory can reach, have a different order of magnitude. Furthermore, the expressions presented above assumed the real tetrad with ξ = +1. When one chooses ξ = −1 or the complex tetrad, these observables are different. It was found that they lead to corrections to GR at lower order than the the real tetrad with ξ = +1 [641]. Let us stress again that all of these observables can be used in a direct way with any solution (exact or perturbed) to constrain gravitational theories by using, for example, Solar System observations or any other astrophysical system where these effects occur. There are also other ways to contrast Solar System observations with modified gravitational models. One very important method is the so-called PPN analysis, that we will explain in the next section. ds 2 = e2Ψ(r) (dt − e−2Ψ(r) ωr 2 dφ̃) 2 − A(r)r 2 e−2Ψ(r) dφ̃ 2 . (9.47) Here, e2Ψ(r) = A(r) − r 2 ω 2 . One can further find that the angular frequency of the gyroscope would be given by [657] 1/2 eΨ  Ω = √ kik kjl (ωi,j − ωj,i )(ωk,l − ωl,k ) , 2 2 (9.48) where ωi = e−2Ψ ωr 2 δi3 and kij is the spatial 3-metric. By using several identities, one finds that the precession over one orbit (per year) becomes s  r " # A′ 1 rA ′ ΩdS = 1− 1− , (9.49) 2r B 2A where B is the grr component of the metric. The geodetic effect has been measured by the Gravity Probe B experiment which was a satellite that orbited the Earth with a gyroscope precessing due to the rotation of the Earth. It was found a geodetic precession of −6601.8 ± 18.3 mas yr−1 at around 642 km from the Earth. In [264], the authors calculated this effect for the power-law (9.33) finding that αmax ∼ 1032 km 2 . The last observable that we will show here is the LenseThirring effect which measures a relativistic precession of a rotating body (e.g. a gyroscope) near a massive compact body. This effect can be found by analyzing the geodesic equation of a freely falling gyroscope which is initially at rest and has a spin characterized by the spin vector Sµ . By using the force-like equation (3.73) which is equivalent as the geodesic equation (3.71), we arrive at dSµ + Γµσρ S σ uρ = Kµσρ S σ uρ . dτ One of the most useful analytical tools to test modified theories of gravity on smaller scales is by the so-called ‘Parametrized Post-Newtoninan’ (PPN) formalism [174] which essentially is encoded in ten parameters. These parameters have been tested with high precision (for example in the Solar System) and essentially they are obtained by considering the lowest orders deviations from the PPN expansion of a given gravitational theory (weak field limit). The standard formalism has been developed in curvature-based theories containing a metric as a fundamental field. Since the fundamental variable in TG is the tetrads, the formalism needs to be adapted by assuming perturbations in the tetrad. This study was first performed in [225] and we will review its most important results here. One key ingredients to the PPN formalism is assuming that the matter acting as a source of a gravitational field is given by a perfect fluid whose velocity is small compared with the velocity of light in a specific frame. Then, one can expand in orders of velocity. The energy-momentum tensor is then 9.3.2. Parametrized post-Newtoninan formalism. (9.50) Here we have assumed that uµ is the gyroscope’s velocity measured when it is at rest, i.e. uµ Sµ = 0. If we again assume that the metric can be expressed as gµν = ηµν + δgµν and |hµν |  1 (as in equation (7.1)), which is a weak field approximation, the above equation can be written as dSi = ϵikl Ωk Sl , dτ Θµν = (ρ + ρΠ + p)uµ uν − pgµν , (9.52) where uν is the 4-velocity and is normalized by uµ uµ = +1 and ρ, p and Π are the energy density, pressure and specific internal energy, respectively. To proceed, we need to expand all the important fields in orders of velocity O(n) ∝ |⃗v|n . For the tetrads, the expansion can be written as (9.51) where Ωk ≡ − 21 ϵkmn ∂m h0n . Physically speaking, the quantity Ωk represents the angular velocity precession vector of the gyroscope. In [264], it was found that the Lense-Thirring effect for f(T, B) gravity in the weak field regime is the same as 0 1 2 3 4 e A µ = e A µ + e A µ + e A µ + e A µ + e A µ + O(5), 145 (9.53) Rep. Prog. Phys. 86 (2023) 026901 Review where we have assumed that the expansion is taken around consisting of PPN potentials and some constant coefficients. One can take the following ansatz for the tetrad at each velocity order: 0A a flat diagonal background tetrad e µ = diag(1, 1, 1, 1). Here, n overscripts denote the velocity order considered, i.e. e A µ ∼ O(n). We will only consider expansions up to the fourth order in the velocity. It is useful to lower the Lorentz index, and this can be done by contracting the above perturbation with the Minkowski metric, yielding 0A eµν = e µ ηAB e B ν , 0A n nB eµν = e µ ηAB e ν . 2 2 e00 = a1 U, eij = a2 Uδij + a3 Uij , 3 (9.58a) 3 e0i = b3 Vi + b4 Wi , ei0 = b1 Vi + b2 Wi , (9.58b) (9.54) 4 e00 = c1 Φ1 + c2 Φ2 + c3 Φ3 + c4 Φ4 + c5 U 2 + c6 ΦW + c7 A, (9.58c) One important remark is that after doing the expansions one finds that not all the tetrad perturbation components are relevant. If we assume the Newtonian energy conservation and a time reversal symmetry, one finds that the only non-trivial components are 2 e00 , 2 3 eij , e0i , 3 4 ei0 , e00 . with the potentials satisfying [174, 614] ∇ 2 U = −4πρ, 1 2 3 4 Uij = χ,ij + Uδij , (9.59a) (9.55) ∇ 2 Vi = −4πρvi , Since the tetrads are the fundamental variables of all teleparallel theories (we assume the Weitzenböck gauge), we can then use the above perturbation in a specific theory and expand all the geometrical objects in the underlying theory in the velocity orders. If in our studied theory we have a scalar field ϕ (for example in teleparallel Horndeski, see section 5.8.2), we also need to expand it in velocity orders, namely, ϕ = Φ + ψ = Φ + ψ + ψ + ψ + ψ +O(5). ∇ 2 χ = −2U, ∇ 2 Wi = −4πρvi + 2U,0i , ∇ 2 Φ1 = −4πρv 2 , ∇ 2 Φ2 = −4πρU, ∇ 2 Φ3 = −4πρΠ, ∇ 2 Φ4 = −4πp, (9.59b) (9.59c) where ∇ 2 = δ ij ∂i ∂j . Using these ansatz, the tetrad can be identified with the standard PPN parameters as follows [659] 2 (9.56) Moreover, if one has arbitrary functions of a geometrical variable, as in the teleparallel analogue of Horndeski gravity, one would need to expand the corresponding Lagrangian function. For example for this theory, this can be done by taking the following Taylor series expansion e00 = U, (9.60a) e(ij) = γUδij , (9.60b) 2 1 3 e(0i) = − (3 + 4γ + α1 − α2 + ζ1 − 2ξ)Vi 4 1 − (1 + α2 − ζ1 + 2ξ)Wi , (9.60c) 4 Gi (ϕ, X) = Gi (Φ, 0) + Gi,ϕ (Φ, 0)ψ + Gi,X (Φ, 0)X 1 + Gi,ϕϕ (Φ, 0)ψ 2 + Gi,ϕX (Φ, 0)ψX 2 1 + Gi,XX (Φ, 0)X 2 + . . . (9.57a) 2 1 1 4 e00 = (1 − 2β)U 2 + (2 + 2γ + α3 + ζ1 − 2ξ)Φ1 2 2 + (1 + 3γ − 2β + ζ2 + ξ)Φ2 + (1 + ζ3 )Φ3 1 + (3γ + 3ζ4 − 2ξ)Φ4 − ξΦW − (ζ1 − 2ξ)A. (9.60d) 2 1 = Gi + Gi,ϕ ψ + Gi,X X + Gi,ϕϕ ψ 2 + Gi,ϕX ψX 2 1 2 + Gi,XX X + . . . , (9.57b) 2 where ˆ ˆ 0 ΦW = where Φ = ϕ, bold letters denote the constant Taylor coefficients at the background level and Gi are the functions related to teleparallel Horndeski (see equations (5.122a)–(5.122d)). This is just an example where the Lagrangian involves a function of an arbitrary scalar, but the expansion should be in the same way for other theories different to the one written above. These are all the ingredients needed to start the computations. Now, one needs to insert all of these expansions in the field equations, expand each quantity accordingly in the PPN approximation and then solve the equations order by order. To do this easily, one introduces some suitable potential ansatz ρ ′ρ ′ ′   ⃗x −⃗x ′ ⃗x ′ −⃗x ′ ′ ⃗x −⃗x ′ ′ · − ′ d3 x ′ d3 x ′ ′ , ′ ′ ′ 3 |⃗ x −⃗ x | |⃗ x −⃗x ′ ′ | |⃗x −⃗x | (9.61a) ˆ A= d3 x ′ ρ(t,⃗x ′ ) [vi (t,⃗x ′ )(xi − xi′ )] . |⃗x −⃗x ′ |3 2 (9.61b) For more details regarding these potentials, see section 4.2.1 in [614]. The so-called PPN parameters are the ten parameters appearing in the above equation γ, β, α1 , α2 , α3 , ζ1 , ζ2 , ζ3 , ζ4 , ξ. 146 (9.62) Rep. Prog. Phys. 86 (2023) 026901 Review Table 20. Summary of PPN parameters for different teleparallel theories reported in the literature. In all of these theories, all the other PPN parameters are zero (the same as GR). The terms Hi and β̃ are written in the appendix C. Theory β −1 γ −1 f (T)/GR [225] f(T1 , T2 , T3 ) [225] 0 2f +f +f − 4(2f,,11 +f,,22 +2f,3,3 ) 0 2f +f +f −2 2f,,11+f,,22+2f,,33 f(T, ϕ, X, Y) [664] fT fX f Y2 (16fTϕ − 7f Y ) + 3f 4Y ( f Y − 2fTϕ ) i −8f T 2 f X2 fTϕ + 2f T 2 f Y 2f X2 + f Y fXϕ − 2f X fYϕ f Y2 2fT f X −2f Y2 Teleparallel Horndeski [666] − 8 H H +H 2 2 3Hβ̃2 −2H H −2H ( ,1 ,5 ,3 ) ( ,3 ,1 ( ,4 ,5 )) − 2 H H +H,23 ( ,1 ,5 ,3 ) fY 2 8(4fT fX −3f Y2 )( f Y2 −fT fX ) h The parameter ξ measures preferred-location effects, αi is related to preferred-frame effects while ζ i measure possible violations of the conservation of the total momentum. Finally, the parameters γ and β are related to how much spacetime curvature is produced (by unit rest mass) and the amount of non-linearity in the gravitational model, respectively. For GR (or equivalently TEGR), we have αi = ζi = ξ = 0, γ = β = 1. such as at Solar Systems. Later in [664, 665], the authors studied a more general teleparallel scalar-tensor theory composed of f(T, X, Y, ϕ) where Y = gµν Tρ ρµ ϕ,ν (see section 5.8). This theory can be understood as a generalization of the model studied in [663]. They also found that only γ and β differ from their GR value. Finally, recently this study was further generalized to the teleparallel Horndeski case, again finding that only γ and β differ from GR [666]. One interesting remark is the fact that there are some classes of theories in teleparallel Horndeski having γ = β = 1, and then, all of those model will have the same predictions as GR at smaller scales. Two recent papers [667, 668] found that for a TG parity violating model which is similar to a Nieh–Yan model and also for a higherorder derivative TG (see section 5.7), all the PPN parameters are the same as GR. All of the most important results are summarized in table 20. Let us finish here by mentioning that a viable modified theory of gravity must pass all the Solar System tests, and one possible route that has not been explored in teleparallel theories are by considering some screening mechanisms [660]. (9.63) The strongest observations bounds for γ and β are set by the Cassini–Huygens tracking experiment [652] and measurements provided by the perihelion precession of Mercury [174], which are γ − 1 = 4(β − 1) ⩽ (2.1 ± 2.3) · 10−5 . 2H,1 H,4 +H 2 (9.64) GR works very well at Solar System scales, so using these bounds one can set constrains on modified theories of gravity on smaller scales. Some scalar-tensor theories can have screening mechanisms which allow us to have a non-trivial density profile for the scalar field and then, modifications of gravity become negligible at smaller scales and only become important on other scales such as at cosmological scales [660]. Still, the screening mechanisms for teleparallel theories have not been studied in the literature. There are few studies in teleparallel concerning the PPN analysis explained above. The first study doing preliminary analysis under these directions was done in [661] for teleparallel DE (see section 5.8). The authors concluded that all the PPN parameters of this theory are the same as GR. The same result in a slightly more general teleparallel scalar-tensor theory was obtained in [662]. Later, the case of a teleparallel scalar-tensor theory with a scalar field non-minimally coupled to both the boundary term B and the torsion scalar T was studied, finding some deviations for γ and β [663]. Some years later Ualikhanova and Hohmann presented a formal and general way to construct the PPN formalism in TG in a more rigorous way [225], which is the one presented above. They specifically studied a teleparallel theory constructed from an arbitrary function f(T1 , T2 , T3 ), where T i are the 3 non-parity violating scalars that can be constructed from the torsion tensor (see section 5.5). They found that only γ and β are different to the GR value and for the specific case of f (T) gravity, the PPN parameters are identical to the GR ones. Therefore, f (T) gravity has the same predictions as GR at smaller scales 9.4. Astrophysical compact objects Soon after the development of f (T) gravity, different papers studied different compact objects in TG. The main conclusion of the first works is that the majority of them suffered from the lack of understanding the covariant formulation of TG, leading some authors to totally ignore the fact that they were choosing the incorrect good-tetrad spin connection pair. This created an enormous number of papers with mistakes since many authors ignored the fact that their choice contained some non-trivial antisymmetric field equation. The standard mistake was to consider a diagonal tetrad e A µ = (A, B, r, r sin ϑ) with a vanishing spin-connection. As was explained in section 5.12, this choice is not correct since it constrain theories to be just GR. Thus, any paper using that diagonal tetrad in spherical coordinates with a vanishing spin connection would be incorrect. Taking this into account, in the following we will review some references which applied correctly the covariant formulation, without forgetting the role of the spin-connection (and the antisymmetric equations). The majority of the works related to stars in modified TG have been carried out in f (T) gravity. The first paper where these astrophysical systems were studied properly 9.4.1. Stars. 147 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 29. Particle number–to–stellar radius curves for f(T) = −T − (a/2)T 2 gravity with a polytropic index equal to Γ = 2. The dashed line a = 0 corresponds to GR. Reprinted figure with permission from [670], Copyright (2018) by the American Physical Society. was done in [669] for relativistic spherically symmetric stars in f (T) gravity. Essentially, they used the tetrad-spin connection pair given by equation (5.147) for f (T) gravity. The equations for this case are written in equations (9.2a)–(9.2c) if one sets f(T, B) = f(T). Then, the set of equations (9.2a)–(9.2c) and the conservation equation (9.5) describe the internal structure of self-gravitating fluids in f(T, B) gravity. As happens in GR, we need to choose an EoS to close the system. Later in [670], the authors studied compact stars in f(T) = −T − (a/2)T 2 gravity for an isotropic fluid (pr = pl = p) with a polytropic EoS given by ρ = Kp1/Γ + p K 1 nΓ = K2 n + , Γ−1 Γ−1 stellar system which is related with particle number density defined as dN n := , (9.67) dV where N is the total√ particle number enclosed in the differential volume dV = 4πr 2 Bdr. Then, we can integrate this equation from its center r = 0 to its radius R, yielding that the particle number is ˆ √ N = 4π n(r) Br 2 dr. (9.68) This equation must be used with p = knΓ , (9.65) and the field equations (9.2a)–(9.2c) to describe the internal structure of the star. From these equations, one can also derive a modified version of the standard Tolman–Oppenheimer–Volkoff (TOV) equation that helps us describing compact stars in an easier way. The numerical procedure used in [670] works as follows: there are three field equations in spherical symmetry for f (T) gravity and four variables (two metric functions A, B and the energy density ρ and pressure p for the fluid). To close the system one must assume an EoS for the fluid, that as we pointed out above, one can assume to be a polytropic one. Then, some normalized variables are introduced in order to have all functions and constants dimensionless. To have a realistic star, one must impose the correct boundary conditions. At the center of the star, one can impose that the pressure satisfies p(r = 0) = p0 where p0 is a constant and the metric functions B(r = 0) = 1 and (A ′ /(2A))kr=0 = 0. At the radius of the star, one imposes that p(r = R) = 0. Having all of that, one can numerically solve the equations. Figure 29 shows numerical solutions for the system which are the particle-to-stellar radius (9.65) with the adiabatic exponent being Γ = 2, K 1 and K 2 are constants and we have introduced the particle number density n. Then, the star would be sourced similarly as it is usually described by GR since its source will be a fluid with a polytropic EoS. However, since the field equations of the f (T) (or other modifications) are different to GR, the structure of the star could be different (for example its mass and radius). As has been discussed in the review in [671], the definition of the mass of the star in modified gravity has been vastly debated. In GR, this mass coincides with the ADM mass of the spacetime defined as ˆ M := 4π ρr 2 dr, (9.66) but in a more general gravitational theory, this equation is not known to be true. Therefore, in [670], the authors used a more general way to understand the matter contained in a 148 Rep. Prog. Phys. 86 (2023) 026901 Review curves for different values of the modifications of GR coming from −(a/2)T 2 . This graph would be the analogous of the well known mass–radius plot usually displayed in GR and its extensions. Then, it can be seen from this figure that it is possible to construct compact stars in squared-f (T) gravity having different properties as GR. One can notice that if a > 0(a < 0), then the amount of matter supported against gravity is less(more) than in GR. Moreover, in [670] they also claimed that the energy density of the star over the central region is roughly constant (n incompressible core) for positive values of a. It can be also conclude that for large positive values of a, there is a phase transition where the energy density and pressure change its sign within the interior of the star. Recently, this study was further generalized for boson stars [672] by adding a non-interacting complex scalar field. They found that the mass of the boson stars increases when a is negative and large enough in absolute value. This means that in f (T) gravity, the mass is no longer bounded as in GR and then, the situation is quite different to GR since the boson stars will not be necessarily a mini boson star. A similar study related to boson stars was done in [673] in a theory where the torsion scalar is non-minimally coupled to a complex scalar field. In this case, the authors found that for a sufficiently large coupling constant, the energy density increases outwardly, surrounded by a thick shell, following an abrupt decrease in its profile ending in the so-called asymptotic tail. It is interesting to mention that this feature does not occur in a theory where one has the Ricci scalar non-minimally (or minimally) coupled to a complex scalar field. Later, the same authors [674] generalized this study by adding a U(1) gauge field, leading to an increase in the maximal mass and the particle number. Recently, in a series of papers [675, 676], a similar result for f (T)-power law gravity was found. In [677], with the Raychaudhuri equation, collapsing stellar distributions were analyzed in f (T) gravity. The authors considered a star with an interior behaving as a FLRW metric. They found a critical condition for f (T), which tells us that depending on its form, one could model collapsing stars or not. Regarding other modified teleparallel theories, some studies have been performed with stars in f(T, Θ) where Θ is the trace of the energy-momentum tensor. In [678], the authors used a perturbative approach and derived the TOV equation for this model and later, in [679], the same authors analyzed the case of quark stars finding the corresponding mass-radius and mass-central density relations. More studies regarding neutron stars are needed. Using pulsars observations, it was recently reported in [680] that in power-law f (T) gravity one can construct realistic neutron stars by considering certain bounds for the parameter of the theory. Other works that are important to mention are related to junction conditions which are needed to have a correct matching of different regions of spacetime. These conditions are important to construct a viable collapsing star model. In [681], they analyzed the junction conditions for f (T) gravity by using the ADM decomposition and the thin shell formulation finding that one requires more junction conditions than in GR. In this formulation, one imposes conditions in such a way to remove delta-like distributions in the field equations. The extra conditions in f (T) gravity are related to the tetrads and also with the continuity in the torsion scalar. This study was then generalized in [682] in a more general approach by considering the variational principle for a covariant version of f (T) gravity. The conclusion of this work states that for highly symmetric spacetimes such as in spherical symmetry, one has the same standard junction conditions as in GR but for more general situations, one obtains different junction conditions that are the same as the ones obtained in [681] in a specific limit. Recently, in [683] it was shown that the remnant group of local Lorentz transformations for teleparallel theories play a crucial role to match tetrads in order to ensure that scalars constructed from teleparallel quantities are continuous. As a final remark, it is interesting to mention that there are no works related to the thermodynamics or Hayashi tracks of stars [684] in modified teleparallel theories or in any other theory different to f (T). Further, the complex tetrad (5.153) has not been studied to construct stars. This could lead to some new phenomenology that could be important to study in the future. The study of black holes in teleparallel theories is an on-going research stream in the literature. Several studies like [685–687] have analyzed different types of black holes but in TEGR which in the end, is equivalent to GR. Different attempts have been made in order to find exact black hole solutions in modifications of TEGR, like Noether’s symmetry approach [355, 633], or directly by assuming different forms of the function for f (T) [688], but none of them are neither correct nor were able to find exact black hole solutions for the real tetrad. Recall that this tetrad provides a smooth limit to Minkowski for the scalars when we switch off gravity. Unfortunately, only perturbed methods (as the one mentioned in section 9.1) or numerical approaches have been used to study the real tetrad. For example, in [689], the authors studied the evolution of primordial black holes in f (T) gravity by assuming a Chaplygin gas. They studied different properties of such black holes like its accretion mass by using a numerical approach. In [690], the authors studied the matter conditions for regular black holes in f (T) gravity. To do this, they used a power-law form of f and study the equations of motion of the interior of the black hole by determining the structure of matter in the vicinity of a possible singular point. Basically, the authors used perturbed approaches to determine the conditions for which it is possible to find non-singular black holes. Later, in [691], p the authors studied a Born-infield like form of f(T) = λ( 1 + 2T/λ − 1) and find the existence of a regular black hole interior which has a new length scale related to the Born–Infeld parameter λ. Due to this new length scale, they showed that the central curvature singularity is replaced by an infinitely long cosmic string (with constant curvature invariants). This means that the interior of the black hole is free from singularities. They used numerical and dynamical system techniques to show this, and then they used perturbation techniques 9.4.2. Black holes. 149 Rep. Prog. Phys. 86 (2023) 026901 Review to obtain an analytical perturbed solution for the interior of the black hole. Later on, the same authors noticed that this perturbed solution is of the form of a non-rotating BTZ black hole [692]. Using numerical techniques it was found that a squared power law f (T) gravity case does for the real tetrad with ξ = +1 does not contain asymptotically flat black hole solutions beyond TEGR [693]. Also in [694] the study for Schwarzschild in TEGR was reanalyzed to understand the role of the spin connection. As we discussed in section 9.1, perturbation theory allows us to obtain spherically symmetric solutions such as for example the ones expressed in equations (9.9a) and (9.9b). Since this solution contains a logarithmic term, it is not so easy to check if this solution can represent a black hole solution since we need A(rh ) = B −1 (rh ) = 0 and detgµν |r=rh 6= 0, with rh being the event horizon, to describe a black hole perturbed solution. A recent paper [658] showed that when the Ci are given by equation (9.11) and β = 0, the solution is not a black hole. However, one can also set Ci in a different way, which is α  − 6 + 12µh − 21µh2 − 108µ3h + 66µ4h C2 = 3µh2  + 20µ5h − 39µh6 + 12µ7h + 12µh2 ln(µh ) + ζ1 , (9.69a) where rc is the radial circular orbit. For the solution equations (9.7a) and (9.7b), we can find the photon sphere predicted by f(T, B) gravity. To do this, we need to expand rc = r0 + ϵr1 and solve V(r = rc ) = V ′ (r = rc ) = 0 order by order. By doing this, we get that the zeroth-order which is related to the Schwarzschild contribution, gives the standard√ photon sphere result r0 = 3M and the momentum h0± = ±3 3k0 M. This observable is related to the shadow of the black hole. Then, the first order correction of the photon sphere will depend on the perturbed solution found for the specific teleparallel model. It is interesting to mention that the prediction of the photon sphere is equivalent for the integration constants (9.69a) and (9.69b) than for (9.11) for f (T) gravity, and it becomes rph ≈ (1.5 + 0.282 675 β) 2M r . Another point that is worth mentioning is that black hole singularities in TG might have a different description as it is usually described in theories like GR. To analyze this, one usually uses the Hawking–Penrose singularities theorems which essentially defines singularities as being the points/paths where the geodesic is incomplete for moving particles. In TG the so-called Raychaudhuri’s equation might be different, so that, it is still unclear if the evolution equations for the congruence of curves would be the same. Moreover, a useful way to realize if one has a coordinate singularity is by constructing several scalars such as the Kretschmann scalar. It is also unclear which teleparallel scalars would give a hint whether one has a coordinate singularity. 32 C1 = 2MC2 − α, (9.69b) As pointed out in section 9.1, if one allows the torsion scalar 3 to be nonzero for the Minkowski limit, one can take the results of the complex tetrad (5.153) as being the first attempts µh − 1 2 3 4 5 6 to construct black hole configurations in modified TG. The ζ1 = − (6 − 6µ − 49µ + 59µ − 7µ − 27µ + 12µ ) h h h h h h 3µh2 recent solutions found in [641, 642] might help us to under(9.69c) stand more about black holes in these theories as well as the − 4 ln(µh ), role of the torsion scalar being nonzero for the Minkowski 1/2 where µh = (1 − 2M/rh ) , to have a black hole solution. limit. One remarkable aspect of the complex tetrad is that it is After showing that by choosing the above constants one has that it is much simpler to obtain exact spherically symmetric a perturbed black hole solution, Pfeifer and Schuster [658] black hole solutions. Further, there is a much richer family of analyzed several observables such as perihelion shift, Shapiro exact scalarized asymptotically flat black hole solutions in TG delay or light deflection finding that the influence of the per- than in the standard Riemannian case. Let us finish this section turbed solution is much smaller than the one derived from the by emphasizing that we are still very far away of understandsolution with the constants (9.11). Then, they analyzed several ing how black holes will be in modified teleparallel theories. thermodynamics quantities such as the surface gravity and the sparsity finding corrections of the Hawking effect. 9.4.3. Wormholes. A wormhole is usually described by the An important remark about black holes in f (T) gravity is well known Morris–Thorne metric which is given by [695] that it is not possible to have any solution beyond TEGR plus a cosmological constant with the metric functions satisfying 1 dr2 − r2 (dϑ2 + sin2 ϑdφ2 ), (9.71) ds2 = e2Φ(r) dt2 − A(r) = B(1r) . In other words, the only solution with A(r) = 1 − b(rr) 1 B(r) in f (T) gravity is the Reissner–Nordström solution and the form of f is always constrain to be −T + Λ. This conclusion is where Φ(r) and b(r) are the redshift and shape function, valid for any of the two tetrads. respectively. One then needs to have different properties for One interesting feature that black holes produce is the so- these functions to be able to describe a wormhole, like the called photon sphere, which is the radius that defines a region non-monotonically decrease of b(r) from infinity to a minimal where photons are forced to travel in orbits, that can be derived value r0 at the throat (b(r0 ) = r0 ), and the so-called flaringby taking circular photon orbits, which is obtained by setting out condition stating that (b − b ′ r)/b2 > 0 at the throat. For a more detailed description about wormholes, see [696–698]. photon sphere: σ = 0, and V(r = rc ) = V ′ (r = rc ) = 0, In order to study wormholes in TG, one needs to use a good (9.70) tetrad-spin connection pair. Since they are usually described 150 Rep. Prog. Phys. 86 (2023) 026901 Review as spherically symmetric (as the metric above), one can then choose a zero spin connection but a tetrad being of the form     A e µ=    eΦ(r) 0 0 0 0 0 0  −1/2  sin ϑ cos φ r cos ϑ cos φ −r sin ϑ sin φ  1 − b(rr)   −1/2 , b(r) 1− r sin ϑ sin φ r cos ϑ sin φ r sin ϑ cos φ     −1/2 1 − b(rr) cos ϑ −r sin ϑ 0  which is the same tetrad (5.147) with the specific metric coefficients chosen accordingly to reproduce the Morris–Thorne metric. The great majority of studies related to wormholes in teleparallel theories have been done in f (T) gravity. The first paper studying this using a good tetrad-spin connection pair found that spherically symmetric wormholes can be supported in f (T) gravity by matter satisfying the standard energy conditions [699]. This conclusion cannot be made in GR since in this case, one requires exotic matter to maintain the form of the throat of the wormhole. The main point here is that the modification of gravity, which in this case is related to the form of f (T), can create a repulsive gravitational effect to then leads to the possibility of having wormholes supported by standard matter without violating the energy conditions. This result is not unique in modifications of TG, since in many modified theories of gravity one can achieve this situation [700, 701]. In [699], they assumed a squared power-law form of f (T), and later, other authors assumed other models to studying wormholes. For example in [702] they used exponential and logarithmic forms of f finding that for the first case, the null energy condition must be violated to construct wormholes and for the later, one can get wormholes satisfying all the energy conditions. Similar studies like in [703–706] also found similar results but in a non-commutative background by setting a specific form of the energy density. In these different papers, different authors found that it is possible to construct stable wormholes exhibiting a conformal motion. Concerning other teleparallel theories different to f (T), there is only one paper where the authors studied a scalar field model non-minimally coupled with the scalar-torsion and the boundary term [707]. In this paper, they found exact wormhole solutions using the Noether’s symmetry approach. The main conclusion that one could point out then is that in modified TG, it is possible to construct wormholes satisfying the standard matter energy conditions. (9.72) which is a type of matter that has not been detected yet. Modified gravity also has tried to tackle this issue with the aim of not introducing dark matter but to fully describe the motion of galaxies using the new corrections appearing in the underlying theory. We here distinguish between the astrophysical dark matter responsible for galactic dynamics and the cosmological dark matter which is related to the stability of structure formation in the early Universe and the impact that has on the energy budget of the Universe in the CMB. In this context, we can attempt to explain astrophysical dark matter in galactic systems but leave open the problem of cosmological dark matter [524] which we tackle in section 10. In f (T) gravity, this can be achieved in a satisfactory way as it was explored in [358]. In this paper, the authors used a spherically symmetric perturbed solution around Minkowski for power-law f(T) = −T − αT n gravity by assuming the tetrad (5.147) with ξ = −1, which is the branch where T does not vanish in the Minkowski limit, which may be theoretically problematic. This does not happen for the ξ = 1 branch which is defined in equation (5.147), and may be an interesting analysis to perform and compare. In this case, the solution is explicitly given by the metric (9.1) with  A(r) = 1 − ϵ  B(r) = 1 + ϵ  2M r2−2n 3n−1 +α 2 , r 2n − 3 (9.73a)  2−2n 2M 2 r 3n−1 + α(1 − 3n + 2n ) 2 , r 2n − 3 (9.73b) with n 6= 3/2 and ϵ  1 being the tracking parameter. In essence, this solution behaves as a weakly Schwarzschild plus an additional small correction coming from α. To determine the rotational curve profile, one can consider a test particle following orbits around the galactic core, which is described by the effective potential (9.27). In this work, the authors used the perturbed solution around a Minkowski background in equation (9.27) which would mean setting M to being a small parameter, and choosing a(r) and b(r) described by the perturbations around this background, and with the corresponding integration constant related to a small mass. Furthermore, they assumed h = 0, which effectively means that there are no meaningful angular momentum contributions from the 9.5. Galaxy rotation curves Besides the problem of explaining the late-time acceleration of the Universe in a theoretically satisfactory way, GR also suffers from the problem in explaining galaxy rotation curves. The only way to cure this in GR is to introduce dark matter, 151 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 30. Best fit for a power-law f(T) = −T − αT n with the constants n = 1.00001 and α = −4.505 731 524 05 × 10−34 km2n kg−1 s−2 , bulge surface mass density of 5.94 × 10 6 kg km−2 and disk mass 4.08 × 1010 M⊙ (these are the best fit parameter presented in [358]). Reproduced from [358]. CC BY 4.0. galactic system model. Note that the definition used in [358] differs from ours. In our definitions, we absorbed the energy contribution in the potential whereas in this work the potential does not have the energy. Let us consider their form of the potential (equation (23) in [358])   1 M r2−2n 3n−2 Veff = − ϵ +α 2 . (9.74) 2 r 2n − 3 to fit the observations. Figure 30 shows the best-fit model for the α parameter for the rotation profile for the Milky way for a power-law f(T) = −T − αT n where one can notice that the profile fits the data with a good accuracy. As a final remark, let us again mention that the authors used a tetrad with T 6= 0 for Minkowski, which may be problematic foundationally. If one uses the same tetrad in equation (5.147) with ξ = 1, one avoids this issue but there are no perturbations around Minkowski for f (T) for this tetrad. Thus, to analyze the behavior of galaxy rotation curves in the case ξ = 1, one would need to use the perturbed solutions around Schwarzschild (see equations (9.9a) and (9.9b)) and assume M  1 which is a valid approximation in this astrophysical scenario. Next, we need to find the velocity profile that can be obtained by considering the following centripetal and gravitational acceleration ac = v2 , r ag = − dVeff , dr (9.75) 9.6. Overview and future work which gives us 2 veff = −r dVeff . dr As was stressed through this entire section, there are many unsolved and unexplored problems in modified teleparallel theories of gravity in the context of astrophysics. Some important open problems are listed below: (9.76) The simplest model of a galaxy is composed by two regions, namely, the disk and the bulge and each of them must be taken in a separated way. The standard approach for the disk is to use a velocity profile composed by the sum of the combined potential of the individual sources within the galaxy as described in [708]. For the bulk, one assumes that this region is described as a spherical mass with a Vaucouleurs profile [709]. Following this analysis, the authors in [358] used data from the Milky way and three other galaxies to fit the parameters α and n finding different sets of constants in such a way to describe the motion of the galaxy rotation curves in a satisfactory way without introducing any dark matter component. The main constraint that they found is that one requires n to be small (a) Exact or numerical black hole solutions—As was pointed out in section 9.4.2, for the real tetrad, there are only perturbed solutions or numerical solutions, but there are no analytical exact solutions describing black holes. Furthermore, numerical black holes solutions have not been analyzed in detail yet for theories that differ from f (T) gravity. It would be interesting to analyze the solutions for the complex tetrad in more depth. (b) Quasi-normal modes—One interesting characteristic of the waveform of a merger of black holes is its final stage which is known as its ringdown phase. In this phase, the 152 Rep. Prog. Phys. 86 (2023) 026901 Review system is radiating GW in the form of quasi-normal radiation. A recent paper [710] studied the case of TEGR for Schwarzschild which can help as a tool to extend it for modified TG. (c) Time-dependent spherically symmetric good tetradspin connection pair—Thus far, it has been quite hard to perform a similar approach as in section 5.12 for a generic time-dependent spherically symmetric spacetime beyond TEGR. This result could be very important for describing collapse of stars or if there is a way of finding theories beyond GR satisfying the Birkhoff’s theorem. Furthermore, one could also study time-dependent spacetimes such as cosmological wormholes. (d) Most general axially symmetric good tetrad-spin connection pair—As discussed in section 5.12.3, only some specific cases were derived. For example, a Taub-NUT-like form or a slowly rotating Kerr case has been found but, what is missing is a general axially symmetric good tetrad spin-connection pair which can include non-slowly rotating Kerr as a special case. If one has this, one could study axial symmetry in a generic way for any astrophysical system. (e) Galaxy rotation curves—As discussed in the previous section, in [358], the authors investigated how galactic rotation curves are described in f (T) gravity. They considered two different characteristics of galaxies: disk and the bulge. It would be interesting to re-analyze this problem by using other perturbed solutions (see equations (9.9a) and (9.9b)) for f(T, B) gravity (and other theories or models) and then use Milky way data to fit the model with the observations which will add to the consistency of theory. More ambitiously, it would be ideal to constrain these models again a number of galactic rotation curve data sets. (f) White dwarfs and neutron stars—As discussed in section 9.4.1, there are some studies regarding compact stars in TG but the majority of them are only focused on solving the TOV equations numerically to then find the mass-radius plot. There are no studies regarding other aspects of compact stars such as their early evolution that can gives us information about other characteristics of stars as their effective temperature, luminosity, radiative core or how stars appear in the Hayashi tracks [684]. In addition, all the works in TG have been only focused on regular stars and then, more studies regarding neutron stars or white dwarfs are needed. Further, the complex tetrad have not been used to study such configurations. (g) Exotic objects and GW echos—As we discussed in section 9.4.3, there are some studies about wormholes in TG. Nevertheless, there are no studies about other exotic objects such as quark stars, gravastars or boson stars. Further, the GW echos of exotic objects are important observational features (see [711]) that have not been computed in TG. (h) Black hole singularities—As discussed in section 9.4.2, the Raychaudhuri’s equation in TG might have a different behavior as the standard one. This might lead to a different interpretation of singularities in TG. Further, coordinatelike singularities are usually understood by constructing several scalars such as the Kretschmann scalars. In TG, one would need to construct other scalars to understand these types of singularities. (i) 3 + 1 numerical simulations for GW observations— The 3 + 1 split formalism helps us to write down the Einstein’s field equations in a particular form (Gauss-like equations) which is useful for numerical studies [712]. We know that TEGR has the same field equations as GR. However, as we have seen in section 4.1.1, they are expressed in terms of teleparallel quantities. What can happen if one uses the 3 + 1 formalism for the TEGR field equations? could it be that this formalism leads to an easier numerical description of the Einstein’s field equations? A first attempt toward understanding these questions were addressed in [713] and further followups studied are needed. (j) Weak lensing for galaxy distribution—There are no studies performing analyzes of weak lensing for galaxy distributions. This might gives us some hints toward understanding dark matter in a more satisfactory way and to understand better the physics of galaxies. The above open problems are just some possible examples that have not been addressed in the literature. To fully describe astrophysical objects in TG beyond TEGR, one would need to start solving some of them. 10. Observational and precision cosmology in TG This section explored the pivotal regime of observational and precision cosmology of TG and the various models that emerge from the earlier analyses. Here, we review the plethora of works in the literature on the topic, and explain how future works may open new possibilities for the most promising teleparallel models being explored in this Review. More than using data to constrain the cosmological parameters inherent in these theories, when we discuss about precision cosmology, more than a qualitative (and statistics) endeavor, this is about a deeper understanding of the physics behind these theories of gravity (i.e. we want to go toward the inverse cosmology problem, that is to say, we want to know if the observations at hand can give us information about the nature of gravity). The ΛCDM model has been extremely successful in meeting most of the observational data up to now. Assuming only GR and the well-understood linear perturbations about a homogeneous and isotropic background cosmological model, with just six parameters, this model accounts for all cosmological observations on a wide range of scales. To achieve this we require the cosmological constant Λ, a constant usually associated with the vacuum energy, and dark matter, yet-undetected particle (or particles) which are predicted [714], for example, by supersymmetric extensions of the standard model of particle physics [715]. While the theoretical backbone for both the cosmological constant 153 Rep. Prog. Phys. 86 (2023) 026901 Review and dark matter may be justly questioned, the ΛCDM model has, hard-headed, maintained its hegemony; this is due to its performance in reproducing the observational data along with its numerical simplicity. However, since the first release of the CMB observations by the Planck Collaboration in 2013 in [716], the determination of the Hubble constant H 0 based on the concordance model of cosmology started to be in tension with the cosmologyindependent determination of this constant, e.g. via local Supernovae Ia (SN) calibrated by the Hubble Space telescope in 2011 in [717]. The initial tension of 2.4σ has been exacerbated over the years. On the one hand, systematic uncertainties have become better understood and CMB data has accumulated. On the other hand, the survey of local SN has increased and the anchors used to calibrate them significantly improved. The current status is that the two inferences of the Hubble– Lemaı̂tre constant, one by the Planck Collaboration in 2018 [372] assuming the ΛCDM model, and the other by the SH0ES Collaboration in 2019 [718], are in tension at the 4.4σ level. Furthermore, using complementary sources of Cepheid calibration [719] we now reach a 4.2σ difference with the prediction from Planck and recently, a 5σ using observations from HST of Cepheids [720]. Many articles have been written trying to understand the implications of this tension: so far, it is the most severe problem confronting the standard model. The effect of local structure—the so-called cosmic variance on H 0 —has been thoroughly analyzed, along with thorough inspections of the error budget. Furthermore, physics beyond the concordance model has been investigated, with the aim that this tension could reveal alternatives to the highly tuned cosmological constant and the yet-undetected dark matter. So far, there is no well-posed panorama beyond the standard model that could solve this issue [721]. Moreover, the ΛCDM model successfully describes a range of observations from several epochs of the Universe: from primordial nucleosynthesis to the accelerating late-time cosmic expansion. As the standard cosmological parameters of this model are being inferred with increasing precision, there is no guarantee that the same model will fit more precise surveys from different cosmic epochs. The current cosmological field set on the ΛCDM model is supported by observational evidence as a description of the cosmic evolution on all cosmological scales [2, 22], which is achieved by considering matter, the so-called dark matter, beyond the standard model of particle physics. Modeling itself this way, dark matter plays an important role in galactic structures [722, 723] and is manifest as CDM particle(s), while DE in its simplest description is denoted by the cosmological constant [316, 724] capable of producing the late-time cosmic accelerated expansion [19, 20]. However, despite great efforts, some consistency problems persist with this constant [21], as well as a strong lack of direct observations of dark matter particle(s) [725, 726]. Furthermore, this standard cosmological model was designed to describe the Hubble flow, but the so-called H 0 tension problem now casts a shadow over this—given that the observational discrepancy between cosmology-independent measurements [718, 727–734] and predicted ones [372, 376, 735, 736] from the early-time Universe (assuming the ΛCDM model) seems to be growing. While measurements based on, for example, the tip of the red giant branch [737] (TRGB, Carnegie–Chicago Hubble Program) or strong lensing measurements TDCOSMO+SLACS [738] have a relatively small mean value of H 0 , their much larger error bars mean that they are currently unable to distinguish between the two faces of the tension. This issue could be resolved by future observations which involve new measuring techniques such as the use of GW sources [739, 740] observed with ground-based detectors (such as the Einstein Telescope [741]) or by the LISA mission [742, 743]. Let us consider first two questions for the early-time Universe probes: (i) what kind of smoking-gun signature can be used to identify unknown systematic errors that could affect the predictions for H 0 for the late Universe that is independent of the model of cosmology? (ii) does there exist any indication of tension in the early-time data that could reveal systematic errors of the standard six-parameter ΛCDM model that is exhibited purely in the early Universe? Aside from the well known small difference between the inference of H 0 from low and high angular resolution Planck data, all of the earlytime data seem to be consistently predicting a low value of H 0 . Experiments like ACT and SPT, which use CMB data to calibrate the sound horizon and baryonic acoustic oscillations (BAO), are in agreement with Planck. These methods lead to low H 0 values of ∼67–68.5 km s−1 Mpc−1 . Another independent, consistent value of H 0 can also be obtained by using light element abundances to calibrate the sound horizon, BAOs and lower redshift probes. A compilation of the H 0 estimates up to 2019 is detailed in figure 31. Our second question concerns anomalies with significant tension among high-redshift probes [744]. Some of them appear as a departure from unity, at ∼2.8σ, of the inferred parameter Alens , a phenomenological parameter firstly introduced in [745], which is the amplitude of the gravitational lensing in the angular power spectra and can be used to rescale the lensing effect in the CMB 2-point correlation function. If this deviation is confirmed, it will be clear evidence that something is not well understood in the relation between CMB anisotropies and the growth of structure. Up to now, we also have other 3σ tensions: (i) the tension between the σ 8 -Ωm0 contours inferred from the CMB in a ΛCDM scenario and those inferred by cosmic shear data [746–751]; (ii) the tension between the cosmological parameters from the BAO signal in galaxies at z < 1 and those of Lyα at higher redshift [752] and (iii) time derivative of the cosmological parameters with the CMB fluctuation scale p used to determine it [753]. A compilation of the S8 := σ8 Ωm0 /0.3, discussed in section 7.7 and estimates up to 2020 is detailed in figure 32. The statistical errors of the above methods are expected to decrease in the next few years, and will reveal whether the tension is a statistical accident that we can expect when considering of order dozens of nuisance 154 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 31. Hubble Constant estimates according to the publication year. The CMB-based, early Universe mean value for H 0 is 67 km s Mpc−1 . The Cepheid-based, late Universe mean value is 74 km s Mpc−1 . A new alternative to Cepheids, i.e red giant stars that flare with a known intrinsic brightness, has only complicated the tension. They indicate a H 0 mean value of about 70 km s Mpc−1 , a value that is midway between the other two, but in agreement with both the measurements within 2σ. Other measurements, e.g. ACTPol 2020 (with H0 = 67.9 ± 1.5 km s Mpc−1 ) [376], ACT DR4 + Planck 2018 (with H0 = 67.53 ± 0.56 km s Mpc−1 ) [760], HST photometry + Gaia EDR3 (with H0 = 73.2 ± 1.3 km s Mpc−1 ) [719] have also been considered. Reproduced from [761]. CC BY 4.0. Figure 32. S8 estimates according to the publication year from four different sources: WL, x-ray clusters, Sunyaev-Zel’dovich (SZ) clusters and CMB surveys. The error bars denote 1σ CL and the year of publication for each analysis is denoted below the estimate, respectively: [372, 376, 762–772]. The orange bars denote constraints from DES Y1 2018 with Planck, BAO from SDSS, 6dF, and BOSS and JLA SNIa 0.024 up to 2σ [749]. The dashed lines denote the 2σ limit of KiDS-1000 [765]. Notice that for the case S8 = 0.759+ −0.021 from KiDS-1000 according to [748], while the quoted number in reference [765] is a combination with BOSS. parameters, or whether it is hinting at yet to be discovered new physics. For example, the latest DES-y3 release [750, 754, 755] is in agreement with the Planck ΛCDM predictions, but it is showing an inconsistency of the amplitudes of the galaxy clustering with the galaxy–galaxy lensing. Even so, many have questioned if a solution to the late versus early-times discrepancy could be more credible if it also relaxed one or more of these additional tensions [744, 756–759]. 10.1. State-of-the-art Over the last two years, TG has started to take advantage of its analytical and physical description to gravity in order to provide a solution to the H 0 and S8 tension problems (see [721, 757, 758, 773, 774] for an overview of the tensions) and describe the observed and well-tested late-time dynamics with current cosmological data [775, 776]. As we have 155 Rep. Prog. Phys. 86 (2023) 026901 Review discussed, in these kind of theories the curvature description is torsion-based one by the replacement of the Levi-Civita connection with a curvatureless connection labeled as the teleparallel connection [777]. It turns out, as discussed in detail in section 2.3, at the level of the action GR and TEGR differ just by a boundary term which renders their field equations identical. There are several ways to modify the TEGR proposal (see section 5 for a review of them). With this methodology, we can generalize the Lagrangian to arbitrary functions of the torsion scalar in order to produce f (T) gravity [38, 39, 568], the other hand, massive binary black holes as standard sirens have the disadvantage that the redshift is difficult to measure without an independently EM signal of the event. However, recently some progress has been made in determining the redshift from GW observations alone [782]. If we identify the host galaxy then it should be possible to determine with joint events whether long-wavelength GW and short-wavelength EM radiation undergo the same number of spacetime dimensions [783]. In section 8 we gave an overview of GW in teleparallel theories. In this setting, another interesting feature of GW astronomy is its potential in probing higher dimensional theories where spacetime leaks into the extra dimensions, leading to the effect that cosmologically distant sources appear dimmer than they truly are, resulting in a systematic error in the inferred distance to the GW source [784]. Also, there have been studies considering the propagation from the inspiraling of compact binary systems within the context of TG theories in [596, 785], meanwhile it is important to mention that these studies are in the context of a scenario where the GR cosmological model is considered as an effective fluid in the standard perturbations equations. These studies lead open the idea of explaining these effects using non-standard extensions of GR, as it is in TG theories. Moreover, unless higher precision probes of the Hubble flow H 0 and the growth of structure S8 will be available without tension issues, these theories cannot be discriminated. The issue about the theoretical playground suggests that an approach to the problem of cosmic acceleration should rely on as few model-dependent quantities as possible. Since the number of viable gravitational theories at hand is large, on both astrophysical and cosmological scales, as well as in stronggravity regimes, some of them could be good tools to compare theoretical predictions with survey observations, extracted from the CMB, cosmological expansion, large-scale structures, astrophysical configurations and GW signals. Moreover, these comparisons suffer from the so-called degeneracy problem, which means that several gravitational theories are capable of describing certain observations with the same statistical confidence. A standard way of relaxing this issue consists of combining several observational measurements to reduce the phase space of free parameters. The search for viable priors may result in numerical issues, where a possible solution could be to consider a cosmography approach rather than determining model-specific solutions from Friedmann equations and confronting them with observations. In this approach, cosmographic solutions are expressed, in the standard way, as derivatives of the scale factor, introducing the so-called cosmographic parameters which makes possible to fit the observations to the distance-redshift relation without prior theoretical assumptions. In the standard approach to cosmography, we start with the FLRW metric and extend its parameters beyond the value of H 0 , which evokes, at least, a fifth order Taylor expansion of the a(t) in order to obtain a reliable approximation of the distanceredshift relation [786–789]. This is needed when one is dealing with certain theories containing higher-order derivatives in the Hubble parameter. ◦ which follows the same idea as f(R) gravity. However, unlike ◦ f(R) theory, f (T) gravity produces generally second-order field equations, which makes them more easily open to test with observations. Furthermore, given the importance of the boundary term in relating GR and TEGR, f(T, B) gravity has also been well-studied in [157, 263, 446] as a possible extension to TEGR. This theory has been studied throughout this Review to describe FLRW cosmology at the background level and at the perturbation level (see sections 6.2 and 7). Moreover, applying consistency tests with current cosmological data allows one to identify a viable gravity theory and deal at the same time with systematic effects in the data or any problems with the underlying cosmological model. Some data samples are sensitive to the theory and the dynamics of the Universe and some other samples are sensitive to the growth of large-scale structure [778], e.g. in the early-time route of the distance ladder, BBN allows a technique to measure precise abundances of hydrogen, helium, lithium and deuterium, and hence to test the cosmological model through being extremely sensitive at early-times [779]. In these cases, these two sets of observations must be consistent with one another in order to solve the cosmological tension inside a viable theory of gravity. At late-times, any deviations between different theories can be measured through an effective EoS (mimicking a DE component) close to the standard ΛCDM EoS value (w = −1). Although the ΛCDM model is very successful in explaining almost all observations [372, 376, 762–766, 768–770, 772], it has some theoretical issues within its backbone structure. These include the mysterious physical origin of the two largest contributions to the energy content of the late-time Universe: CDM and the cosmological constant, together with the too varied forms of inflation in the literature and their as yet unobserved nature. Furthermore, through their mapping of the distanceredshift relation up to a redshift of z ∼ 1, SNIa measurements have been a sensitive probe of the late-time cosmic acceleration. Another very powerful probe are GW, e.g. those emitted from merging binary black holes and neutron stars, since the propagation of gravitons and photons can change at a fundamental level [780]. To achieve this, GW observations are sometimes coupled with a multimessenger signal through an electromagentic (EM) counterpart. These kind of sources are treated as standard sirens (gravitational analogues of standard candles such as SN) in the sense that the GWs emitted by a compact binary contains information about the GW luminosity of the binary system, and thus its luminosity distance [781]. On 156 Rep. Prog. Phys. 86 (2023) 026901 A similar issue was tackled in [790], where the cosmographic equations can be written down in an inverted version to study the dynamics through kinematics. In this approach, a homogeneous SN observational sample was used to perform the statistical analysis of specific DE parameterizations, to directly obtain the kinematics of the cosmographic parameters. This approach relies on the fact that it is possible to set constraints over the cosmographic parameter values using statistical analyzes of the dynamics imposed in DE. In the literature, there are many studies that attempt to use these SN observational samples to constrain the cosmography of models beyond ΛCDM. However, due to the low redshift nature of the data sets in question, it may be misleading to consider such cases since it is unlikely to reveal any new dynamics without higher corrective terms. A study in that direction was developed in [791], where an improved version of the standard cosmography approach was shown to be a viable solution to these issues. However, the approach is plagued with several shortcomings which limit its applicability in certain circumstances, and even more in extended theories of gravity. The main problems concern the inability of the proposed polynomials to be constrained with the currently available cosmological data so as to fix the kinematic expansion of the Universe, especially at early-times. Also, the arbitrary order of truncation of polynomials might compromise the predictive power of cosmography. The impossibility to measure them separately but only their sum, leads to different results depending on the probability distribution associated with each coefficient. To relax these issues, in [790, 792] a mathematical expression to derive the EoS for a specific dynamical model was proposed. This expression allows us to obtain the cosmographic parameters without directly assuming a cosmography-dependent polynomial series over them and avoid dealing with the aforementioned problems. To apply these ideas to extended theories of gravity, a first step was performed in [793], where a novel computational tool based on machine learning for SN, called Recurrent-Bayesian Neural Networks, was proposed to deal with different kinds of modified gravity models. A deep learning architecture can develop a trained homogeneous SN sample, where the resulting dynamics of DE could lead to the necessity of another cosmological model different from ΛCDM. In this scenario, the inverse cosmography approach can fit statistically very well since it is not necessary to consider higher-order corrections of the cosmography series to obtain a convergent best-fit in comparison to the standard DE EoS. Over the last few years an avalanche of observations and numerical techniques have been bringing new ways to test extended theories of gravity [794–796]. Our purpose in this section is to set the optimal conditions and recipes to treat them at the same level as the cosmological viable models derived from GR. 10.2. Viability of teleparallel cosmology In order to handle the analyses related to specific TG models we require a strict recipe of the tools, theoretical and numerical, to study the cosmology derived from these models. To Review achieve this goal in the following sections (including information on standard statistical cosmology in the supplementary annexes (supplementary 3 and 4), we present several discussions on how to adapt TG models at numerical level and study their cosmological constraints. We can enlist these tools as follows: • Numerical codes and gravity model algorithms—In the supplementary annexes (supplementary 3) we present three general algorithms to explain in detail how to implement some important TG models using the standard numerical codes rewritten to analyze background and perturbative cosmology. Furthermore, we describe how to proceed with the Boltzmann code in order to adapt TG models. • Surveys and cosmological data—In the supplementary annexes (supplementary 4) we describe the current observational compilations available in the literature and how to handle the statistical analyses using χ2 methods and Bayesian inference. We will use these compilations in the description and computation of the best-fit cosmological parameters for several TG scenarios. • Cosmological compendium of Teleparallel scenarios— Using the information described in the supplementary annexes (supplementary 3–4) we are ready to present in section 10.3 the cosmologically viable TG scenarios in order to study the current cosmic issues as described in the previous section. In table 21 we present the most important extended theories where observational cosmology was studied, along with their references. As an extension, in tables 22 and 23 we describe in detail the motivation, viability, maturity and late/early tests related to each theory under consideration. • Precision cosmology for f (T) and f(T, B) gravity theories—Using the viable models described in the compendium, we present complete analyses for both theories using the respective modified version of Boltzmann codes and latetime data in section 10.3.1. We further analyze the possibility of relaxing the H 0 tension at the background level. • Cosmography for TG—This approach can be considered as a possible model-independent approach to tackle extensions and modifications of GR in view of the current observational constraints, while breaking the degeneracy among models. In section 10.3.4 we will revise the proposals already available in the literature and confront them with precision cosmology tools, to present an update of the results. • Beyond data-driven analysis for TG—Reconstruction techniques that are model independent have since become more robust and better understood in terms of cosmological data sets. These nonparametric approaches, as they are commonly called in the literature, allow us to extract from the data the best cosmological model. Using this nonparametric view, and as a perspective for future TG scenarios analysis, in section 10.4 we describe two current approaches on that path: (i) Reconstruction of data using Gaussian processes (GP) that can offer an avenue by which gravitational theories can be reconstructed, guided by observational data without imposing stringent physical model assumptions. In 157 Rep. Prog. Phys. 86 (2023) 026901 Review Table 21. Compilation of cosmological models from TG theories. The cosmological models labeled at the second column correspond to the ones discussed in the references indicated at the first column. All these models involve statistical tests and constraining cosmological parameters analyses with observations. The type of observables and cosmological behaviors are detailed specifically in tables 22 and 23. Theory Cosmological models constrained by observations f (T) [359, 401, 422, 489, 803–811] (1) Exponential Law: −T + 1 − e−T (2) Power Law T α (3) Linear-Power: −T + β(−T) n (4) Logarithmic: −T + γ log (δT) (5) Hybrid: −T +√eϵT (−T) α (6) Linder: 1 − e T/ζ (7) Trigonometric: −T + (−T)n tanh   T0 T (8) Model independent: ηT + (T − T0 ) coshh(T − T0 ) + sinh(T − T0 )  (9) Specific w-EoS (Hypergeometric):α × β 2 F1 f(T, B) [263, 627, 812] f(T, ϕ) [38, 536, 813–815] f(T, Θ) [45, 323] ◦ ◦ f(T, (∇T) 2 , □T) [286] ◦ ◦ Tf(□−1 T, □−1 B) [335, 495] −T 1 2,− 6 + γ 2 F1 1 −T 2,− 6 i (1) Power Law: −T + αBn + β(−T)m (2) Mixed Power Law: −T + αBn (−T)m (3) Logarithmic: −T + α log B (4) Taylor polynomial: f(0) + fT (0)T + fB (0)B + fTB (0)TB + fTT (0)T 2 + fBB (0)B 2 + O(TB, T 2 , B 2 ) Mixed Power Law: −T − G(T)G(ϕ) − V(ϕ), with G(T) = Ts , F(ϕ) = ξϕc and V(ϕ) = λϕd (1) Power Law: −T + α(−T)n Θ + Λ (2) Power Law (second order): −T + αΘ + γT 2 ◦ 2 Model I: T + α1 (T∇2T) + α2 e ◦ ◦ ◦ δ(∇T) 2 T4 ◦ σ □T 2 Model 2: T + β1T□T + β2 (T□3T) + β3 e T3 ◦ Model I: f = −1 − □−1 T ◦ ◦ Model II: f = −1 − □−1 T + □−1 B modified TG, this method has been applied to f (T) models using compilations of galaxy ages and observations of BAO calibrated with Hubble data. (ii) Machine learning tools for TG through deep learning techniques, which is a field of machine learning that uses several layers of non-linear processing neurons to obtain and transform at each successive layer an output from the previous layer. A new deep learning method to classify f (T) models is presented in this section. 10.3. Compendium of teleparallel cosmologies In order to confront different cosmological models with observations, and in view of the diversity of cosmological models to be confronted with observations, in this section we present a theoretical and observational state-of-art of the current TG scenarios. An overview of different cosmological models considered in the literature for different TG theories is presented in table 21. This includes a compilation of constrained cosmological models discussed in the references indicated at the first column. The second column denotes the cosmological models already constrained by observations in the references mentioned. All these studies involve statistical tests on the family of models labeled and analytically detailed there. Notice that these equations were rewritten in comparison to the expressions in the reconstruction tables in the supplementary annexes (supplementary 1), in order to reduce the free parameter phasespace to be constrained. This is to ensure the convergence of the statistical test over the cosmological parameters of relevance and related directly with observable, e.g. the modulus distance luminosity for supernovae. Together with the analysis of the cosmological models described, table 22 is detailed the following 2 different characteristics for each theory: • Cosmological behaviors—These involve theories that can reproduce scenarios as early/late accelerated expansion, the dark sector and phantom/quintessence. • Maturity—Models that are already constrained by observations. It is indicated also which of them are not analyzed with observations yet. All the models described in each theory are cosmological viable since they can reproduce several epoch of the Universe: from non-accelerating, matter dominated and late-time expansion era. To complete the description of both tables 21 and 22, in table 23 we included the observations used to constraint the models described. Notice that the analyses were made with late-time samples and early-time samples. For this we include the following catalogues described in detail in the supplementary annexes (supplementary 4): • SNeIa—Pantheon compilation. Also, JLA 2015 from [797] and Union 2 (2010) compilation from [798]. • BAO—Sample calibrated as inverse distance ladder. Also, it includes: SDSS from [799], BOSS from [800], the 2dF 158 Rep. Prog. Phys. 86 (2023) 026901 Review Table 22. A table of cosmological behavior, which is a base for further observational constraints, and maturity, that indicates whether the particular theory can describe both early- and late-time behaviors in principle for the models reported in table 21. To that end, we list: (i) Early/Late accel. means that the theory models can explain early inflation and late-time accelerated expansion; (ii) Dark sector refers to the ability of this theory to explain both the cosmological dark matter and dark energy sector in terms of an effective dark energy; (iii) Phantom/Quintessence indicates whether the models can produce EoS values above and below −1; (iv) Perturbations (scalar/tensor) means that it produces reasonable scalar perturbations and tensor perturbations that predict GW propagation speeds equal or very close to that of light; (v/vi) Early/Late constraints indicates whether these models have been constrained against late-time data sets (e.g. SN, BAO or H(z)) and early-time data sets (e.g. CMB, BBN, BAO or WL) (see table 23 for further details). Cosmological behaviors Maturity Theories Early/ Late accel. Dark sector Phantom/Quintessence Perturbations (scalar/tensor) Early constraints Late constraints f (T) f(T, B) f(T, ϕ) f(T, Θ) 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 f(T, ∇T, □T) 3 8 3 8 8 8 −1 3 8 3 8 8 8 ◦ ◦ Tf(□ ◦ T) Table 23. A table of early- and late-time constraints in the literature for the models reported in table 21. Here, NAY means not analyzed yet in the literature which is different to the indication provided by the theoretical work highlighted in table 22. The asterisk indicates that the theories are constrained using CMB at background level which refers to whether the corrected Boltzmann equations have been used to analyze observational data in the literature. Observables Late-time Early-time Theories SN BAO H(z) CMB BBN BAO WL f (T) f(T, B) f(T, ϕ) f(T, Θ) 3 [Pantheon] 3 [Pantheon] 3 [Union 2] NAY 3 3 3 NAY 3 [CC-2018] 3 [CC-2018] NAY NAY 3∗ 3∗ 3∗ NAY 3 NAY NAY NAY 3 NAY NAY NAY 3 NAY NAY NAY f(T, ∇T, □T) NAY NAY NAY NAY NAY NAY NAY Tf(□−1 T) 3 [JLA] 3 3 [CC-2016] NAY NAY NAY NAY ◦ ◦ ◦ Galaxy Redshift Survey from [588] and BOSS, CMASS, Lyman-α [801]. • H(z)-CC—This includes the samples described in the supplementary annexes (supplementary 4) and also other one from [802]. • CMB—From Planck legacy 2018. • WL—Optical and infrarred samples from KiDS 450 + Viking (see figure 32). Therefore, one needs to take these results with caution. However, for more general TG theories, this issue has not been reported or studied in the literature since the Hamiltonian analysis has not been performed for them. 10.3.1. Precision cosmology for teleparallel scenarios. There are numerous alternative models to ΛCDM, as there are numerous problems with which they tend to struggle with. Some models are plagued by theoretical instabilities and others require at least some degree of fine-tuning of the model parameters in order to meet observational constraints. However, numerical analyses have their own limitations in practice. A major drawback is their lack of efficiency, as is evident in the effort poured into making existing numerical codes faster to cope with the ever increasing demand for numbers of independent realizations, volume and resolution. This is performed in order to match the characteristics of current and future observational surveys. Unlike the standard cosmological model, which is unique and is widely accepted as a concordance model, there are many modified theories, which makes it more difficult to allocate much effort to forecasting for individual models or performing a continuous parameter search. Furthermore, simulations are often used as a black Also we noted which of the theories are not tested with these observations yet. While there are reported constraints of these theories using early-time measurements, it is relevant to mention that all of them are performed by assuming the cosmological model derived from TG as an effective dark energy fluid. Therefore, for future precision cosmology analyses of them it should be consider also the modified perturbed version of Boltzmann in the supplementary annexes (supplementary 3). One important remark to note here is that the results reviewed in table 23 depend on the cosmological perturbations of the TG theory. As it was pointed out in section 7, there is a strong coupling problem in f (T) gravity that makes the conclusions regarding perturbations unclear or incomplete. 159 Rep. Prog. Phys. 86 (2023) 026901 Review box, making the underlying physics untreatable, which does not help in developing reliable theoretical templates used in model constraints. Although a simulation can predict structure formation at very small scales, this information is not entirely trustworthy due to the uncertainties in modeling baryonic physics. In such cases we need accurate model predictions to further analyze the non-linear regime and when the physics governing the structure formation needs to considered. However, if approximation methods can be developed for the gravity models, and demonstrated to be valid in certain regimes, then it would be realistic to use observations in those regimes to determine model constraints much more efficiently. On this path, in this Review we will devoted to possible TG models that can offers a possibility to solve (or alleviate) the cosmological problems described. To proceed with the implementation of a specific TG model in the numerical cosmological codes, we start by describing the required cosmological evolution equations. In f (T) and f(T, B) theories it was found in section 6 that the EoS can be our starting equation to employ in constriction analyses. These EoS equations for each theory are linked to a specific form of f (T) and f(T, B), whose can be related to specific cosmological models in order to investigate the effects of a late-time cosmic accelerated expansion without the influence of an exotic DE or extra fields. With the background evolution and EoS equations, we can proceed with their implementation in the numerical code for each specific theory. The architecture structures to achieve this task are detailed in the supplementary annexes (supplementary 3), where for the background analysis we need to modify a Markov Chain Monte Carlo (MCMC) code and the information detailed in the supplementary annexes (supplementary 3–5). [822] and the SNeIa data corresponds to the SNLS-SDSS JLA [799]. This information is managed using the concepts described in the supplementary annexes (supplementary 4 and 5). The results described in the following sections are based on [816], were an adopted H 0 prior corresponds to HR16 0 = 73.24 ± 1.74 km s−1 Mpc−1 [823]. It is important to mention that such results, tighter to a prior, could have a larger significance which can derive in a worse constrain on the parameters in the f (T) models. However, we should also point out that the methodology described here will allow a better control on the flat prior in future analysis on this matter. On the issue of priors in the MCMC analysis, there are strong concerns about the statistical coherence of such analyses since they may put undue bias toward such values. This is especially poignant when putting priors on the Hubble constant. Putting prior values on in these analysis may spoil the posterior outputs and thus the estimates on the model parameters [824]. On the other hand, there have been numerous studies in the literature that assume a prior on the Hubble constant. These analyses may prove useful in understanding viable parameter values. However, caution should be exercised when interpreting these studies. In the standard f (T) literature we deal with the following models: (a) Power law model. We consider the power law model [273] (f1 (T)) as the first f (T) model, specified by f1 (T) = −T + α1 (−T)b , (10.2) where α1 and b are constant model parameters, such 1−b Ωm0 −1 that α1 = 6H02 2b−1 (this is found by evaluating the Friedmann equation at current times). We depict the inferred confidence regions for the power law model in figure 33 and the corresponding model parameter constraints in table 24. It is clear that the ΛCDM model is within the 1σ confidence level (CL) when the CMB data sets are not considered, while the f1 (T) model becomes consistent with the concordance model of cosmology at 2σ when including the CMB data. (b) Square–root–exponential model. The second f (T) model (labeled as (f2 (T))) is the square–root–exponential model given by [39]   √ f2 (T) = −T + α2 T0 1 − e−p T/T0 , (10.3) Observing the form of the first Friedmann equations (6.7a), we deduce that in f (T) cosmology we acquire an effective DE sector of gravitational origin. In particular, we can define the effective DE density as [45, 817]   3 f TFT ρDE ≡ 2 − + . (10.1) κ 6 3 10.3.2. Precision cosmology for f(T). Now, we are ready to analyze a set of viable f (T) cosmological models. According to tables 22 and 23, there are several f (T) models proposed to explain the present accelerating cosmic expansion. In this Review, we consider specific forms of f (T) and confront them with BAO, CMB, SNeIa, and Hubble expansion data by using the publicly available MCMC code CosmoMC32 [818, 819] which is interfaced with the Boltzmann integrator CAMB33 [820]. For the analyses considered in this section, the Planck 2015 CMB distance priors [821] are adopted, while the H(z) data consists of 38 data points from m0 with model parameters α2 and p, while α2 = − 1−(1−Ω 1+p)e−p . From the left panel of figure 34 as well ad from the inferred model parameter constraints of table 25, the Base data set was not able to constrain the model parameter p. However, the inclusion of the other data sets led to a constraint on p, with the tightest constraint being derived when considering the CMB data set. (c) Exponential model. We now consider the exponential ◦ model [39] (f3 (T)), which is also motivated by f(R) gravity [825], and is specified by   f3 (T) = −T + α3 T0 1 − e−qT/T0 , (10.4) 32 Cosmological montecarlo documentation (available at: https://cosmologist.info/cosmomc/) 33 Code for anisotropies in the microwave background documentation (available at: https://camb.info) 160 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 33. Left: CL plots of H 0 and Ωm0 in the ΛCDM. Base denotes the combination of the JLA SNeIa and the local value of the Hubble constant. Right: Confidence regions of H 0 , Ωm0 and p ≡ b in the f1 (T) model. Base denotes the combination of the JLA SNeIa and the local value of the Hubble constant. Reproduced from [816]. © IOP Publishing Ltd. 2018. The American Astronomical Society. Table 24. The mean value and the corresponding 68% limits [816] tightly constrain DE models because of the lack of low redshift data. However, it has been repeatedly shown that when the forthcoming redshift drift data is jointly combined with any other existing cosmological data set, this enhances the current model parameter constraints and leads to fewer parameter degeneracies. In [826], the authors first constrain the power law and exponential f (T) models using a combination of current data sets from SNeIa, CMB, BAO, and a direct measurement of the Hubble constant H 0 , and then choose the best-fit models as fiducial models by producing 30 mock redshift drift data points, which are then used to jointly constrain the f (T) models with this simulated data. The inferred constraints are illustrated in the panels of figure 35 for the power law f (T) and exponential f (T) models. Furthermore, advances related to the current scenario have been done using early-time data. In [828, 829] it has been reported that perturbed exponential infrared f (T) produces a CMB spectrum identical to ΛCDM. of the model parameters of the f1 (T) model. Base refers to the JLA SNeIa data and H 0 prior. Data set H 0 (km s Mpc−1 ) Ωm0 b Base 73.2 ± 1.8 0.100 0.284+ −0.077 0.64 −0.02+ −0.37 Base+BAO 73.3 ± 1.7 0.30 0.13+ −0.24 Base+BAO+H(z) 70.0 ± 1.1 0.040 0.269+ −0.030 0.020 0.255+ −0.016 Base+BAO+CMB 69.4 ± 0.9 0.10 0.298 ± 0.007 −0.11+ −0.08 Base+BAO+ H(z)+CMB 69.4 ± 0.8 0.09 0.298 ± 0.007 −0.10+ −0.07 0.19 0.10+ −0.17 1−Ωm0 with α3 = −1+( 1+2q)e−q , and q is the remaining model parameter. Similar to the previously considered model. The derived model parameter constraints are reported in table 26, where the model parameter p ≡ 1/q has always been constrained by the data sets being considered. The corresponding confidence regions are illustrated in the right panel of figure 34. The generic EoS provided in equation (6.28) which we now consider for four possible models in an f(T, B) gravity setting. Our goal is to study the effects of dynamical DE for each models. 10.3.3. Precision cosmology for f(T, B) gravity. There have been other studies of a similar nature in the literature such as [827]. Upcoming constraints on the model parameters of f (T) gravity are expected to be improved, for instance, [826] explores the confrontation of f (T) with forthcoming redshift drift data via the Sandage–Loeb test. It is expected that the thirty–nine metre European Extremely Large Telescope (E-ELT) will be equipped with a high–resolution spectrograph called CODEX (COsmic Dynamics EXperiment), which is designed to collect such Sandage–Loeb test signals. It was found that the redshift drift data alone cannot (a) General Taylor expansion model. Taking the general Taylor model that is introduced in equation (6.43), the EoS for the (background) effective fluid component can be determined through equation (6.28) giving34 : 34 We represented here a (4) = d4 t/dt4 . 161 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 34. Left: CL plots of H 0 , Ωm0 and p in the square–root–exponential f2 (T) model. Right: Confidence regions of H 0 , Ωm0 and p ≡ 1/q in the f3 (T) model. Base denotes the combination of the JLA SNeIa and the local value of the Hubble constant. Reproduced from [816]. © IOP Publishing Ltd. 2018. The American Astronomical Society. Table 25. The mean value and the corresponding 68% limits [816] of the model parameters of the f2 (T) model. Base refers to the JLA SNeIa data and H 0 prior. Data set H 0 (km s Mpc−1 ) Base Ωm0 p 73.2 ± 1.7 0.056 0.278+ −0.070 <0.23 Base+BAO 73.4 ± 1.7 0.279 ± 0.025 0.67 −0.39+ −0.80 Base+BAO+H(z) 70.0 ± 1.1 0.260 ± 0.015 0.49 −0.26+ −0.57 Base+BAO+CMB Base+BAO+H(z)+CMB 69.5 ± 0.9 69.5 ± 0.8 0.298 ± 0.007 0.297 ± 0.007 0.44 ± 0.32 0.41 ± 0.31 −24A3 + 6(H ′ + H 2 ) + 24H3 wx1 − 6wx2 − 24H5 wx7 + wx9 −6(H ′ + H 2 ) + 72H3 wx4 + wx5 + wx6   −24wx7 ȧ(t)5 + 6a(t)4 ä(t) − 4A3 a (4) (t) + 24a(t)ȧ(t)3 wx1 − 6a(t)3 wx2 + wx9 = , −6a(t)4 ä(t) + wx5 + 72a(t)ȧ(t)3 wx4 + wx6 w(a) = (10.5a) (10.5b) where each wi (z) function is also defined in appendix D. To solve this equation we analyze the behavior of the (10.6) EoS associated with the fluid representation of f(T, B) denoted by w considering four kind of cases (cf figure 37). The specific cases that we can consider in order to develop a constraint analysis in the redshift range of observations include aspects when the contribution from T is larger that B (T  B), and vice versa (T  B): where the wxi definitions can be found in appendix D. To perform the numerical analysis, we rewrite the above expression in terms of redshift z = a0 /a − 1, where z = 0 corresponds to the present time. The EoS in this setting can be expressed w(z) = w1 (z) + 6 [w2 (z) − w6 (z)] + 24 [w3 (z) − w4 (z) − w5 (z)] , −w1 (z) − 6w7 (z) + 72 [w8 (z) − w9 (z) − w10 (z)] − (z+121)7 • Case 1.1: Domination of the boundary term over the torsion scalar. (10.6) 162 Rep. Prog. Phys. 86 (2023) 026901 Review Table 26. The mean value and the corresponding 68% limits [816] of the model parameters of the f3 (T) model. Base refers to the JLA SNeIa data and H 0 prior. Data set H 0 (km s Mpc−1 ) Base 73.3 ± 1.7 Base+BAO Ωm0 p 73.3 ± 1.8 0.084 0.295+ −0.070 0.022 0.285+ −0.025 0.22 0.10+ −0.51 0.11 −0.03+ −0.21 Base+BAO+H(z) 70.2 ± 1.1 0.263 ± 0.014 0.09 −0.01+ −0.14 Base+BAO+CMB 69.6 ± 0.9 0.296 ± 0.008 0.08 0.15+ −0.11 Base+BAO+H(z)+CMB 69.6 ± 0.9 0.296 ± 0.007 0.09 0.13+ −0.11 Figure 35. Constraints in the Ωm0 − H0 plane and in the Ωm0 − n/p plane for power law f (T) (left) and exponential f (T) (right) models with current only and current + Sandage–Loeb 30–year data. Reproduced from [826], with permission from Springer Nature. Figure 36. General Taylor expansion model contours plots for Cases 1.1 (orange color) and 1.2 (blue color) using CC+Pantheon+BAO samplers. Reproduced from [263]. © IOP Publishing Ltd. All rights reserved. • Case 1.2: Domination of the torsion scalar over the boundary term. • Case 2.1: Domination of the boundary term over the torsion scalar. • Case 2.2: Domination of the torsion scalar over the boundary term. phantom solution in agreement with Planck data for the density of matter. (b) Power law model. Now we consider a Lagrangian of separated power law style models for the torsion and boundary scalars such that f(T, B) = b0 Bk + t0 (−T)m . (10.7) Since we are interested in understanding whether this power law model can reproduce a DE-like behavior, we compute the EoS for the model in equation (10.7) and obtain For equation (10.6), we perform the fitting using the completed compilation of the observations described. In figure 36, we see that the model has a preference for a 163 Rep. Prog. Phys. 86 (2023) 026901 Review 6k b0 (2H 2 +H ′ +H 2 )k (k−1)kwx1 − 22+m 3m (H 2 )m+1 (m − 1)mt0 wx2 (2H 2 +H ′ +H 2 )3 w(a) = −1 + , 3(−Bk b0 + 6H ′ − Tm t0 + wx4 + wx5 )  b0 6k (k−1)k = −1 + a(t)ä(t)+2ȧ(t) 2 a(t) 2 k wx1 [a(t)ä(t)+2ȧ(t) 2 ]3 n 3 − (10.8a)   2 m+1 t0 2m+2 3m (m−1)ma(t) ȧ(t) 2 wx2 a(t) ȧ(t)5 6[a(t)ä(t)−ȧ(t) 2 ] + wx4 + wx5 − b0 Bk − t0 Tm a(t) 2 o , (10.8b) corresponds to the present time. The power law model EoS can be expressed as where the functions wxi are given in appendix D. To perform the numerical analysis, we rewrite the above expression in terms of the redshift z = a0 /a − 1, where z = 0 b0 3k 8k−2 (k − 1)k(z + 1)12 h 1 (z+1) 2 ik w(z)1 w(z)4 , w(z) = −1 +  h ik 3 −b0 Bk − (z + 1)3 w(z)3 − b0 23k−1 3k (k − 1)k (z+11) 2 − t0 Tm + (z+61) 2 (10.9) behavior to ΛCDM between z = 2 and z = 4. Afterwards starts to grow asymptotically to the first model at large z. Cases 3.1 and 3.2 cross the phantom divided-line, and below z = 2 they are similar. Below z = 2.5 both models can start with an EoS with w < −1/3, where both have an asymptotic behavior at large z, which can mimic a matter phase with w = 0 (cf with figure 37(bottom left)). This is a case of an oscillating fluid representation of f(T, B) EoS below z = 6. Case 4 (cf with figure 37(bottom right)) has an oscillating particularity, but it experiences a divergence point due to the corresponding energy-density becoming zero. Thus, the possible cases of the different phases of the cosmic evolution are shown individually with all possible dominance terms being shown. The confidence regions of this model are shown in figure 38, where we can infer the constraints on the theory against (CC+Pantheon+BAO) observational data for free model parameters. The precise values are shown in table 27. the very large value of these parameters may be problematic for other sectors of phenomenology. where the functions w(z)i are given in appendix D. Notice that equation (10.9) reduces to the standard ΛCDM model w = −1 when k, m are zero, as expected. As a first strategy, we are going to analyze the behavior of the EoS in equation (10.9) which is associated with the fluid representation of f(T, B) considering seven cases (cf figure 37) as: • Case 1.1: Domination of the boundary term over the torsion scalar. • Case 1.2: Domination of the torsion scalar over the boundary term. • Case 2.1: m and k as free parameters with the condition that m > k. • Case 2.2: m and k as free parameters with the condition that k > m. • Case 3.1: b0 and t0 as free parameters with the condition that b0 < t0 . • Case 3.2: b0 and t0 as free parameters with the condition that b0 > t0 . • Case 4: t0 and m as free parameters and negative values. Observations from SNeIa and gamma ray bursts are optimal tools in precision cosmology since with them it is possible to reconstruct the Hubble flow. Moreover, due to the degeneracy of model parameterizations, a viable path to overcome these issues is by using cosmography. To achieve this goal we expand the scale factor in Taylor series wrt the cosmic time. Such an expansion leads to a distance-redshift relation which only relies on the assumption of the FLRW metric being model-independent since it does not depend on the specific form of the solution of the 10.3.4. Cosmography for TG gravity. Cases 1.1 and 1.2 (cf with figure 37(top left)) at z < 2 show an accelerating cosmic solution. Case 1.1 decelerate at z = 3, while Case 1.2 preserves this acceleration with EoS w < −1/3. It is important to point that that since the two scalars, namely the torsion scalar and boundary term, can interchange dominance over the evolution of the Universe, these cases will turn out to be mutually exclusive in most models. Cases 2.1 and 2.2 (cf with figure 37(top right)) have an EoS with w < −1/3, but shows an asymptotic 164 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 37. Evolution of Power Law EoS (10.9). Top left: Case 1: solving T and B, with T ≪ B (solid line) and T ≫ B (dashed line). Top right: Case 2: m and k as free parameters, with m < k (solid line) and m > k (dashed line). Bottom left: Case 3: b0 and t0 as free parameters, with b0 negative and f 0 positive (solid line) and vice versa (dashed line). Bottom right: Case 4: t0 and m as free parameters and with negative values. The values considered are: m = 78.93, k = 49.62, b0 , t0 ≈ 1015 . Reproduced from [263]. © IOP Publishing Ltd. All rights reserved. Figure 38. One-dimensional marginalized distribution, and two-dimensional contours with 68% and 95% confidence level for the free parameters of the Power Law model using the constrained solutions for T and B scalars and CC + Pantheon + BAO total sampler. Reproduced from [263]. © IOP Publishing Ltd. All rights reserved. cosmic equations. It is convenient to introduce the following definitions which denotes the Hubble, deceleration, jerk, snap and lerk parameters, respectively. Their present day values (denoted with a subscript 0) can be used to characterize the cosmic evolution, e.g. if q0 < 0, this denotes an accelerated expansion, while j0 allows to discriminate among different accelerating models. Using these definitions straightforward we can obtain the following equations [791, 830] 1 da 1 d 2 a −2 1 d3 a −3 , q=− H , j = H , a dt a dt 2 a dt3 1 d5 a −5 1 d4 a −4 H , l = H , s= (10.10) a dt4 a dt5 H= 165 Rep. Prog. Phys. 86 (2023) 026901 Review Table 27. Parameters and mean values for the Power law model. Parameter Best-fit Mean ±σ H 0 (km s Mpc−1 ) 67.74 +1.1 67.74− 1. 1 65.54 69.89 m 78.93 4 79.19+ −6.1 70.17 88.64 k 49.62 +0.73 49.81− 1 47.82 51.73 8.16 × 10+15 +7×10+14 1.099 × 10+16 −1.1×10+16 9.981 × 10+11 1.640 × 10+15 1.169 × 10+16 1.074 × 10+16 b0 8.949 × 10+15 c0 +15 +2.8×10 7.974 × 10+15 −6.3×10+15 Ḣ = −H 2 (1 + q), (10.11a) Ḧ = H3 (j + 3q + 2), (10.11b) ... H = H4 [s − 4j − 3q(q + 4) − 6], (10.11c) where a dot denotes derivative wrt the cosmic time t and H (4) = d4 H/dt4 . Equations (10.11a)–(10.11d) make it possible to relate the derivative of the Hubble parameter to the other cosmographic parameters. Instead of choosing a parameterized form for a f (T) model and then numerically solving the modified version of Friedmann equations for given values of the boundary conditions, it is possible to relate the present day values (z = 0) of its derivatives to the cosmographic parameters (q0 , j0 , s0 , l0 ) in order to constraint them in a model-independent way and obtain a form of f (T) that can be able to fit with the observational data at hand. As a preliminary step, it is worth considering again that T = −6H 2 . Differentiating wrt t, we get the following expressions T̈ = −12(Ḣ 2 + HḦ), ... ... T = −12(3ḢḦ + HH), ... T (4) = −12(3Ḧ 2 + 4ḢH + HH (4) ). H (4) = (10.12a) (10.12c) where the fourth and fifth derivative are denoted by the quantities f (4) (T) = d4 f(T)/dT 4 and f(v) (T) = d5 f(T)/dT5 , respectively. At this level, it is possible to derive a generic cosmography, moreover, the restriction of a specific f (T) is still present. In [831], an f (T) Taylor series model was studied up to fifth order to investigate a priori a given f (T) theory by simply comparing the theoretically predicted (f2 , f3 ) models with the observed ones. Unfortunately, this analysis was unable to predict the observationally values of (f2 , f3 ) models due the degeneracy over the cosmographic parameters and the difficulties concurrently constraining with current surveys. Furthermore, [791] explored cosmographic constraints following the same methodology of exotic polynomials, e.g. Padé polynomial. Moreover, the current stateof-the-art suggests that it is not yet possible to falsify the ΛCDM model by robust statements adopting only this kind of cosmography. At present day (z = 0) values of f (T) and its derivatives up to the fifth order. After some algebra, we have (10.12d) 1 [−TΩm + f(T)], 12fT (10.13a) 1 [−TΩm − 4HṪfTT ]. 4fT (10.13b) Ḣ = Once with this quantities, we can derive up to fourth order in H to derive our cosmographic parameters as Ḧ = Ωm [HṪ + T(3H 2 + 2Ḣ)] 4HfT 1 − [(2ḢṪ + HT̈)fTT + HṪ 2 fTTT ], fT Ωm h  −T 10HḢḦ + 12H3 Ḧ − 27H 6 − 12H 2 Ḣ 2 − 8Ḣ3 4H3 fT   ... ... −2H 2 H + H3 T + H 2 Ṫ 9HḢ + 27H3 − 5Ḧ   i −3H 2 T̈ 3H 2 + Ḣ + 7HḢ 2 Ṫ   ... 1 h − 2 3HḢH + Ḣ 2 Ḧ + HḦ 2 fT + H 2 Ṫ 2 7ḢṪ + 6HT̈ H fT  ... ... × fTTTT + 4H 2 HṪ + 2Ḣ3 Ṫ + 7H 2 Ḣ T  +10HḢ 2 T̈ + 7H 2 ḦT̈ + 11HḢḦṪ + H3 T (4) fTT  + H 10Ḣ 2 Ṫ 2 + 7HḦṪ 2 + 21HḢṪT̈ i ... +3H 2 T̈ 2 + 4H 2 Ṫ T fTTT + H3 Ṫ 4 fTTTT , (10.14c) (10.12b) The modified Friedmann given by equations (6.7a) and (6.7b) can be rewritten for the pressureless case p = 0 as H2 = − 95% upper   ... Ωm  −T 9H4 + 6H 2 Ḣ + 4Ḣ 2 − HṪ 3Ḣ + 6H 2 H= 4H 2 fT  +H HT̈ − 2ḦT ... 1  ḢḦfT + 2Ḣ 2 Ṫ + 3HḦṪ + 4HḢT̈ + H 2 T fTT − HfT   +H 2 Ṫ3 fTTTT + HṪ 4ḢṪ + 3HT̈ fTTT , (10.14b) H (4) = H5 [l − 5s + 10(q + 2)j + 30(q + 2)q + 24], (10.11d) Ṫ = −12HḢ, 95% lower (10.14a) 166 Rep. Prog. Phys. 86 (2023) 026901 Review f(T0 ) = Ωm0 − 2, fT (T0 ) = 1, 6H02 fTT (T0 ) 3Ωm0 1 =− + , 2 − 1 4(1 + q0 ) 2 (6H0 ) It is well–known that a non-minimal coupling between a scalar field and matter fields would break the Einstein equivalence principle, and would further lead to the variation of fundamental constants of Nature [844–846]. For instance, a scalar field coupling with the electromagnetic Lagrangian would lead to a variation of the fine–structure constant, or Sommerfeld’s constant, which characterizes the strength of the electromagnetic field and appears as a coupling constant in the electromagnetic action. A variation in the fundamental constants of Nature [847], which could be conservatively defined as those theoretical free parameters that could not be calculated with our present knowledge of physics, has been a long–established intriguing question [848, 849] with pertinent consequences for fundamental physics and cosmology (see, for instance, [840, 844, 845]). Interestingly, when Dirac’s numerological principle [848, 849] was encapsulated in a field–theoretical framework, this led to the birth of the Jordan–Fierz–Brans–Dicke scalar–tensor theory of gravitation [287, 850, 851]. A number of theoretical models have been proposed in order to explore the possibility of a dynamical fine–structure constant α ≡ e 2 /ℏc. These models have been primarily formulated as Lagrangian theories with explicit variation of the velocity of light c (varying speed of light theories) [852], or of the charge on the electron e (varying electric charge theories) [853, 854]. These models have been first formulated by Bekenstein [853] from a generalization of Maxwell’s equations, that led to the construction of the cosmological varying–e Bekenstein–Sandvik–Barrow–Magueijo theory of varying α [855–860]. Other frameworks include, for instance, a runaway dilaton [861, 862], supersymmetric generalization of Bekenstein’s model [863] and a disformally coupled electromagnetic sector [864]. Each theoretical framework has its own cosmological signatures, and we should remark that the spacetime dependence of fundamental constants has also been linked [865] with the currently reported Hubble tension via the inferred effects in the ionization history and profile of CMB anisotropies. For example, in f (T) gravity, the form of this fine–structure constant dependence can be obtained by a conformal transformation (refer to section 5.3.2 for further details) of the tetrad where ẽAµ = Ω eAµ , and ẼAµ = Ω−1 EAµ , which results in the regular conformal transformation g̃µν = Ω 2 gµν , as expanded upon in [256], where Ω 2 = −fT = |fT | (note that since in our conventions T > 0 and fT < 0, we have replaced −fT by |fT |), and that tilde denotes conformally transformed quantities. It is well known that f (T) gravity cannot be written in the Einstein frame through conformal transformations, which implies that it will induce a dependence in its associated fine–structure constant [866] (as will be shown later in equation (10.17)) . In fact, this produces an  extra 2Ω−2 ∂˜µ Ω 2 T̃ν νµ term which cannot be removed (see equation (5.70)). The remainder of the√scalar field becomes a phantom field with the choice of ϕ = 3 ln fT [256], which is partially favored by the 2018 Planck data [372]. The result of a conformal transformation is the introduction of a new DoF, ϕ, that arises from the transformation Aaµ → ẽAµ , where ẽAµ represents the Einstein frame tetrad. This then induces an electromagnetic coupling which takes on the form (10.15a) fTTT (T0 ) −3Ωm0 (3q02 + 6q0 + j0 + 2) 3 + , = 2 8(1 + q0 )3 4 (6H0 )−2 (10.15b) fTTTT (T0 ) 3Ωm0 =− 2 − 3 16(1 + q0 )5 (6H0 ) × [−s0 (1 + q0 ) + j0 (6q02 + 17q0 + 3j0 + 5) 15 + 3q0 (5q30 + 20q02 + 29q0 + 16) + 9] + , 8 (10.15c) fTTTT (T0 ) 3Ωm0 =− 2 − 4 32(1 + q0 )7 (6H0 ) × [l0 (1 + q0 ) 2 + s0 (10q30 + 43q02 + 46q0 + 13) − 10j0 s0 (1 + q0 ) + 5j02 (6q02 + 22q0 + 3j0 + 7) + j0 (45q40 + 225q30 + 412q02 + 219q0 + 32) + 3q0 (35q05 + 210q40 + 518q30 + 666q02 + 448q0 + 150) + 60] + 105 . 16 (10.15d) As we can notice from the latter equations, there is still a dependency of the model directly of the f (T) form, which lead us to a model-dependent cosmography related to the Taylor f (T) series form. As a perspective in this direction, the consideration of f (T) models can help to understand the convergence of the series imposed by cosmography, even thought some of these models should be aligned to a particular form of Taylor series or extension of them. Due to the excessive number of free parameters in the equations of motion, [791] a degeneracy creeks into the system and so it becomes difficult to find a unique best-fit curve for these parameters. As a result, the order of the expansion does not have statistical relevance, which is a consequence of the low density of data points used to get this conclusion. Moreover, increasing the order of the expansion shows a deviation from the standard cosmological values. 10.3.5. Probing f(T) gravity with the induced variation of fundamental constants. Another cosmological probe of TG involves measurements on the variation of the fine–structure constant α, denoted by ∆α/α. It predominantly consists of an archival astrophysical data set of quasar absorption lines observed at the Keck observatory [832] and with the VLT (V) [833], along with a set of 21 dedicated new measurements (N) [834–842], and the constraint from the Oklo (O) natural nuclear reactor at an effective redshift of z = 0.14 [843]. The measurements contained in the N data set were reported from the ESO Ultraviolet and Visual Echelle Spectrograph (UVES) Large Program which was specifically developed for such measurements. For simplicity, we will henceforth be denoting the joint data set of K + V + N + O by KVNO. 167 Rep. Prog. Phys. 86 (2023) 026901 1 SEM = 4 Review ˆ d4 x eBF (ϕ) Fµν Fµν , (a) Power law model. The derived constraint on the power law model [273] (f1 (T)) parameter b of (a) was found to be 0.24 R b = −0.16+ −0.49 [869] when adopting the SN + CC + H0 joint data set, which is consistent with ΛCDM and in line with the findings in previous studies [816, 876–878]. As illustrated in the left panel of figure 39, the SN + CC + KVNO and SN + CC + KVNO + HR0 joint data sets also improve the constraints on the model parameter b, which was found to be consistent with zero. Similar constraints have been reported in [868], although the presented constraints are tighter. The reported constraints of table 28, also show that βF is compatible with zero. Consequently, there is a negligible deviation from the f (T) distance– duality relation in this model. (b) Square–root–exponential model. In the square–root– exponential model [39] (f2 (T)) of (b), the derived constraints on 1/p from the SN + CC + HR0 and SN + CC + KVNO + HR0 data sets were found to be consistent with zero at around 1σ [869]. This model was also found to be in agreement with the ΛCDM model at around 2σ when the SN + CC + KVNO data set was adopted (see, for instance, [816, 868, 876–879] for similar conclusions). The marginalized confidence contours are depicted in the right panel of figure 39, while the list of all the derived parameter constraints is reported in the second panel of table 28. (c) Exponential model. Similar to the previously considered model, in the f3 (T) [39] model we observe that the ΛCDM model is recovered when q → +∞, or equivalently 1/q → 0+ . The derived parameter constraints [869] are reported in the third panel of table 28, whereas the marginalized confidence contours are depicted in top-left panel of figure 40. Similar to the previous exponential f (T) model, the model parameter 1/q is found to be consistent with the ΛCDM limit at around 1σ, where such a result is compatible with the results of [816, 876, 878, 879]. Furthermore, this exponential f (T) model was found to be characterized by a null variation in the fine–structure constant, since βF was always found to be consistent with zero. (10.16) where Fµν = Aν,µ − Aµ,ν is the standard Faraday tensor, and BF (ϕ) represents the nonvanishing ϕ−coupling. The consequence of this induced coupling is that the fine–structure constant and the luminosity distance will be altered comparing to GR [2, 867]. As in [290, 863, 866, 868], this can be expanded about ϕ(t = t0 ), which is suitably small, to give BF (ϕ) ≃ 1 + βF κ2 ϕ, (10.17) where βF = O(1) is a constant (βF ϕ  κ−2 ). Given an initially uncoupled Jordan–frame electromagnetic action, the fine–structure constant turns out to be given by [863] αE (ϕ) = αJ (ϕ) , BF (ϕ) (10.18) where αE and αJ are the fine–structure constants in the Einstein and Jordan frames respectively.  Given that Ω 2 = |fT |, this would imply that the ∂˜µ Ω 2 would be very small rendering the additional term negligible. We will revisit this reasoning against the results of the analysis. With this approximation to the Einstein frame, the variation of the fine–structure constant takes the form √ ∆α κ−2 + 3βF [ln |fT (T0 )|] √ = − 1, α κ−2 + 3βF [ln |fT |] (10.19) which vanishes for the ΛCDM case of f(T) = −T + Λ, as expected. Equation (10.19) embodies the redshift dependence of the fine–structure constant in TG, since the torsion scalar depends on redshift. Another consequence of a nonvanishing scalar field coupling to the electromagnetic action is that the luminosity distance will be altered [846]. By considering the standard derivation of luminosity distance [2] with this new action, [867] shows that this leads to r dL = c (1 + z) s = c (1 + z) BF0 BF ˆ z dz ′ ′ 0 H(z ) √ κ−2 + 3βF [ln |fT (T0 )|] √ κ−2 + 3βF [ln |fT |] (10.20a) 10.4. Beyond data-driven analysis for TG cosmologies We have been discussing several proposals that can be cosmologically viable and well constrained by the current observational data. However, due the avalanche of these observational catalogues from different species, we require the implementation of new numerical technique to perform reconstruction that can be model independent in order to solve the inverse cosmology problem, i.e. going from the data to the best (or optimal) theory/model. These nonparametric approaches, as they are commonly called in the literature can be seen as a perspective for future TG scenarios analysis, one of the can lies in the classification of reconstruction of data using Gaussian processes (GP), and the second one as machine learning tools for TG through deep learning techniques [880]. This section is devoted to explain how we can implement TG scenarios in these ambiance. While the methodology behind each approach is wide, we present in the supplementary annexes ˆ z dz ′ , (10.20b) ′ 0 H(z ) as the luminosity distance for f (T) gravity, which limits to the GR formula for BF = 1. We now consider the following f (T) models, in which we primarily focus on the respective induced variation in the electromagnetic coupling and the inferred cosmological parameter constraints. These models were confronted with the KVNO varying fine–structure constant data set, along with the SNeIa Pantheon compilation [870] and a CC data set [802, 871–875]. Occasionally, a prior on the Hubble constant is assumed in the following analyses, which is specified by HR0 = 74.03 ± 1.42 km s−1 Mpc−1 [718]. 168 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 39. Marginalized two-dimensional likelihood constraints on the parameters of the f1 (T) (left) and f2 (T) (right) models of equation (10.2) and equation (10.3), respectively. Reproduced from [869]. © 2020 IOP Publishing Ltd and Sissa Medialab srl. Table 28. The mean value and the corresponding 68% limits [869] of the model parameters of three fi (T) models (i ∈ {1, 2, 3}), as described in section 10.3.5. We also report the minimum χ2 value. f1 (T) Model Parameter SN + CC+ HR0 SN + CC + KVNO SN + CC + KVNO + HR0 b 0.24 −0.16+ −0.49 0.053 0.003+ −0.059 0.050 −0.001+ −0.048 +0.036 0.281− 0.035 0.026 0.300+ −0.024 0.023 0.283+ −0.021 2. 0 68.9+ −1.9 0.063 −0.003+ −0.056 1. 2 72.2+ −1.2 βF +1.4 72.8− 1. 3 +0.32 0.28− 0.32 0.074 −0.003+ −0.067 2 χmin 1042.578 1365.427 1370.014 Ωm0 H0 (km s Mpc −1 ) f2 (T) Model Parameter SN + CC+ HR0 +0.171 0.093− 0.079 SN + CC + KVNO   35.9 10.8+ × 10−3 −4.9 SN + CC + KVNO + HR0   9. 3 −3 41.7+ −30.8 × 10 1/p +0.025 0.279− 0.031 0.021 0.300+ −0.020 0.020 0.283+ −0.019 1. 8 69.0+ −1.9 0.45 −0.01+ −0.75 1. 3 72.2+ −1.2 βF +1.3 72.2− 1.2 0.49 −0.10+ −0.56 0.58 −0.07+ −0.55 2 χmin 1045.741 1366.626 1371.221 Ωm0 H0 (km s Mpc −1 ) f3 (T) Model Parameter SN + CC+ HR0 1/q SN + CC + KVNO + HR0 +0.088 0.065− 0.045 SN + CC + KVNO   28.7 15.7+ × 10−3 −9.5 0.018 0.029+ −0.020 +0.021 0.279− 0.020 0.022 0.302+ −0.023 0.019 0.283+ −0.019 2. 0 69.0+ −1.9 0.61 −0.05+ −0.53 1. 2 72.2+ −1.3 βF +1.3 72.2− 1.2 0.76 −0.22+ −0.40 0.61 0.00+ −0.52 2 χmin 1046.074 1366.661 1371.241 Ωm0 H0 (km s Mpc −1 ) 169 Rep. Prog. Phys. 86 (2023) 026901 Review can be determined the distance (in units of σ) between the GP determined value against the literature priors discussed above, so that this distance is defined as follows H0,i − H0,j , d (H0,i , H0,j ) = q σi2 + σj2 (10.21) where H0,i and H0,j are two respective values of the present value of the Hubble parameter together with their respective 1σ uncertainties σ i and σ j . The derived results are presented in figure 41. The GP approach is used for each set of priors for H 0 as shown in the sub-figures. Moreover, for every GP reconstruction, the 1σ and 2σ regions are illustrated. As a reference point, it is present the ΛCDM behavior in all instances. There also exist diagnostic tools to assess preferences in the reconstructions toward deviations from ΛCDM [885, 886]. Considering the GR first Friedmann equation for a cosmos filled with a dark fluid EoS w(z) E 2 (z) := Figure 40. Marginalized two-dimensional likelihood constraints on the parameters of the f3 (T) model of equation (10.4). Reproduced from [869]. © 2020 IOP Publishing Ltd and Sissa Medialab srl. H 2 (z) 3 2 = Ωm0 (1 + z) + Ωk0 (1 + z) H02   ˆ z 1 + w(z ′ ) ′ dz , + ΩΛ0 exp 3 (10.22) 1 + z′ 0 where E(z) is the reduced Hubble parameter, and which can be rearranged to reconstruct the EoS of the dark fluid via (supplementary 6 and 7) the main ideas behind the numerical architecture. w(z) = 10.4.1. Gaussian processes and the reconstruction of f(T) gravity. The discussed GP approach in the supplementary 2(1 + z)E(z)E ′ (z) − 3E 2 (z) , 3 [E 2 (z) − Ωm0 (1 + z)3 ] (10.23) where it has been assumed that the Universe observes spatial flatness. The GP reconstructions of w(z) are reported in figure 42, where it is clear that w = −1 is not excluded by the currently available data. However, notice that this reconstruction is dependent on the matter density parameter, which restricts us from constructing physical models. On the other hand, one could test the flat ΛCDM model by considering the following diagnostic redshift function [869] annexes (supplementary 6) can be applied to a number of different H(z) data sources, from which it is possible to reconstruct H 0 . In the considered analyses, three sources of H(z) data have been considered, namely CC, SNeIA and BAO. From the presented H(z) data in the supplementary annexes (supplementary 4), only the CC data as reported in [802] are adopted in the following analyses [776]. Regarding the SNeIa data which was introduced in the supplementary annexes (supplementary 4), the combination of the compressed Pantheon compilation is used [870] together with the CANDELS and CLASH Multi-cycle Treasury data [881]. For this data set, the Hubble rate parameter measurements of E(z) = H(z)/H0 were added along with the corresponding correlation matrix, such that only five of the reported six data points could be taken into consideration, since the z = 1.5 data point is not Gaussian– distributed. Moreover, a set of eight BAO data points from the SDSS (Data Release 12 and 14) [882–884] along with the corresponding correlation matrices were adopted in the following GP analyses. We refer the reader to [776] for further details on the inclusion of the BAO data set and SNeIa measurements via an iterative technique. The GP approach is now applied to the various sources of Hubble data together with the priors described. The results pertaining to the value of H 0 are presented in table 29 which contains the principal results for the square exponential kernel. We here present the GP inferred value of H 0 for the case of taking no prior, and the HR0 , HTRGB and HHW priors. In every case, it 0 0 Om(1) (z) := E 2 (z) − 1 , z(3 + 3z + z 2 ) (10.24) (1) which reduces to Om (z) = Ωm0 in the ΛCDM scenario. Thus, this diagnostic can measure how close a data set is to GR. (1) We should remark that Om (z) is not so much dependent on ΛCDM but a characteristic parameter that can test how consistent a data set is with this model. As reported in figure 43, one could clearly notice that the considered data sets are in a very good agreement with the ΛCDM predictions, although the H 0 prior has a significant effect on the reconstruction and hence the viability of the concordance model of cosmology. (1) Complementary to the Om (z), its derivative can also offer a useful way to quantify deviations from ΛCDM, defined through L(1) (z) = 3(1 − E 2 (z))(1 + z) 2 + 2z(3 + 3z + z 2 )E(z)E ′ (z). (10.25) 170 Rep. Prog. Phys. 86 (2023) 026901 Review Table 29. Different GP reconstructions of H 0 [869] with the square exponential kernel function expressed in the supplementary annexes (supplementary 6). The reconstructed values of H 0 are complemented by their distance (in units of σ) from literature priors. Data set(s) H 0 (km s Mpc−1 ) d(H0 , HR0 ) d(H0 , HTRGB ) 0 d(H0 , HHW 0 ) d(H0 , HP18 0 ) CC CC+SN CC+SN+BAO CC+HR0 CC+SN+HR0 CC+SN+BAO+HR0 CC+HTRGB 0 CC+SN+HTRGB 0 CC+SN+BAO+HTRGB 0 CC+HHW 0 CC+SN+HHW 0 CC+SN+BAO+HHW 0 67.539 ± 4.7720 67.001 ± 1.6531 66.197 ± 1.4639 73.782 ± 1.3743 72.022 ± 1.0756 71.180 ± 1.0245 69.604 ± 1.7557 68.468 ± 1.2212 67.811 ± 1.1470 72.966 ± 1.6636 70.850 ± 1.1991 69.911 ± 1.1276 −1.3037 −3.2253 −3.8407 −0.12556 −1.1271 −1.6279 −1.9599 −2.9695 −3.4070 −0.4863 −1.7111 −2.2717 −0.4408 −1.1183 −1.5127 1.7106 1.0265 0.6447 −0.0760 −0.5942 −0.9036 1.2617 0.4710 0.0506 −1.1334 −2.6165 −3.1132 0.2166 −0.6220 −1.0457 −1.4908 −2.2641 −2.6233 −0.1382 −1.1550 −1.6280 0.0290 −0.2309 −0.7776 4.3640 3.8969 3.3155 1.2076 0.8096 0.3284 3.2043 2.6555 2.0355 Figure 41. GP reconstructions of H(z) with the squared exponential kernel function expressed in the supplementary annexes (supplementary 6). The data sets along with the different H0 priors are indicated in each respective panel. Reproduced from [776]. © IOP Publishing Ltd. All rights reserved. Any deviation from L(1) (z) = 0 represents a deviation from ΛCDM, which makes L(1) (z) a good diagnostic over which to assess the behavior of the concordance model. The L(1) (z) diagnostic is presented in figure 44. In figure 41, the square exponential kernel GP reconstructions are shown for the redshift range of the full data set. In all cases, the BAO data reduce the 1σ and 2σ uncertainties at higher redshifts since the other data sets do not feature points in that regime. In fact, in the cases of CC and CC+SN, the ΛCDM theoretical prediction only deviates into the 2σ uncertainty region for these high 171 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 42. GP reconstructions of w(z) with the squared exponential kernel functions, along with the ΛCDM prediction. Reproduced from [776]. © IOP Publishing Ltd. All rights reserved. values of redshift, and only outside of both when the BAO data are included. The BAO data are dependent on the concordance model of cosmology, and so one would expect it to produce issues of this kind since the other two data sets are independent of cosmological models. The remainder of the reconstruction regions remain in close range of the ΛCDM prediction. This is further exposed by the diagnostic consistency test shown in figure 44 where a number of regions mark slight deviations from ΛCDM. However, it is the BAO data set that exposes this deviation at high redshifts. Another interesting feature of these diagnostic tests is that as with the H 0 reconstructions in table 29, the HR0 prior brings about the largest deviation of the reconstructions. One could also infer the GP reconstructed deceleration parameter, given by q(z) = (1 + z) H ′ (z) − 1. H(z) specific form of the Lagrangian. We start by expressing the Friedmann equation in terms of redshift dependence alone. We therefore convert the Lagrangian derivative term so that fT = df/dz f ′ (z) = ′ , dT/dz T (z) (10.27) where f ′ (z) = df/dz and T ′ (z) = −12HH ′ are redshift derivatives. The immediate follow up becomes the issue of handling f ′ (z) in the analysis which is tackled in this analysis through the central difference method as f ′ (zi ) ≃ lim ∆z→0 (10.26) f(zi+1 ) − f(zi−1 ) , zi+1 − zi−1 (10.28) since this will produce smaller uncertainties O(∆z 2 ) rather than O(∆z) which occur for the forward and backward difference methods, where ∆z = zi+1 − zi−1 . By substituting into the Friedmann equation (6.7a) (taking the mapping f(z) → −T + f(z), as in [869, 887]) and simplifying, this method produces a numerical propagation equation for f (z) given by We illustrate the GP reconstruction of q(z) in figure 45. We finally discuss and present the application of the above cosmologically model-independent data for the GP reconstruction of a general f (T) function without assuming a 172 Rep. Prog. Phys. 86 (2023) 026901 Review (1 ) Figure 43. GP reconstructions of Om (z) with the squared exponential kernel function, along with the ΛCDM prediction. Reproduced from [776]. © IOP Publishing Ltd. All rights reserved. H ′ (zi ) f(zi+1 ) = f(zi−1 ) + 2 (zi+1 − zi−1 ) H(zi )   f(z ) i 3 2 2 × 3H (zi ) + − 3H0 Ωm0 (1 + zi ) , 2 (10.29) (b) The second boundary condition can be obtained by using the forward differencing method through f ′ (zi ) ≃ lim ∆z→0 f(zi+1 ) − f(zi ) , zi+1 − zi (10.31) that leads to the equation where the propagation equation parameters H 0 and Ωm0 are selected from the corresponding GP reconstruction within the P18 and H 0 priors, respectively. While superior in terms of having lower associated uncertainties, the central differencing requires two initial conditions to be employed which we form as follows: H ′ (zi ) f(zi+1 ) = f(zi ) + 2 (zi+1 − zi ) H(zi )   f(zi ) 3 2 2 − 3H0 Ωm0 (1 + zi ) , × 3H (zi ) + 2 (10.32) (a) Friedmann equation boundary condition: Evaluating the Friedmann equations (6.7a) and (6.7b) at z = 0 gives which straightforwardly leads to the necessary second boundary condition. f(z = 0) ≃ −16πGρ0m + 6H02 = −6H02 (Ωm0 − 1) , (10.30) The propagation of the f (T) function is complemented by its associated MCMC error propagation which produces the 1σ and 2σ uncertainties. Together the propagation equation in equation (10.29) along with the boundary conditions in (i) and (ii) can express the redshift dependent Lagrangian f (z) in terms of where we have imposed that ΛCDM dominates at present times, i.e. fT (z = 0) ≃ 0. This again relies on the same parameters as the propagation equation itself; 173 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 44. GP reconstructions of L(1) (z)/(1 + z) 6 with the squared exponential kernel function expressed in the supplementary annexes (supplementary 6). The data sets along with the different H0 priors are indicated in each respective panel. Reproduced from [776]. © IOP Publishing Ltd. All rights reserved. ΛCDM for the breadth of the evolution interval, and mostly within the 1σ region for the data set combinations. The furthest propagated line to the ΛCDM is the Hubble parameter reconstructed from the combined data that includes the BAO data set. On the other hand, once the priors are included the situation changes drastically, with the HR0 prior favoring a slight deviation from ΛCDM over a portion of the cosmic evolution for all data sets. The same situation, but to a lesser extent, occurs for the HHW prior, with the HTRGB prior being 0 0 the only one to not affect this propagation in a significant way. These f (T) propagations show that the data sets alone favor an f (T) that only slightly deviates from ΛCDM within the redshift region being probed, while the propagations that do contain literature priors prefer a stronger deviation from ΛCDM, however the latter is inside the allowed regions. z in a (cosmological) model-independent way. Similarly, the corresponding torsion scalar can be associated with each of the redshift values in question through the Hubble parameter relation in equation (6.5). In this way, the Lagrangian function f (T) can be plotted as a function of the torsion scalar T. The f (T) reconstructions against the torsion scalar are illustrated35 in figure 46 where the 1σ and 2σ regions are shown in every case. We mention that the ΛCDM paradigm appears as a constant in these plots with a value of f(T) → −6H02 (Ωm0 − 1), which is denoted by horizontal lines in figure 46. One could easily observe that the ΛCDM scenario lies inside each and every reconstructed region. Nevertheless, the GP reconstruction procedure shows a slight tendency of f T to negative values, i.e. to f (T) forms that are slightly decreasing functions of T. This is the main result of this section since the aforementioned feature needs to be taken into account in the f (T) model building. In the case where no prior is used for the GP reconstruction, the f (T) evolution remains within the 2σ confidence region of In order to use machine learning in a specific deep learning neural network, we need to classify TG cosmological models using the surveys described in section 6. To perfom these analyses we require the architecture described in the supplementary annexes (supplementary 7). Following this, we begin 10.4.2. Machine learning for teleparallel cosmology. 35 We here reproduce the same figure as reported in [869], therefore f(T) < 0 in this figure. 174 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 45. GP reconstructions of q(z) with the squared exponential kernel functions, along with the ΛCDM prediction. Reproduced from [776]. © IOP Publishing Ltd. All rights reserved. model with the optimizer RMSprop38 with a standard value of learning rate of 0.003 for this type of algorithm. According to the natural language processing on Adam optimiser, and the loss is the mean squared error. The training of this data can be done with a small number of epochs, say 100. To generate non-linear data we must use a different architecture. By including layers with a certain activation function we can train non-linear models. The architecture for the following networks consists of two hidden layers, one with 20 neurons and the other with 12. Each of them have the same activation function, and an output layer with a single neuron. This kind of architecture with several hidden layers is also known as Deep Neural Networks. The first non-linear model that we train uses the rectified linear unit (ReLU) as activation function. This function is defined as f(x) = max(0, x). Due to the nature of our data we do not expect very different predictions compared to the linear model. For the rest of the neural networks we will use the Adam optimizer39 with a learning rate of 0.003 and the mean squared error as loss. We train the network for 300 epochs and by training several neural networks with different data and different activation functions. Next, we will generate new data with the trained networks and merge this data in a single data frame. Once we have all our new data in a single data frame, we split it in a training set and a test set. By evaluating the model with the test data, we can decide whether the neural network architecture is appropriate for this task or not. For the first training we use a small data set of CC, which has information about the redshift (z) and the Hubble parameter (H(z)). In this coding we use Tensorflow V2 36 and Keras37 . The data consists of two columns and 17 rows. Due the small size of this data set, it is standard to not split them in test and train sets. First, we extract the data and normalize the values of H(z). This is done because it is easier for the training to work with small numbers. The architecture of our first network consists in a sequential model with a single layer of just one neuron. This is because we want the output to be a single number for each prediction. We compile this sequential 36 www.tensorflow.org/?hl=es-419. 38 https://keras.io/api/optimizers/rmsprop/. 37 https://keras.io. 39 https://keras.io/api/optimizers/. 175 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 46. GP reconstructions of f (T) with the squared exponential kernel function expressed in the supplementary annexes (supplementary 6). The data sets along with the different H0 priors are indicated in each respective panel. Reproduced from [776]. © IOP Publishing Ltd. All rights reserved. again make store predictions for values up to 5 to avoid missing less obvious behaviors such as appear in partial time-series problems. We repeat the previous process for different activation functions. We decided to use the hyperbolic tangent, the Scaled Exponential Linear Unit and the sigmoid function as our activation functions. The predictions made in each case could be different for each training because we are predicting for values of z far from the highest value used for training. The predictions for all the trained neural networks are shown in figure 47. Building on this base, we can introduce our cosmological models and analyze them through the lens of the deep neural network that has since been constructed. To do this, we consider another deep neural network consisting of two hidden layers each with 32 neurons using the ReLU activation function, together with an output layer with the softmax activation function and 5 neurons. This architecture is widely used for nonbinary classification problems such as the current one where we want to measure the degree to which a model is approximated. Between the last hidden layer and the output layer we include a dropout of 0.2, in order to reduce the over fitting for regularization. This is important since overfitting Figure 47. Deep learning training for CC compilation up to z = 5. The observations from the sample are denote by ‘+’. can become a serious issue otherwise. We use the Adam optimizer with a learning rate of 0.003 and for the loss the sparse categorical cross entropy. For this example we use the accuracy 176 Rep. Prog. Phys. 86 (2023) 026901 Review modified theories helps to identify these flaws and make the method of comparison between theory and observations as model independent as much as possible. Studying the predictions of different cosmological models and constraining them with observations provides a way to validate the foundations of the ΛCDM model and quantify deviations from the constituent suppositions. The analysis of self-consistent models allows the examination of a given physical set-up within computational restrictions. This supposition will become important as the cosmological and astrophysical wealth and quality of data increase. In the near future, there will be next-generation experiments, space-missions and facilities on Earth, that will be of key importance to confirm or to falsify models beyond the standard one proposed at present and in the past years [744]. Between these models, the TG theories are starting to gain ground because they can easily address the H 0 and S8 tensions and simultaneously explain the acceleration of the Universe due to the DE. Unfortunately, at the moment, these theories have only been used for cosmology at the background level and very poorly explored with the current available data and, for example, a complete full analysis, taking into account the effect of the perturbations, is completely missing. The topic of the TG models is therefore a ‘work in progress’ field, that is quickly growing and trying to fill the gap with the other beyond-the-standard-model scenarios extensively studied and compared to the data, we have in the literature. Many works appeared in the last few years going in this direction (see table 21), and we expect many more coming out in the near future. For this reason it is important to list here which are the experiments expected in the next few years that could have a big impact in the DE/DM field, improving significantly the sensitivity we have on the current cosmological parameters, and helping us shaping plans for the kind of work we need to do in the future to make this TG field more robust and supported by the data. For example, a 50% improvement on the current Hubble constant measurements is expected by the local distance ladder observations, that will reach a precision of 1% [888]. A better H 0 measurement is predicted also for the gravitational time delays experiments, expected to achieve a ∼1.5% precision [889], and for the standard sirens (GWSS) [781, 890–893], the GW analogue of astronomical standard candles, that will reach a 2% precision. New improved CMB data will be possibly released by the ground-based experiments like CMB-S4 [894, 895] and Simons Observatory [896], and satellite experiments like LiteBIRD [897], CORE [898], PIXIE [899], PRISM [900], and PICO [901]. Of great importance will be the Euclid spacebased survey mission [902], that will be able to release an important variety of cosmological probes, starting from the gravitational lensing, to the BAO, to the galaxy clustering, and the multi-purpose radio-interferometer SKA40 , making measurements of the 21 cm line in neutral hydrogen (HI). Figure 48. Example of classifications of f (T) models described in the compendium table 21 using CC data. as metrics. We train our model with a batch size of 80 and a validation split of 0.2 to evaluate if we are over-fitting. The fitting is done for 500 epochs. The result of this process is shown in figure 48 where the classification output is more clearly shown. The lower value of the loss for the validation set and the higher accuracy is a good signal that we are not over fitting the data. To test the classifier we use a confusion matrix. The rows of the matrix are the true labels and the columns are the predicted labels. The closer the confusion matrix is to the identity, the better the classifier is. We make predictions with all the values from the initial data-set and insert the results in the confusion matrix. This matrix that we study contains five different f (T) models. In figure 49 we show the percentage between the predicted and true model. We notice that Linder model (see table 21) seems to be better identify and preferred by CC observations forecasted up to z = 5, in comparison to the other four models. 10.5. Perspectives in observational cosmology Over the last few years, we have witnessed great advances in precision cosmology in terms of our understanding of the Universe. In addition to known forms of matter, GR and an ansatz for the space-time metric as FLRW spacetime, two contributions remain a mystery and need to be added accounted for in observations, namely: the CDM to explain the formation of cosmic structures, and a cosmological constant to account for the late-time cosmic acceleration. The drastically increasing precision and wealth of observations and the surprising findings may call for a revision of the cosmological model hypothesis. On this matter, addressing the dynamics of the cosmic acceleration can be achieved by postulating modifications of GR acting only on the largest cosmological scales. Comparing these modified models against surveys does not only provide potential hints to test new physics: it is itself a way to validate the assumptions on which the standard paradigm is formulated. Model dependent supposition are usually made in the analysis of raw data, and certain curtailment which might be compatible with the usual supposition might not be fulfilled in modified scenarios. The study of the different 40 www.skatelescope.org. 177 Rep. Prog. Phys. 86 (2023) 026901 Review Figure 49. Metrics of the trained classifier for CC compilation up to z = 5. Moreover, starting in 2022 it is planned the galaxy survey Rubin Observatory Legacy Survey of Space and Time (LSST [903]), and in 2025 the powerful heir of the HST, i.e. the Roman Space Telescope (formerly known as WFIRST [904]). Future BAO data are expected also from galaxy spectroscopic surveys such as DESI [905], or radio telescopes such as BINGO [906]. All these experiments combined together are expected to shed light on the current cosmological tensions, giving us the possibility of ruling out large portions of the parameters space, and therefore, of the scenarios beyond-the-standard-model, that now compete for fitting the data. as the symmetries and transformations within the formulation. We close this section with a new discussion on the relationship of teleparallel geometries to other branches of physics such as continuum mechanics where we lay out the way in which this has been used to describe crystal structures. After laying out the geometric foundations to TG and its relationship to the broader class of metric-affine theories, we apply these concepts to the teleparallel equivalent of GR in section 4 also, similar to GR, holds a special place in teleparallel theories as the best behaved gravitational model in TG. Indeed, we also discuss some aspects related to attempts to quantize TEGR and their relationship to similar approaches in GR. On the other hand, there have been a plethora of proposals to modify TEGR which we examine in section 5 where we present the best studied and most promising teleparallel theories of gravity. We here tackle a number of issues related to fully describing the ideas behind good tetrad-spin connection pair which is crucial for determining the correct field equations of a theory. We also point out progress in some of these modifications and extensions to TEGR in terms of their spin connection contribution. In particular, we flesh out the impact of the antisymmetric field equations which are related to the local Lorentz transformation DoF within these theories. Finally, we expand on the scalar-tensor sector of the landscape of teleparallel theories and focus our attention on the new and interesting BDLS theory (teleparallel analogue of Horndeski gravity) which have been gaining popularity as a way to revive standard Horndeski gravity, which has become severely constrained due to its prediction on the speed of propagation of GWs (in many of its models). Taking these theories, we explore possible cosmologies in their backgrounds through various approaches in section 6. By and large, we explore flat homogeneous and isotropic backgrounds but are not limited to this. Moreover, we also focus on f (T) cosmology since this is the most studied in the literature, but again we also analyze other scenarios. It is worth noting that the majority of works in the literature in TG have been associated with works in this regime, namely background cosmology. In this context, we first explore possible reconstruction methods, which have shown promise in many works in the literature (even outside of TG). This is a convenient way 11. Conclusion In this Review of TG, we have discussed the foundations of this approach to constructing gravitational theories, together with the observational and precision cosmology consequences of some of the most popular forms of teleparallel theories. We started with a survey of the historical context in which TG was born (into from the early work by Einstein, Weitzenböck and others). In section 1, we also point out new results that are first being reported here together with corrections to the literature where incorrect results may have appeared. This section also contains our review convention which would be advantageous to use in future works since we have made a significant effort to represent all theories and the vast majority of their results in this convention. We begin the technical details of the formulation of teleparallel geometry through the position of metric-affine geometry in section 2 which was formulated in the later years after Einstein’s original work. In this part of the Review, we see how TG forms part of a much larger trinity of gravity that involves curvature but also non-metricity. Together, these three geometries, through their respective connections, can fully utilize the components of a general gravitational connection. We then focus our attention solely on the teleparallel geometry manifestation of this trinity in section 3 where we delve into the details of how TG can be sourced through a gauge theory of translations. We also go into how the minimal coupling prescription is needed to couple matter to this geometry, as well 178 Rep. Prog. Phys. 86 (2023) 026901 of designing models in the various classes of gravity theories which can then be probed against other phenomena. We also explore the wide range of works that make use of Noether symmetries as an approach to again design particular models of gravity that adhere to particular symmetry setups. At a higher level, we present dynamical systems and its approach in determining important generic behaviors of particular classes of cosmology models and whether they are consistent with our Universe. In this section, we apply these approaches to some of the theories of gravity being surveyed in section 5. The level to which we probe these regimes of theories depends on the amount of work in the literature on these branches of theories with TG. TG poses new challenges in the perturbative sector as well since the tetrad and spin connection approach means that we cannot utilize the traditional mechanics of perturbation theory in the metrical formulation of gravity theories. In section 7 we discuss these crucial points by opening on the traditional 3+1 decomposition of the metric and how this can be translated to teleparallel geometry. Together with a discussion about the SVT decomposition, we then explore how this can be applied within this framework. We also discuss the idea of gauge transformations which are important for relating results between gauge setups. While we use this pathway to cosmological perturbation theory in a number of theoretical setups, the analysis is much more general and lays out the general approach one should take to this kind of study. Indeed, in section 8, we use this approach to probe the polarization structure of GW in the various scenarios that it has been studied in the literature. Here, we find a rich structure of polarizations which manifest some of the propagating DoF of each theory, and their subclasses. In the final part of our Review on TG, we explore the observational sector of the theory. Firstly, in section 9 we analyze works related to astrophysics and how the theories from section 5 can impact those physical process and observational measurements. Indeed, to probe this regime of physics we first need to understand very well how the tetrad spin connection pair can be formulated in a good way to best describe the physical systems that we know from metrical approaches to gravity. We open with long discussion on these technical details together with an explanation of the problem of how a generalized Birkhoff theorem may be conceptualized in TG. On the observational side of things, we work out both the classic Solar System tests and their extension to the so-called gravitomagnetic effect, as well as some details of the PPN analysis as applied to TG. We then move to the idea of compact objects in TG which is poorly understood. By this, we mean that the literature contains some analyses on the topic but which could be better expanded to explore other regime more related to astrophysical measurements. Moreover, due to the good tetrad-spin connection pair paradigm, some early works may need correcting, as we explain in this section. In addition, we discuss the open problems related to this topic such as the issue about interpreting black holes singularities and the issue of wormholes. The section also contains a discussion of the impact of some gravitational models in galactic rotation curve analyses which are important to understanding well the issue of astrophysical dark matter. Review Finally, we explore the all important topic of observational cosmology together with the results and general investigation of precision cosmology in section 10. We do this by discussing firstly the state-of-the-art of cosmology and the growing issues being posed by cosmological tensions. We also discuss the state-of-the-art in terms of cosmography. Interestingly, we have one of the few surveys that details the way in which community codes need to be used to correctly explore modified gravity theories (rather than interpreting modified gravity as an effective fluid in a dynamical DE fashion). In this background, we then collect the most promising models being introduced in the theories in section 5 and then promoted in the background and perturbative sections that ensue in a compendium survey for convenience. We explore the literature for these models both in terms of observational constraints and also their restrictions coming from precision cosmology in the community codes we discussed earlier. This section also entails a significant discussion on possible applications of machine learning in modified gravity in general, and TG in particular, such as in the use of Gaussian processes to find confidence regions for cosmological proposals for f (T) gravity models. In this work, we have expanded on proposed solutions to the problems related to finding good tetrad-spin connection pairs that allow us to probe modified TG. We have also corrected a number of small mistakes in the literature, and discussed places where the literature has been incorrect in more significant ways. We explain a lot of this in section 1.1, and through the remainder of the Review. Below, we also point out open problems and issues in the teleparallel community that need further attention for future work. • Strong coupling in f (T) Gravity—Strong coupling is a pathology of the disparity between DoF that appear in a full Hamiltonian analysis and the DoF that appear at linear perturbative level for particular spacetimes. These pathologies may cause perturbations that are strongly coupled to potentially arbitrarily higher order perturbations which is problematic for making finite calculations. We describe the issue further in section 5.3.1 as well as in other places throughout the f (T) gravity analysis. It is not clear without a full Hamiltonian analysis of other theories whether this issue pervades other formulations of TG for these specific spacetimes. However, we know that certain more nuanced setups which can, with certainty, escape this issue such as ◦ f(T, B) = f(−T + B) = f(R) gravity. It is important to highlight that strong coupling is not a problem of a theory or the model that populate it but of the solutions that can be probed using traditional perturbation theory. It may be the case, that the issue of strong coupling points to the need for a new approach to perturbation theory. TG is a real alternative to the curvature-based geometry of GR, and thus it is not totally surprising that certain approaches breakdown such as the way that the linearized equations are derived. Thus, new approaches to taking perturbations in f (T) should be investigated for potential curing of the strong coupling problem, particularly in the case of cosmological perturbations. 179 Rep. Prog. Phys. 86 (2023) 026901 • Good tetrad-spin connection pairs—In this Review, we have made an active effort to fully immerse readers into the covariant formulation of TG. To this end, we introduce the spin connection and its integral importance in sections 2, 3, and 5.1. Here, we also describe how the antisymmetric field equations are related to the spin connection components. Thus, in TG we find additional six equations from these antisymmetric equations of motion which are related to the Local Lorentz DoF. We give examples of this in section 5.12 but more work needs to be done in order to better understand other examples in the plethora of TG theories, as well as to ground a more routine procedure to finding these good tetrad-spin connection components. • Quantum gravity—The quantization of GR and other gravitational theories has been a long-standing programme in gravitational research. As is well known in the community, there are many attempts at constructing a quantum completion of GR, which had had various successes. In section 4.4, we detail some possible insights that TEGR may have in resolving the problems of quantizing GR. However, many details remain to be worked out and a fully fledged quantum completion of TEGR remains unknown in the literature. Moreover, attempts at using the mechanisms of modified TG to construct teleparallel quantum theories of gravity remain poorly studied and would be an interesting problem to have tackled. In this vein, loop quantum gravity also uses very similar analytical tools as TG to approach the quantum gravity problem. It would be interesting to use similar routes in TG and to explore the differences that may emerge between the two approaches. • Black holes—Black holes are a fundamental part of astrophysics research from individual black hole systems to understanding the dynamics of galaxies. In section 9.4.2, we discuss the foundational concepts behind black hole physics in TG from their definition to their impact on the photon sphere which has become a detectable phenomenon in recent years. However, a fully fleshed out description continues to remain an open problem. It is crucial that a fully fledged route is opened to black hole physics for the theory to continue to progress beyond toy model solutions and into more complicated and physical scenarios. • Compact stars—Another important aspect to astrophysics is that of compact stars which are ubiquitous throughout astronomy and have a huge impact on observational astronomy such as in the measurements of pulsars. Better exploring the inner mechanics of these stars in TG, both in terms of foundations and in the context of simulations, is crucial to having a competitive description of these all important stars in TG. • Galactic simulations—GR necessitates CDM to accurately describe the dynamics of galaxies and clusters of galaxies. In section 9.5, we detail how the galactic dynamics may be explained using a particular form of f (T) gravity. It is not clear if this can also be achieved in other f (T) models or other TG theories. It would be interesting to pursue this problem further and to also expand the analysis to include simulations for clusters of galaxies. Review • Teleparallel cosmology codes—Many cosmological surveys interpret extended theories of gravity as an effective fluid in the cosmological context. This is accurate at background level and can help in many implementations of modified gravity in community codes related to cosmology. However, the perturbations that one determines from the gravitational section, for these terms, and that determined by the code through the fluid interpretation do not equate to each other. While interesting a dynamical dark energy approach, a full implementation of perturbations from TG in these codes in terms of modified Boltzmann equations and perturbation modules would be extremely advantageous to moving the framework forward in the observational sector, and particularly for the precision cosmology approach. We explain this prospect further in section 10. In this Review, we have attempted to give a snapshot of the current state-of-the-art of TG together with possible future directions which are necessary for the topic to progress forward both in terms of its foundations as well as in terms of observational predictions. TG is a very novel approach to modified gravity which naturally produces problems that require novel solutions. It will be interesting to see how the community advances over the next decade of research. Data availability statement No new data were created or analysed in this study. Acknowledgments S B and M H were supported by the Estonian Research Council grants PRG356 ‘Gauge Gravity’ and by the European Regional Development Fund through the Center of Excellence TK133 ‘The Dark Side of the Universe’. S B is also supported by JSPS Postdoctoral Fellowships for Research in Japan and KAKENHI Grant-in-Aid for Scientific Research No. JP21F21789. K F D acknowledges support by the Hellenic Foundation for Research and Innovation (H.F.R.I.) under the ‘First Call for H.F.R.I. Research Projects to support Faculty members and Researchers and the procurement of high-cost research equipment grant’ (Project Number: 2251). The work was also supported by Nazarbayev University Faculty Development Competitive Research Grant No. 11022021FD2926. C E-R acknowledges the Royal Astronomical Society as FRAS 10147 and DGAPA-PAPIIT-UNAM Project IA100220. J L S and J M would also like to acknowledge funding support from Cosmology@MALTA which is supported by the University of Malta. J L S would also like to acknowledge funding from ‘The Malta Council for Science and Technology’ in Project IPAS-2020-007. E D V acknowledges the support of the Addison-Wheeler Fellowship awarded by the Institute of Advanced Study at Durham University. This research was partially carried out using computational facilities procured through the European Regional Development Fund, Project 180 Rep. Prog. Phys. 86 (2023) 026901 Review No. ERDF-080 ‘A supercomputing laboratory for the University of Malta’ and the ‘CosmoNag project’ through Tochtli ICN-UNAM cluster. The authors would like to acknowledge Laur Järv for kindly preparing figure 1. The authors would like to thank José Beltran Jimenez for very useful discussions regarding the strong coupling problem in f (T) gravity. All the authors would like to acknowledge the thorough and significant work performed by the anonymous referees during the preparation of this review. J L S and J M would like to acknowledge support from the Malta Digital Innovation Authority through the IntelliVerse grant. This paper is 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). α δe Tαµν = −EA α Tβ µν δe A β + 2EA α δe Γ A [νµ] (A.2e) = −EA α Tβ µν δe A β + EA α h i × ∂µ δe A ν − ∂ν δe A µ + ω A Bµ δeB ν − ω A Bν δeB µ , (A.2f )   δe Tµ = − EA µ Tλ + gµλ TA + Tλ A µ δe A λ   + gµν EA λ ∂λ δe A ν − ∂ν δe A λ + ω A Bλ δeB ν − ω A Bν δeB λ . (A.2g) In what follows, we consider these variations familiar, in order to move to more complicated expressions. It is crucial to see how the torsion scalar varies with respect to the tetrad. Thus, using the above quantities, we have 1 1 δe T = δ(Tµνα Tµνα ) + δ(Tµνα Tνµα ) − δ(Tµ Tµ ), (A.3) 4 2 Appendix A. Variations in TG where In this section we will briefly collect the variations of some basic and some more complicated quantities of teleparallel theory, in order for the reader to familiarize themselves with the notation. Apart from that we will calculate the gauge current and finally we will discuss the variations of specific theories, i.e. the f(T, B, ϕ, X), f(T, B, TG , BG ) and the teleparallel analogue of Horndeski gravity (BDLS). δ(Tµνα Tµνα ) = 4Tµ να EA µ (∂ν δe A α + ω A Bν δe B α ) − 4Tµνα Tµνβ EA β δe A α , (A.4a) δ(Tµνα Tνµα ) = 2(Tβνµ − Tµνβ )Tνµα EA α δ e A β + 2(Tµ ν β − Tβ ν µ )EA ν (∂µ δ e A β + ω A Bµ δ e B β ), (A.4b) δ(Tµ Tµ ) = −2(−Tβ Tα βµ + Tα Tµ )EA µ δe A α A.1. Basic teleparallel quantities − 2(Tµ EA β − Tβ EA µ )(∂µ δe A β + ω A Bµ δe B β ). (A.4c) Let us start by showing some necessary and useful identities ∂e B ν = δAB δνµ , ∂e A µ (A.1a) ∂EB ν = −EB µ EA ν , ∂e A µ (A.1b) ∂e = eEA µ , ∂e A µ (A.1c) By replacing equations (A.4a)–(A.4c) in equation (A.3), integrating by parts and neglecting the boundary terms, we get   1 λµ σ µλ B νλ e δe T = 2e ∂µ (eSA ) − T µA Sσ + ω Aν SB δe A λ . e (A.5) ∂g αβ = −gµβ EA α − gµα EA β . ∂e A µ (A.1d) Similarly, the variation of the torsion tensor with respect to the spin connection ω A Bµ reads from equation (1.6a) δω Tα µν = δω (EA α T A µν ) = EA α δω T A µν Using the above identities, one can easily obtain the variations of basic quantities with respect to the tetrad, δ e , such the inverse tetrad, EA µ , the determinant of the tetrad, e, a general metric and its inverse, (gµν , gµν ), as well as the torsion tensor, Tα µν and the torsion vector, T µ . All of the above vary as follows δe EA = −EB EA δe ν , (A.2a) δe e = δ det(e A µ ) = e EA µ δe A µ , (A.2b)  δe gµν = ηAB e A µ δe B ν + e A ν δe B µ , (A.2c) δe gµν = − (gµα EA ν + gνα EA µ ) δe A α , (A.2d) µ µ ν B = EA α e B ν δω A Bµ − EA α e B µ δω A Bν , (A.6) and thus, the torsion vector becomes δω Tµ = δω Tα αµ = EA α e B ν δω A Bα − δω B Bν . (A.7) It is worth noticing that the spin connection has to be purely inertial, since if one varies the teleparallel action with respect to arbitrary spin connection coefficients, the associated equations of motion will lead to the vanishing of the torsion tensor [123]. However, if one assumes a spin connection in the inertial class only, ω A Bµ = Λ A C ∂µ (Λ−1 )C B , effectively it means that there exists a frame in which ω = 0 (Weitzenböck gauge), but simultaneously one is allowed to perform a local Lorentz transformation Λ A B that belongs to the Lorentz group and which produces the above connection. 181 Rep. Prog. Phys. 86 (2023) 026901 Review A.2. Calculation of the gauge current A.3. Example teleparallel theories By definition, the gauge current can be written as Here we will give the variations of teleparallel quantities that appear in three well-known modifications of TG; the f(T, B, ϕ, X), the f(T, B, TG , BG ) theory presented in section 5.6 and the BDLS theory in section 5.8.2. JAµ = − 1 ∂ (eT) ∂T = −TEAµ − A , A e ∂e µ ∂e µ (A.8) where equation (A.1c) was used for the tetrad determinant derivative. Now, using equation (1.6d), this can be expanded to JAµ = −TEAµ − Apart from the bare quantities that we discussed so far, one can encounter those in arbitrary functions, such as f(T), f(T, B) and so on. For this reason, we thought it would be useful to consider a general function of the form f(T, B, ϕ, X) in order to discuss its variations. Thus, varying the Lagrangian A.3.1. f(T,B,ϕ,X). B σ µ̄ B 1 ∂ T σµ̄ σµ̄ 1 ∂ TB 1 ∂ T σρ ρσ TB − TBσµ̄ − T B A A 4 ∂e µ 4 ∂e µ 2 ∂ eAµ ∂ TρσB ∂ Tρσρ ∂ Tµ̄σµ̄ 1 − TBσρ + Tµσµ + Tρσρ , A A 2 ∂e µ ∂e µ ∂ eAµ (A.9) L = ef(T, B, ϕ, X), (A.13) with respect to the tetrad we get where the torsion tensor expression in equation (3.28) means that the derivatives in the second and fourth terms can be determined by δe (ef(T, B, ϕ, X)) = f(T, B, ϕ, X)δe e + efT δe T + efB δe B + efX δe X, (A.14) B ∂T σµ̄ = ω BAσ δµ̄µ − ω BAµ̄ δσµ , ∂eAµ where f i with i = T, B, X is the partial derivative of f with respect to i. The first term is easily obtained by equation (A.2c) and the second one by (A.10) while the derivatives in the third, fifth, sixth and seventh term turn out to respectively be represented by h   i efT δ T = −2 e (∂µ fT ) SA µβ + ∂µ eSA µβ fT − efT TµσA Sσ βµ δ e A β .  ∂TBσµ̄ ∂  σ σ̄ µ̄λ B̄ = η T g g B B̄ σ̄λ ∂eAµ ∂eAµ (A.15) For the third term we have B = 2∂µ (eTµ )/e and thus = −TBµµ̄ EAσ − TBAµ̄ gµσ − TBσµ EAµ̄ − TBσA gµµ̄ + ηBB̄ ω B̄Aσ̄ gσσ̄ gµµ̄ − ηBB̄ ω B̄Aλ gσµ gµλ , (A.11a) ∂TρσB ∂eAµ = efB δe B = − [BfB + 2Tµ ∂µ fB ] δe e − 2e(∂µ fB )δe Tµ . The first term of the right-hand side is trivial; the second term gives ∂  µ̄ σσ̄ ρ B̄  E g EB̄ T σ̄µ̄ ∂eAµ B h    i − 2∂ν eEA β (∂ ν fB ) δe A β . (A.17) Thus, equation (A.16) will be h  e fB δe B = e (∂µ fB ) SA µβ + Kµβ A + EC β EA σ gλµ ∂σ eC λ  − EC σ EA λ gµβ ∂λ eC σ i ◦ ◦ ◦ + EA µ ∇µ ∇β fB − EA β □fB − BfB EA β δe A β , = −TµσA − Tρµρ EAσ − TρAρ gµσ + EBµ gσσ̄ ω BAσ̄ (A.11c)  ∂Tµ̄σµ̄ ∂ = A gσσ̄ Tµ̄σ̄µ̄ ∂eAµ ∂e µ (A.18) ◦ where the relation e∇µ Yµ = ∂µ (eYµ ) and the definitions of the superpotential and the contortion were used. Using several identities, it is possible to show that all the terms appearing in the first line of the above equation is 2(∂µ fB )SA λµ (see [157] for further details). Thus, the variations of the boundary term contribution are  ◦ ◦ ◦ 1 e fB δe B = 2e EA ν ∇λ ∇µ fB − EA λ □fB − BfB EA λ 2  + (∂µ fB )SA λµ δeA λ . (A.19) (A.11d) where the derivatives in equations (A.1a), (A.1b) and (A.1d) were used throughout. By combining these derivative contributions in equation (A.9) JAµ = −TEAµ + 2ω BAσ SBµσ + 2TBσA SBσµ ,  + ω β A µ − ω σ Aσ gµβ + 2∂µ eEA µ (∂ β fB )  ∂Tρσρ ∂  = A EBρ gσσ̄ TBσ̄ρ A ∂e µ ∂e µ = −TµσA + EBµ ω BAσ − δσµ EBρ ω BAρ ,  −2e(∂µ fB )δe Tµ = 2e(∂µ fB ) TA gµβ + Tβ EA µ + Tβ A µ = −TρσA EBµ − TµσB EAρ − TρµB EAσ − TρAB gµσ   + EB̄ρ EBµ gσσ̄ ω B̄Aσ̄ − EBµ̄ gσµ ω B̄Aµ̄ , (A.11b) − EBρ gσµ ω BAρ , (A.16) (A.12) where a long but straightforward series of simplifications were taken together with the expressions for the contortion tensor in equation (4.149) and superpotential in equation (4.163). 182 Rep. Prog. Phys. 86 (2023) 026901 Review Finally, the kinetic term in equation (A.14) will give (A.20) The equations of motion are given in section 5.6 equation (5.91). The action of the theory is given by 1 Sf(T,B,TG ,BG ) = 2κ 2 h  MBCD (KL C K ),D + KCD P KL P Q δIJKL i MDPB C Q L K MPCD L K + ηPJ δQCKL K D K I + δKLIJ K P KCD B  ACBD (A.26) − δKLIJ ∂σ eED σ EA µ (∂µ fBG )KL C K . P B IJ = eEM µ (∂µ fBG ) 1 efX δe X = − efX (∂µ ϕ)(∂ν ϕ)δe gµν 2 = efX (∂µ ϕ)(∂ν ϕ)gµβ EA ν δe A β . A.3.2. f(T,B,TG ,BG ) theory. and ˆ d4 x e f(T, B, TG , BG ) + Sm , (A.21) By varying the action (5.126) with respect to the tetrads, we find that A.3.3. Teleparallel analogue of Horndeski theory. and its variation will be δSf(T,B,TG ,BG ) = ˆ 1 d4 x [fδe + efT δT + efB δB 2κ 2 +efTG δTG + efBG δBG ] + δSm , (A.22) δe SBDLS = eLTele EA µ δe A µ + eδe LTele + e 5 X Li EA µ δe A µ i=2 + eδe where 5 X Li + 2κ2 eΘA µ δeA µ = 0. (A.27) i=2 fδ e = efEA β δ e A β , (A.23a) h  i  efT δ T = −2 e (∂µ fT ) SA µβ + ∂µ eSA µβ fT − efT T σ µA Sσ βµ δ e A β , (A.23b) h ◦ ◦ ◦ efB δB = 2eEA ν ∇β ∇µ fB − 2eEA β □fB − BefB EA β  (A.23c) −2e (∂µ fB ) SA µβ δe A β , 2(∂λ GTele,T )SA µλ + 2e−1 ∂λ (eSA µλ )GTele,T − 2GTele,T T σ λA Sσ λµ + 2GTele,T ω B Aν SB µν h − ϕ;A GTele,X ϕ;µ − GTele,I2 vµ − 2GTele,J1 aµ aJ ϕ;J h   efTG δTG = ∂µ EH µ EB β (Y B A H − YH A B + YA [BH] )   + TI AB EH β Y B I H − YH I B + YI [BH] i MBCD − 2efTG δIJKL ED β KJ M I KEB K ∂A (KL C E ) δe A β , + GTele,J3 vI tK µI ϕ;K − 2GTele,J5 tIµK tIJK ϕ;J + 2GTele,J6 tILK tµ M I ϕ;K ϕ;L ϕ;M − 2GTele,J8 tIJK tIJ µ ϕ;K  i − GTele,J10 aJ ϕ;I ϵµJCD tI CD + ϵIJCD tµCD 1h + MI (ϵIB CD EC µ TB AD − ϵIB CD ED µ ω B AC ) 3  i + e−1 ∂ν eMI ϵIA CD EC ν ED µ (A.23d) h   efBG δ BG = − ∂µ (P B A H − PH A B + PA [BH] )EH µ EB β   + TI AB EH β P B I H − PH I B + PI [BH] MBCD − δIJKL eED β (∂M fBG )KJ B I (∂A KL C K ) − NI (EI µ ω ρ Aρ − ω µ AI − Tµ AI − vA EI µ )   + e−1 ∂ν eNI (EA ν EI µ − EA µ EI ν )   − OIJK HIJKA µ + e−1 ∂ν eOIJK LIJKA µν − LTele EA µ i µ β β A + e(∂µ fBG )(EA β Bµ G − EA BG ) + efBG BG EA δ e β . (A.23e) In the above, we have defined X A IJ = ∂TG , ∂KA IJ + 2EA ν gµα as well as  5 X G (i) αν = 2κ2 ΘA µ , (A.28) i=2 CABD Y B IJ = efTG X B IJ − 2δELKJ ∂µ  × efTG ED µ KL C E KIA K , P5 The variations of δe i=2 Li give the standard Horndeski field equations whereas the variations δe LTele are related to the extra terms coming from TG. After doing several computations, one finds that the field equations can be written as where we have defined (A.24) MI = 2GTele,Tax aI + 2GTele,J1 ϕ;I ϕ;J aJ + GTele,J10 ϵA I CD ϕ;A ϕ;J tJ CD ,  1 J I KL K B R CD + KJ B I KFC K KL D F 2  ABCD J I = δIJKL K B (KL C K ),D + KCD M KL M K (A.25) BGA = ϵIJKL ϵBCDA (A.29a) NI = 2GTele,Tvec vI + GTele,I2 ϕ;I + 2GTele,J2 ϕ;I ϕ;J vJ + GTele,J3 ϕ;K ϕ;J T i KJ , 183 (A.29b) Rep. Prog. Phys. 86 (2023) 026901 Review OIJK = GTele,J3 ϕ;J ϕ;K vI + 2GTele,J5 ϕ;L ϕ;J T i L K where Jµ−Tele and Pϕ−Tele are defined as ;J ;K ;L ;M i + 2GTele,J6 ϕ ϕ ϕ ϕ T LM ◦ ;L ;K IJ ◦ Jµ−Tele = −GTele,X (∇µ ϕ) + GTele,I2 vµ + 2GTele,J1 aµ aν ∇ν ϕ JK ;A ;B ;I + 2GTele,J8 ϕ ϕ t L + GTele,J10 ϵAB ϕ ϕ ϕ , (A.29c) ◦ ◦ − GTele,J3 vα tµ να (∇ν ϕ) − 2GTele,J5 tβνα tβµα (∇ν ϕ) ◦ + 2GTele,J8 t αν µ tαν β (∇β ϕ) ∂tIJK HIJKA µ = A , ∂e µ ◦ (A.29d) ◦ ◦ − 2GTele,J6 tναβ tµ σ ν (∇α ϕ)(∇β ϕ)(∇σ ϕ), ◦ − GTele,J10 aν (∇α ϕ)(ϵµ νρσ t αρσ + ϵ α νρσ tµρσ ), ∂tIJK LIJKA µν = A . ∂e µ,ν (A.32a) (A.29e) Pϕ−Tele = GTele,ϕ . In addition, the variation of the scalars appearing in LTele is given by (A.32b) For more details about the derivation of these equations, see ◦ eGTele,Tax δe Tax = 2eGTele,Tax aI δe aI , [666]. Using R = −T + B, one finds that Piϕ is given by [305] (A.30a) P2ϕ = G2,ϕ , I eGTele,Tvec δe Tvec = 2eGTele,Tvec v δe vI , (A.30b) ◦ (A.33a) ◦ P3ϕ = ∇µ G3,ϕ ∇µ ϕ, eGTele,I2 δe I2 = eGTele,I2 ϕ;I δe vI − eGTele,I2 vµ ϕ;A δe eA µ , (A.30c) h ◦ i ◦ ◦ P4ϕ = G4,ϕ (−T + B) + G4,ϕX (□ϕ) 2 − (∇µ ∇ν ϕ) 2 , (A.33c) eGTele,J1 δe J1 = 2eGTele,J1 ϕ;I ϕ;J aJ δe aI − 2eGTele,J1 aJ aµ ϕ;J ϕ;A δe eA µ , ◦ ◦ ◦ 1 P5ϕ = −∇µ G5,ϕ Gµν ∇ν ϕ − G5,ϕX 6 h i ◦ ◦ ◦ ◦ 3 × (□ϕ) − 3□ϕ(∇µ ∇ν ϕ) 2 + 2(∇µ ∇ν ϕ)3 , (A.33d) (A.30d) eGTele,J3 δe J3 = eGTele,J3 ϕ;K ϕ;J T i KJ δe vI + eGTele,J3 ϕ;J ϕ;K vI δe tIJK + eGTele,J3 vI tK µI ϕ;K ϕ;A δe eA µ , (A.33b) (A.30e) and Jiµ is defined as [305] eGTele,J5 δe J5 = 2eGTele,J5 ϕ;L ϕ;J T i L K δe tIJK − 2eGTele,J5 tIµK tIJK ϕ;J ϕ;A δe eA µ , ◦ J2µ = −L2,X ∇µ ϕ, (A.30f ) ◦ eGTele,J6 δe J6 = 2eGTele,J6 ϕ ϕ ϕ ϕ T LM δe tIJK + 2eGTele,J6 tILK tµ M I ϕ;K ϕ;L ϕ;M ϕ;A δe eA µ , (A.30g) ◦ ;J GTele,J10 δe J10 = eGTele,J10 ϵA CD ϕ ϕ tJ JK CD δe aI B ;A ; I (A.30i) By varying the action (5.126) with respect to the scalar field we obtain the modified Klein Gordon equation ∇ µ Jµ−Tele + 5 X i=2 ! Jiµ = Pϕ−Tele + 5 X ◦ ◦ + eGTele,J10 ϵAB A ϕ ϕ δe tIJK   − eGTele,J10 aJ ϕ;I ϕ;A ϵµJCD tI CD + ϵIJCD tµCD δe eA µ . ◦ Piϕ , (A.34b) ◦ ◦ ◦ − 2G4,ϕX (□ϕ∇µ ϕ + ∇µ X), (A.30h) ◦ ;A ◦ J4µ = −L4,X ∇µ ϕ + 2G4,X Rµν ∇ν ϕ ◦ ◦  ◦ ◦ ◦ − 2G4,XX □ϕ∇µ X − ∇ν X∇µ ∇ν ϕ eGTele,J8 δe J8 = 2eGTele,J8 ϕ;L ϕ;K tIJ L δe tIJK I ◦ J3µ = −L3,X ∇µ ϕ + G3,X ∇µ X + 2G3,ϕ ∇µ ϕ, ;J ;K ;L ;M i − 2eGTele,J8 tIJK tIJ µ ϕ;K ϕ;A δe eA µ , (A.34a) (A.31) i=2 184 ◦ ◦ (A.34c) J5µ = −L5,X ∇µ ϕ − 2G5,ϕ Gµν ∇ν ϕ h◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ − G5,X Gµν ∇ν X + Rµν □ϕ∇ν ϕ − Rνλ ∇ν ϕ∇λ ∇µ ϕ i ◦ ◦ ◦ ◦ −Rαµβν ∇ν ϕ∇ α ∇β ϕ n1 ◦ h ◦ i ◦ ◦ + G5,XX ∇µ X (□ϕ) 2 − (∇α ∇β ϕ) 2 2 o ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ − ∇ ν X □ ϕ∇ µ ∇ ν ϕ − ∇ α ∇ µ ϕ∇ α ∇ ν ϕ n1 ◦ h ◦ i ◦ ◦ + G5,ϕX ∇µ ϕ (□ϕ) 2 − (∇α ∇β ϕ) 2 2 o ◦ ◦ ◦ ◦ ◦ + □ ϕ∇ µ X − ∇ ν X∇ ν ∇ µ ϕ . (A.34d) Rep. Prog. Phys. 86 (2023) 026901 Review A3 = −16H2 ΞΥ, It is easy to see that the equations of motion of well-studied theories can be recovered by choosing the form of the Gi functions; e.g. Brans–Dicke theory can be obtained by setting GTele = 0, G2 = ωϕ X, G3 = 0, G4 = ϕ, and G5 = 0. "  A4 = −3a 6 Π̇(ΠB + ΠT ) Π ′ 6Ḣ2 (ΠB − fTB )  + 6H2 Ḣ(3fTB − ΠB ) + 18H4 (ΠT − fTB ) − H2 fT − HḟT  + 6H2 Ḣ(−24H3 (fTB − ΠB )(fTB − ΠT ) + ΠB ḟT ) + H Π̇2 Appendix B. Geff calculation in f(T,B) gravity Here we present all the coefficients appearing in section 7.7.2 regarding the effective Newton’s constant in f(T, B) gravity and subcases (for more details see [158]) A1 = −a4 ΥΠ̇(Π̇ + 12H3 fTT ), − 12HḢ2 (fTB (ΠB − fTB ) + Ξ) #  2 + H (2HfT + ḟT )(fTB − ΠT ) , (B.1a) A2 = −4a2 HΥ(2ΠB Π̇ − 12H3 fBB fTB − HΥfT ), (B.1c) (B.1d) (B.1b)  A5 = 3a4 −4H4 fT 2 (ΠB + ΠT )2 + 48H4 fTB 3 −12H2 (5ΠB + 4ΠT )Ḣ + 24(ΠB + ΠT )Ḣ2 + HΠ̇ + 24H3 fTB 2     × 24H3 2Ξ + (4ΠB + ΠT )(ΠB + 2ΠT ) Ḣ − 12H 3Ξ + 2 ΠB 2 + 4ΠB ΠT + ΠT 2 Ḣ2   − H 2(ΠB + 2ΠT )Π̇ + (ΠB + ΠT )f˙T − Ḣ (11ΠB + 10ΠT )Π̇ + (ΠB + ΠT )f˙T  2  + 2HfT −24H3 fTB 2 (ΠB + ΠT )(3H2 − 8Ḣ) + (ΠB + ΠT ) −H2 ΠT + ΠB (7H2 − 3Ḣ) Π̇ + 4H2 fTB   × 6H −ΠB + 3ΠB ΠT + 2 Ξ + ΠT 3  2 + 3H 16H ΠB −Ξ + ΠT 4 2  2    − 12H 4Ξ + 2ΠB + 5ΠB ΠT − ΠT Ḣ + (ΠB + ΠT )Π̇ 2 2  + 16H ΠB 4Ξ + 3ΠB ΠT − ΠT Ḣ − 4Ξ(ΠB + ΠT )Ḣ 2 2 2 − H(ΠB + ΠT )2 f˙T  !     + 3ΠB 192H 6 ΞΠB + Ξ + ΠB 2 ΠT Ḣ − 96H4 ΠT (3Ξ + 2ΠB ΠT )Ḣ2 + (ΠB + ΠT )ḢΠ̇2 + 3H2 (ΠB + ΠT )Π̇(−Π̇ + f˙T ) − 8H5 − 4H3 Ḣ     −2Ξ + 3ΠT (ΠB + ΠT ) Π̇ + 2(Ξ + ΠB ΠT )f˙T       29Ξ + 20ΠB ΠT − 7ΠT 2 Π̇ + 2 Ξ − ΠT 2 f˙T + fTB −12H3 H2 4Ξ − 8ΠB 2 − 6ΠB ΠT − 6ΠT 2   − 29Ξ + 11ΠB 2 + 33ΠB ΠT − 7ΠT 2 Ḣ Π̇ + (ΠB + ΠT )(4H2 − 3Ḣ)Π̇2  + 24H3 12H(ΠB + ΠT )(3Ξ + 4ΠB ΠT )Ḣ2 + H2 2Ξ + ΠB 2 + 3ΠB ΠT f˙T   + Ḣ −24H ΠB + ΞΠT + 4ΠB ΠT + ΠB 3Ξ + 2ΠT 3 3 2 185 2  + Ξ − ΠT 2  !!! f˙T , (B.1e) Rep. Prog. Phys. 86 (2023) 026901 Review ∆9 = 2a4 A7 − Π̇(−Π̇ − 12H3 fTT ) + 32A5 H2 Ξ  + 8a2 A6 H −(HΥfT − 2ΠB Π̇) A6 = a2 −4H2 fT 2 (ΠB + ΠT )2 + 864H 6 (fTB )3 (3ΠB + 2ΠT ) + 24H3 ΠB (7Ξ + 6ΠB ΠT )Π̇ − ΠB (ΠB + ΠT )Π̇2   + 4HfT −6H3 (fTB − ΠB ) 2Ξ + (ΠB − ΠT )ΠT   − 9H (fTB )2 (ΠB + ΠT ) + ΠB Ξ − ΠT 2   + fTB −Ξ + ΠT 2 Ḣ + ΠB (ΠB + ΠT )Π̇  +24H3 (fBB fTT + Ξ) + 12H3 fTB (fTT − Υ) ,   ∆10 = 8H a2 A7 −(HΥfT − 2ΠB Π̇) + 24H3 (fBB fTT + Ξ)   (B.1q) +12H3 fTB (fTT − Υ) + 4A6 HΞ , − 12H3 fTB 2 72H3 ΠB (4ΠB + 5ΠT ) + 60HΞḢ   − (7ΠB + 5ΠT )Π̇ + fTB 864H 6 3ΞΠB + ΠB 3 − ΞΠT  + 4ΠB 2 ΠT + 720H4 Ξ(ΠB + ΠT )Ḣ   − 12H3 10Ξ + 7ΠB 2 + 17ΠB ΠT Π̇ + (ΠB + ΠT )Π̇2     + 12H3 −72H3 ΠB ΞΠB + Ξ + ΠB 2 ΠT ! ˙ − 60HΞ(2Ξ + ΠB ΠT )Ḣ + 5Ξ(ΠB + ΠT )fT , (B.1f ) 2 2 A7 = 12H Ξ(12H (fBB fTT + 2Ξ) + ΥfT ), (B.1g) ∆1 = −a4 A1 Π̇(12HΠB Ḣ − Π̇), (B.1h) ∆11 = 32A7 H2 Ξ. The terms appearing in table 20 are H,1 = G2,X − 2G3,ϕ + GTele,X , (C.1a) H,2 = GTele,ϕT , (C.1b) H,3 = GTele,I2 − 2G4,ϕ , (C.1c) H,4 = GTele,Tvec , (C.1d) H,5 = 2GTele,Tvec − GTele,T + G4 , (C.1e)   H,6 = 2 G2,X − 2G3,ϕ + GTele,X (GTele,ϕT − GTele,ϕTvec ) (B.1j) × (G2,X − 2G3,ϕ + GTele,X ) + (GTele,I2 − 2G4,ϕ )  × (2G4,ϕϕ − GTele,ϕI2 ) + (GTele,I2 − 2G4,ϕ ) 2    ∆4 = 2H 4H a2 A3 (−ΥfT − 6ΞḢ) + 3A2 Ξ + 5a2 A3 ΠB Π̇ , × (G2,ϕX − 2G3,ϕϕ + GTele,ϕX ) , (B.1k) ∆5 = 24A3 H2 Ξ, (C.1f ) and (B.1l)   2 2 2 − 3H,5 β̃ = 4H,1 H,3 H,2 4H,4 H,5 − 2H,4 ∆6 = 2a4 A4 Π̇(Π̇ + 12H3 fTT ), (B.1r) Appendix C. PPN parameters for teleparallel analogue of Horndeski  ∆2 = a2 −2HΠB Π̇(6a2 A2 Ḣ − 5A1 ) + a2 A2 Π̇2  +8A1 H2 (−ΥfT − 6ΞḢ) , (B.1i) ∆3 = a4 A3 Π̇2 − 2a2 HΠB Π̇(6a2 A3 Ḣ − 5A2 )  + 8H2 a2 A2 (−ΥfT − 6ΞḢ) + 3A1 Ξ ,  (B.1m) 2 2 + H,5 + H,3 3H,4 H,5 − H,4     ∆7 = 2a Π̇ a2 A5 (Π̇ + 12H3 fTT ) + 8A4 HΠB   − 4H3,1 H,4 H,5 H,4 − 2H,5  + H,1 H3,3 8H,2 H,4 − 2H,5 + H,3 4H,4 + 7H,5 2  + 96A4 H (fBB fTT + Ξ) + 48A4 H fTB (fTT − Υ) + H,3 −6H,2 H4,3 + 3H5,3 + 2H,6 H,5 − 2H,4 − 4A4 H2 ΥfT , + G4,ϕ − 8H,1 H3,3 H,4 − 2H,5 4 (B.1p) 4  (B.1n)     2   2 2 2 + 6H5,3 . + 8H,1 H,3 H,5 − H,4 ∆8 = −2a4 A6 Π̇(−Π̇ − 12H3 fTT ) + 8a2 A5 H  + −(HΥfT − 2ΠB Π̇) + 24H3 (fBB fTT + Ξ)  (B.1o) +12H3 fTB (fTT − Υ) + 32A4 H2 Ξ,  (C.2) Here, Gi and GTele denote the constant Taylor coefficients at the background level for Gi and GTele (see equations (5.122a)– (5.122d) and (5.125)). 186 Rep. Prog. Phys. 86 (2023) 026901 Review Appendix D. EoS terms for f(T,B) gravity w(z)4 = We include some definitions that are used in the derivation of f(T, B) cosmology in section 10.3.3 which originated in [263]. General Taylor expansion model:  wx1 = [(4A2 + 27A3 + 11A4 ) ä(t) − (4A2 + 12A3 + 7A4 ) ȧ(t)] , (D.1) h 2 2 wx2 = 4 (3A3 + A4 ) ä(t) + ȧ(t) + 4a (3) 2(4A2 +27A3 +11A4 ) 3 +7A4 + 4A2 +(z12A (z+1)3 +1) 2 , (z + 1)7 w(z)5 = 6(9A3 +2A4 ) + (z+1) 6 (D.2) h wx3 = (9A3 + 2A4 ) a(3) (t)ȧ(t) + ä(t) ((4A2 + 3A4 ) ä(t) (D.3) h i wx6 = 72 (3A3 + A4 ) a(t) 2 ȧ(t) 2 ä(t) − a(3) (t)   + a(t)5 A3 B 2 + T (A4 B + A2 T + A1 ) + A0 , w(z)8 = wx 4 = wx 5 = t0 2m+2 3m (m − 1)m (D.11) ȧ(t) 2 a(t) 2 (D.15) i (z + 1) 6 , 4A2 + 12A3 + 7A4 , (z + 1)10 (D.16) (D.17) (D.18) h i ... wx3 = b0 3k+1 (k − 1)kȧ(t)3 a(t) 2 a (t) − 4ȧ(t)3 + 3a(t)ȧ(t)ä(t)  k 2a(t)ä(t) + 4ȧ(t) 2 × , (D.21) a(t) 2 (D.10) h (D.14)   wx2 = ȧ(t) 2 − a(t)ä(t) ȧ(t) 2 − a(t) [ä(t) + ȧ(t)] , (D.20) (D.9) (8A2 + 36A3 + 17A4 ) , (z + 1)10 2 6 (z+1)3 + (z+1)4 (D.13) Power law model:  ... wx1 = (k − 1)a(t)5 a (t) + 48ȧ(t) 6 − 30a(t)ȧ(t)4 ä(t)   .... + a(t)4 ä(t) a (t) + 3(k − 1)ȧ(t)   ... − a(t) 2 ȧ(t) 2 21ä(t) 2 + 2 a (t)ȧ(t)   .... + a(t)3 3ä(t)3 + ȧ(t) 2ȧ(t) a (t) − 2(k − 1)ȧ(t) ... − a (t)ä(t))]} , (D.19)   wx8 = a(t)5 − A3 B 2 + T (A4 B + A2 T + A1 ) + A0 , (D.8) 96A3 2 (z+1)3 − (z+1)5 w(z)2 = , (z + 1)4  , 2(4A2 +9A3 +6A4 ) 4A3 +A4 (z+1) 2 − (z+1)3 w(z)10 = . (z + 1)7 (D.7) A3 B 2 + T (A4 B + A2 T + A1 ) + A0 , (z + 1)5 (3A3 + A4 ) w(z)9 = (D.6) wx7 = 8A2 + 36A3 + 17A4 , 72A3 1 (z+1)4 + (z+1) 2 , (z + 1)5 h wx5 = 72 (4A2 + 12A3 + 7A4 ) ȧ(t)5 + 6a(t)3 ȧ(t)   × 12A3 a(3) (t) + ȧ(t) , (D.5) w(z)3 = (z+1)3 4 (z + 1) w(z)7 = − wx4 = − [(4A2 + 9A3 + 6A4 ) ä(t) + (4A3 + A4 ) ȧ(t)] , (D.4) w(z)1 = − 2(4A2 +3A4 ) 2(2A2 +9A3 +4A4 ) − (z+1)3 (z+1) 2   2A3 2A4 24 − − 16(3A3 +A4 ) (z+1) 2 (z+1)3 − + (z+11)4 (z+1)4 (z+1) 6 w(z)6 = , (z + 1)3 i (t) (2A3 ȧ(t) − A4 ä(t)) , +2 (2A2 + 9A3 + 4A4 ) ȧ(t))] , 2 (D.12) im   w a(t)ä(t) − ȧ(t) 2 + [a(t)ä(t)+x32ȧ(t) 2 ] 2 a(t)ȧ(t) , ik  h ... 2ȧ(t) 2 b0 6k (k − 1)kȧ(t) −a(t) 2 a (t) + 4ȧ(t)3 − 3a(t)ȧ(t)ä(t) a(t)ä(at()+ t) 2 [a(t)ä(t) + 2ȧ(t) 2 ] 187 2 (D.22) . (D.23) Rep. Prog. Phys. 86 (2023) 026901 Review h i  k−1)  6(k − 1) 2 24 5 − 3(z(+ 1) 2 (z+1) w(z)1 = − +  (z + 1)9 (z + 1)7 + 24 (z+1)9 − 12 − (z+1)7   2(k−1) 24 2 + (z+1) 2 (z+1)5 (z+1) 2 (z+1) 2 (z + 1)3 − [11] Zwicky F 1937 On the Masses of nebulae and of clusters of nebulae Astrophys. J. 86 217–46 [12] Silk J et al 2010 Particle Dark Matter: Observations, Models and Searches (Cambridge: Cambridge University Press) [13] Mukhanov V 2005 Physical Foundations of Cosmology (Cambridge: Cambridge University Press) [14] Guth A H 1981 The inflationary universe: a possible solution to the horizon and flatness problems Phys. Rev. D 23 347–56 [15] Linde A D 1982 A new inflationary universe scenario: a possible solution of the horizon, flatness, homogeneity, isotropy and primordial monopole problems Phys. Lett. B 108 389–93 [16] Starobinsky A A 1980 A new type of isotropic cosmological models without singularity Phys. Lett. B 91 99–102 [17] Sporea C A, Borowiec A and Wojnar A 2018 Galaxy rotation curves via conformal factors Eur. Phys. J. C 78 308 [18] Mannheim P D and O’Brien J G 2013 Galactic rotation curves in conformal gravity J. Phys.: Conf. Ser. 437 012002 [19] Riess A G et al 1998 Observational evidence from supernovae for an accelerating universe and a cosmological constant Astron. J. 116 1009–38 [20] Perlmutter S et al (The Supernova Cosmology Project Collaboration) 1999 Measurements of Omega and Lambda from 42 high redshift supernovae Astrophys. J. 517 565–86 [21] Weinberg S 1989 The cosmological constant problem Rev. Mod. Phys. 61 1–23 [22] Clifton T, Ferreira P G, Padilla A and Skordis C 2012 Modified gravity and cosmology Phys. Rep. 513 1–189 [23] Bull P et al 2016 Beyond ΛCDM: problems, solutions and the road ahead Phys. Dark Universe 12 56–99 [24] Capozziello S and De Laurentis M 2011 Extended theories of gravity Phys. Rep. 509 167–321 [25] Saridakis E N et al (CANTATA Collaboration) 2021 Modified gravity and cosmology: an update by the CANTATA network (arXiv:2105.12582 [gr-qc]) [26] Nojiri S, Odintsov S D and Oikonomou V K 2017 Modified gravity theories on a nutshell: inflation, bounce and late-time evolution Phys. Rep. 692 1–104 [27] Goenner H F M 2014 On the history of unified field theories. Part II. (ca. 1930–ca. 1965) Living Rev. Relativ. 17 5 [28] Møller C 1961 Conservation laws and absolute parallelism in general relativity K. Dan. Vidensk. Selsk. Mat. Fys. Skr. 1 1–50 [29] Pellegrini C and Plebanski J 1963 Tetrad fields and gravitational fields Kgl. Danske Videnskab. Selskab, Mat. Fys. Skrifter 2 4 [30] Møller C 1961 Further remarks on the localization of the energy in the general theory of relativity Ann. Phys., NY 12 118–33 [31] Brown J D and York J W Jr 1993 Quasilocal energy and conserved charges derived from the gravitational action Phys. Rev. D 47 1407–19 [32] Hehl F W, von der Heyde P, Kerlick G D and Nester J M 1976 General relativity with spin and torsion: foundations and prospects Rev. Mod. Phys. 48 393–416 [33] Hayashi K and Nakano T 1967 Extended translation invariance and associated gauge fields Prog. Theor. Phys. 38 491–507 [34] Hayashi K and Shirafuji T 1979 New general relativity Phys. Rev. D 19 3524–53 Hayashi K and Shirafuji T 1982 Phys. Rev. D 24 3312–4 (addendum) [35] Aldrovandi R and Pereira J 1995 An Introduction to Geometrical Physics (Singapore: World Scientific)      108 , (z + 1)12     (D.24) " # 2 1 1 (z+1)3 − (z+1) 2 w(z)2 = − , (z + 1)4 z+1 w(z)3 = (D.25)  ik h   b0 23k−1 3k+1 (k − 1)k (z+11) 2   + (z + 1)4 t0 2 3 (m − 1)m m +2 m (z + 1)4 h 1 (z+1) 2 im      . (D.26) ORCID iDs Sebastian Bahamonde  https://orcid.org/0000-0001-6235120X Konstantinos F Dialektopoulos  https://orcid.org/00000002-0672-1496 Celia Escamilla-Rivera  https://orcid.org/0000-0002-8929250X Jackson Levi Said  https://orcid.org/0000-0002-7835-4365 Eleonora Di Valentino  https://orcid.org/0000-0001-84086961 References [1] van Dongen J 2018 Einstein’s Unification (Cambridge: Cambridge University Press) [2] Misner C, Thorne K and Wheeler J 1973 Gravitation ed W H Freeman (Princeton, NJ: Princeton University Press) [3] Einstein A 1917 Cosmological considerations in the general theory of relativity Sitzungsber. Preuss. Akad. Wiss. Berlin (Math. Phys.) 1917 142–52 [4] Sauer T 2007 Einstein’s unified field theory program (available at: http://philsci-archive.pitt.edu/3293/1/uft.pdf) [5] Einstein A and Schucking E 1997 The Berlin Years: Writings, 1914–1917 (English translation supplement) (Princeton, NJ: Princeton University Press) [6] Barack L et al 2019 Black holes, gravitational waves and fundamental physics: a roadmap Class. Quantum Grav. 36 143001 [7] Glashow S L 1961 Partial symmetries of weak interactions Nucl. Phys. 22 579–88 [8] Weinberg S 1967 A model of leptons Phys. Rev. Lett. 19 1264–6 [9] Salam A 1968 Weak and electromagnetic interactions Conf. Proc. C 680519 367–77 [10] Rubin V C and Ford W K Jr 1970 Rotation of the Andromeda nebula from a spectroscopic survey of emission regions Astrophys. J. 159 379–403 188 Rep. Prog. Phys. 86 (2023) 026901 Review [36] Aldrovandi R and Pereira J G 2013 Teleparallel Gravity vol 173 (Dordrecht: Springer) [37] Maluf J W 2013 The teleparallel equivalent of general relativity Ann. Phys., Lpz. 525 339–57 [38] Ferraro R and Fiorini F 2007 Modified teleparallel gravity: inflation without inflation Phys. Rev. D 75 084031 [39] Linder E V 2010 Einstein’s other gravity and the acceleration of the universe Phys. Rev. D 81 127301 Linder E V 2010 Phys. Rev. D 82 109902 (erratum) [40] Sotiriou T P and Faraoni V 2010 f (R) theories of gravity Rev. Mod. Phys. 82 451–97 [41] Capozziello S, De Laurentis M and Faraoni V 2010 A Bird’s eye view of f (R)-gravity Open Astron. J. 3 49 [42] De Felice A and Tsujikawa S 2010 f (R) theories Living Rev. Relativ. 13 3 [43] Krššák M, van den Hoogen R, Pereira J, Böhmer C and Coley A 2019 Teleparallel theories of gravity: illuminating a fully invariant approach Class. Quantum Grav. 36 183001 [44] Krššák M and Saridakis E N 2016 The covariant formulation of f (T) gravity Class. Quantum Grav. 33 115009 [45] Cai Y-F, Capozziello S, De Laurentis M and Saridakis E N 2016 f (T) teleparallel gravity and cosmology Rep. Prog. Phys. 79 106901 [46] Bahamonde S, Böhmer C G and Krššák M 2017 New classes of modified teleparallel gravity models Phys. Lett. B 775 37–43 [47] Hehl F W, McCrea J D, Mielke E W and Ne’eman Y 1995 Metric affine gauge theory of gravity: field equations, Noether identities, world spinors and breaking of dilation invariance Phys. Rep. 258 1–171 [48] Järv L, Hohmann M and Saal M 2020 Laiendatud geomeetrilised gravitatsiooniteooriad Eesti Vabariigi Preemiad ed T Soomere (Tallinn: Teaduste Akadeemia Kirjastus) pp 57–73 [49] Kobayashi S and Nomizu K 1963 Foundations of Differential Geometry. Foundations of Differential Geometry (New York: Interscience Publishers) [50] Hohmann M 2021 A geometric view on local Lorentz transformations in teleparallel gravity Int. J. Geom. Methods Mod. Phys. 19 2240001 [51] Nester J M and Yo H-J 1999 Symmetric teleparallel general relativity Chin. J. Phys. 37 113 [52] Jiménez J B, Heisenberg L and Koivisto T S 2019 The geometrical trinity of gravity Universe 5 173 [53] Beltrán Jiménez J, Heisenberg L and Koivisto T 2020 The coupling of matter and spacetime geometry Class. Quantum Grav. 37 195013 [54] Beltrán Jiménez J, Heisenberg L, Iosifidis D, Jiménez-Cano A and Koivisto T S 2020 General teleparallel quadratic gravity Phys. Lett. B 805 135422 [55] Böhmer C G and Jensko E 2021 Modified gravity: a unified approach (arXiv:2103.15906 [gr-qc]) [56] Utiyama R 1956 Invariant theoretical interpretation of interaction Phys. Rev. 101 1597–607 [57] Koivisto T, Hohmann M and Złośnik T 2019 The general linear Cartan Khronon Universe 5 168 [58] Blagojević M 2002 Gravitation and Gauge Symmetries (Series in High Energy Physics, Cosmology and Gravitation) (Bristol: IOP Publishing) [59] Ortín T 2004 Gravity and Strings (Cambridge Monographs on Mathematical Physics) (Cambridge: Cambridge University Press) [60] Jackson J 2007 Classical Electrodynamics 3rd edn (New York: Wiley) [61] Mach E, McCormack T and Menger K 1960 The Science of Mechanics: A Critical and Historical Account of Its Development (Open Court Classics) (Chicago, IL: Open Court Publishing Company) [62] Barbour J and Pfister H 1995 Mach’s Principle: From Newton’s Bucket to Quantum Gravity (Einstein Studies) (Boston, MA: Birkhäuser) [63] Einstein A 1996 The Meaning of Relativity (New York: MJF Books) [64] Iorio L 2007 The Measurement of Gravitomagnetism: A Challenging Enterprise (New York: Nova Science Publisher) [65] Sternberg S 1999 Lectures on Differential Geometry (AMS Chelsea Publishing Series) (Providence, RI: American Mathematical Society) [66] Carroll S M 2019 Spacetime and Geometry (Cambridge: Cambridge University Press) [67] Weinberg S and de Campos T 1995 The Quantum Theory of Fields (Modern Applications vol 2) (Cambridge: Cambridge University Press) [68] Huguet E, Le Delliou M, Fontanini M and Lin Z C 2021 Teleparallel gravity as a gauge theory: coupling to matter using the Cartan connection Phys. Rev. D 103 044061 [69] Fontanini M, Huguet E and Le Delliou M 2019 Teleparallel gravity equivalent of general relativity as a gauge theory: translation or Cartan connection? Phys. Rev. D 99 064006 [70] Mosna R A and Pereira J 2004 Some remarks on the coupling prescription of teleparallel gravity Gen. Relativ. Gravit. 36 2525–38 [71] Weldon H A 2001 Fermions without vierbeins in curved space-time Phys. Rev. D 63 104010 [72] Vinckers U B, de la Cruz-Dombriz A and Maldonado Torralba F 2020 Focusing conditions for extended teleparallel gravity theories J. Cosmol. Astropart. Phys. JCAP12(2020)020 [73] Wald R M 1984 General Relativity (Chicago, IL: Chicago University Press) [74] Hohmann M, Järv L, Krššák M and Pfeifer C 2019 Modified teleparallel theories of gravity in symmetric spacetimes Phys. Rev. D 100 084002 [75] Pfeifer C 2022 A quick guide to spacetime symmetry and symmetric solutions in teleparallel gravity (arXiv: 2201.04691 [gr-qc]) [76] Hamermesh M 2012 Group Theory and Its Application to Physical Problems (Chicago, IL: University of Chicago Press) [77] Hohmann M 2021 Complete classification of cosmological teleparallel geometries Int. J. Geom. Methods Mod. Phys. 18 2140005 [78] Weyl H 1921 Raum, Zeit, Materie (Berlin: Springer) [79] Bekenstein J D 1993 The relation between physical and gravitational geometry Phys. Rev. D 48 3641–7 [80] Ezquiaga J M, García-Bellido J and Zumalacárregui M 2017 Field redefinitions in theories beyond Einstein gravity using the language of differential forms Phys. Rev. D 95 084039 [81] Iosifidis D and Koivisto T 2018 Scale transformations in metric-affine geometry Universe 5 82 [82] Golovnev A and Guzmán M J 2020 Disformal transformations in modified teleparallel gravity Symmetry 12 152 [83] Hohmann M 2019 Disformal transformations in scalar-torsion gravity Universe 5 167 [84] Nakahara M 2003 Geometry, Topology and Physics (Graduate Student Series in Physics) 2nd edn (London: Taylor and Francis) [85] Bahamonde S and Böhmer C G 2016 Modified teleparallel theories of gravity: Gauss–Bonnet and trace extensions Eur. Phys. J. C 76 578 [86] de la Cruz-Dombriz A, Farrugia G, Said J L and Saez-Gomez D 2017 Cosmological reconstructed solutions in extended teleparallel gravity theories with a 189 Rep. Prog. Phys. 86 (2023) 026901 Review teleparallel Gauss–Bonnet term Class. Quantum Grav. 34 235011 [87] Kofinas G and Saridakis E N 2014 Teleparallel equivalent of Gauss–Bonnet gravity and its modifications Phys. Rev. D 90 084044 [88] Gasperini M 2013 Theory of Gravitational Interactions (Undergraduate Lecture Notes in Physics) (Milano: Springer) [89] Kofinas G and Saridakis E N 2014 Cosmological applications of F(T, TG ) gravity Phys. Rev. D 90 084045 [90] de la Cruz-Dombriz A, Farrugia G, Said J L and Sáez-Chillón Gómez D 2018 Cosmological bouncing solutions in extended teleparallel gravity theories Phys. Rev. D 97 104040 [91] Zubair M and Jawad A 2015 Generalized second law of thermodynamics in f(T, TG ) gravity Astrophys. Space Sci. 360 11 [92] Bilby B A, Bullough R, Smith E and Whittaker J M 1955 Continuous distributions of dislocations: a new application of the methods of non-Riemannian geometry Proc. R. Soc. A 231 263–73 [93] Jizba P, Kleinert H and Scardigli F 2010 Uncertainty relation on world crystal and its applications to micro black holes Phys. Rev. D 81 084030 [94] Klinkhamer F R and Volovik G E 2019 Tetrads and q-theory JETP Lett. 109 364–7 [95] Kleinert H and Zaanen J 2004 World nematic crystal model of gravity explaining the absence of torsion Phys. Lett. A 324 361–5 [96] Nissinen J and Volovik G E 2018 Tetrads in solids: from elasticity theory to topological quantum Hall systems and Weyl fermions Sov. Phys.-JETP 127 948–57 [97] Nissinen J and Volovik G E 2019 Elasticity tetrads, mixed axial-gravitational anomalies and 3+1d quantum Hall effect Phys. Rev. Res. 1 023007 [98] Boöhmer C G and Obukhov Y N 2012 A gauge-theoretical approach to elasticity with microrotations Proc. R. Soc. A 468 1391–407 [99] Hull D and Bacon D 2011 Introduction to Dislocations (Materials Science and Technology) (Amsterdam: Elsevier) [100] Frank F C 1949 On the equations of motion of crystal dislocations Proc. R. Soc. A 62 131–4 [101] Kondo K 1952 Raag Memoirs of the Unifying Study of the Basic Problems in Physics and Engineering Science by Means of Geometry vol 1 (Tokio: Gakujutsu Bunken Fukyu-Kay) [102] Bilby B A, Bullough R and Smith E 1955 Continuous distributions of dislocations: a new application of the methods of non-Riemannian geometry Proc. R. Soc. A 231 263–73 [103] Unzicker A 2000 What can physics learn from continuum mechanics? (arXiv:gr-qc/0011064) [104] Zubkov M A 2019 Emergent gravity in superplastic crystals and cosmological constant problem (arXiv:1909.08412 [gr-qc]). [105] Lazar M 2000 Dislocation theory as a 3-dimensional translation gauge theory Ann. Phys., Lpz. 9 461–73 [106] Caselle M, D’Adda A and Magnea L 1989 Regge calculus as a local theory of the poincaré group Phys. Lett. B 232 457–61 [107] Schmidt J and Kohler C 2001 Torsion degrees of freedom in the Regge calculus as dislocations on the simplicial lattice Gen. Relativ. Gravit. 33 1799–808 [108] Pereira J and Vargas T 2002 Regge calculus in teleparallel gravity Class. Quantum Grav. 19 4807–16 [109] Zubkov M A 2004 Teleparallel gravity on the lattice Phys. Lett. B 582 243–8 [110] Cosserat E M P and Cosserat F 1909 Théóorie des Corps Défformables (Paris: A. Hermann et fils) [111] Burnett J, Chervova O and Vassiliev D 2009 Dirac equation as a special case of cosserat elasticity Analysis, Partial Differential Equations and Applications ed A Cialdea, P E Ricci and F Lanzara (Basel: Birkhäuser) pp 15–29 [112] Burnett J and Vassiliev D 2009 Weyl’s Lagrangian in teleparallel form J. Math. Phys. 50 102501 [113] Chervova O and Vassiliev D 2009 Massless Dirac equation as a special case of Cosserat elasticity Int. Conf. on Recent Trends in Mathematical Sciences (ICRMS 2008) (arXiv:0902.1268 [gr-qc]) [114] Chervova O and Vassiliev D 2010 The stationary Weyl equation and Cosserat elasticity J. Phys. A: Math. Theor. 43 335203 [115] Burnett J and Vassiliev D 2012 Modeling the electron with Cosserat elasticity Mathematika 58 349–70 [116] Böhmer C G, Downes R J and Vassiliev D 2011 Rotational elasticity Q. J. Mech. Appl. Math. 64 415–39 [117] Böhmer C G and Obukhov Y N 2012 A gauge-theoretical approach to elasticity with microrotations Proc. R. Soc. A 468 1391–407 [118] Böhmer C G and Lee Y 2021 Compatibility conditions of continua using Riemann–Cartan geometry Math. Mech. Solids 26 513–29 [119] Nissinen J 2020 Emergent spacetime and gravitational Nieh–Yan anomaly in chiral p + ip Weyl superfluids and superconductors Phys. Rev. Lett. 124 117002 [120] Nissinen J and Volovik G E 2020 Thermal Nieh–Yan anomaly in Weyl superfluids Phys. Rev. Res. 2 033269 [121] Laurila S and Nissinen J 2020 Torsional Landau levels and geometric anomalies in condensed matter Weyl systems Phys. Rev. B 102 235163 [122] Koivisto T 2018 An integrable geometrical foundation of gravity Int. J. Geom. Methods Mod. Phys. 15 1840006 [123] Golovnev A, Koivisto T and Sandstad M 2017 On the covariance of teleparallel gravity theories Class. Quantum Grav. 34 145013 [124] Hohmann M 2021 Variational principles in teleparallel gravity theories Universe 7 114 [125] Hohmann M, Järv L, Krššák M and Pfeifer C 2018 Teleparallel theories of gravity as analogue of nonlinear electrodynamics Phys. Rev. D 97 104042 [126] Obukhov Y N and Rubilar G F 2006 Invariant conserved currents in gravity theories with local Lorentz and diffeomorphism symmetry Phys. Rev. D 74 064002 [127] Hohmann M 2018 Scalar-torsion theories of gravity III: analogue of scalar-tensor gravity and conformal invariants Phys. Rev. D 98 064004 [128] Blixt D, Ferraro R, Golovnev A and Guzmán M-J 2022 Lorentz gauge-invariant variables in torsion-based theories of gravity (arXiv:2201.11102 [gr-qc]) [129] Beltrán Jiménez J, Heisenberg L and Koivisto T S 2018 Teleparallel Palatini theories J. Cosmol. Astropart. Phys. JCAP08(2018)039 [130] Itin Y, Hehl F W and Obukhov Y N 2017 Premetric equivalent of general relativity: teleparallelism Phys. Rev. D 95 084020 [131] Itin Y, Obukhov Y N, Boos J and Hehl F W 2018 Premetric teleparallel theory of gravity and its local and linear constitutive law Eur. Phys. J. C 78 907 [132] Hohmann M 2018 Scalar-torsion theories of gravity I: general formalism and conformal transformations Phys. Rev. D 98 064002 [133] Hehl F W and Obukhov Y 2001 Foundations of Classical Electrodynamics (Boston, MA: Birkhäuser) [134] Wise D K 2010 MacDowell–Mansouri gravity and Cartan geometry Class. Quantum Grav. 27 155010 190 Rep. Prog. Phys. 86 (2023) 026901 Review [135] Wise D K 2009 Symmetric space Cartan connections and gravity in three and four dimensions SIGMA 5 080 [136] Maluf J 1994 Hamiltonian formulation of the teleparallel description of general relativity J. Math. Phys. 35 335–43 [137] Maluf J W and da Rocha-Neto J F 1999 General relativity on a null surface: Hamiltonian formulation in the teleparallel geometry Gen. Relativ. Gravit. 31 173–85 [138] Maluf J and Sousa A 2000 Hamiltonian formulation of teleparallel theories of gravity in the time gauge (arXiv:gr-qc/0002060 [gr-qc]) [139] Maluf J W and da Rocha-Neto J F 2001 Hamiltonian formulation of general relativity in the teleparallel geometry Phys. Rev. D 64 084014 [140] da Rocha Neto J F, Maluf J W and Ulhoa S C 2010 Hamiltonian formulation of unimodular gravity in the teleparallel geometry Phys. Rev. D 82 124035 da Rocha Neto J F, Maluf J W and Ulhoa S C 2013 Phys. Rev. D 87 069909 (erratum) [141] Okołów A and Swiezewski J 2012 Hamiltonian formulation of a simple theory of the teleparallel geometry Class. Quantum Grav. 29 045008 [142] Okołów A 2013 ADM-like Hamiltonian formulation of gravity in the teleparallel geometry Gen. Relativ. Gravit. 45 2569–2610 [143] Okołów A 2014 ADM-like Hamiltonian formulation of gravity in the teleparallel geometry: derivation of constraint algebra Gen. Relativ. Gravit. 46 1636 [144] Ferraro R and Guzmán M J 2016 Hamiltonian formulation of teleparallel gravity Phys. Rev. D 94 104045 [145] Ferraro R and Guzmán M J 2018 Hamiltonian formalism for f (T) gravity Phys. Rev. D 97 104028 [146] Guzmán M-J and Ibraheem S K 2020 Hamiltonian formalism for new general relativity in the premetric approach (arXiv:2009.13430 [gr-qc]) [147] Blixt D, Hohmann M and Pfeifer C 2019 Hamiltonian and primary constraints of new general relativity Phys. Rev. D 99 084025 [148] Blixt D, Hohmann M and Pfeifer C 2019 On the gauge fixing in the Hamiltonian analysis of general teleparallel theories Universe 5 143 [149] Hohmann M 2020 Hamiltonian of new general relativity using differential forms Int. J. Mod. Phys. A 35 2040014 [150] Cheng W-H, Chern D-C and Nester J M 1988 Canonical analysis of the one parameter teleparallel theory Phys. Rev. D 38 2656–8 [151] Blagojević M and Nikolic I A 2000 Hamiltonian structure of the teleparallel formulation of GR Phys. Rev. D 62 024021 [152] Blagojević M 2001 Hamiltonian structure and gauge symmetries of Poincare gauge theory Ann. Phys., Lpz. 10 367–91 [153] Mitrić P 2019 Canonical structure of the teleparallel equivalent of general relativity (arXiv:1910.02810 [gr-qc]) [154] Blagojević M and Nester J M 2020 Local symmetries and physical degrees of freedom in f (T) gravity: a Dirac Hamiltonian constraint analysis Phys. Rev. D 102 064025 [155] Golovnev A and Guzman M-J 2021 Lorentz symmetries and primary constraints in covariant teleparallel gravity Phys. Rev. D 104 124074 [156] Unzicker A and Case T 2005 Translation of Einstein’s attempt of a unified field theory with teleparallelism (arXiv:physics/0503046 [physics.hist-ph]) [157] Bahamonde S, Böhmer C G and Wright M 2015 Modified teleparallel theories of gravity Phys. Rev. D 92 104042 [158] Bahamonde S, Gakis V, Kiorpelidi S, Koivisto T, Levi Said J and Saridakis E N 2021 Cosmological perturbations in modified teleparallel gravity models: boundary term extension Eur. Phys. J. C 81 53 [159] de Andrade V C and Pereira J G 1997 Gravitational Lorentz force and the description of the gravitational interaction Phys. Rev. D 56 4689–95 [160] Krššák M 2017 Holographic renormalization in teleparallel gravity Eur. Phys. J. C 77 44 [161] de Andrade V C, Guillen L C T and Pereira J G 2000 Gravitational energy momentum density in teleparallel gravity Phys. Rev. Lett. 84 4533–6 [162] de Andrade V and Pereira J 1999 Torsion and the electromagnetic field Int. J. Mod. Phys. D 8 141–51 [163] Krššák M and Pereira J G 2015 Spin connection and renormalization of teleparallel action Eur. Phys. J. C 75 519 [164] Shirafuji T, Nashed G G L and Kobayashi Y 1996 Equivalence principle in the new general relativity Prog. Theor. Phys. 96 933–48 [165] Combi L and Romero G E 2018 Is teleparallel gravity really equivalent to general relativity? Ann. Phys., Lpz. 530 1700175 [166] Beltrán Jiménez J and Koivisto T S 2022 Noether charges in the geometrical trinity of gravity Phys. Rev. D 105 L021502 [167] Ringström H 2009 The Cauchy Problem in General Relativity (ESI Lectures in Mathematics and Physics) (European Mathematical Society) [168] Motohashi H and Suyama T 2015 Third order equations of motion and the Ostrogradsky instability Phys. Rev. D 91 085009 [169] Wipf A W 1994 Hamilton’s formalism for systems with constraints Canonical Gravity: From Classical to Quantum (Lecture Notes in Physics vol 434) (Berlin: Springer) pp 22–58 [170] Maluf J W and da Rocha-Neto J F 2000 Hamiltonian formulation of the teleparallel equivalent of general relativity without gauge fixing (arXiv:gr-qc/0002059 [gr-qc]) [171] Calcagni G, Kuroyanagi S, Marsat S, Sakellariadou M, Tamanini N and Tasinato G 2019 Quantum gravity and gravitational-wave astronomy J. Cosmol. Astropart. Phys. JCAP10(2019)012 [172] Amelino-Camelia G, Ellis J R, Mavromatos N E, Nanopoulos D V and Sarkar S 1998 Tests of quantum gravity from observations of gamma-ray bursts Nature 393 763–5 [173] Formaggio J A, Kaiser D I, Murskyj M M and Weiss T E 2016 Violation of the Leggett–Garg inequality in neutrino oscillations Phys. Rev. Lett. 117 050402 [174] Will C M 2001 The confrontation between general relativity and experiment Living Rev. Relativ. 4 4 [175] Aldrovandi R, Pereira J and Vu K 2006 Gravity and the quantum: are they reconcilable? AIP Conf. Proc. 810 217–28 [176] Aldrovandi R, Guillen L, Pereira J and Vu K 2006 Bringing together gravity and the quanta AIP Conf. Proc. 861 277–84 [177] Ho P-M 2016 Generalized Yang–Mills theory and gravity Phys. Rev. D 93 044062 [178] Oshita N and Wu Y-P 2017 Role of spacetime boundaries in Einstein’s other gravity Phys. Rev. D 96 044042 [179] Aldrovandi R, Pereira J and Vu K 2004 Gravitation: global formulation and quantum effects Class. Quantum Grav. 21 51–62 [180] Pereira J G and López D F 2020 An improved framework for quantum gravity Universe 6 243 [181] Jennen H and Pereira J 2016 Dark energy as a kinematic effect Phys. Dark Universe 11 49–53 [182] Koivisto T S and Zheng L 2021 Scale-invariant cosmology in de Sitter gauge theory Phys. Rev. D 103 124063 191 Rep. Prog. Phys. 86 (2023) 026901 Review [183] Dupuis M, Girelli F, Osumanu A and Wieland W 2020 First-order formulation of teleparallel gravity and dual loop gravity Class. Quantum Grav. 37 085023 [184] Okołów A 2013 Construction of spaces of kinematic quantum states for field theories via projective techniques Class. Quantum Grav. 30 195003 [185] Okoł ów A 2014 Variables suitable for constructing quantum states for the teleparallel equivalent of general relativity I Gen. Relativ. Gravit. 46 1620 [186] Okołów A 2014 Variables suitable for constructing quantum states for the teleparallel equivalent of general relativity II Gen. Relativ. Gravit. 46 1638 [187] Okołów A 2014 Kinematic quantum states for the teleparallel equivalent of general relativity Gen. Relativ. Gravit. 46 1653 [188] Ulhoa S and Amorim R 2014 On teleparallel quantum gravity in Schwarzschild space-time Adv. High Energy Phys. 2014 812691 [189] Ulhoa S C, Spaniol E P and Amorim R G G 2019 On quantization of a slowly rotating Kerr black hole in teleparallel gravity Universe 5 29 [190] Ulhoa S 2011 On special re-quantization of a black hole Braz. J. Phys. 41 309–13 [191] Behboodi A, Akhshabi S and Nozari K 2013 Braneworld teleparallel gravity Phys. Lett. B 723 201–6 [192] Behboodi A, Akhshabi S and Nozari K 2015 Braneworld setup and embedding in teleparallel gravity Phys. Lett. B 740 291–7 [193] Aschieri P, Blohmann C, Dimitrijevic M, Meyer F, Schupp P and Wess J 2005 A Gravity theory on noncommutative spaces Class. Quantum Grav. 22 3511–32 [194] Deruelle N and Farina-Busto L 1990 The Lovelock gravitational field equations in cosmology Phys. Rev. D 41 3696 [195] Harko T, Lobo F S N, Nojiri S and Odintsov S D 2011 f(R, T) gravity Phys. Rev. D 84 024020 [196] Nojiri S and Odintsov S D 2005 Modified Gauss–Bonnet theory as gravitational alternative for dark energy Phys. Lett. B 631 1–6 [197] Deser S and Woodard R P 2007 Nonlocal cosmology Phys. Rev. Lett. 99 111301 [198] Mannheim P D and Kazanas D 1989 Exact vacuum solution to conformal weyl gravity and galactic rotation curves Astrophys. J. 342 635–8 [199] Jacobson T 2007 Einstein–Aether gravity: a status report Proc. Sci. 043 020 [200] De Felice A, Heisenberg L, Kase R, Mukohyama S, Tsujikawa S and Zhang Y-l 2016 Cosmology in generalized Proca theories J. Cosmol. Astropart. Phys. JCAP06(2016)048 [201] Hassan S F and Rosen R A 2012 Bimetric gravity from ghost-free massive gravity J. High Energy Phys. JHEP02(2012)126 [202] Kobayashi T 2019 Horndeski theory and beyond: a review Rep. Prog. Phys. 82 086901 [203] Gleyzes J, Langlois D, Piazza F and Vernizzi F 2015 Healthy theories beyond Horndeski Phys. Rev. Lett. 114 211101 [204] de Rham C, Gabadadze G and Tolley A J 2011 Resummation of massive gravity Phys. Rev. Lett. 106 231101 [205] Sahni V and Shtanov Y 2003 Brane world models of dark energy J. Cosmol. Astropart. Phys. JCAP11(2003)014 [206] Randall L and Sundrum R 1999 A large mass hierarchy from a small extra dimension Phys. Rev. Lett. 83 3370–3 [207] Dvali G R and Gabadadze G 2001 Gravity on a brane in infinite volume extra space Phys. Rev. D 63 065007 [208] Klein O 1926 Quantum theory and five-dimensional theory of relativity Z. Phys. 37 895–906 (In German and English) [209] Kanti P, Mavromatos N E, Rizos J, Tamvakis K and Winstanley E 1996 Dilatonic black holes in higher curvature string gravity Phys. Rev. D 54 5049–58 [210] Maldacena J M 1998 The large N limit of superconformal field theories and supergravity Adv. Theor. Math. Phys. 2 231–52 [211] Mukohyama S 2010 Horava–Lifshitz cosmology: a review Class. Quantum Grav. 27 223101 [212] Polchinski J 1998 String Theory (Cambridge Monographs on Mathematical Physics vol 1) (Cambridge: Cambridge University Press) [213] Amelino-Camelia G, Arzano M, Gubitosi G and Magueijo J 2013 Rainbow gravity and scale-invariant fluctuations Phys. Rev. D 88 041303 [214] Sahlmann H 2010 Loop quantum gravity—a short review Foundations of Space and Time: Reflections on Quantum Gravity (arXiv:1001.4188 [gr-qc]) [215] Niedermaier M and Reuter M 2006 The asymptotic safety scenario in quantum gravity Living Rev. Relativ. 9 5–173 [216] Yamaguchi M 2011 Supergravity based inflation models: a review Class. Quantum Grav. 28 103001 [217] Padmanabhan T 2010 Thermodynamical aspects of gravity: new insights Rep. Prog. Phys. 73 046901 [218] Witten E 1998 Anti-de Sitter space and holography Adv. Theor. Math. Phys. 2 253–91 [219] Verlinde E P 2011 On the origin of gravity and the laws of Newton J. High Energy Phys. JHEP04(2011)029 [220] Barcelo C, Liberati S and Visser M 2005 Analogue gravity Living Rev. Relativ. 8 12 [221] Weinberg S 1972 Gravitation and Cosmology: Principles and Applications of the General Theory of Relativity (New York: Wiley) [222] Cho Y M 1976 Einstein Lagrangian as the translational Yang–Mills Lagrangian Phys. Rev. D 14 2521 [223] Tamanini N and Böhmer C G 2012 Good and bad tetrads in f (T) gravity Phys. Rev. D 86 044009 [224] Casadio R, Kuntz I and Paci G 2022 Quantum fields in teleparallel gravity: renormalization at one-loop Eur. Phys. J. C 82 186 [225] Ualikhanova U and Hohmann M 2019 Parametrized post-Newtonian limit of general teleparallel gravity theories Phys. Rev. D 100 104011 [226] Van Nieuwenhuizen P 1973 On ghost-free tensor Lagrangians and linearized gravitation Nucl. Phys. B 60 478–92 [227] Kuhfuss R and Nitsch J 1986 Propagating modes in gauge field theories of gravity Gen. Relativ. Gravit. 18 1207 [228] Beltrán Jiménez J and Dialektopoulos K F 2020 Non-linear obstructions for consistent new general relativity J. Cosmol. Astropart. Phys. JCAP01(2020)18 [229] Blixt D, Guzmán M-J, Hohmann M and Pfeifer C 2020 Review of the Hamiltonian analysis in teleparallel gravity (arXiv:2012.09180 [gr-qc]) [230] Kopczynski W 1982 Problems with metric-teleparallel theories of gravitation J. Phys. A: Math. Gen. 15 493 [231] Jiménez J B, Golovnev A, Koivisto T and Veermäe H 2021 Minkowski space in f (T) gravity Phys. Rev. D 103 024054 [232] Golovnev A and Guzmán M-J 2021 Nontrivial Minkowski backgrounds in f (T) gravity Phys. Rev. D 103 044009 [233] Møller C 1978 On the crisis in the theory of gravitation and a possible solution Kong. Dan. Vid. Sel. Mat. Fys. Med. 39 13 [234] Mueller-Hoissen F and Nitsch J 1983 Teleparallelism—a viable theory of gravity? Phys. Rev. D 28 718–28 [235] Nester J 1988 Is there really a problem with the teleparallel theory? Class. Quantum Grav. 5 1003 [236] Hayashi K and Shirafuji T 1981 Addendum to “new general relativity” Phys. Rev. D 24 3312–14 192 Rep. Prog. Phys. 86 (2023) 026901 Review [237] Blixt D, Hohmann M, Krššák M and Pfeifer C 2019 Hamiltonian analysis in new general relativity 15th Marcel Grossmann Meeting on Recent Developments in Theoretical and Experimental General Relativity, Astrophysics and Relativistic Field Theories (arXiv:1905.11919 [gr-qc]) [238] Chen H-H, Chern D-C and Nester J M 1987 Positive energy test of teleparallel theory Chin. J. Phys. 25 481 [239] Järv L, Rünkla M, Saal M and Vilson O 2018 Nonmetricity formulation of general relativity and its scalar-tensor extension Phys. Rev. D 97 124025 [240] Beltrán Jiménez J, Heisenberg L, Koivisto T S and Pekar S 2020 Cosmology in f (Q) geometry Phys. Rev. D 101 103507 [241] Krššák M 2017 Variational problem and bigravity nature of modified teleparallel theories (arXiv:1705.01072 [gr-qc]) [242] Izumi K and Ong Y C 2013 Cosmological perturbation in f (T) gravity revisited J. Cosmol. Astropart. Phys. JCAP06(2013)029 [243] Golovnev A and Koivisto T 2018 Cosmological perturbations in modified teleparallel gravity models J. Cosmol. Astropart. Phys. JCAP11(2018)012 [244] Bahamonde S, Dialektopoulos K F, Hohmann M, Said J L, Pfeifer C and Saridakis E N 2022 Perturbations in non-flat cosmology for f (T) gravity (arXiv:2203.00619 [gr-qc]) [245] Golovnev A and Guzmán M-J 2020 Foundational issues in f (T) gravity theory Int. J. Geom. Methods Mod. Phys. 0 2140007 [246] Li M, Miao R-X and Miao Y-G 2011 Degrees of freedom of f (T) gravity J. High Energy Phys. JHEP07(2011)108 [247] Ferraro R and Guzmán M J 2018 Quest for the extra degree of freedom in f (T) gravity Phys. Rev. D 98 124037 [248] Guzmán M J and Ferraro R 2019 Degrees of freedom and local Lorentz invariance in f (T) gravity 15th Marcel Grossmann Meeting on Recent Developments in Theoretical and Experimental General Relativity, Astrophysics and Relativistic Field Theories vol 3 (arXiv:1903.06774 [gr-qc]) [249] Ferraro R and Guzmán M J 2020 Pseudo-invariance and the extra degree of freedom in f (T) gravity Phys. Rev. D 101 084017 [250] Nassur S B, Ainamon C, Houndjo M J S and Tossa J 2016 Unimodular f (T) gravity Eur. Phys. J. Plus 131 420 [251] Bamba K, Odintsov S D and Saridakis E N 2017 Inflationary cosmology in unimodular F(T) gravity Mod. Phys. Lett. A 32 1750114 [252] Godonou A E, Houndjo M J S and Tossa J 2018 Holographic dark energy model in unimodular f (T) gravity Int. J. Geom. Methods Mod. Phys. 16 1950003 [253] Yang R-J 2011 Conformal transformation in f (T) theories Europhys. Lett. 93 60001 [254] Bamba K, Odintsov S D and Sáez-Gómez D 2013 Conformal symmetry and accelerating cosmology in teleparallel gravity Phys. Rev. D 88 084042 [255] Obukhov Y N 1982 Conformal invariance and space-time torsion Phys. Lett. A 90 13–16 [256] Wright M 2016 Conformal transformations in modified teleparallel theories of gravity revisited Phys. Rev. D 93 103002 [257] Paliathanasis A 2017 de Sitter and Scaling solutions in a higher-order modified teleparallel theory J. Cosmol. Astropart. Phys. JCAP08(2017)027 [258] Sahlu S, Ntahompagaze J, Abebe A and Mota D F 2019 Accelerating universe in modified teleparallel gravity theory IAU Symp. 356 397–9 [259] Bhattacharjee S 2021 Constraining f(T, B) teleparallel gravity from energy conditions New Astron. 83 101495 [260] Zubair M, Waheed S, Atif Fayyaz M and Ahmad I 2018 Energy constraints and the phenomenon of cosmic evolution in the f(T, B) framework Eur. Phys. J. Plus 133 452 [261] Bahamonde S, Zubair M and Abbas G 2018 Thermodynamics and cosmological reconstruction in f(T, B) gravity Phys. Dark Universe 19 78–90 [262] Pourbagher A and Amani A 2019 Thermodynamics and stability of f(T, B) gravity with viscous fluid by observational constraints Astrophys. Space Sci. 364 140 [263] Escamilla-Rivera C and Levi Said J 2020 Cosmological viable models in f(T, B) theory as solutions to the H 0 tension Class. Quantum Grav. 37 165002 [264] Farrugia G, Levi Said J and Finch A 2020 Gravitoelectromagnetism, solar system test and weak-field solutions in f(T, B) gravity with observational constraints Universe 6 34 [265] Capozziello S, Capriolo M and Caso L 2020 Gravitational waves in higher order teleparallel gravity Class. Quantum Grav. 37 235013 [266] Franco G A R, Escamilla-Rivera C and Levi Said J 2020 Stability analysis for cosmological models in f(T, B) gravity Eur. Phys. J. C 80 677 [267] Caruana M, Farrugia G and Levi Said J 2020 Cosmological bouncing solutions in f(T, B) gravity Eur. Phys. J. C 80 640 [268] Bahamonde S and Capozziello S 2017 Noether symmetry approach in f(T, B) teleparallel cosmology Eur. Phys. J. C 77 107 [269] Paliathanasis A 2021 Minisuperspace quantization of f(T, B) cosmology Universe 7 150 [270] Myrzakulov R 2012 FRW cosmology in F(R, T) gravity Eur. Phys. J. C 72 2203 [271] Salti M, Korunur M, Acikgoz I, Pirinccioglu N and Binbay F 2018 f(T, R) theory of gravity Int. J. Mod. Phys. D 27 1850062 [272] Ferraro R and Fiorini F 2011 Non trivial frames for f (T) theories of gravity and beyond Phys. Lett. B 702 75–80 [273] Bengochea G R and Ferraro R 2009 Dark torsion as the cosmic speed-up Phys. Rev. D 79 124019 [274] Ferraro R and Fiorini F 2008 On Born–Infeld gravity in Weitzenböck spacetime Phys. Rev. D 78 124019 [275] Koivisto T and Tsimperis G 2019 The spectrum of teleparallel gravity Universe 5 80 [276] Conroy A and Koivisto T 2018 The spectrum of symmetric teleparallel gravity Eur. Phys. J. C 78 923 [277] Heisenberg L 2019 A systematic approach to generalisations of general relativity and their cosmological implications Phys. Rep. 796 1–113 [278] Biswas T, Gerwick E, Koivisto T and Mazumdar A 2012 Towards singularity and ghost free theories of gravity Phys. Rev. Lett. 108 031101 [279] Geng C-Q and Luo L-W 2017 Teleparallel conformal invariant models induced by Kaluza–Klein reduction Class. Quantum Grav. 34 115012 [280] Formiga J 2014 Equivalence between an extension of teleparallelism to a Weyl geometry and general relativity Int. J. Theor. Phys. 53 1971–7 [281] Capozziello S, De Laurentis M and Dialektopoulos K F 2016 Noether symmetries in Gauss–Bonnet-teleparallel cosmology Eur. Phys. J. C 76 629 [282] Gonzalez P and Vásquez Y 2015 Teleparallel equivalent of Lovelock gravity Phys. Rev. D 92 124023 [283] González P, Reyes S and Vásquez Y 2019 Teleparallel equivalent of Lovelock gravity, generalizations and cosmological applications J. Cosmol. Astropart. Phys. JCAP07(2019)040 [284] Bahamonde S, Marciu M, Odintsov S D and Rudra P 2021 String-inspired Teleparallel cosmology Nucl. Phys. B 962 115238 193 Rep. Prog. Phys. 86 (2023) 026901 Review [285] Addazi A et al 2021 Quantum gravity phenomenology at the dawn of the multi-messenger era—a review (arXiv:2111.05659 [hep-ph]) [286] Otalora G and Saridakis E N 2016 Modified teleparallel gravity with higher-derivative torsion terms Phys. Rev. D 94 084021 [287] Brans C and Dicke R H 1961 Mach’s principle and a relativistic theory of gravitation Phys. Rev. 124 925–35 [288] Geng C-Q, Lee C-C, Saridakis E N and Wu Y-P 2011 “Teleparallel” dark energy Phys. Lett. B 704 384–7 [289] Capozziello S, Cardone V F, Carloni S and Troisi A 2003 Curvature quintessence matched with observational data Int. J. Mod. Phys. D 12 1969–82 [290] Copeland E J, Nunes N J and Pospelov M 2004 Models of quintessence coupled to the electromagnetic field and the cosmological evolution of alpha Phys. Rev. D 69 023501 [291] Capozziello S 2002 Curvature quintessence Int. J. Mod. Phys. D 11 483–92 [292] Paliathanasis A 2021 O (d, d) symmetry in teleparallel dark energy Eur. Phys. J. Plus 136 674 [293] Paliathanasis A 2021 Dynamics in interacting scalar-torsion cosmology Universe 7 244 [294] Otalora G 2015 A novel teleparallel dark energy model Int. J. Mod. Phys. D 25 1650025 [295] Bahamonde S and Wright M 2015 Teleparallel quintessence with a nonminimal coupling to a boundary term Phys. Rev. D 92 084034 Bahamonde S and Wright M 2016 Phys. Rev. D 93 109901 (erratum) [296] Zubair M, Bahamonde S and Jamil M 2017 Generalized second law of thermodynamic in modified teleparallel theory Eur. Phys. J. C 77 472 [297] Bahamonde S, Marciu M and Said J L 2019 Generalized tachyonic teleparallel cosmology Eur. Phys. J. C 79 324 [298] Banijamali A and Fazlpour B 2012 Tachyonic teleparallel dark energy Astrophys. Space Sci. 342 229–35 [299] Fazlpour B and Banijamali A 2015 Non-minimally coupled tachyon field in teleparallel gravity J. Cosmol. Astropart. Phys. JCAP04(2015)030 [300] Hohmann M, Järv L and Ualikhanova U 2018 Covariant formulation of scalar-torsion gravity Phys. Rev. D 97 104011 [301] Hohmann M and Pfeifer C 2018 Scalar-torsion theories of gravity II: L(T, X, Y, ϕ) theory Phys. Rev. D 98 064003 [302] Bahamonde S, Marciu M and Rudra P 2018 Generalised teleparallel quintom dark energy non-minimally coupled with the scalar torsion and a boundary term J. Cosmol. Astropart. Phys. JCAP04(2018)056 [303] Abedi H, Capozziello S, D’Agostino R and Luongo O 2018 Effective gravitational coupling in modified teleparallel theories Phys. Rev. D 97 084008 [304] Deffayet C, Esposito-Farese G and Vikman A 2009 Covariant Galileon Phys. Rev. D 79 084003 [305] Capozziello S, Dialektopoulos K F and Sushkov S V 2018 Classification of the Horndeski cosmologies via Noether symmetries Eur. Phys. J. C 78 447 [306] Bhattacharya S, Dialektopoulos K F and Tomaras T N 2016 Large scale structures and the cubic galileon model J. Cosmol. Astropart. Phys. JCAP05(2016)036 [307] Bahamonde S, Dialektopoulos K F and Levi Said J 2019 Can Horndeski theory be recast using teleparallel gravity? Phys. Rev. D 100 064018 [308] Deffayet C, Gao X, Steer D and Zahariade G 2011 From k-essence to generalised Galileons Phys. Rev. D 84 064039 [309] Bahamonde S, Dialektopoulos K F, Gakis V and Levi Said J 2020 Reviving Horndeski theory using teleparallel gravity after GW170817 Phys. Rev. D 101 084060 [310] Bernardo R C, Said J L, Caruana M and Appleby S 2021 Well-tempered teleparallel Horndeski cosmology: a teleparallel variation to the cosmological constant problem J. Cosmol. Astropart. Phys. JCAP10(2021)078 [311] Bernardo R C, Said J L, Caruana M and Appleby S 2022 Well-tempered Minkowski solutions in teleparallel Horndeski theory Class. Quantum Grav. 39 015013 [312] Dialektopoulos K F, Said J L and Oikonomopoulou Z 2022 Classification of teleparallel Horndeski cosmology via Noether symmetries Eur. Phys. J. C 82 259 [313] Nicosia G-P, Levi Said J and Gakis V 2021 Generalised Proca theories in teleparallel gravity Eur. Phys. J. Plus 136 191 [314] Deffayet C, Pujolas O, Sawicki I and Vikman A 2010 Imperfect dark energy from kinetic gravity braiding J. Cosmol. Astropart. Phys. JCAP10(2010)026 [315] Perrotta F, Baccigalupi C and Matarrese S 1999 Extended quintessence Phys. Rev. D 61 023507 [316] Copeland E J, Sami M and Tsujikawa S 2006 Dynamics of dark energy Int. J. Mod. Phys. D 15 1753–936 [317] Horndeski G W 1974 Second-order scalar-tensor field equations in a four-dimensional space Int. J. Theor. Phys. 10 363–84 [318] Kobayashi T, Yamaguchi M and Yokoyama J 2011 Generalized G-inflation: inflation with the most general second-order field equations Prog. Theor. Phys. 126 511–29 [319] Gonzalez-Espinoza M, Otalora G, Videla N and Saavedra J 2019 Slow-roll inflation in generalized scalar-torsion gravity J. Cosmol. Astropart. Phys. JCAP08(2019)029 [320] Maluf J W and Faria F F 2012 Conformally invariant teleparallel theories of gravity Phys. Rev. D 85 027502 [321] Abedi H and Salti M 2015 Multiple field modified gravity and localized energy in teleparallel framework Gen. Relativ. Gravit. 47 93 [322] Kiani F and Nozari K 2014 Energy conditions in F(T, Θ) gravity and compatibility with a stable de Sitter solution Phys. Lett. B 728 554–61 [323] Harko T, Lobo F S N, Otalora G and Saridakis E N 2014 f(T, T ) gravity and cosmology J. Cosmol. Astropart. Phys. JCAP12(2014)021 [324] Saez-Gomez D, Carvalho C S, Lobo F S N and Tereno I 2016 Constraining f(T, T ) gravity models using type Ia supernovae Phys. Rev. D 94 024034 [325] Chen C-Y and Kung Y-H 2022 Modified teleparallel gravity induced by quantum fluctuations Phys. Dark Universe 35 100956 [326] Bernardo R C, Chen C-Y, Said J L and Kung Y-H 2021 Confronting quantum-corrected teleparallel cosmology with observations (arXiv:2111.11761 [gr-qc]) [327] Harko T, Lobo F S N, Otalora G and Saridakis E N 2014 Nonminimal torsion-matter coupling extension of f (T) gravity Phys. Rev. D 89 124036 [328] Bahamonde S 2018 Generalised nonminimally gravity-matter coupled theory Eur. Phys. J. C 78 326 [329] Feng C-J, Ge F-F, Li X-Z, Lin R-H and Zhai X-H 2015 Towards realistic f (T) models with nonminimal torsion-matter coupling extension Phys. Rev. D 92 104038 [330] Bertolami O, Böhmer C G, Harko T and Lobo F S N 2007 Extra force in f (R) modified theories of gravity Phys. Rev. D 75 104016 [331] Harko T and Lobo F S 2010 f(R, Lm ) gravity Eur. Phys. J. C 70 373–9 [332] Harko T, Lobo F S N and Saridakis E N 2021 Gravitationally induced particle production through a nonminimal torsion–matter coupling Universe 7 227 [333] Amadei L and Perez A 2019 Hawking’s information puzzle: a solution realized in loop quantum cosmology (arXiv:1911.00306 [gr-qc]) 194 Rep. Prog. Phys. 86 (2023) 026901 Review [334] Stoica O C 2018 Revisiting the black hole entropy and the information paradox Adv. High Energy Phys. 2018 4130417 [335] Bahamonde S, Capozziello S and Dialektopoulos K F 2017 Constraining generalized non-local cosmology from Noether symmetries Eur. Phys. J. C 77 722 [336] de la Cruz-Dombriz A, Maldonado Torralba F J and Mazumdar A 2019 Nonsingular and ghost-free infinite derivative gravity with torsion Phys. Rev. D 99 104021 [337] de la Cruz-Dombriz A, Maldonado Torralba F J and Mazumdar A 2021 Ghost-free higher-order theories of gravity with torsion Eur. Phys. J. C 81 240 [338] Bahamonde S, Capozziello S, Faizal M and Nunes R C 2017 Nonlocal teleparallel cosmology Eur. Phys. J. C 77 628 [339] Nojiri S and Odintsov S D 2008 Modified non-local-F(R) gravity as the key for the inflation and dark energy Phys. Lett. B 659 821–6 [340] De Felice A and Sasaki M 2015 Ghosts in classes of non-local gravity Phys. Lett. B 743 189–97 [341] Nojiri S, Odintsov S D and Oikonomou V K 2020 Ghost-free non-local F(R) Gravity Cosmology Phys. Dark Universe 28 100541 [342] de Andrade V, Guillen L and Pereira J 2000 Teleparallel equivalent of Kaluza–Klein Phys. Rev. D 61 084031 [343] Barbosa A, Guillen L and Pereira J 2002 Teleparallel equivalent of nonAbelian Kaluza–Klein theory Phys. Rev. D 66 064028 [344] Geng C-Q, Lai C, Luo L-W and Tseng H-H 2014 Kaluza–Klein theory for teleparallel gravity Phys. Lett. B 737 248–50 [345] Sebastiani L, Vagnozzi S and Myrzakulov R 2017 Mimetic gravity: a review of recent developments and applications to cosmology and astrophysics Adv. High Energy Phys. 2017 3156915 [346] Chamseddine A H and Mukhanov V 2013 Mimetic dark matter J. High Energy Phys. JHEP11(2013)135 [347] Mirza B and Oboudiat F 2019 Mimetic f (T) teleparallel gravity and cosmology Gen. Relativ. Gravit. 51 96 [348] Hohmann M and Pfeifer C 2020 Teleparallel axions and cosmology Eur. Phys. J. C 81 376 [349] Li M, Rao H and Zhao D 2020 A simple parity violating gravity model without ghost instability J. Cosmol. Astropart. Phys. JCAP11(2020)023 [350] Nieh H T and Yan M L 1982 An identity in Riemann–Cartan geometry J. Math. Phys. 23 373 [351] Li M, Rao H and Tong Y 2021 Revisiting a parity violating gravity model without ghost instability: local Lorentz covariance Phys. Rev. D 104 084077 [352] Li M, Li Z and Rao H 2022 Ghost instability in the teleparallel gravity model with parity violations (arXiv:2201.02357 [gr-qc]) [353] Bahamonde S and Pfeifer C 2020 General teleparallel modifications of Schwarzschild geometry (arXiv:2010.02161 [gr-qc]) [354] Bahamonde S, Levi Said J and Zubair M 2020 Solar system tests in modified teleparallel gravity J. Cosmol. Astropart. Phys. JCAP10(2020)024 [355] Bahamonde S and Camci U 2019 Exact spherically symmetric solutions in modified teleparallel gravity Symmetry 11 1462 [356] Bahamonde S, Flathmann K and Pfeifer C 2019 Photon sphere and perihelion shift in weak f (T) gravity Phys. Rev. D 100 084064 [357] Ruggiero M L and Radicella N 2015 Weak-field spherically symmetric solutions in f (T) gravity Phys. Rev. D 91 104014 [358] Finch A and Levi Said J 2018 Galactic rotation dynamics in f (T) gravity Eur. Phys. J. C 78 560 [359] Farrugia G, Said J L and Ruggiero M L 2016 Solar system tests in f (T) gravity Phys. Rev. D 93 104034 [360] Iorio L, Radicella N and Ruggiero M L 2015 Constraining f (T) gravity in the solar system J. Cosmol. Astropart. Phys. JCAP1508(2015)021 [361] Gonzalez P A, Saridakis E N and Vásquez Y 2012 Circularly symmetric solutions in three-dimensional teleparallel, f (T) and Maxwell-f (T) gravity J. High Energy Phys. JHEP07(2012)053 [362] Bejarano C, Ferraro R and Guzmán M J 2015 Kerr geometry in f (T) gravity Eur. Phys. J. C 75 77 [363] Järv L, Hohmann M, Krššák M and Pfeifer C 2019 Flat connection for rotating spacetimes in extended teleparallel gravity theories Universe 5 142 [364] Bahamonde S, Valcarcel J G, Järv L and Pfeifer C 2021 Exploring axial symmetry in modified teleparallel gravity Phys. Rev. D 103 044058 [365] Houndjo M J S, Momeni D and Myrzakulov R 2012 Cylindrical solutions in modified f (T) gravity Int. J. Mod. Phys. D 21 1250093 [366] Stephani H, Kramer D, MacCallum M A, Hoenselaers C and Herlt E 2003 Exact Solutions of Einstein’s Field Equations (Cambridge Monographs on Mathematical Physics) (Cambridge: Cambridge University Press) [367] Nurbaki A N, Capozziello S and Deliduman C 2020 Spherical and cylindrical solutions in f (T) gravity by Noether Symmetry Approach Eur. Phys. J. C 80 108 [368] Jawad A, Momeni D, Rani S and Myrzakulov R 2016 Dynamical instability of cylindrical symmetric collapsing star in generalized teleparallel gravity Astrophys. Space Sci. 361 141 [369] Jawad A and Azam M 2018 f (T) corrected instability of cylindrical collapsing object with Harrison-Wheeler equation of state Adv. High Energy Phys. 2018 7265785 [370] Rodrigues M E, Houndjo M, Momeni D and Myrzakulov R 2013 Planar symmetry in f (T) gravity Int. J. Mod. Phys. D 22 1350043 [371] Dodelson S 2003 Modern Cosmology (New York: Academic) [372] Aghanim N et al (Planck Collaboration) 2020 Planck 2018 results. VI. Cosmological parameters Astron. Astrophys. 641 A6 [373] Peacock J 1999 Cosmological Physics (Cambridge Astrophysics) (Cambridge: Cambridge University Press) [374] Melchiorri A et al 2000 A measurement of omega from the North American test flight of BOOMERANG Astrophys. J. Lett. 536 L63–L66 [375] Hinshaw G et al 2013 Nine-year Wilkinson microwave anisotropy probe (WMAP) observations: cosmological parameter results Astrophys. J. Suppl. 208 19 [376] Aiola S et al (ACT Collaboration) 2020 The Atacama cosmology telescope: DR4 maps and cosmological parameters J. Cosmol. Astropart. Phys. JCAP12(2020)047 [377] Balkenhol L et al 2021 Constraints on ΛCDM extensions from the SPT-3G 2018 EE and TE power spectra (arXiv:2103.13618 [astro-ph.CO]) [378] Di Valentino E, Melchiorri A and Silk J 2019 Planck evidence for a closed Universe and a possible crisis for cosmology Nat. Astron. 4 196–203 [379] Handley W 2021 Curvature tension: evidence for a closed universe Phys. Rev. D 103 L041301 [380] Di Valentino E, Melchiorri A and Silk J 2021 Investigating cosmic discordance Astrophys. J. Lett. 908 L9 [381] Dent J B, Dutta S and Saridakis E N 2011 f (T) gravity mimicking dynamical dark energy. Background and perturbation analysis J. Cosmol. Astropart. Phys. JCAP01(2011)009 [382] Hamani Daouda M, Rodrigues M E and Houndjo M J S 2012 Reconstruction of f (T) gravity according to holographic dark energy Eur. Phys. J. C 72 1893 195 Rep. Prog. Phys. 86 (2023) 026901 Review [383] Farooq M U, Jamil M, Momeni D and Myrzakulov R 2013 Reconstruction of f (T) and f (R) gravity according to (m, n)-type holographic dark energy Can. J. Phys. 91 703–8 [384] Debnath U 2015 Reconstructing f(R), f(G), f(T), and Einstein–Aether gravities from entropy-corrected (m, n) type pilgrim dark energy Astrophys. Space Sci. 355 405–11 [385] Huang P, Huang Y-C and Yuan F-F 2013 f (T) gravity from holographic Ricci dark energy model with new boundary conditions Mod. Phys. Lett. A 28 1350171 [386] Chattopadhyay S and Karmakar S 2019 Reconstruction of f (T) gravity in the context of standard Chaplygin gas as tachyon scalar field and study of the stability against gravitational perturbation Int. J. Geom. Methods Mod. Phys. 16 1950101 [387] Jawad A, Rani S and Saleem M 2017 Cosmological study of reconstructed f (T) models Astrophys. Space Sci. 362 63 [388] Waheed S 2020 Reconstruction paradigm in a class of extended teleparallel theories using Tsallis holographic dark energy Eur. Phys. J. Plus 135 11 [389] Said J L 2017 Reconstruction from scalar–tensor theory and the inhomogeneous equation of state in f (T) gravity Eur. Phys. J. C 77 883 [390] Nojiri S, Odintsov S D, Oikonomou V K and Paul T 2020 Unifying holographic inflation with holographic dark energy: a covariant approach Phys. Rev. D 102 023540 [391] Myrzakulov R, Saez-Gomez D and Tsyba P 2014 Cosmological solutions in F(T) gravity with the presence of spinor fields Int. J. Geom. Methods Mod. Phys. 12 1550023 [392] Salako I G, Rodrigues M E, Kpadonou A V, Houndjo M J S and Tossa J 2013 Λ CDM model in f (T) gravity: reconstruction, thermodynamics and stability J. Cosmol. Astropart. Phys. JCAP11(2013)060 [393] El Hanafy W and Nashed G G L 2019 Phenomenological reconstruction of f (T) teleparallel gravity Phys. Rev. D 100 083535 [394] Myrzakulov R 2011 Accelerating universe from F(T) gravity Eur. Phys. J. C 71 1752 [395] Buchholz H, Lichtblau H and Wetzel K 2013 The Confluent Hypergeometric Function: With Special Emphasis on Its Applications (Springer Tracts in Natural Philosophy) (Berlin: Springer) [396] Chattopadhyay S and Pasqua A 2013 Reconstruction of f (T) gravity from the Holographic dark energy Astrophys. Space Sci. 344 269–74 [397] Chattopadhyay S 2014 QCD ghost reconstruction of f (T) gravity in flat FRW universe Eur. Phys. J. Plus 129 82 [398] Al Mamon A and Das S 2016 A divergence free parametrization of deceleration parameter for scalar field dark energy Int. J. Mod. Phys. D 25 1650032 [399] Mamon A A and Das S 2017 A parametric reconstruction of the deceleration parameter Eur. Phys. J. C 77 495 [400] Mukherjee A 2016 Acceleration of the universe: a reconstruction of the effective equation of state Mon. Not. R. Astron. Soc. 460 273–82 [401] Chakrabarti S, Said J L and Bamba K 2019 On reconstruction of extended teleparallel gravity from the cosmological jerk parameter Eur. Phys. J. C 79 454 [402] Suzuki N et al 2012 The Hubble space telescope cluster supernova survey. V. Improving the dark-energy constraints above z > 1 and building an early-type-hosted supernova sample Astrophys. J. 746 85 [403] Paliathanasis A 2021 f (T) cosmology with nonzero curvature Mod. Phys. Lett. A 36 2150261 [404] El Hanafy W and Nashed G G L 2015 The hidden flat like universe: Starobinsky-like inflation induced by f (T) gravity Eur. Phys. J. C 75 279 [405] El Hanafy W and Nashed G G L 2016 The hidden flat like universe II: quasi inverse power law inflation by f (T) gravity Astrophys. Space Sci. 361 266 [406] Nashed G G L and El Hanafy W 2014 A built-in inflation in the f (T)-cosmology Eur. Phys. J. C 74 3099 [407] El Hanafy W and Nashed G L 2016 Reconstruction of f (T)-gravity in the absence of matter Astrophys. Space Sci. 361 197 [408] Wei H, Guo X-J and Wang L-F 2012 Noether symmetry in f (T) theory Phys. Lett. B 707 298–304 [409] Myrzakulov R 2012 Cosmology of F(T) gravity and k-essence Entropy 14 1627–51 [410] Sk N 2017 Noether symmetry in f (T) teleparallel gravity Phys. Lett. B 775 100–4 [411] Basilakos S, Capozziello S, De Laurentis M, Paliathanasis A and Tsamparlis M 2013 Noether symmetries and analytical solutions in f (T)-cosmology: a complete study Phys. Rev. D 88 103526 [412] Atazadeh K and Darabi F 2012 f (T) cosmology via Noether symmetry Eur. Phys. J. C 72 2016 [413] Mohseni Sadjadi H 2012 Generalized Noether symmetry in f (T) gravity Phys. Lett. B 718 270–5 [414] Dong H, Wang J and Meng X 2013 The distinctions between ΛCDM and f (T) gravity according Noether symmetry Eur. Phys. J. C 73 2543 [415] Fazlollahi H R 2020 Quantum cosmology f (T) and full Noether symmetries Phys. Dark Universe 28 100523 [416] Bajardi F and Capozziello S 2021 Noether symmetries and quantum cosmology in extended teleparallel gravity Int. J. Geom. Methods Mod. Phys. 18 2140002 [417] Sarkar K, Sk N, Ruz S, Debnath S and Sanyal A K 2013 Why Noether symmetry of F(R) theory yields three-half power law? Int. J. Theor. Phys. 52 1515–31 [418] Capozziello S, Martin-Moruno P and Rubano C 2008 Dark energy and dust matter phases from an exact f (R)-cosmology model Phys. Lett. B 664 12–15 [419] Sk N 2017 Noether symmetry of Palatini F(ℜ) gravity Int. J. Geom. Methods Mod. Phys. 15 1850046 [420] Jamil M, Yesmakhanova K, Momeni D and Myrzakulov R 2012 Phase space analysis of interacting dark energy in f (T) cosmology Cent. Eur. J. Phys. 10 1065–71 [421] Hohmann M, Järv L and Ualikhanova U 2017 Dynamical systems approach and generic properties of f (T) cosmology Phys. Rev. D 96 043508 [422] El Hanafy W and Nashed G G L 2017 Lorenz gauge fixing of f (T) teleparallel cosmology Int. J. Mod. Phys. D 26 1750154 [423] Awad A, El Hanafy W, Nashed G G L, Odintsov S D and Oikonomou V K 2018 Constant-roll inflation in f (T) teleparallel gravity J. Cosmol. Astropart. Phys. JCAP07(2018)026 [424] Awad A, El Hanafy W, Nashed G G L and Saridakis E N 2018 Phase portraits of general f (T) cosmology J. Cosmol. Astropart. Phys. JCAP02(2018)052 [425] Awad A and Nashed G 2017 Generalized teleparallel cosmology and initial singularity crossing J. Cosmol. Astropart. Phys. JCAP02(2017)046 [426] Jamil M, Momeni D and Myrzakulov R 2012 Attractor solutions in f (T) cosmology Eur. Phys. J. C 72 1959 [427] Wu P and Yu H W 2010 The dynamical behavior of f (T) theory Phys. Lett. B 692 176–9 [428] Setare M R and Mohammadipour N 2012 Cosmological viability conditions for f (T) dark energy models J. Cosmol. Astropart. Phys. JCAP11(2012)030 [429] Setare M R and Mohammadipour N 2013 Can f (T) gravity theories mimic ΛCDM cosmic history J. Cosmol. Astropart. Phys. JCAP01(2013)015 [430] Feng C-J, Li X-Z and Liu L-Y 2014 Bifurcation and global dynamical behavior of the f (T) theory Mod. Phys. Lett. A 29 1450033 196 Rep. Prog. Phys. 86 (2023) 026901 Review [431] Mirza B and Oboudiat F 2017 Constraining f (T) gravity by dynamical system analysis J. Cosmol. Astropart. Phys. JCAP11(2017)011 [432] Ganiou M G, Logbo P H, Houndjo M J S and Tossa J 2019 Cosmological study of autonomous dynamical systems in modified tele-parallel gravity Eur. Phys. J. Plus 134 45 [433] Böhmer C G, Fiorini F, González P A and Vásquez Y 2019 D = 11 cosmologies with teleparallel structure Phys. Rev. D 100 084007 [434] Fiorini F, Gonzalez P A and Vasquez Y 2014 Compact extra dimensions in cosmologies with f(T) structure Phys. Rev. D 89 024028 [435] Escamilla-Rivera C, Obregon O and Urena-Lopez L 2010 Supersymmetric classical cosmology J. Cosmol. Astropart. Phys. JCAP12(2010)011 [436] Escamilla-Rivera C, Rave-Franco G A and Said J L 2021 f(T, B) cosmography for high redshifts Universe 7 441 [437] Paliathanasis A and Leon G 2022 f(T, B) gravity in a Friedmann–Lemaı̂tre–Robertson–Walker universe with nonzero spatial curvature (arXiv:2201.12189 [gr-qc]) [438] Bahamonde S, Camci U and Capozziello S 2019 Noether symmetries and boundary terms in extended Teleparallel gravity cosmology Class. Quantum Grav. 36 065013 [439] Capozziello S, De Laurentis M and Myrzakulov R 2015 Noether symmetry approach for teleparallel-curvature cosmology Int. J. Geom. Methods Mod. Phys. 12 1550095 [440] Paliathanasis A 2017 Cosmological evolution and exact solutions in a fourth-order theory of gravity Phys. Rev. D 95 064062 [441] Paliathanasis A and Leon G 2021 Cosmological evolution in f(T, B) gravity Eur. Phys. J. Plus 136 1092 [442] Karpathopoulos L, Basilakos S, Leon G, Paliathanasis A and Tsamparlis M 2018 Cartan symmetries and global dynamical systems analysis in a higher-order modified teleparallel theory Gen. Relativ. Gravit. 50 79 [443] Shah P and Samanta G C 2019 Stability analysis for cosmological models in f (R) gravity using dynamical system analysis Eur. Phys. J. C 79 414 [444] Odintsov S and Oikonomou V 2018 Dynamical systems perspective of cosmological finite-time singularities in f (R) gravity and interacting multifluid cosmology Phys. Rev. D 98 024013 [445] Rave-Franco G A, Escamilla-Rivera C and Levi Said J 2021 Dynamical complexity of the teleparallel gravity cosmology Phys. Rev. D 103 084017 [446] Farrugia G, Levi Said J, Gakis V and Saridakis E N 2018 Gravitational waves in modified teleparallel theories Phys. Rev. D 97 124064 [447] Jawad A 2015 Cosmological reconstruction of pilgrim dark energy model in f(T, TG ) gravity Astrophys. Space Sci. 356 119–27 [448] Chattopadhyay S, Jawad A, Momeni D and Myrzakulov R 2014 Pilgrim dark energy in f(T, TG ) cosmology Astrophys. Space Sci. 353 279–92 [449] Kofinas G, Leon G and Saridakis E N 2014 Dynamical behavior in f(T, TG ) cosmology Class. Quantum Grav. 31 175011 [450] Tretyakov P V 2016 Dynamical stability of extended teleparallel gravity Mod. Phys. Lett. A 31 1650085 [451] Tretyakov P 2015 Dynamical stability of Minkowski space in higher order gravity Int. J. Geom. Methods Mod. Phys. 12 1550094 [452] Faraoni V 2004 De Sitter attractors in generalized gravity Phys. Rev. D 70 044037 [453] Momeni D and Myrzakulov R 2014 Cosmological reconstruction of f(T, T ) gravity Int. J. Geom. Methods Mod. Phys. 11 1450077 [454] Shahzad M U, Azhar N, Jawad A and Rani S 2019 Viability of specific reconstructed f(T, T ) models Int. J. Mod. Phys. A 34 1950184 [455] Salako I G, Jawad A and Chattopadhyay S 2015 Holographic dark energy reconstruction in f(T, T ) gravity Astrophys. Space Sci. 358 13 [456] Junior E L B, Rodrigues M E, Salako I G and Houndjo M J S 2016 Reconstruction, thermodynamics and stability of ΛCDM model in f(T, T ) gravity Class. Quantum Grav. 33 125006 [457] Nassur S B, Houndjo M J S, Rodrigues M E, Kpadonou A V and Tossa J 2015 From the early to the late time universe within f(T, T ) gravity Astrophys. Space Sci. 360 60 [458] Carloni S, Lobo F S N, Otalora G and Saridakis E N 2016 Dynamical system analysis for a nonminimal torsion-matter coupled gravity Phys. Rev. D 93 024034 [459] Ribeiro R and Páramos J 2014 Dynamical analysis of nonminimal coupled theories Phys. Rev. D 90 124065 [460] Azizi T and Yaraie E 2014 Cosmological dynamics of modified gravity with a nonminimal curvature-matter coupling Int. J. Mod. Phys. D 23 1450021 [461] An R, Xu X, Wang B and Gong Y 2016 Dynamical analysis of modified gravity with nonminimal gravitational coupling to matter Phys. Rev. D 93 103505 [462] Azevedo R and Páramos J 2016 Dynamical analysis of generalized f(R, L) theories Phys. Rev. D 94 064036 [463] Chakrabarti S, Said J L and Farrugia G 2017 Some aspects of reconstruction using a scalar field in f (T) gravity Eur. Phys. J. C 77 815 [464] Euler N 1997 Transformation properties of ẍ + f1 (t)ẋ + f2 (t)x + f3 (t)xn = 0 J. Nonlinear Math. Phys. 4 310–37 [465] Harko T, Lobo F S N and Mak M K 2013 Integrability cases for the anharmonic oscillator equation J. Nonlinear Math. Phys. 10 115–29 [466] Aslam A, Jamil M, Momeni D and Myrzakulov R 2013 Noether gauge symmetry of modified teleparallel gravity minimally coupled with a canonical scalar field Can. J. Phys. 91 93–97 [467] Sáez D and Ballester V J 1986 A simple coupling with cosmological implications Phys. Lett. A 113 467–70 [468] Jamil M, Ali S, Momeni D and Myrzakulov R 2012 Bianchi type I cosmology in generalized Saez–Ballester theory via Noether gauge symmetry Eur. Phys. J. C 72 1998 [469] Rao V, Vijaya Santhi M and Vinutha T 2008 Exact Bianchi type II, VIII and IX string cosmological models in Saez–Ballester theory of gravitation Astrophys. Space Sci. 314 73–77 [470] Adhav K S, Ugale M R, Kale C B and Bhende M P 2007 Bianchi type VI string cosmological model in Saez–Ballester’s scalar-tensor theory of gravitation Int. J. Theor. Phys. 46 3122–7 [471] Ram S, Zeyauddin M and Singh C P 2009 Bianchi type-V cosmological models with perfect fluid and heat flow in Saez–Ballester theory Pramana 72 415–27 [472] Rao V U M, Vinutha T and Vijaya Shanthi M 2007 An exact Bianchi Type-V cosmological model in Saez–Ballester theory of gravitation Astrophys. Space Sci. 312 189–91 [473] Jamil M, Momeni D and Myrzakulov R 2012 Noether symmetry of F(T) cosmology with quintessence and phantom scalar fields Eur. Phys. J. C 72 2137 [474] Tajahmad B 2017 Noether symmetries of a modified model in teleparallel gravity and a new approach for exact solutions Eur. Phys. J. C 77 211 [475] Maleknejad A, Sheikh-Jabbari M M and Soda J 2013 Gauge fields and inflation Phys. Rep. 528 161–261 [476] Vakili B 2014 A late time accelerated FRW model with scalar and vector fields via Noether symmetry Phys. Lett. B 738 488–92 197 Rep. Prog. Phys. 86 (2023) 026901 Review [477] Kucukakca Y 2013 Scalar tensor teleparallel dark gravity via Noether symmetry Eur. Phys. J. C 73 2327 [478] Gecim G and Kucukakca Y 2018 Scalar–tensor teleparallel gravity with boundary term by Noether symmetries Int. J. Geom. Methods Mod. Phys. 15 1850151 [479] Tajahmad B 2017 Studying the intervention of an unusual term in f (T) gravity via the Noether symmetry approach Eur. Phys. J. C 77 510 [480] Myrzakulov N, Bekov S, Myrzakulova S and Myrzakulov R 2019 Cosmological model of F(T) gravity with fermion fields via Noether symmetry J. Phys.: Conf. Ser. 1391 012165 [481] Kucukakca Y 2014 Teleparallel dark energy model with a fermionic field via Noether symmetry Eur. Phys. J. C 74 3086 [482] Gecim G and Sucu Y 2017 Dirac field as a source of the inflation in 2 + 1 dimensional teleparallel gravity Adv. High Energy Phys. 2017 2056131 [483] Biswas S K and Chakraborty S 2015 Interacting dark energy in f (T) cosmology: a dynamical system analysis Int. J. Mod. Phys. D 24 1550046 [484] Järv L and Toporensky A 2016 General relativity as an attractor for scalar-torsion cosmology Phys. Rev. D 93 024051 [485] Wei H 2012 Dynamics of teleparallel dark energy Phys. Lett. B 712 430–6 [486] Skugoreva M A, Saridakis E N and Toporensky A V 2015 Dynamical features of scalar-torsion theories Phys. Rev. D 91 044023 [487] Xu C, Saridakis E N and Leon G 2012 Phase-space analysis of teleparallel dark energy J. Cosmol. Astropart. Phys. JCAP07(2012)005 [488] Jamil M, Momeni D and Myrzakulov R 2012 Stability of a non-minimally conformally coupled scalar field in F(T) cosmology Eur. Phys. J. C 72 2075 [489] D’Agostino R and Luongo O 2018 Growth of matter perturbations in nonminimal teleparallel dark energy Phys. Rev. D 98 124013 [490] Channuie P and Momeni D 2018 Noether symmetry in a nonlocal f (T) gravity Nucl. Phys. B 935 256–70 [491] Tomboulis E T 2015 Renormalization and unitarity in higher derivative and nonlocal gravity theories Mod. Phys. Lett. A 30 1540005 [492] Modesto L and Rachwał L 2017 Nonlocal quantum gravity: a review Int. J. Mod. Phys. D 26 1730020 [493] Modesto L, Rachwał L and Shapiro I L 2018 Renormalization group in super-renormalizable quantum gravity Eur. Phys. J. C 78 555 [494] Bajardi F, Capozziello S and Vernieri D 2020 Non-local curvature and Gauss–Bonnet cosmologies by Noether symmetries Eur. Phys. J. Plus 135 942 [495] Bamba K, Momeni D and Ajmi M A 2018 Phase space description of nonlocal teleparallel gravity Eur. Phys. J. C 78 771 [496] Baumann D and McAllister L 2015 Inflation and String Theory (Cambridge Monographs on Mathematical Physics) (Cambridge: Cambridge University Press) [497] Romania M G, Tsamis N C and Woodard R P 2013 Quantum gravity and inflation Quantum Gravity and Quantum Cosmology (Lecture Notes in Physics vol 863) (Berlin: Springer) pp 375–95 [498] Born M and Fowler R H 1934 On the quantum theory of the electromagnetic field Proc. R. Soc. A 143 410–37 [499] Born M and Infeld L 1933 Foundations of the new field theory Nature 132 1004 [500] Born M 1933 Modified field equations with a finite radius of the electron Nature 132 282.1 [501] Deser S and Gibbons G W 1998 Born–Infeld–Einstein actions? Class. Quantum Grav. 15 L35–L39 [502] Beltrán Jiménez J, Heisenberg L, Olmo G J and Rubiera-Garcia D 2018 Born–Infeld inspired modifications of gravity Phys. Rep. 727 1–129 [503] Vollick D N 2004 Palatini approach to Born–Infeld–Einstein theory and a geometric description of electrodynamics Phys. Rev. D 69 064030 [504] Banados M, Ferreira P G and Skordis C 2009 Eddington-Born–Infeld gravity and the large scale structure of the Universe Phys. Rev. D 79 063511 [505] Fiorini F and Ferraro R 2009 A Type of Born–Infeld regular gravity and its cosmological consequences Int. J. Mod. Phys. A 24 1686–9 [506] Jana S 2014 Cosmology in a reduced Born–Infeld f (T) theory of gravity Phys. Rev. D 90 124007 [507] Nesseris S and Perivolaropoulos L 2004 A comparison of cosmological models using recent supernova data Phys. Rev. D 70 043531 [508] Fiorini F 2013 Nonsingular promises from Born–Infeld gravity Phys. Rev. Lett. 111 041104 [509] Bouhmadi-Lopez M, Chen C-Y and Chen P 2014 Cosmological singularities in Born–Infeld determinantal gravity Phys. Rev. D 90 123518 [510] Fiorini F 2016 Primordial brusque bounce in Born–Infeld determinantal gravity Phys. Rev. D 94 024030 [511] Wilson-Ewing E 2013 The matter bounce scenario in loop quantum cosmology J. Cosmol. Astropart. Phys. JCAP03(2013)026 [512] Cai Y-F, Chen S-H, Dent J B, Dutta S and Saridakis E N 2011 Matter bounce cosmology with the f (T) gravity Class. Quantum Grav. 28 215011 [513] Bamba K, de Haro J and Odintsov S D 2013 Future singularities and teleparallelism in loop quantum cosmology J. Cosmol. Astropart. Phys. JCAP02(2013)008 [514] Amorós J, de Haro J and Odintsov S D 2013 Bouncing loop quantum cosmology from F(T) gravity Phys. Rev. D 87 104037 [515] Casalino A, Sanna B, Sebastiani L and Zerbini S 2021 Bounce models within teleparallel modified gravity Phys. Rev. D 103 023514 [516] Akrami Y et al (Planck Collaboration) 2020 Planck 2018 results. X. Constraints on inflation Astron. Astrophys. 641 A10 [517] Haro J and Amoros J 2014 Viability of the matter bounce scenario in F(T) gravity and loop quantum cosmology for general potentials J. Cosmol. Astropart. Phys. JCAP12(2014)031 [518] de Haro J and Amorós J 2015 Viability of the matter bounce scenario J. Phys.: Conf. Ser. 600 012024 [519] Haro J and Amorós J 2016 Matter bounce scenario in F(T) gravity Proc. Sci. 224 163 [520] de Haro J and Cai Y-F 2015 An extended matter bounce scenario: current status and challenges Gen. Relativ. Gravit. 47 95 [521] Haro J 2013 Cosmological perturbations in teleparallel loop quantum cosmology J. Cosmol. Astropart. Phys. JCAP11(2013)068 Haro J 2014 J. Cosmol. Astropart. Phys. 05 E01 [522] de Haro J and Amoros J 2014 Viability of the matter bounce scenario in loop quantum cosmology from BICEP2 last data J. Cosmol. Astropart. Phys. JCAP08(2014)025 [523] Haro J, Amorós J and Aresté Saló L 2017 The matter-ekpyrotic bounce scenario in Loop Quantum Cosmology J. Cosmol. Astropart. Phys. JCAP09(2017)002 [524] Peebles P J E 2020 Principles of Physical Cosmology (Princeton, NJ: Princeton University Press) [525] Rubio J 2019 Higgs inflation Front. Astron. Space Sci. 5 50 198 Rep. Prog. Phys. 86 (2023) 026901 Review [526] Bezrukov F 2013 The Higgs field as an inflaton Class. Quantum Grav. 30 214001 [527] Raatikainen S and Rasanen S 2019 Higgs inflation and teleparallel gravity J. Cosmol. Astropart. Phys. JCAP12(2019)021 [528] Mohseni Sadjadi H 2017 Symmetron and de Sitter attractor in a teleparallel model of cosmology J. Cosmol. Astropart. Phys. JCAP01(2017)031 [529] Mohseni Sadjadi H 2015 Onset of acceleration in a universe initially filled by dark and baryonic matters in a nonminimally coupled teleparallel model Phys. Rev. D 92 123538 [530] Benisty D, Guendelman E I, Nissimov E and Pacheva S 2020 Quintessential inflation with dynamical Higgs generation as an affine gravity Symmetry 12 734 [531] Hamada Y, Kawai H and Oda K-Y 2014 Minimal Higgs inflation Prog. Theor. Exp. Phys. 2014 023B02 [532] Gundhi A and Steinwachs C F 2020 Scalaron-Higgs inflation Nucl. Phys. B 954 114989 [533] Schlögel S 2016 Particle like solutions in modified gravity: the Higgs monopoles Proc. Sci. 224 096 [534] Järv L and Lember J 2021 Global portraits of nonminimal teleparallel inflation Universe 7 179 [535] Farrugia G and Levi Said J 2016 Stability of the flat FLRW metric in f (T) gravity Phys. Rev. D 94 124054 [536] Gonzalez-Espinoza M and Otalora G 2020 Generating primordial fluctuations from modified teleparallel gravity with local Lorentz-symmetry breaking Phys. Lett. B 809 135696 [537] Gonzalez-Espinoza M, Herrera R, Otalora G and Saavedra J 2021 Reconstructing inflation in scalar-torsion f(T, ϕ) gravity Eur. Phys. J. C 81 731 [538] Goodarzi P and Mohseni Sadjadi H 2019 Reheating in a modified teleparallel model of inflation Eur. Phys. J. C 79 193 [539] Rezaei Akbarieh A and Izadi Y 2019 Tachyon inflation in teleparallel gravity Eur. Phys. J. C 79 366 [540] Woodard R P 2015 Ostrogradsky’s theorem on Hamiltonian instability Scholarpedia 10 32243 [541] De Felice A, Hindmarsh M and Trodden M 2006 Ghosts, instabilities and superluminal propagation in modified gravity models J. Cosmol. Astropart. Phys. JCAP08(2006)005 [542] Barth N H and Christensen S M 1983 Quantizing fourth order gravity theories. 1. The functional integral Phys. Rev. D 28 1876 [543] Stelle K S 1978 Classical gravity with higher derivatives Gen. Relativ. Gravit. 9 353–71 [544] Deffayet C, Mukohyama S and Vikman A 2022 Ghosts without runaway instabilities Phys. Rev. Lett. 128 041301 [545] De Felice A, Frusciante N and Papadomanolakis G 2017 On the stability conditions for theories of modified gravity in the presence of matter fields J. Cosmol. Astropart. Phys. JCAP03(2017)027 [546] Rubakov V A 2014 The null energy condition and its violation Phys.-Usp. 57 128–42 [547] Perivolaropoulos L and Skara F 2020 Scalar tachyonic instabilities in gravitational backgrounds: existence and growth rate Phys. Rev. D 102 104034 [548] Gsponer R and Noller J 2022 Tachyonic stability priors for dark energy Phys. Rev. D 105 064002 [549] El Hanafy W and Nashed G G L 2017 Generic phase portrait analysis of finite-time singularities and generalized Teleparallel gravity Chin. Phys. C 41 125103 [550] Bamba K, Nashed G G L, El Hanafy W and Ibraheem S K 2016 Bounce inflation in f (T) cosmology: a unified inflaton-quintessence field Phys. Rev. D 94 083513 [551] Lifshitz E 2017 Republication of: on the gravitational stability of the expanding universe Gen. Relativ. Gravit. 49 18 [552] Malik K A and Wands D 2009 Cosmological perturbations Phys. Rep. 475 1–51 [553] Kodama H and Sasaki M 1984 Cosmological perturbation theory Prog. Theor. Phys. Suppl. 78 1–166 [554] Jiménez J B and Koivisto T S 2021 Accidental gauge symmetries of Minkowski spacetime in teleparallel theories Universe 7 143 [555] Chen H, Nester J M and Yo H-J 1998 Acausal PGT modes and the nonlinear constraint effect Acta Phys. Polon. B 29 961–70 [556] Ong Y C, Izumi K, Nester J M and Chen P 2013 Problems with propagation and time evolution in f (T) gravity Phys. Rev. D 88 024019 [557] Chen P, Izumi K, Nester J M and Ong Y C 2015 Remnant symmetry, propagation and evolution in f (T) gravity Phys. Rev. D 91 064003 [558] Böhmer C G, Hollenstein L and Lobo F S 2007 Stability of the Einstein static universe in f (R) gravity Phys. Rev. D 76 084005 [559] Boehmer C G and Lobo F S N 2010 Stability of the Einstein static universe in IR modified Hořava gravity Eur. Phys. J. C 70 1111–8 [560] Böhmer C G and Lobo F S 2009 Stability of the Einstein static universe in modified Gauss–Bonnet gravity Phys. Rev. D 79 067504 [561] Wu P and Yu H 2011 The Stability of the Einstein static state in f (T) gravity Phys. Lett. B 703 223–7 [562] Hohmann M 2021 General cosmological perturbations in teleparallel gravity Eur. Phys. J. Plus 136 65 [563] Gourgoulhon E 2007 3 + 1 formalism and bases of numerical relativity (arXiv:gr-qc/0703035) [564] Bojowald M 2010 Canonical Gravity and Applications: Cosmology, Black Holes and Quantum Gravity (Cambridge: Cambridge University Press) [565] Giulini D 2015 Dynamical and Hamiltonian Formulation of General Relativity (Berlin: Springer) pp 323–62 [566] Mukhanov V F, Feldman H A and Brandenberger R H 1992 Theory of cosmological perturbations. Part 1. Classical perturbations. Part 2. Quantum theory of perturbations. Part 3. Extensions Phys. Rep. 215 203–333 [567] Dodelson S 2003 Modern Cosmology (Amsterdam: Academic) [568] Chen S-H, Dent J B, Dutta S and Saridakis E N 2011 Cosmological perturbations in f (T) gravity Phys. Rev. D 83 023508 [569] Zheng R and Huang Q-G 2011 Growth factor in f (T) gravity J. Cosmol. Astropart. Phys. JCAP03(2011)002 [570] Wu Y-P and Geng C-Q 2012 Matter density perturbations in modified teleparallel theories J. High Energy Phys. JHEP11(2012)142 [571] Bruni M, Matarrese S, Mollerach S and Sonego S 1997 Perturbations of space-time: gauge transformations and gauge invariance at second order and beyond Class. Quantum Grav. 14 2585–606 [572] Malik K A and Matravers D R 2008 A concise introduction to perturbation theory in cosmology Class. Quantum Grav. 25 193001 [573] Bardeen J M 1980 Gauge invariant cosmological perturbations Phys. Rev. D 22 1882–905 [574] Saltas I D, Sawicki I, Amendola L and Kunz M 2014 Anisotropic stress as a signature of nonstandard propagation of gravitational waves Phys. Rev. Lett. 113 191101 [575] Abbott B P et al (LIGO Scientific, Virgo, Fermi-GBM and INTEGRAL Collaboration) 2017 Gravitational waves and gamma-rays from a binary neutron star merger: 199 Rep. Prog. Phys. 86 (2023) 026901 Review GW170817 and GRB 170817A Astrophys. J. Lett. 848 L13 [576] Abbott B P et al (LIGO Scientific and Virgo Collaboration) 2016 Observation of gravitational waves from a binary black hole merger Phys. Rev. Lett. 116 061102 [577] Will C M 2018 Solar system versus gravitational-wave bounds on the graviton mass Class. Quantum Grav. 35 17LT01 [578] Will C M 2018 New general relativistic contribution to mercury’s perihelion advance Phys. Rev. Lett. 120 191101 [579] de Rham C, Deskins J T, Tolley A J and Zhou S-Y 2017 Graviton mass bounds Rev. Mod. Phys. 89 025004 [580] Ezquiaga J M and Zumalacárregui M 2017 Dark energy after GW170817: dead ends and the road ahead Phys. Rev. Lett. 119 251304 [581] Riazuelo A and Uzan J-P 2000 Quintessence and gravitational waves Phys. Rev. D 62 083506 [582] Copeland E J, Kopp M, Padilla A, Saffin P M and Skordis C 2019 Dark energy after GW170817 revisited Phys. Rev. Lett. 122 061301 [583] De Felice A, Gerard J-M and Suyama T 2010 Cosmological perturbation in f(R,G) theories with a perfect fluid Phys. Rev. D 82 063526 [584] Hohmann M and Pfeifer C 2022 Gravitational wave birefringence in spatially curved teleparallel cosmology (arXiv:2203.01856 [gr-qc]) [585] Wu Q, Zhu T, Niu R, Zhao W and Wang A 2022 Constraints on the Nieh–Yan modified teleparallel gravity with gravitational waves Phys. Rev. D 105 024035 [586] Farrugia G and Levi Said J 2016 Growth factor in f(T, T ) gravity Phys. Rev. D 94 124004 [587] Basilakos S and Pouri A 2012 The growth index of matter perturbations and modified gravity Mon. Not. R. Astron. Soc. 423 3761 [588] Colless M et al 2003 The 2dF galaxy redshift survey: final data release (arXiv:astro-ph/0306581) [589] Fu X, Wu P and Yu H 2011 The growth of matter perturbations in f(T) gravity Int. J. Mod. Phys. D 20 1301–11 [590] Geng C-Q and Wu Y-P 2013 Density perturbation growth in teleparallel cosmology J. Cosmol. Astropart. Phys. JCAP04(2013)033 [591] Basilakos S 2016 Linear growth in power law f (T) gravity Phys. Rev. D 93 083007 [592] Gonzalez-Espinoza M, Otalora G, Saavedra J and Videla N 2018 Growth of matter overdensities in non-minimal torsion-matter coupling theories Eur. Phys. J. C 78 799 [593] Kazantzidis L and Perivolaropoulos L 2018 Evolution of the fσ8 tension with the Planck15/ΛCDM determination and implications for modified gravity theories Phys. Rev. D 97 103503 [594] Levi Said J, Mifsud J, Sultana J and Adami K Z 2021 Reconstructing teleparallel gravity with cosmic structure growth and expansion rate data (arXiv:2103.05021 [astro-ph.CO]) [595] Nunes R C 2018 Structure formation in f (T) gravity and a solution for H 0 tension J. Cosmol. Astropart. Phys. JCAP05(2018)052 [596] Nunes R C, Pan S and Saridakis E N 2018 New observational constraints on f (T) gravity through gravitational-wave astronomy Phys. Rev. D 98 104055 [597] Tsujikawa S 2007 Matter density perturbations and effective gravitational constant in modified gravity models of dark energy Phys. Rev. D 76 023514 [598] Alvarenga F G, de la Cruz-Dombriz A, Houndjo M J S, Rodrigues M E and Sáez-Gómez D 2013 Dynamics of scalar perturbations in f(R, T) gravity Phys. Rev. D 87 103526 Alvarenga F G, de la Cruz-Dombriz A, Houndjo M J S, Rodrigues M E and Sáez-Gómez D 2013 Phys. Rev. D 87 129905 (erratum) [599] Eardley D, Lee D, Lightman A, Wagoner R and Will C 1973 Gravitational-wave observations as a tool for testing relativistic gravity Phys. Rev. Lett. 30 884–6 [600] Eardley D, Lee D and Lightman A 1973 Gravitational-wave observations as a tool for testing relativistic gravity Phys. Rev. D 8 3308–21 [601] Nakao K-I, Harada T, Shibata M, Kawamura S and Nakamura T 2001 Response of interferometric detectors to scalar gravitational waves Phys. Rev. D 63 082001 [602] Takeda H, Morisaki S and Nishizawa A 2021 Pure polarization test of GW170814 and GW170817 using waveforms consistent with modified theories of gravity Phys. Rev. D 103 064037 [603] Tobar M E, Suzuki T and Kuroda K 1999 Detecting free mass common mode motion induced by incident gravitational waves Phys. Rev. D 59 102002 [604] Abbott R et al (The LIGO Scientific Collaboration, The Virgo Collaboration and The KAGRA Collaboration) 2021 Tests of general relativity with GWTC-3 (arXiv:2112.06861 [gr-qc]) [605] Abbott B P et al (LIGO Scientific Collaboration and Virgo Collaboration) 2017 GW170814: a three-detector observation of gravitational waves from a binary black hole coalescence Phys. Rev. Lett. 119 141101 [606] Abbott B P et al (LIGO Scientific Collaboration and Virgo Collaboration) 2019 Tests of general relativity with GW170817 Phys. Rev. Lett. 123 011102 [607] Abbott R et al (The LIGO Scientific Collaboration and The Virgo Collaboration) 2020 Tests of general relativity with binary black holes from the second LIGO-Virgo gravitational-wave transient catalog (arXiv:2010.14529 [gr-qc]) [608] Abbott R et al (The LIGO Scientific Collaboration and The Virgo Collaboration) 2021 GWTC-2: compact binary coalescences observed by LIGO and Virgo during the first half of the third observing run (arXiv:2010.14527 [gr-qc]) [609] Miller J, Barsotti L, Vitale S, Fritschel P, Evans M and Sigg D 2015 Prospects for doubling the range of Advanced LIGO Phys. Rev. D 91 062005 [610] Nguyen C 2021 Status of the Advanced Virgo gravitational-wave detector (arXiv:2105.09247 [astro-ph.IM]) [611] Michimura Y et al 2020 Prospects for improving the sensitivity of the cryogenic gravitational wave detector KAGRA Phys. Rev. D 102 022008 [612] Unnikrishnan C S 2013 IndIGO and LIGO-India: scope and plans for gravitational wave research and precision metrology in India Int. J. Mod. Phys. D 22 1341010 [613] Abbott B P et al (KAGRA Collaboration, LIGO Scientific Collaboration and Virgo Collaboration) 2020 Prospects for observing and localizing gravitational-wave transients with Advanced LIGO, Advanced Virgo and KAGRA Living Rev. Relativ. 23 3 [614] Will C M 2018 Theory and Experiment in Gravitational Physics (Cambridge: Cambridge University Press) [615] Bahamonde S, Caruana M, Dialektopoulos K F, Gakis V, Hohmann M, Levi Said J, Saridakis E N and Sultana J 2021 Gravitational wave propagation and polarizations in the teleparallel analog of Horndeski gravity (arXiv:2105.13243 [gr-qc]) [616] Hohmann M 2018 Polarization of gravitational waves in general teleparallel theories of gravity Astron. Rep. 62 890–7 [617] Hohmann M, Krššák M, Pfeifer C and Ualikhanova U 2018 Propagation of gravitational waves in teleparallel gravity theories Phys. Rev. D 98 124004 200 Rep. Prog. Phys. 86 (2023) 026901 Review [618] Obukhov Y N and Pereira J G 2003 Teleparallel origin of the Fierz picture for spin-2 particle Phys. Rev. D 67 044008 [619] Hörmander L 2015 The Analysis of Linear Partial Differential Operators I: Distribution Theory and Fourier Analysis (Classics in Mathematics) (Berlin: Springer) [620] Hörmander L 2007 The Analysis of Linear Partial Differential Operators III: Pseudo-Differential Operators (Classics in Mathematics) (Berlin: Springer) [621] Pfeifer C and Siemssen D 2016 Electromagnetic potential in pre-metric electrodynamics: causal structure, propagators and quantization Phys. Rev. D 93 105046 [622] Liang D, Gong Y, Hou S and Liu Y 2017 Polarizations of gravitational waves in f (R) gravity Phys. Rev. D 95 104034 [623] Yang L, Lee C-C and Geng C-Q 2011 Gravitational waves in viable f (R) models J. Cosmol. Astropart. Phys. JCAP08(2011)029 [624] Capozziello S, Corda C and De Laurentis M F 2008 Massive gravitational waves from f (R) theories of gravity: potential detection with LISA Phys. Lett. B 669 255–9 [625] Berry C P and Gair J R 2011 Linearized f (R) gravity: gravitational radiation and solar system tests Phys. Rev. D 83 104022 Berry C P and Gair J R 2012 Phys. Rev. D 85 089906 (erratum) [626] Hyun Y-H, Kim Y and Lee S 2019 Exact amplitudes of six polarization modes for gravitational waves Phys. Rev. D 99 124002 [627] Capozziello S, Capriolo M and Caso L 2020 Weak field limit and gravitational waves in f(T, B) teleparallel gravity Eur. Phys. J. C 80 156 [628] Abedi H and Capozziello S 2018 Gravitational waves in modified teleparallel theories of gravity Eur. Phys. J. C 78 474 [629] Bamba K, Capozziello S, De Laurentis M, Nojiri S and Sáez-Gómez D 2013 No further gravitational wave modes in F(T) gravity Phys. Lett. B 727 194–8 [630] Hou S, Gong Y and Liu Y 2018 Polarizations of gravitational waves in Horndeski theory Eur. Phys. J. C 78 378 [631] Gong Y and Hou S 2018 The polarizations of gravitational waves Universe 4 85 [632] Ferraro R and Fiorini F 2011 Spherically symmetric static spacetimes in vacuum f (T) gravity Phys. Rev. D 84 083518 [633] Paliathanasis A, Basilakos S, Saridakis E N, Capozziello S, Atazadeh K, Darabi F and Tsamparlis M 2014 New Schwarzschild-like solutions in f (T) gravity through Noether symmetries Phys. Rev. D 89 104042 [634] Bejarano C, Ferraro R and Guzmán M J 2017 McVittie solution in f (T) gravity Eur. Phys. J. C 77 825 [635] Daouda M H, Rodrigues M E and Houndjo M J S 2012 Anisotropic fluid for a set of non-diagonal tetrads in f (T) gravity Phys. Lett. B 715 241–5 [636] Hamani Daouda M, Rodrigues M E and Houndjo M 2011 New static solutions in f (T) theory Eur. Phys. J. C 71 1817 [637] Meng X-H and Wang Y-B 2011 Birkhoff’s theorem in the f (T) gravity Eur. Phys. J. C 71 1755 [638] Capozziello S, Gonzalez P A, Saridakis E N and Vásquez Y 2013 Exact charged black-hole solutions in D-dimensional f (T) gravity: torsion vs curvature analysis J. High Energy Phys. JHEP02(2013)039 [639] Houndjo M, Momeni D, Myrzakulov R and Rodrigues M 2015 Evaporation phenomena in f (T) gravity Can. J. Phys. 93 377–83 [640] Kofinas G, Papantonopoulos E and Saridakis E N 2015 Self-gravitating spherically symmetric solutions in scalar-torsion theories Phys. Rev. D 91 104034 [641] Bahamonde S, Golovnev A, Guzmán M-J, Said J L and Pfeifer C 2022 Black holes in f(T,B) gravity: exact and perturbed solutions J. Cosmol. Astropart. Phys. JCAP01(2022)037 [642] Bahamonde S, Ducobu L and Pfeifer C 2022 Scalarized black holes in teleparallel gravity (arXiv:2201.11445 [gr-qc]) [643] DeBenedictis A and Ilijić S 2018 Regular solutions in f (T)-Yang–Mills theory Phys. Rev. D 98 064056 [644] Golovnev A and Guzmán M-J 2021 Approaches to spherically symmetric solutions in f (T) gravity (arXiv:2103.16970 [gr-qc]) [645] Golovnev A and Guzmán M-J 2020 Bianchi identities in f (T) gravity: paving the way to confrontation with astrophysics Phys. Lett. B 810 135806 [646] DeBenedictis A and Ilijic S 2016 Spherically symmetric vacuum in covariant F(T) = T + α2 T 2 + O(Tγ ) gravity theory Phys. Rev. D 94 124025 [647] Abuter R et al 2020 Detection of the Schwarzschild precession in the orbit of the star S2 near the Galactic centre massive black hole Astron. Astrophys. 636 L5 [648] Ishak M and Rindler W 2010 The relevance of the cosmological constant for lensing Gen. Relativ. Gravit. 42 2247–68 [649] Azreg-Aïnou M, Bahamonde S and Jamil M 2017 Strong gravitational lensing by a charged Kiselev black hole Eur. Phys. J. C 77 414 [650] Robertson D S, Carter W E and Dillinger W 1991 New measurement of solar gravitational deflection of radio signals using vlbi Nature 349 768–70 [651] Shapiro S, Davis J, Lebach D and Gregory J 2004 Measurement of the solar gravitational deflection of radio waves using geodetic very-long-baseline interferometry data, 1979–1999 Phys. Rev. Lett. 92 121101 [652] Bertotti B, Iess L and Tortora P 2003 A test of general relativity using radio links with the Cassini spacecraft Nature 425 374–6 [653] Reasenberg R D, Shapiro I I, MacNeil P E, Goldstein R B, Breidenthal J C, Brenkle J P, Cain D L, Kaufman T M, Komarek T A and Zygielbaum A I 1979 Viking relativity experiment: verification of signal retardation by solar gravity Astrophys. J. Lett. 234 L219–21 [654] Demorest P, Pennucci T, Ransom S, Roberts M and Hessels J 2010 Shapiro delay measurement of a two solar mass neutron star Nature 467 1081–3 [655] Cromartie H T et al 2019 Relativistic Shapiro delay measurements of an extremely massive millisecond pulsar Nat. Astron. 4 72–76 [656] Vessot R F C et al 1980 Test of relativistic gravitation with a space-borne hydrogen maser Phys. Rev. Lett. 45 2081–4 [657] Rindler W 2003 Relativity: special, general and cosmological Am. J. Phys. 71 1085–6 [658] Pfeifer C and Schuster S 2021 Static spherically symmetric black holes in weak f (T)-gravity (arXiv:2104.00116 [gr-qc]) [659] Hohmann M 2020 Gauge-invariant approach to the parametrized post-Newtonian formalism Phys. Rev. D 101 024061 [660] Burrage C and Sakstein J 2018 Tests of chameleon gravity Living Rev. Relativ. 21 1 [661] Li J-T, Wu Y-P and Geng C-Q 2014 Parametrized post-Newtonian limit of the teleparallel dark energy model Phys. Rev. D 89 044040 [662] Chen Z-C, Wu Y and Wei H 2015 Post-Newtonian approximation of teleparallel gravity coupled with a scalar field Nucl. Phys. B 894 422–38 [663] Mohseni Sadjadi H 2017 Parameterized post-Newtonian approximation in a teleparallel model of dark energy with a boundary term Eur. Phys. J. C 77 191 [664] Flathmann K and Hohmann M 2020 Post-Newtonian limit of generalized scalar-torsion theories of gravity Phys. Rev. D 101 024005 201 Rep. Prog. Phys. 86 (2023) 026901 Review [665] Emtsova E D and Hohmann M 2020 Post-Newtonian limit of scalar-torsion theories of gravity as analogue to scalar-curvature theories Phys. Rev. D 101 024017 [666] Bahamonde S, Dialektopoulos K F, Hohmann M and Levi Said J 2020 Post-Newtonian limit of teleparallel Horndeski gravity Class. Quantum Grav. 38 025006 [667] Rao H 2021 Parametrized post-Newtonian limit of the Nieh–Yan modified teleparallel gravity Phys. Rev. D 104 124084 [668] Gonzalez-Espinoza M, Otalora G, Kraiselburd L and Landau S 2021 Parametrized post-newtonian formalism in higher-order teleparallel gravity (arXiv:2112.06117 [gr-qc]) [669] Böhmer C G, Mussa A and Tamanini N 2011 Existence of relativistic stars in f (T) gravity Class. Quantum Grav. 28 245020 [670] Ilijic S and Sossich M 2018 Compact stars in f (T) extended theory of gravity Phys. Rev. D 98 064047 [671] Olmo G J, Rubiera-Garcia D and Wojnar A 2020 Stellar structure models in modified theories of gravity: lessons and challenges Phys. Rep. 876 1–75 [672] Ilijić S and Sossich M 2020 Boson stars in f (T) extended theory of gravity Phys. Rev. D 102 084019 [673] Horvat D, Ilijić S, Kirin A and Narančić Z 2015 Nonminimally coupled scalar field in teleparallel gravity: boson stars Class. Quantum Grav. 32 035023 [674] Horvat D, Ilijic S, Kirin A and Narancic Z 2015 Note on the charged boson stars with torsion-coupled field Phys. Rev. D 92 024045 [675] de Araujo J C N and Fortes H G M 2021 Solving Tolman–Oppenheimer–Volkoff equations in f (T) gravity: a novel approach applied to polytropic equations of state (arXiv:2105.09118 [gr-qc]) [676] Fortes H G M and de Araujo J C N 2021 Solving Tolman-Oppenheimer-Volkoff equations in f (T) gravity: a novel approach (arXiv:2105.04473 [gr-qc]) [677] Chakrabarti S and Levi Said J 2020 Geodesic congruences and a collapsing stellar distribution in f (T) theories Phys. Rev. D 101 124044 [678] Pace M and Said J L 2017 A perturbative approach to neutron stars in f(T, T ) gravity Eur. Phys. J. C 77 283 [679] Pace M and Said J L 2017 Quark stars in f(T, T ) gravity Eur. Phys. J. C 77 62 [680] Lin R-H, Chen X-N and Zhai X-H 2021 Neutron stars in f (T) gravity with realistic models of matter (arXiv:2109.00191 [gr-qc]) [681] de la Cruz-Dombriz A, Dunsby P K S and Saez-Gomez D 2014 Junction conditions in extended teleparallel gravities J. Cosmol. Astropart. Phys. JCAP12(2014)048 [682] Velay-Vitow J and DeBenedictis A 2017 Junction conditions for f (T) gravity from a variational principle Phys. Rev. D 96 024055 [683] Fiorini F and Onetto M 2021 Matching tetrads in f (T) gravity Phys. Rev. D 103 084051 [684] Wojnar A 2020 Early evolutionary tracks of low-mass stellar objects in modified gravity Phys. Rev. D 102 124045 [685] Sousa A A and Maluf J W 2002 Black holes in 2 + 1 teleparallel theories of gravity Prog. Theor. Phys. 108 457–70 [686] Castello-Branco K and da Rocha-Neto J 2012 Gravitational energy of a magnetized Schwarzschild black hole: a teleparallel approach Gen. Relativ. Gravit. 44 3017–28 [687] Castello-Branco K and da Rocha-Neto J 2013 Gravitational energy, gravitational pressure and the thermodynamics of a charged black hole in teleparallel gravity Phys. Rev. D 88 024045 [688] Junior E L B, Rodrigues M E and Houndjo M J S 2015 Born–Infeld and charged black holes with non-linear source in f (T) gravity J. Cosmol. Astropart. Phys. JCAP1506(2015)37 [689] Debnath U and Paul B C 2015 Evolution of primordial black hole in f (T) gravity with modified Chaplygin gas Astrophys. Space Sci. 355 147–53 [690] Aftergood J and DeBenedictis A 2014 Matter conditions for regular black holes in f (T) gravity Phys. Rev. D 90 124006 [691] Böhmer C G and Fiorini F 2019 The regular black hole in four dimensional Born–Infeld gravity Class. Quantum Grav. 36 12LT01 [692] Böhmer C G and Fiorini F 2020 BTZ gems inside regular Born–Infeld black holes Class. Quantum Grav. 37 185002 [693] DeBenedictis A, Ilijić S and Sossich M 2022 On spherically symmetric vacuum solutions and horizons in covariant f (T) gravity theory (arXiv:2202.08958 [gr-qc]) [694] Emtsova E D, Krššák M, Petrov A N and Toporensky A V 2021 On conserved quantities for the Schwarzschild black hole in teleparallel gravity Eur. Phys. J. C 81 743 [695] Morris M S and Thorne K S 1988 Wormholes in space-time and their use for interstellar travel: a tool for teaching general relativity Am. J. Phys. 56 395–412 [696] Visser M 1989 Traversable wormholes: some simple examples Phys. Rev. D 39 3182–4 [697] Morris M, Thorne K and Yurtsever U 1988 Wormholes, time machines and the weak energy condition Phys. Rev. Lett. 61 1446–9 [698] Visser M 1995 Lorentzian Wormholes: From Einstein to Hawking (Computational and Mathematical Physics) (New York: American Institute of Physics) [699] Böhmer C G, Harko T and Lobo F S N 2012 Wormhole geometries in modified teleparallel gravity and the energy conditions Phys. Rev. D 85 044033 [700] Lobo F S and Oliveira M A 2009 Wormhole geometries in f (R) modified theories of gravity Phys. Rev. D 80 104012 [701] Harko T, Lobo F S, Mak M and Sushkov S V 2013 Modified-gravity wormholes without exotic matter Phys. Rev. D 87 067504 [702] Rani S, Bilal Amin M and Jawad A 2016 Exponential and logarithmic f (T) wormhole solutions in Lorentzian noncommutative background Eur. Phys. J. Plus 131 436 [703] Sharif M and Rani S 2014 Charged noncommutative wormhole solutions in f (T) gravity Eur. Phys. J. Plus 129 237 [704] Jawad A and Rani S 2015 Lorentz distributed noncommutative wormhole solutions in extended teleparallel gravity Eur. Phys. J. C 75 173 [705] Rani S, Jawad A and Amin M 2016 Charged noncommutative wormhole solutions via power-law f (T) models Commun. Theor. Phys. 66 411–22 [706] Mustafa G, Waheed S, Zubair M and Xia T 2019 Gaussian distributed wormholes exhibiting conformal motion in f (T) gravity Int. J. Geom. Methods Mod. Phys. 16 1950143 [707] Bahamonde S, Camci U, Capozziello S and Jamil M 2016 Scalar-tensor teleparallel wormholes by Noether symmetries Phys. Rev. D 94 084042 [708] Casertano S 1983 Rotation curve of the edge-on spiral galaxy NGC 5907: disc and halo masses Mon. Not. R. Astron. Soc. 203 735–47 [709] Sofue Y, Honma M and Omodaka T 2009 Unified rotation curve of the galaxy—decomposition into de Vaucouleurs bulge, disk, dark halo and the 9-kpc rotation dip— Publ. Astron. Soc. Japan 61 227 [710] Maluf J W, Ulhoa S, Carneiro F L and Castello-Branco K H C 2021 Perturbations of the gravitational energy in the TEGR: quasinormal modes of the Schwarzschild black hole Universe 7 100 202 Rep. Prog. Phys. 86 (2023) 026901 Review [711] Cardoso V and Pani P 2019 Testing the nature of dark compact objects: a status report Living Rev. Relativ. 22 4 [712] Gourgoulhon E and Jaramillo J L 2006 A 3 + 1 perspective on null hypersurfaces and isolated horizons Phys. Rep. 423 159–294 [713] Capozziello S, Finch A, Said J L and Magro A 2021 The 3 + 1 formalism in teleparallel and symmetric teleparallel gravity Eur. Phys. J. C 81 1141 [714] Alves Batista R et al 2021 EuCAPT white paper: opportunities and challenges for theoretical astroparticle physics in the next decade (arXiv:2110.10074 [astro-ph.HE]) [715] Kazakov D I 2000 Beyond the standard model: in search of supersymmetry 2000 European School of High-Energy Physics (arXiv:hep-ph/0012288) [716] Ade P et al (Planck Collaboration) 2014 Planck 2013 results. XVI. Cosmological parameters Astron. Astrophys. 571 16 [717] Riess A G, Macri L, Casertano S, Lampeitl H, Ferguson H C, Filippenko A V, Jha S W, Li W and Chornock R 2011 A 3% solution: determination of the hubble constant with the hubble space telescope and wide field camera 3 Astrophys. J. 730 119 [718] Riess A G, Casertano S, Yuan W, Macri L M and Scolnic D 2019 Large magellanic cloud cepheid standards provide a 1% foundation for the determination of the hubble constant and stronger evidence for physics beyond ΛCDM Astrophys. J. 876 85 [719] Riess A G, Casertano S, Yuan W, Bowers J B, Macri L, Zinn J C and Scolnic D 2021 Cosmic distances calibrated to 1% precision with Gaia EDR3 parallaxes and hubble space telescope photometry of 75 milky way cepheids confirm tension with ΛCDM Astrophys. J. Lett. 908 L6 [720] Riess A G et al 2021 A comprehensive measurement of the local value of the hubble constant with 1 km/s/Mpc uncertainty from the hubble space telescope and the SH0ES team (arXiv:2112.04510 [astro-ph.CO]) [721] Di Valentino E, Mena O, Pan S, Visinelli L, Yang W, Melchiorri A, Mota D F, Riess A G and Silk J 2021 In the realm of the hubble tension − a review of solutions (arXiv:2103.01183 [astro-ph.CO]) [722] Baudis L 2016 Dark matter detection J. Phys. G: Nucl. Part. Phys. 43 044001 [723] Bertone G, Hooper D and Silk J 2005 Particle dark matter: evidence, candidates and constraints Phys. Rep. 405 279–390 [724] Peebles P J E and Ratra B 2003 The cosmological constant and dark energy Rev. Mod. Phys. 75 559–606 [725] Schumann M 2019 Direct detection of WIMP dark matter: concepts and status J. Phys. G: Nucl. Part. Phys. 46 103003 [726] Billard J et al 2021 Direct detection of dark matter—APPEC Committee report (arXiv:2104.07634 [hep-ex]) [727] Wong K C et al 2020 H0LiCOW—XIII. A 2.4 per cent measurement of H 0 from lensed quasars: 5.3σ tension between early- and late-universe probes Mon. Not. R. Astron. Soc. 498 1420–39 [728] Breuval L et al 2020 The Milky Way Cepheid Leavitt law based on Gaia DR2 parallaxes of companion stars and host open cluster populations Astron. Astrophys. 643 A115 [729] Reid M, Pesce D and Riess A 2019 An improved distance to NGC 4258 and its implications for the Hubble constant Astrophys. J. Lett. 886 27 [730] Kourkchi E, Tully R B, Anand G S, Courtois H M, Dupuy A, Neill J D, Rizzi L and Seibert M 2020 Cosmicflows-4: the calibration of optical and infrared Tully–Fisher relations Astrophys. J. 896 3 [731] Schombert J, McGaugh S and Lelli F 2020 Using the baryonic Tully–Fisher relation to measure H 0 Astron. J. 160 71 [732] de Jaeger T, Stahl B E, Zheng W, Filippenko A V, Riess A G and Galbany L 2020 A measurement of the Hubble constant from Type II supernovae Mon. Not. R. Astron. Soc. 496 3402–11 [733] Shajib A J et al 2020 STRIDES: a 3.9 per cent measurement of the Hubble constant from the strong lens system DES J0408–5354 Mon. Not. R. Astron. Soc. 494 6072–102 [734] Blakeslee J P, Jensen J B, Ma C-P, Milne P A and Greene J E 2021 The Hubble constant from infrared surface brightness fluctuation distances Astrophys. J. 911 65 [735] Ade P et al (Planck Collaboration) 2016 Planck 2015 results. XIII. cosmological parameters Astron. Astrophys. 594 A13 [736] Dutcher D et al 2021 Measurements of the E-mode polarization and temperature-E-mode correlation of the CMB from SPT-3G 2018 data (arXiv:2101.01684 [astro-ph.CO]) [737] Freedman W L, Madore B F, Hoyt T, Jang I S, Beaton R, Lee M G, Monson A, Neeley J and Rich J 2020 Calibration of the tip of the red giant branch (TRGB) Astrophys. J. 891 57 [738] Birrer S et al 2020 TDCOSMO - IV. Hierarchical time-delay cosmography—joint inference of the Hubble constant and galaxy density profiles Astron. Astrophys. 643 A165 [739] Graef L L, Benetti M and Alcaniz J S 2019 Primordial gravitational waves and the h0 -tension problem Phys. Rev. D 99 043519 [740] The LIGO Scientific Collaboration and The Virgo Collaboration, The 1M2H Collaboration, The Dark Energy Camera GW-EM Collaboration and the DES Collaboration, The DLT40 Collaboration, The Las Cumbres Observatory Collaboration, The VINROUGE Collaboration and The MASTER Collaboration 2017 A gravitational-wave standard siren measurement of the hubble constant Nature 551 85–88 [741] Maggiore M et al 2020 Science case for the Einstein telescope J. Cosmol. Astropart. Phys. JCAP03(2020)050 [742] Baker J et al 2019 The laser interferometer space antenna: unveiling the millihertz gravitational wave sky (arXiv:1907.06482) [743] Amaro-Seoane P et al (LISA Collaboration) 2017 Laser interferometer space antenna (arXiv:1702.00786 [astro-ph.IM]) [744] Abdalla E et al 2022 Cosmology intertwined: a review of the particle physics, astrophysics, and cosmology associated with the cosmological tensions and anomalies (arXiv:2203.06142 [astro-ph.CO]) [745] Calabrese E, Slosar A, Melchiorri A, Smoot G F and Zahn O 2008 Cosmic microwave weak lensing data as a test for the dark universe Phys. Rev. D 77 123531 [746] Joudaki S et al 2017 KiDS-450: testing extensions to the standard cosmological model Mon. Not. R. Astron. Soc. 471 1259–79 [747] Hildebrandt H et al 2020 KiDS+VIKING-450: cosmic shear tomography with optical and infrared data Astron. Astrophys. 633 A69 [748] Asgari M et al 2021 KiDS-1000 cosmology: cosmic shear constraints and comparison between two point statistics Astron. Astrophys. 645 A104 [749] Abbott T M C et al (DES Collaboration) 2018 Dark energy survey year 1 results: cosmological constraints from galaxy clustering and weak lensing Phys. Rev. D 98 043526 [750] Abbott T M C et al (DES Collaboration) 2021 Dark energy survey year 3 results: cosmological constraints from galaxy clustering and weak lensing (arXiv:2105.13549 [astro-ph.CO]) [751] Nunes R C and Vagnozzi S 2021 Arbitrating the S8 discrepancy with growth rate measurements from 203 Rep. Prog. Phys. 86 (2023) 026901 Review Redshift-Space Distortions Mon. Not. R. Astron. Soc. 505 5427 [752] Cuceu A, Farr J, Lemos P and Font-Ribera A 2019 Baryon acoustic oscillations and the Hubble constant: past, present and future J. Cosmol. Astropart. Phys. JCAP10(2019)044 [753] Lombriser L 2020 Consistency of the local Hubble constant with the cosmic microwave background Phys. Lett. B 803 135303 [754] Lombriser L 2021 Dark energy survey year 3 results: cosmology from cosmic shear and robustness to data calibration (arXiv:2105.13543 [astro-ph.CO]) [755] Lombriser L 2021 Dark energy survey year 3 results: constraints on cosmological parameters and galaxy bias models from galaxy clustering and galaxy-galaxy lensing using the redMaGiC sample (arXiv:2105.13545 [astro-ph.CO]) [756] Di Valentino E et al 2021 Snowmass2021—letter of interest cosmology intertwined I: perspectives for the next decade Astropart. Phys. 131 102606 [757] Di Valentino E et al 2021 Snowmass2021—letter of interest cosmology intertwined II: the hubble constant tension Astropart. Phys. 131 102605 [758] Di Valentino E et al 2021 Cosmology intertwined III: fσ8 and S8 Astropart. Phys. 131 102604 [759] Di Valentino E et al 2021 Snowmass2021—letter of interest cosmology intertwined IV: the age of the universe and its curvature Astropart. Phys. 131 102607 [760] Naess S et al 2020 The Atacama cosmology telescope: arcminute-resolution maps of 18 000 square degrees of the microwave sky from ACT 2008–2018 data combined with Planck J. Cosmol. Astropart. Phys. JCAP12(2020)046 [761] Escamilla-Rivera C 2021 A dynamical dark energy solution to Hubble–Lemaı̂tre tension in the light of the multimessenger era (arXiv:2103.01318 [astro-ph.CO]) [762] Heymans C et al 2013 CFHTLenS tomographic weak lensing cosmological parameter constraints: mitigating the impact of intrinsic galaxy alignments Mon. Not. R. Astron. Soc. 432 2433 [763] Choi A, Heymans C, Blake C, Hildebrandt H, Duncan C A J, Erben T, Nakajima R, Van Waerbeke L and Viola M 2016 CFHTLenS and RCSLenS: testing photometric redshift distributions using angular cross-correlations with spectroscopic galaxy surveys Mon. Not. R. Astron. Soc. 463 3737–54 [764] Hildebrandt H et al 2017 KiDS-450: Cosmological parameter constraints from tomographic weak gravitational lensing Mon. Not. R. Astron. Soc. 465 1454 [765] Heymans C et al 2021 KiDS-1000 cosmology: multi-probe weak gravitational lensing and spectroscopic galaxy clustering constraints Astron. Astrophys. 646 A140 [766] van Uitert E et al 2018 KiDS+GAMA: cosmology constraints from a joint analysis of cosmic shear, galaxy–galaxy lensing and angular clustering Mon. Not. R. Astron. Soc. 476 4662–89 [767] Feigelson E D et al 2011 X-ray star clusters in the carina complex Astrophys. J. Suppl. 194 9 [768] Böhringer H, Chon G and Collins C A 2014 The extended ROSAT-ESO flux limited x-ray galaxy cluster survey (REFLEX II) IV. X-ray luminosity function and first constraints on cosmological parameters Astron. Astrophys. 570 A31 [769] von der Linden A et al 2014 Robust weak-lensing mass calibration of Planck galaxy clusters Mon. Not. R. Astron. Soc. 443 1973–8 [770] Böhringer H, Chon G and Kronberg P P 2016 The cosmic large-scale structure in x-rays (CLASSIX) cluster survey I: probing galaxy cluster magnetic fields with line of sight rotation measures Astron. Astrophys. 596 A22 [771] Hilton M et al 2018 The Atacama cosmology telescope: the two-season ACTPol Sunyaev–Zel’dovich effect selected cluster catalog Astrophys. J. Suppl. 235 20 [772] Bleem L E et al 2015 Galaxy clusters discovered via the Sunyaev-Zel’dovich effect in the 2500-square-degree SPT-SZ survey Astrophys. J. Suppl. 216 27 [773] Verde L, Treu T and Riess A 2019 Tensions between the early and the late universe Nat. Astron. 3 891 [774] Perivolaropoulos L and Skara F 2021 Challenges for ΛCDM: an update (arXiv:2105.05208 [astro-ph.CO]) [775] Escamilla-Rivera C 2021 Precision cosmology in modified and extended theories of gravity: an insightful test Astron. Nachr. 342 63–68 [776] Briffa R, Capozziello S, Levi Said J, Mifsud J and Saridakis E N 2020 Constraining teleparallel gravity through Gaussian processes Class. Quantum Grav. 38 055007 [777] Weitzenböck R 1923 Invariantentheorie (Gronningen: Noordhoff) [778] Jiménez Cruz N M and Escamilla-Rivera C 2021 Late-time and Big Bang Nucleosynthesis constraints for generic modified gravity surveys Eur. Phys. J. Plus 136 51 [779] Aver E, Olive K A and Skillman E D 2015 The effects of He I λ10830 on helium abundance determinations J. Cosmol. Astropart. Phys. JCAP07(2015)011 [780] Abbott R et al 2021 Tests of general relativity with binary black holes from the second LIGO-Virgo gravitational-wave transient catalog Phys. Rev. D 103 122002 [781] Schutz B F 1986 Determining the hubble constant from gravitational wave observations Nature 323 310–1 [782] Abbott R et al 2021 Constraints on the cosmic expansion history from GWTC-3 (arXiv:2111.03604 [astro-ph.CO]) [783] Pardo K, Fishbach M, Holz D E and Spergel D N 2018 Limits on the number of spacetime dimensions from GW170817 J. Cosmol. Astropart. Phys. JCAP07(2018)048 [784] Corman M, Escamilla-Rivera C and Hendry M A 2021 Constraining extra dimensions on cosmological scales with LISA future gravitational wave siren data J. Cosmol. Astropart. Phys. JCAP02(2021)005 [785] Nunes R C, Alves M E S and de Araujo J C N 2019 Forecast constraints on f (T) gravity with gravitational waves from compact binary coalescences Phys. Rev. D 100 064012 [786] Demianski M, Piedipalumbo E, Sawant D and Amati L 2017 Cosmology with gamma-ray bursts: II Cosmography challenges and cosmological scenarios for the accelerated Universe Astron. Astrophys. 598 A113 [787] de la Cruz-Dombriz A 2016 Limitations of cosmography in extended theories of gravity Proc. Sci. 268 007 [788] Luongo O, Pisani G B and Troisi A 2016 Cosmological degeneracy versus cosmography: a cosmographic dark energy model Int. J. Mod. Phys. D 26 1750015 [789] Dunsby P K S and Luongo O 2016 On the theory and applications of modern cosmography Int. J. Geom. Methods Mod. Phys. 13 1630002 [790] Escamilla-Rivera C and Capozziello S 2019 Unveiling cosmography from the dark energy equation of state Int. J. Mod. Phys. D 28 1950154 [791] Capozziello S, D’Agostino R and Luongo O 2019 Extended gravity cosmography Int. J. Mod. Phys. D 28 1930016 [792] Munõz C Z and Escamilla-Rivera C 2020 Inverse cosmography: testing the effectiveness of cosmographic polynomials using machine learning J. Cosmol. Astropart. Phys. JCAP12(2020)007 [793] Escamilla-Rivera C, Quintero M A C and Capozziello S 2020 A deep learning approach to cosmological dark energy models J. Cosmol. Astropart. Phys. JCAP03(2020)008 204 Rep. Prog. Phys. 86 (2023) 026901 Review [794] Bernardo R C, Grandón D, Said Levi J and Cárdenas V H 2022 Parametric and nonparametric methods hint dark energy evolution Phys. Dark Universe 36 101017 [795] Dialektopoulos K, Said J L, Mifsud J, Sultana J and Adami K Z 2022 Neural network reconstruction of late-time cosmology and null tests J. Cosmol. Astropart. Phys. JCAP02(2022)023 [796] Bernardo R C and Levi Said J 2021 Towards a model-independent reconstruction approach for late-time Hubble data J. Cosmol. Astropart. Phys. JCAP08(2021)027 [797] Betoule M et al 2014 Improved cosmological constraints from a joint analysis of the sdss-ii and snls supernova samples Astron. Astrophys. 568 A22 [798] Supernova Cosmology Project T 2012 The Hubble space telescope cluster supernova survey. V. Improving the dark-energy constraints above z > 1 and building an early-type-hosted supernova sample Astrophys. J. 746 85 [799] Supernova Cosmology Project T 2014 Improved cosmological constraints from a joint analysis of the SDSS-II and SNLS supernova samples Astron. Astrophys. 568 A22 [800] Supernova Cosmology Project T 2017 The clustering of galaxies in the completed SDSS-III Baryon oscillation spectroscopic survey: tomographic BAO analysis of DR12 combined sample in Fourier space Mon. Not. R. Astron. Soc. 466 762–79 [801] Archidiacono M, Hooper D C, Murgia R, Bohr S, Lesgourgues J and Viel M 2019 Constraining dark matter-dark radiation interactions with CMB, BAO and Lyman-α J. Cosmol. Astropart. Phys. JCAP10(2019)055 [802] Moresco M, Pozzetti L, Cimatti A, Jimenez R, Maraston C, Verde L, Thomas D, Citro A, Tojeiro R and Wilkinson D 2016 A 6% measurement of the Hubble parameter at z ∼ 0.45: direct evidence of the epoch of cosmic re-acceleration J. Cosmol. Astropart. Phys. JCAP1605(2016)014 [803] Shaikh A, Shaikh A and Wankhade K 2020 Panoroma behaviour of domain walls cosmological models in teleparallel gravity (arXiv:2006.12300 [gr-qc]) [804] Sahlu S, Ntahompagaze J, Elmardi M and Abebe A 2019 The Chaplygin gas as a model for modified teleparallel gravity? Eur. Phys. J. C 79 749 [805] Mandal S and Sahoo P K 2020 A complete cosmological scenario in teleparallel gravity Eur. Phys. J. Plus 135 706 [806] Li C, Cai Y, Cai Y-F and Saridakis E N 2018 The effective field theory approach of teleparallel gravity, f (T) gravity and beyond J. Cosmol. Astropart. Phys. JCAP10(2018)001 [807] D’Agostino R and Nunes R C 2020 Measurements of H 0 in modified gravity theories: the role of lensed quasars in the late-time universe Phys. Rev. D 101 103505 [808] Karimzadeh S and Shojaee R 2019 Phantom-like behavior in modified teleparallel gravity Adv. High Energy Phys. 2019 4026856 [809] Capozziello S, Lambiase G and Saridakis E N 2017 Constraining f (T) teleparallel gravity by big bang nucleosynthesis Eur. Phys. J. C 77 576 [810] Capozziello S, D’Agostino R and Luongo O 2017 Model-independent reconstruction of f (T) teleparallel cosmology Gen. Relativ. Gravit. 49 141 [811] Benetti M, Capozziello S and Lambiase G 2020 Updating constraints on f (T) teleparallel cosmology and the consistency with big bang nucleosynthesis Mon. Not. R. Astron. Soc. 500 1795–805 [812] Bhattacharjee S 2020 Gravitational baryogenesis in extended teleparallel theories of gravity Phys. Dark Universe 30 100612 [813] Geng C-Q, Gu J-A and Lee C-C 2013 Singularity problem in teleparallel dark energy models Phys. Rev. D 88 024030 [814] Gu J-A, Lee C-C and Geng C-Q 2013 Teleparallel dark energy with purely non-minimal coupling to gravity Phys. Lett. B 718 722–6 [815] Geng C-Q, Lee C-C and Saridakis E N 2012 Observational constraints on teleparallel dark energy J. Cosmol. Astropart. Phys. JCAP01(2012)002 [816] Xu B, Yu H and Wu P 2018 Testing viable f (T) models with current observations Astrophys. J. 855 89 [817] de Santos F B M, Gonzalez J E and Silva R 2021 Observational constraints on f (T) gravity from model-independent data (arXiv:2112.15249 [astro-ph.CO]) [818] Lewis A and Bridle S 2002 Cosmological parameters from CMB and other data: a Monte Carlo approach Phys. Rev. D 66 103511 [819] Lewis A 2013 Efficient sampling of fast and slow cosmological parameters Phys. Rev. D 87 103529 [820] Lewis A, Challinor A and Lasenby A 2000 Efficient computation of CMB anisotropies in closed FRW models Astrophys. J. 538 473–6 [821] Huang P and Yuan F-F 2016 Generalized Hodge dual for torsion in teleparallel gravity Int. J. Geom. Methods Mod. Phys. 13 1650106 [822] Farooq O, Madiyar F R, Crandall S and Ratra B 2017 Hubble parameter measurement constraints on the redshift of the deceleration–acceleration transition, dynamical dark energy and space curvature Astrophys. J. 835 26 [823] Riess A G et al 2016 A 2.4% determination of the local value of the Hubble constant Astrophys. J. 826 56 [824] Efstathiou G 2021 To H0 or not to H0? Mon. Not. R. Astron. Soc. 505 3866–72 [825] Linder E V 2009 Exponential gravity Phys. Rev. D 80 123528 [826] Geng J-J, Guo R-Y, He D-Z, Zhang J-F and Zhang X 2015 Redshift drift constraints on f (T) gravity Front. Phys. 10 109501 [827] Briffa R, Escamilla-Rivera C, Said J L, Mifsud J and Pullicino N L 2021 Impact of H 0 priors on f (T) late time cosmology (arXiv:2108.03853 [astro-ph.CO]) [828] Hashim M, El Hanafy W, Golovnev A and El-Zant A A 2020 Toward a concordance teleparallel cosmology I: background dynamics (arXiv:2010.14964 [astro-ph.CO]) [829] Hashim M, El-Zant A A, El Hanafy W and Golovnev A 2021 Toward a concordance teleparallel cosmology II: linear perturbation (arXiv:2104.08311 [astro-ph.CO]) [830] Faraoni V and Capozziello S 2011 Beyond Einstein Gravity: A Survey of Gravitational Theories for Cosmology and Astrophysics (Dordrecht: Springer) [831] Capozziello S, Cardone V F, Farajollahi H and Ravanpak A 2011 Cosmography in f (T)-gravity Phys. Rev. D 84 043527 [832] Murphy M T, Flambaum V V, Webb J K, Dzuba V V, Prochaska J X and Wolfe A M 2004 Constraining variations in the fine–structure constant, quark masses and the strong interaction Astrophysics, Clocks and Fundamental Constants (Lecture Notes in Physics vol 648) (Berlin: Springer) pp 131–50 [833] King J A, Webb J K, Murphy M T, Flambaum V V, Carswell R F, Bainbridge M B, Wilczynska M R and Koch F E 2012 Spatial variation in the fine-structure constant—new results from VLT/UVES Mon. Not. R. Astron. Soc. 422 3370–413 [834] Agafonova I I, Molaro P, Levshakov S A and Hou J L 2011 First measurement of Mg isotope abundances at high redshifts and accurate estimate of Delta alpha/alpha Astron. Astrophys. 529 A28 [835] Molaro P et al 2013 The UVES large program for testing fundamental physics: I. Bounds on a change in alpha towards quasar HE 2217–2818 Astron. Astrophys. 555 A68 205 Rep. Prog. Phys. 86 (2023) 026901 Review [836] Songaila A and Cowie L L 2014 Constraining the variation of the fine structure constant with observations of narrow quasar absorption lines Astrophys. J. 793 103 [837] Evans T M et al 2014 The UVES large program for testing fundamental physics—III. Constraints on the fine-structure constant from three telescopes Mon. Not. R. Astron. Soc. 445 128–50 [838] Kotuš S M, Murphy M T and Carswell R F 2017 High-precision limit on variation in the fine-structure constant from a single quasar absorption system Mon. Not. R. Astron. Soc. 464 3679–703 [839] Murphy M T, Malec A L and Prochaska J X 2016 Precise limits on cosmological variability of the fine-structure constant with zinc and chromium quasar absorption lines Mon. Not. R. Astron. Soc. 461 2461–79 [840] Martins C J A P 2017 The status of varying constants: a review of the physics, searches and implications Rep. Prog. Phys. 80 126902 [841] Bainbridge M B and Webb J K 2017 Artificial intelligence applied to the automatic analysis of absorption spectra. objective measurement of the fine structure constant Mon. Not. R. Astron. Soc. 468 1639–70 [842] Alves C S, Leite A C O, Martins C J A P, Silva T A, Berge S A and Silva B S A 2018 Current and future constraints on extended Bekenstein–type models for a varying fine–structure constant Phys. Rev. D 97 023522 [843] Petrov Y V, Nazarov A I, Onegin M S, Petrov V Y and Sakhnovsky E G 2006 Natural nuclear reactor Oklo and variation of fundamental constants. Part 1. Computation of neutronic of fresh core Phys. Rev. C 74 064610 [844] Uzan J-P 2011 Varying constants, gravitation and cosmology Living Rev. Relativ. 14 2 [845] Uzan J-P 2003 The fundamental constants and their variation: observational status and theoretical motivations Rev. Mod. Phys. 75 403 [846] Hees A, Minazzoli O and Larena J 2014 Breaking of the equivalence principle in the electromagnetic sector and its cosmological signatures Phys. Rev. D 90 124064 [847] Flowers J L and Petley B W 2001 Progress in our knowledge of the fundamental constants of physics Rep. Prog. Phys. 64 1191–246 [848] Dirac P A M 1937 The cosmological constants Nature 139 323 [849] Dirac P A M 1938 New basis for cosmology Proc. R. Soc. A 165 199–208 [850] Jordan P 1937 Die physikalischen weltkonstanten Naturwissenschaften 25 513–17 [851] Fierz M 1956 On the physical interpretation of P. Jordan’s extended theory of gravitation Helv. Phys. Acta 29 128–34 [852] Barrow J D 1999 Cosmologies with varying light speed Phys. Rev. D 59 043515 [853] Bekenstein J D 1982 Fine structure constant: is it really a constant? Phys. Rev. D 25 1527–39 [854] Livio M and Stiavelli M 1998 Does the fine structure constant really vary in time? Astrophys. J. Lett. 507 L13 [855] Sandvik H B, Barrow J D and Magueijo J 2002 A simple cosmology with a varying fine structure constant Phys. Rev. Lett. 88 031302 [856] Barrow J D, Sandvik H B and Magueijo J 2002 The behavior of varying alpha cosmologies Phys. Rev. D 65 063504 [857] Barrow J D, Magueijo J and Sandvik H B 2002 Variations of alpha in space and time Phys. Rev. D 66 043515 [858] Barrow J D and Mota D F 2002 Qualitative analysis of universes with varying alpha Class. Quantum Grav. 19 6197–212 [859] Barrow J D and Lip S Z W 2012 A generalized theory of varying alpha Phys. Rev. D 85 023514 [860] Barrow J D and Graham A A H 2013 General dynamics of varying-alpha universes Phys. Rev. D 88 103513 [861] Damour T, Piazza F and Veneziano G 2002 Runaway dilaton and equivalence principle violations Phys. Rev. Lett. 89 081601 [862] Damour T, Piazza F and Veneziano G 2002 Violations of the equivalence principle in a dilaton runaway scenario Phys. Rev. D 66 046007 [863] Olive K A and Pospelov M 2002 Evolution of the fine structure constant driven by dark matter and the cosmological constant Phys. Rev. D 65 085044 [864] van de Bruck C, Mifsud J and Nunes N J 2015 The variation of the fine-structure constant from disformal couplings J. Cosmol. Astropart. Phys. JCAP12(2015)018 [865] Hart L and Chluba J 2020 Updated fundamental constant constraints from Planck 2018 data and possible relations to the Hubble tension Mon. Not. R. Astron. Soc. 493 3255–63 [866] Brax P, Davis A-C, Li B and Winther H A 2012 A unified description of screened modified gravity Phys. Rev. D 86 044015 [867] Minazzoli O and Hees A 2014 Late-time cosmology of a scalar-tensor theory with a universal multiplicative coupling between the scalar field and the matter Lagrangian Phys. Rev. D 90 023017 [868] Nunes R C, Bonilla A, Pan S and Saridakis E N 2017 Observational constraints on f (T) gravity from varying fundamental constants Eur. Phys. J. C 77 230 [869] Levi Said J, Mifsud J, Parkinson D, Saridakis E N, Sultana J and Adami K Z 2020 Testing the violation of the equivalence principle in the electromagnetic sector and its consequences in f (T) gravity J. Cosmol. Astropart. Phys. JCAP11(2020)047 [870] Scolnic D M et al 2018 The complete light-curve sample of spectroscopically confirmed SNe Ia from Pan-STARRS1 and cosmological constraints from the combined pantheon sample Astrophys. J. 859 101 [871] Moresco M et al 2012 Improved constraints on the expansion rate of the Universe up to z ∼ 1.1 from the spectroscopic evolution of cosmic chronometers J. Cosmol. Astropart. Phys. JCAP08(2012)006 [872] Simon J, Verde L and Jimenez R 2005 Constraints on the redshift dependence of the dark energy potential Phys. Rev. D 71 123001 [873] Stern D, Jimenez R, Verde L, Kamionkowski M and Stanford S 2010 Cosmic chronometers: constraining the equation of state of dark energy. I: H(z) measurements J. Cosmol. Astropart. Phys. JCAP02(2010)008 [874] Zhang C, Zhang H, Yuan S, Zhang T-J and Sun Y-C 2014 Four new observational H(z) data from luminous red galaxies in the Sloan Digital Sky Survey data release seven Res. Astron. Astrophys. 14 1221–33 [875] Moresco M 2015 Raising the bar: new constraints on the Hubble parameter with cosmic chronometers at zsim2 Mon. Not. R. Astron. Soc. 450 L16–L20 [876] Nesseris S, Basilakos S, Saridakis E N and Perivolaropoulos L 2013 Viable f (T) models are practically indistinguishable from ΛCDM Phys. Rev. D 88 103010 [877] Qi J-Z, Cao S, Biesiada M, Zheng X and Zhu H 2017 New observational constraints on f (T) cosmology from radio quasars Eur. Phys. J. C 77 502 [878] Anagnostopoulos F K, Basilakos S and Saridakis E N 2019 Bayesian analysis of f (T) gravity using fsigma8 data Phys. Rev. D 100 083517 [879] Nunes R C, Pan S and Saridakis E N 2016 New observational constraints on f (T) gravity from cosmic chronometers J. Cosmol. Astropart. Phys. JCAP08(2016)011 [880] Singh C, Ha W, Lanusse F, Boehm V, Liu J and Yu B 2020 Transformation importance with applications to cosmology (arXiv:2003.01926 [stat.ML]) 206 Rep. Prog. Phys. 86 (2023) 026901 Review [881] Riess A G et al 2018 Type Ia supernova distances at redshift >1.5 from the hubble space telescope multi-cycle treasury programs: the early expansion rate Astrophys. J. 853 126 [882] Riess A G et al 2017 The clustering of galaxies in the completed SDSS-III Baryon oscillation spectroscopic survey: cosmological analysis of the DR12 galaxy sample Mon. Not. R. Astron. Soc. 470 2617–52 [883] du Mas des Bourboux H et al 2017 Baryon acoustic oscillations from the complete SDSS-III Lyα-quasar cross-correlation function at z = 2.4 Astron. Astrophys. 608 A130 [884] Zhao G-B et al 2019 The clustering of the SDSS-IV extended Baryon Oscillation Spectroscopic Survey DR14 quasar sample: a tomographic measurement of cosmic structure growth and expansion rate based on optimal redshift weights Mon. Not. R. Astron. Soc. 482 3497–3513 [885] Yahya S, Seikel M, Clarkson C, Maartens R and Smith M 2014 Null tests of the cosmological constant using supernovae Phys. Rev. D 89 023503 [886] Zunckel C and Clarkson C 2008 Consistency tests for the cosmological constant Phys. Rev. Lett. 101 181301 [887] Ren X, Yan S-F, Zhao Y, Cai Y-F and Saridakis E N 2022 Gaussian processes and effective field theory of f (T) gravity under the H 0 tension (arXiv:2203.01926 [astro-ph.CO]) [888] Riess A G, Yuan W, Casertano S, Macri L M and Scolnic D 2020 The accuracy of the Hubble constant measurement verified through cepheid amplitudes Astrophys. J. Lett. 896 L43 [889] Birrer S and Treu T 2020 TDCOSMO V: strategies for precise and accurate measurements of the Hubble constant with strong lensing (arXiv:2008.06157 [astro-ph.CO]) [890] Holz D E and Hughes S A 2005 Using gravitational–wave standard sirens Astrophys. J. 629 15–22 [891] Chen H-Y, Fishbach M and Holz D E 2018 A two per cent Hubble constant measurement from standard sirens within five years Nature 562 545–7 [892] Di Valentino E, Holz D E, Melchiorri A and Renzi F 2018 The cosmological impact of future constraints on H 0 from gravitational-wave standard sirens Phys. Rev. D 98 083523 [893] Palmese A et al 2019 Gravitational wave cosmology and astrophysics with large spectroscopic galaxy surveys Bull. Am. Astron. Soc. 51 3 [894] Abazajian K et al 2019 CMB-S4 science case, reference design, and project plan (arXiv:1907.04473 [astro-ph.IM]) [895] Abazajian K et al 2019 CMB-S4 decadal survey APC white paper Bull. Am. Astron. Soc. 51 209 [896] Koopman B J et al 2020 The Simons observatory: overview of data acquisition, control, monitoring and computer infrastructure Astronomy 6 1145208 [897] Suzuki A, Ade P A R, Akiba Y, Alonso D, Arnold K, Aumont J, Baccigalupi C, Barron D, Basak S, Beckman S and et al 2018 The litebird satellite mission: sub-Kelvin instrument J. Low Temp. Phys. 193 1048–56 [898] Di Valentino E et al (CORE Collaboration) 2018 Exploring cosmic origins with CORE: cosmological parameters J. Cosmol. Astropart. Phys. JCAP04(2018)017 [899] Kogut A, Fixsen D, Chuss D, Dotson J, Dwek E, Halpern M, Hinshaw G, Meyer S, Moseley S and Seiffert M et al 2011 The primordial inflation explorer (pixie): a nulling polarimeter for cosmic microwave background observations J. Cosmol. Astropart. Phys. JCAP2011(2011)025 [900] Delabrouille J et al 2019 Microwave spectro-polarimetry of matter and radiation across space and time (arXiv:1909.01591 [astro-ph.CO]) [901] Hanany S et al 2019 PICO: probe of inflation and cosmic origins (arXiv:1902.10541 [astro-ph.IM]) [902] Laureijs R et al 2011 Euclid definition study report (arXiv:1110.3193 [astro-ph.CO]) [903] Laureijs R et al 2019 LSST: from science drivers to reference design and anticipated data products Astrophys. J. 873 111 [904] Akeson R et al 2019 The wide field infrared survey telescope: 100 hubbles for the 2020s (arXiv:1902.05569 [astro-ph.IM]) [905] Levi M et al (The DESI collaboration) 2013 The DESI experiment, a whitepaper for Snowmass 2013 (arXiv:1308.0847 [astro-ph.CO]) [906] Wuensche C et al 2020 Baryon acoustic oscillations from integrated neutral gas observations: broadband corrugated horn construction and testing Exp. Astron. 50 125–44 207