f(T) Gravity: A Torsion-Based Approach

• f(T) gravity is a modified teleparallel theory that generalizes general relativity by replacing the linear function of torsion T with a nonlinear function, maintaining second-order field equations.
• It alters the coupling between matter and geometry, yielding distinctive predictions in cosmology, structure growth, and astrophysical phenomena such as compact stars and black holes.
• Observational tests, including solar system measurements and cosmological data, help constrain the model parameters, while research continues exploring its degrees of freedom and Lorentz invariance issues.

$f(T)$ gravity is a class of modified teleparallel theories in which the gravitational action is generalized from linear dependence on the torsion scalar $T$—the foundation of the Teleparallel Equivalent of General Relativity (TEGR)—to a nonlinear function $f(T)$. Unlike $f(R)$ extensions of metric gravity, $f(T)$ gravity theories maintain second-order equations of motion and directly modify the coupling between matter and geometry via torsion rather than curvature. This framework is actively investigated for its implications for cosmology, astrophysics, strong gravity, and theoretical foundations, including degrees of freedom, quantum gravity prospects, and the status of local Lorentz invariance.

1. Foundations: Torsion, Action, and Field Equations

$f(T)$ gravity is constructed in the tetrad (vierbein) formalism. The tetrad $e^A{}_\mu(x)$, with $A$ the Lorentz frame index and $\mu$ the spacetime index, defines the metric: $g_{\mu\nu} = \eta_{AB} e^A_\mu e^B_\nu,$

where $\eta_{AB}$ is the Minkowski metric. Gravity is encoded not in curvature but in nontrivial torsion, via the Weitzenböck connection,

$$\Gamma^\lambda_{\mu\nu} = e_A^\lambda \partial_\nu e^A_\mu,$$

whose torsion tensor is $$T^\lambda_{\;\mu\nu} = \Gamma^\lambda_{\nu\mu}-\Gamma^\lambda_{\mu\nu}$$. The superpotential $S_\rho^{\;\;\mu\nu}$ and the contorsion $K^{\mu\nu}{}_\rho$ are used to define the torsion scalar

$T \equiv S_\rho^{\;\;\mu\nu} T^\rho_{\;\mu\nu}.$

The $f(T)$ action with matter fields $\psi$ is

$$S = \frac{1}{16\pi G} \int d^4x\,e\, [T + f(T)] + S_m[g_{\mu\nu}, \psi],$$

where $e = \det(e^A_\mu) = \sqrt{-g}$. Varying with respect to $e^A_\mu$ yields second-order field equations: $$(1+f_T) \big[ e^{-1} \partial_\mu (e\, e_A^\rho S_\rho^{\;\;\mu\nu}) - e_A^\lambda T^\rho_{\;\mu\lambda} S_\rho^{\;\;\nu\mu} \big] + f_{TT} e_A^\rho S_\rho^{\;\;\mu\nu} \partial_\mu T - \frac{e_A^\nu}{4} [T+f(T)] = 4\pi G e_A^\rho \Theta_\rho^{\;\;\nu},$$ where $f_T = df/dT$, $f_{TT}=d^2f/dT^2$, and $\Theta$ is the matter energy-momentum tensor. In the limit $f(T)\rightarrow 0$, one recovers algebraic equivalence with GR via TEGR (Zheng et al., 2010, Cai et al., 2015).

2. Cosmological Backgrounds and Effective Dynamics

For a spatially flat FRW universe (tetrad: $\mathrm{diag}(1,a,a,a)$), $T=-6H^2$ and the modified Friedmann equations are

$$H^2 = \frac{8\pi G}{3}\rho_m - \frac{f}{6} - 2H^2 f_T,$$

with effective dark energy

$$\rho_{\mathrm{DE}} = \frac{1}{16\pi G}(-f + 2T f_T),\quad P_{\mathrm{DE}} = \frac{1}{16\pi G}\left[ f - 2T f_T + 4T^2 f_{TT} \right].$$

The effective total equation-of-state parameter is

$$w_{\mathrm{DE}} = -1 + \frac{(f-2Tf_T)(f_T + 2T f_{TT})}{(-f+2Tf_T)(1+f_T+2T f_{TT})}.$$

Power-law models $f(T)=\alpha (-T)^n$ provide cosmologies with late-time acceleration, with parameters fixed by matching the present Hubble rate and matter fraction. These can closely reproduce the background expansion of $\Lambda$CDM or $w$CDM, with distinctive predictions arising in structure growth (Zheng et al., 2010, Cai et al., 2015).

3. Degrees of Freedom and Lorentz Invariance

Hamiltonian analysis reveals $f(T)$ gravity generically propagates $5$ degrees of freedom in $D=4$ (§ 4 below). In teleparallel GR (TEGR), the 6 Lorentz constraints are first class, eliminating unphysical tetrad rotations; for $f(T)\neq T$, these become second class, leading to $3$ extra physical modes which organize as a massive vector or as a massless vector plus a scalar (Li et al., 2011). This breaking of local Lorentz invariance is a structural aspect of pure-tetrad $f(T)$ models; covariant extensions with nontrivial inertial spin connections have been constructed to restore full invariance (DeBenedictis et al., 2016, Cai et al., 2015).

4. Cosmological Perturbations and Structure Growth

Linear scalar perturbations in Newtonian gauge reveal modifications to the growth of matter fluctuations. The comoving over-density $\delta_m$ satisfies

$$\ddot\delta_m + 2H\dot\delta_m - 4\pi G_\mathrm{eff} \rho_m \delta_m = 0, \qquad G_\mathrm{eff} = \frac{G}{1+f_T}.$$

For background-matched power-law $f(T)$ models with $n>0$ and $\alpha>0$, $f_T>0$ implies $G_\mathrm{eff}<G$, suppressing the growth of large-scale structure relative to general relativity. At small redshift, suppression in the linear growth factor can be a few percent for $n\sim0.1$ (Zheng et al., 2010, Chen et al., 2010). Observational structure formation data are therefore crucial to constraining these theories.

Perturbation sectors:

• Scalar: Stable for suitable $f(T)$ with $1+f'_0-12H^2 f''_0 >0$ and $\omega^2\ge0$.
• Vector: Decoupled and decaying as in GR.
• Tensor: Gravitons remain strictly massless; only friction terms affected (Chen et al., 2010, Cai et al., 2015).

5. Astrophysical and Solar System Tests

$f(T)$ gravity supports a rich phenomenology for compact stars, black holes, and strong-field systems. In spherically symmetric settings, the external vacuum metric remains static and is of Schwarzschild–(A)dS form (Birkhoff's theorem) (Meng et al., 2011). However, the Schwarzschild solution is not a generic vacuum solution in all $f(T)$ models, especially covariant versions, and modifications can be significant in the object's interior (DeBenedictis et al., 2016).

For compact stars with $f(T) = T + \xi T^\beta$, the mass-radius relation, maximum mass, and compactness depend sensitively on the sign and magnitude of $\xi$ and the parity of $\beta$ (Araujo et al., 2023). Observational mass and compactness limits for neutron stars (e.g., PSR J0740+6620) set bounds on $(\xi, \beta)$. Solar system PPN-type tests (perihelion precession, light bending, time delay, gravitational redshift) constrain polynomial couplings $|α|$ in $f(T)=T+αT^n$ to be extremely small for $n=2,3$ (Farrugia et al., 2016, DeBenedictis et al., 2016).

System/Test	Constraint Type	Bound Example (n=2)	Reference
Compact stars	$|\xi T_c^{\beta-1}|$	$|\xi| \lesssim 0.01-0.1$	(Araujo et al., 2023)
Mercury perihelion	$|α|$	$|α|\lesssim 10^{-5}$ km$^2$	(Farrugia et al., 2016)
Binary pulsar	$|α|$	$|α|\lesssim 10^{18}$ km$^2$	(DeBenedictis et al., 2016)

6. Phenomenological Generalizations and Extensions

$f(T)$ gravity admits further generalizations such as

• $f(T,\mathcal{T})$ gravity: Coupling the torsion scalar $T$ to the trace of the matter stress tensor $\mathcal{T}$. This yields novel early and late cosmological acceleration phases and modifies energy-momentum conservation, leading to new signatures in the growth of perturbations and gravitational slip (Harko et al., 2014).
• Unimodular $f(T)$ gravity: Imposing $e = \mathrm{const}$ as a constraint introduces a Lagrange multiplier acting as a dynamical cosmological “constant,” affecting the generalized Friedmann equations (Nassur et al., 2016).
• Reconstruction methods and equivalence with k-essence: $f(T)$ cosmologies can be mapped to purely kinetic k-essence models under suitable field redefinitions (Myrzakulov, 2010).

7. Theoretical and Observational Constraints

• Background cosmology: Fits to SNIa+BAO+CMB data can select viable parameter regimes in $f(T)$ models that closely track $\Lambda$CDM at early and moderate redshift, but typically predict deviations (such as transient acceleration or a return to deceleration) at low $z$ (Qi et al., 2014).
• Energy conditions: The strong, null, and weak energy conditions yield nontrivial bounds on $f(T)$ functions, with WEC imposing $\alpha\lesssim 10^{-2}$ for $n=2$ in the power-law $f(T)$ model (Liu et al., 2012).
• Growth of perturbations: The matter power spectrum and redshift-space distortions remain key discriminants; $f_T>0$ models feature systematically slower structure growth for a given expansion history (Zheng et al., 2010, Chen et al., 2010).

8. Open Issues and Research Directions

• Nonlinear and strong-field structure: Full, non-perturbative and covariant black hole solutions and stability remain under exploration (Cai et al., 2015, DeBenedictis et al., 2016).
• Quantum gravity prospects: Second-order nature of field equations and gauge-theoretic structure may offer advantages over $f(R)$ gravity, but the extra degrees of freedom and Lorentz breaking require resolution.
• Cosmological bounces and nonsingular scenarios: $f(T)$ models admit both inflationary phases and nonsingular bounce solutions, subject to observational and perturbative constraints (Cai et al., 2015).
• Compatibility with local Lorentz invariance: Covariant formulations with inertial spin connection enable restoration of Lorentz symmetry at the cost of a more complex formalism.

$f(T)$ gravity occupies a distinctive position in the landscape of modified gravity, offering a second-order, torsion-based alternative to curvature-driven extensions while yielding rich phenomenology at both cosmological and strong-field scales. Continued confrontation with observational tests, especially those sensitive to growth rates and local gravity, as well as theoretical advances in the covariant and quantum regime, are the critical frontiers for this research area (Cai et al., 2015, Li et al., 2011, Zheng et al., 2010, DeBenedictis et al., 2016, Araujo et al., 2023).