Weak-Field Spherically Symmetric Solutions in f(T) gravity Matteo Luca Ruggiero∗ DISAT, Politecnico di Torino, Corso Duca degli Abruzzi 24, Torino, Italy INFN, Sezione di Torino, Via Pietro Giuria 1, Torino, Italy Ninfa Radicella† 5 1 Dipartimento di Fisica E.R. Caianiello, Universita‘ di Salerno, 0 2 Via Giovanni Paolo II 132, Fisciano (Sa), Italy r p INFN, Sezione di Napoli, Gruppo Collegato di Salerno, Napoli, Italy A (Dated: April 22, 2015) 1 2 Abstract ] c We study weak-field solutions having spherical symmetry in f(T) gravity; to this end, we solve q - r thefieldequationsforanondiagonaltetrad, startingfromLagrangian intheformf(T)=T+αTn, g [ where α is a small constant, parameterizing the departure of the theory from GR. We show that 2 v the classical spherically symmetric solutions of GR, i.e. the Schwarzschild and Schwarzschild-de 8 9 Sitter solutions, are perturbed by terms in the form r2−2n and discuss the impact of these 1 ∝ 2 perturbations in observational tests. 0 . 1 0 5 1 : v i X r a ∗Electronic address: [email protected] †Electronic address: [email protected] 1 I. INTRODUCTION Since the very discovery of the accelerated cosmic expansion [1, 2], and its confirmation due to multiple observations [3–5], it has been customary to investigate theories that extend generalrelativity(GR),inordertogetanagreementwiththeobservations, withoutrequiring the existence of dark entities. Hence, motivated by this instance, which recognizes that GR fails in describing gravity at large scales (consider also the old issue of the rotation curves of spiral galaxies [6]), several theories have been proposed to generalize Einstein’s theory. Some of these modified models of gravity are geometric extensions of GR, in other words they are based on a richer geometric structure, which is supposed to give the required ingredients to support the observations. As a prototype of this approach, one can consider the f(R) theories, where the gravita- tional Lagrangiandepends on a function f of the curvature scalar R (see [7, 8] andreferences therein): when f(R) = R the action reduces to the usual Einstein-Hilbert action, and Ein- stein’s theory is obtained. Another example is given by the so-called f(T) theories, which have similarities and differences with respect to f(R). To begin with, they are based on teleparallel gravity (TEGR) [9], where the gravitational interaction is determined by tor- sion, and the torsion scalar T appears in the Lagrangian instead of the curvature scalar. Furthermore, the underlying Riemann-Cartan space-time is endowed with the Weitzenb¨ock connection (instead of the Levi-Civita connection), which is not commutative under the exchange of the lower indices, and has zero curvature while non-zero torsion. Actually, Ein- stein himself proposed such an alternative point of view on gravitation in terms of torsion and tetrads [10]. In fact, in the TEGR picture the tetrads field is promoted to be the dynamical field instead of the metric tensor. In spite of these differences, TEGR and GR have equivalent dynamics: in other words, every solution of GR is also solution of TEGR. However, when TEGR is generalized to f(T) by considering a gravitational Lagrangian that is a function of the torsion scalar, the equivalence breaks down [11, 12]. As a consequence f(T) theories can be considered potential candidates for explaining (on a purely geometric ground) the accelerated expansion of the universe, without requiring the existence of exotic cosmic fluids (see e.g.[13]). While f(R) theories gives fourth order equations (at least in the metric formalism, while they are still second order in the Palatini approach, see e.g. [8]), the f(T) field equations are 2 second order in the field derivatives since the torsion scalar is a function of the square of the firstderivativesofthetetradsfield. Furthermore, asforf(R)theories, thegeneralizedTEGR displays additional degrees of freedom (whose physical nature is still under investigation [14]) related to the fact that the equations of motion are not invariant under local Lorentz transformations[15]. Inparticular, thisimpliestheexistenceofapreferentialglobalreference frame defined by the autoparallel curves of the manifold that solve the equations of motion. Consequently, even though the symmetry can help in choosing suitable coordinates to write the metric in a simple way, this does not give any hint on the form of the tetrad. As discussed in [16], a diagonal tetrad -that could in principle be a good working-ansatz for dealing with diagonal metrics- is not a good choice to properly parallelize the spacetime both in the context of non-flat homogenous and isotropic cosmologies (Friedman-Lemaitre- Robertson-Walker universes) and in spherically symmetric space-times (Schwarzschild or Schwarzschild-de Sitter solutions). The case of the Schwarzschild solution and, more in general, of the spherically symmetric solutions in f(T) gravity, is particularly important, because these solutions, which describe the gravitational field of point-like sources, allow to test f(T) theories at scales different from the cosmological ones, e.g. in the solar system. Such class of solutions -both with diagonal and non-diagonal tetrads- have been received much attention during the last few years, see for instance [17–23]. Indeed, f(T) theories can be used to explain the cosmic acceleration and observations on large scales (e.g. via galaxy clustering and cosmic shear measurements [24]), but we must remember that since GR is in excellent agreement with solar system and binary pulsar observations [25], every theory that aims at explaining the large scale dynamics of the Universe should reproduce GR in a suitable weak-field limit: the same holds true for f(T) theories. Recently, solar system data [26, 27] have been used to constrain f(T) theories; these results are based on the spherical symmetry solution found by Iorio&Saridakis [26], who used a diagonal tetrad. In this paper we follow the approach described in [16] to define a “good tetrad” in f(T) gravity - that is consistent with the equations of motion without constraining the functional form of the Lagrangian - and solve the field equations to obtain weak-field solutions with a power-law ansatz for an additive term to the TEGR Lagrangian, f(T) = T +αTn. This paper is organized as follows: in Section II we review the theoretical framework of f(T) gravity and write the field equations, whose solutions for spherically symmetric 3 space-times, in weak-field approximation, are given in Section III. Eventually, discussion and conclusions are in Sections IV and V. II. f(T) GRAVITY FIELD EQUATIONS Westartbybrieflydiscussing thef(T)gravityframeworkthatleadstothefieldequations. To begin with, we point out that, in this scenario, the metric tensor can be viewed as a subsidiary field, and the vierbein field is the dynamical object whose components in a given coordinate basis ea are related to the metric tensor by µ g (x) = η ea(x)eb(x) , (1) µν ab µ ν where η = diag(1, 1, 1, 1). Notice that latin indexes refer to the tangent space while ab − − − greek indexes label coordinates on the manifold. Hence, the dynamics is obtained by the action1 1 = f(T)ed4x+ , (2) M S 16πG Z S where e = det ea = det(g ) and is the action for the matter fields2. Here f(T) is a µ − µν SM p differentiable function of the torsion scalar T, which is defined as T = Sρ T µν , (3) µν ρ where the contorsion tensor Sρ is defined by µν 1 1 1 Sρ = Tρ T ρ +T ρ + δρT σ δρT σ , (4) µν 4 µν − µν νµ 2 µ σν − 2 ν σµ (cid:0) (cid:1) and the torsion tensor Tλ is µν Tλ = eλ ∂ ea ∂ ea . (5) µν a ν µ − µ ν (cid:0) (cid:1) Varying the action with respect to the vierbein ea(x), one gets the field equations µ 1 e−1∂ (e e ρS µν)f +e λS νµTρ f +e ρS µν∂ (T)f + eνf = 4πGe µ ν, (6) µ a ρ T a ρ µλ T a ρ µ TT 4 a a Tµ 1 We use units such as c=1. 2 Notice that many authors write the gravitational Lagrangian in the form T +f(T), thus denoting the deviation from GR by means of the function f(T): on the contrary, here f(T) is the whole Lagrangian. 4 where ν is the matter energy-momentum tensor and subscripts T denote differentiation Tµ with respect to T. We look for spherically symmetric solutions of the field equations, so we start from the metric ds2 = eA(r)dt2 eB(r)dr2 r2dΩ2 , (7) − − where dΩ2 = dθ2+sin2θdφ2. Due to the lack of Local Lorentz Invariance, tetrads connected by local Lorentz transformations lead to the same metric - i.e. the same causal structure - but different equations of motions, thus physically inequivalent solutions. This means that, even in the spherically symmetric case, for which symmetry helps us in choosing the coordinates and the metric tensor in a simple form, it is quite complicated to do an ansatz for the tetrad field. In particular, for the symmetry and coordinates we are dealing with, it turns out to be a mistake to choose a diagonal form for ea : it does not properly parallelize µ the static spherically symmetric geometry in the context of f(T) gravity. Then, with this caveat in mind, it is possible to derive the field equations for the non diagonal tetrad: eA/2 0 0 0   0 eB/2sinθcosφ eB/2sinθsinφ eB/2cosθ ea = µ    0 rcosθcosφ rcosθsinφ rsinθ   − −     0 rsinθsinφ rsinθcosφ 0   −  following the approach described in [16]: in doing so, the functional form of the Lagrangian and the specific form of the torsion scalar are not constrained a priori. We remark once more that different choices of tetras, while giving back the same metric represent different physical theories. In this work we are interested in a specific tetras that does not lead to a constant torsion scalar. In such a theory Birkhoff theorem does hold, as shown in [16] while the most general vacuum solution is not of Scwharzschild De Sitter kind, as it happens with tetrads for which T = const. Then, it is worthwhile to investigate the features of the spherically symmetric solutions in this case for a generic Lagrangian, that nevertheless should admit the Schwarzschild solution when it reduces to the teleparallel equivalent of GR. The field 5 equations are f(T) e−B(r) f 2 2eB(r) +r2eB(r)T 2rB′(r) + 4 − T 4r2 − − (cid:0) (cid:1) T′(r)e−B(r) f 1+eB(r)/2 = 4πρ (8) TT − r (cid:0) (cid:1) f(T) e−B(r) +f 2 2eB(r) +r2eB(r)T 2rA′(r) = 4πp (9) − 4 T 4r2 − − (cid:0) (cid:1) f 4+4eB(r) 2rA′(r) 2rB′(r)+r2A′(r)2 r2A′(r)B′(r)+2r2A′′(r) + T − − − − (cid:2) (cid:3) +2rf T′ 2+2eB(r)/2 +rA′(r) = 0 (10) TT (cid:0) (cid:1) where ρ, p are the energy density and pressure of the matter energy-momentum tensor, and the prime denotes differentiation with respect to the radial coordinate r. Moreover, the torsion scalar is 2e−B(r)(1+eB(r)/2) T = 1+eB(r)/2 +rA′(r) . (11) r2 (cid:2) (cid:3) III. WEAK-FIELD SOLUTIONS Exact solutions in vacuum (ρ = p = 0) and in presence of a cosmological constant (ρ = p) of the above field equations are thoroughly discussed in [16]; here we are interested − in weak-field solutions with non constant torsion scalar, i.e. T′ = dT/dr = 0. 6 Indeed, for actual physical situations such as in the solar system, the gravitational field is expected to be just a small perturbation of a flat background Minkowski spacetime. As a consequence, we write eA(r) = 1+A(r), eB(r) = 1+B(r) (12) and confine ourselves to linear perturbations. Moreover, we consider Lagrangians of suffi- cient generality, that we write in the form f(T) = T + αTn, where α is a small constant, parameterizing the departure of these theories from GR, and n = 1. | | 6 To begin with, we consider the case n = 2, that has been already analyzed in [26]. From (8-10) we obtain the following solutions α C1 A(r) = 32 (13) − r2 − r α C1 B(r) = 96 + (14) r2 r 6 where C is an integration constant. Then, on setting C = 2M, we get the weak-field limit 1 1 of the Schwarzschild solution plus a correction due to α: 2M α 2M α ds2 = 1 32 dt2 1+ +96 dr2 r2dΩ2 (15) (cid:18) − r − r2(cid:19) −(cid:18) r r2(cid:19) − Eventually, the torsion scalar turns out to be 8 α T(r) = 128 . (16) r2 − r4 These results can be compared to those obtained in [26], where a Lagrangian in the form f(T) = T + αT2 was considered. While to lowest order approximation in both cases the perturbations are proportional to 1/r2, the numerical coefficients are different: this is not surprising, since the authors in [26] solve different field equations. In particular they use a diagonal tetrad, which constrains the torsion scalar to be constant (see e.g. [16] and references therein): however, the solution given in [26] does not seem to have a constant torsion scalar, which makes it inconsistent. Likewise, if we look for solutions of the equations (8)-(10) with ρ = k, p = k, which − corresponds to a cosmological constant, we obtain 2M α 1 2M α 1 ds2 = 1 32 Λr2 dt2 1+ +96 + Λr2 dr2 r2dΩ2 (17) (cid:18) − r − r2 − 3 (cid:19) −(cid:18) r r2 3 (cid:19) − where we set k = Λ and Λ is the cosmological constant. The torsion scalar is the same 8π as (16). So, the weak-field limit of Schwarzschild - de Sitter solution is perturbed by terms that are proportional to α. The previous results can be generalized to the case of a Lagrangian in the form f(T) = T +αTn, and we get C r2−2n 1 A(r) = 1 α 23n−1 Λr2 (18) − r − 2n 3 − 3 − C r2−2n 1 B(r) = 1 +α 23n−1 3n+1+2n2 + Λr2 (19) r 2n 3 − 3 − (cid:0) (cid:1) In particular, if Λ = 0, we obtain vacuum solutions. Notice that, on setting C = 2M, we 1 obtain a weak-field Schwarzschild - de Sitter solution perturbed by terms that are propor- tional to α and decay with a power of the radial coordinate, the specific value depending on the power-law choosed in the Lagrangian. The torsion scalar is 8 T(r) = +2αr−2n23n(n+1) (20) r2 7 while the perturbation terms due to the deviation from GR are in the form A (r) = αa r2−2n, B (r) = αb r2−2n (21) α n α n 23n−1 23n−1 where a = , b = 2n2 3n+1 . A close inspection of the perturbation n n 2n 3 2n 3 − − − (cid:0) (cid:1) terms reveals that they go to zero both when r with n > 1 and when r 0 with → ∞ → n < 1. In the latter case, in order the keep the perturbative approach self-consistent, a maximum value of r must be defined to consider these terms as perturbations of the flat space-time background. We remark here that our linearized approach can be applied to arbitrary polynomial corrections to the torsion scalar: as a consequence, by writing an arbitrary function as a suitable power series, it is possible to evaluate its impact as a perturbation of the weak- field spherically symmetric solution in GR: the n-th term of the series gives a contribution proportional to r2−2n. It could be interesting to test the impact of the perturbations (21). To this end, we remember that it is possible to obtain the secular variations of the Keplerian orbital ele- ments due to general spherically symmetric perturbations of the GR solution, describing the gravitational field around a point-like mass, as one of us showed in [28]. For instance, the average over one orbital period of the secular precession of pericenter turns out to be: 1 23n−1(2n 2)(1 e2)3−2n 5 3 < ω˙ >= α − − F 2 n, n,2,e2 , for n > (22) 4 n a2n (cid:18) − 2 − (cid:19) 2 b 1 23n−1(2 2n)(3 2n)√1 e2 1 1 < ω˙ >= α − − − F n,n ,2,e2 , for n (23) 4 (2n 3)n a2n (cid:18) − 2 (cid:19) ≤ 2 b − In the above equations n , a, e are, respectively, the mean motion, the semi major axis and b the eccentricity of the unperturbed orbit, while F is the hypergeometric function. These relations can be used to constrain the parameters α,n, on the bases of the ephemerides data. IV. DISCUSSION It is useful to comment on the constraints one can infer for the parameters of our model from solar system data. But, before proceeding, it is important to emphasize a point about thetestsoff(T)gravity. Intheorieswithtorsion,thereisasharpdistinctionbetweenthetest 8 particles trajectories: autoparallels, or affine geodesics, are curves along which the velocity vector is transported parallel to itself, by the space-time connection; extremals, or metric geodesics, are curves of extremal space-time interval with respect to the space-time metric [29]. While in GR autoparallels and extremals curves do coincide and we can simply speak of geodesics, the same is not true when torsion is present. So, it is not trivial to define the actual trajectories of test particles. The results obtained by [26] and [27], together with the expressions (22) and(23) ofthesecular precession ofthepericenter, strictly apply tothecase of metric geodesics. According to us, this is a very important issue, that is often neglected in the literature pertaining to theories alternative to GR based to torsion: we will focus on this issue in a forthcoming publication [30]. In the same publication, we will constrain the parameters α and n, taking into account the recent data of the ephemerides of the Solar System provided by INPOP10a [31, 32] and EPM2011 [33–35]. Actually, perturbations in the form of power-law are present in different models of modified gravity, and their impact on the Solar System dynamics has been analyzed, for instance, in [36–38]. Bearing this in mind, it is possible to comment on our results and compare them to those already available in the literature pertaining to f(T) theories. In particular, because of the different choice of the tetrad, our solution, even in the case of a quadratic deformation of the TEGR Lagrangian, differs from the one found by Iorio&Saridakis. Both corrections are proportional to 1/r2, but they have different numerical coefficients. In particular, on substituting n = 2 in Eq. (22), we obtain 16α < ω˙ >= (24) a4n (1 e2) b − On the contrary, the corresponding expression obtained by Iorio&Saridakis [26] is 3α < ω˙ > = (25) IS a4n (1 e2) b − We see that they differ for a factor 16/3: the same happens to the constraints that can be obtained from our solution, by applying the approach described in [26] and [27]. In particular, Iorio&Saridakis [26] derive constraints from the rate of change of perihelia of the first four inner planets, obtaining Λ 6.1 10−42 m−2 | | ≤ × α 1.8 104 m2. | | ≤ × 9 Tighter results have been obtained in a subsequent paper [27], where the authors consider upper bounds deriving from different phenomena: perihelion advance, light bending, gravi- tational time delay [39–42]. But the strongest constraints come from the perihelion advance, in particular from some supplementary advances constructed by considering that the effects due to the Sun’s quadrupole mass moment might represent possible unexplained parts of perihelion advance in GR [31]. This gives Λ 1.8 10−43 m−2 | | ≤ × α 1.2 102 m2. | | ≤ × The upper bound for the solution in Eqs. (17) would be 3/16 smaller, that is α 2.3 | | ≤ × 10 m2. Eventually, we comment on the issue of the parameterized post-Newtonian formalism (PPN), in the framework of f(T) gravity. In order to test theories of gravity that give rise to detectable torsion effects in the Solar System, a theory-independent formalism that generalizes the PPN formalism when torsion is present was developed in [29] (see also [43]): starting from symmetry arguments, the metric and the connection around a massive body are perturbatively expressed in terms of dimensionless parameters related to the matter- energy content of the source, namely its mass and its angular momentum per unit mass. In doing so, the new parameters, which add to the original PPN ones, can be constrained by the experiments. Our results, however, cannot be directly described in this framework: an inspection of our solutions (21) clearly shows that the perturbations are not related to the matter-energy content of the source, rather they depend on α, which parameterizes the departure of the f(T) theory from GR (see e.g. Eq. (4.2) in [43]). So, a new formalism is required to test the content of the Lagrangian by means of observations: in a sense, α can be considered a new post-Newtonian parameter of this formalism. V. CONCLUSIONS We studied spherically symmetric solutions in the weak-approximation of f(T) gravity. In particular, we started from a Lagrangian in the form f(T) = T + αTn, with n = 1, | | 6 whereαisasmallconstantwhichparameterizesthedepartureofthesetheoriesfromGR,and solved thefield equationsusing anondiagonaltetrad, showing that, tolowest approximation 10