Gravitational resonances on $f(T)$-branes

In this work, we investigate the gravitational resonances in various $f(T)$-brane models with the warp factor $\text{e}^{A(y)}=\tanh\big(k(y+b)\big)-\tanh\big(k(y-b)\big)$. For three kinds of $f(T)$, we give the solutions to the system. Besides, we consider the tensor perturbation of vielbein and obtain the effective potentials by the Kaluza-Klein (KK) decomposition. Then, we analyze what kind of effective potential can produce the gravitational resonances. Effects of different parameters on the gravitational resonances are analysed. The lifetimes of the resonances could be long enough as the age of our universe in some ranges of the parameters. This indicates that the gravitational resonances might be considered as one of the candidates of dark matter. Combining the current experimental observations, we constrain the parameters for these brane models.


I. INTRODUCTION
It is known that extra-dimensional theories have been developed for about one hundred years. In 1920s, in order to unify electromagnetic interaction and gravitational interaction, Kaluza and Klein (KK) proposed a fivedimensional spacetime theory by introducing a compact spatial dimension with Planck scale [1,2]. In 1982, Keiichi Akama presented the ¡°braneworld¡± model which picture that we live in a dynamically localized 3-brane in higher-dimensional spacetime [3]. In 1983, Rubakov and Shaposhnikov proposed the domain wall model in a fivedimensional flat spacetime [4]. In this model, the domain wall is generated by a scalar field with a kink configuration and the extra dimension can be infinite. Fermions can be localized on the domain wall by a Yukawa coupling. However, the four-dimensional effective Newtonian gravity cannot be recovered from this model. Then, more than twenty years ago, in order to solve the huge hierarchical problem between the weak and Planck scales, some brane world models were presented. The two famous ones are the large extra dimension model proposed by Arkani-Hamed, Dimopoulos, and Dvali (ADD) [6] and the warped extra dimension model (RS-I) by Randall and Sundrum (RS) [7]. In these brane models, the sizes of extra dimensions are finite. A great development was achieved in Ref. [8], which shows that the four- * liuyx@lzu.edu.cn, corresponding author dimensional gravity can be recovered on the brane even though the extra dimension is infinite. After that, extradimensional theories attracted a lot of interests .
In this paper, we are interested in thick brane models. In most of these models, branes are generated dynamically by one or more background scalar fields [30][31][32][33][34][35]. Various matter fields and gravity in the higherdimensional spacetime should have the ability to explain the physics in the four-dimensional spacetime. Therefore, in order to recover the standard model and the effective four-dimensional Newtonian potential, the zero modes of these matter fields and tensor fluctuations of gravity should be localized on branes [42]. In addition to the zero modes, we will get massive KK modes, which are new particles predicted by these theories. Generally, for the case of a thick brane embedded in five-dimensional asymptotic Anti-de Sitter (AdS) spacetime, the effective potential felt by KK modes along the extra dimension is volcano-like. In this case, the massive KK modes cannot be localized on the brane, but a finite number of massive KK modes could be quasi-localized on the brane [43]. These quasi-localized KK modes are called resonant KK modes. We can judge whether there are resonances by analyzing the shape of the supersymmetric partner potential of the effective potential [44]. In this paper, we focus on the case of gravitational resonant KK modes, which also contribute to the four-dimensional Newtonian potential [45][46][47]. This provides a possible way to detect the extra dimension. In fact, in the Gregory-Rubakov-Sibiryakov (GRS) model, the four-dimensional Newtonian gravitational force is generated by the quasilocalized gravitons [48]. Furthermore, if the lifetime of the resonance can be long enough as the age of our universe, they might be a candidate for dark matter [50]. Therefore, in braneworld models, the investigation of gravitational resonances is an important topic. In other scenarios, there are also gravitational resonances, such as the quasinormal modes of black hole fluctuation theory [51].
It is well known that general relativity (GR) is a theory of gravity with only curvature. In 1928, Einstein established a gravitational theory with only torsion in spacetime in order to unify gravitational interaction and electromagnetic interaction, called Teleparallel equivalence of general relativity (TEGR) [52,53]. It is in fact equivalent to GR, based on the fact that torsion scalar T differs from the Ricci scalar R only by a boundary term. Although the field equation of TEGR is equivalent to GR, the spacetime geometry described by TEGR is different from GR. In TEGR, the dynamic field is vielbein, which is defined in the tangent space at each point in spacetime. Inspired by f (R) gravity theory, Bengochea and Ferraro firstly proposed the generalization of teleparallel gravity, f (T ) gravity (T is the torsion scalar and f (T ) is an arbitrary function of T ), to explain the acceleration of the universe [54]. Note that f (R) gravity is a higher-order theory, while the field equations of f (T ) gravity still remain second order. Subsequently, f (T ) gravity has been widely investigated in cosmology. The cosmological perturbations in f (T ) gravity were investigated in Ref. [55].
In 2012, the thick brane model in f (T ) gravity was firstly constructed in Ref. [63], the thick brane solutions were obtained, and the corresponding localization of fermions was also investigated. After that, using the superpotential method [64], more thick brane solutions in f (T ) theory were obtained [65]. The tensor perturbations of the vielbein of f (T ) brane and the stability of this system were studied in Ref. [66]. It was found that the zero mode of the perturbation is localized on the brane. Then, in 2018, the braneworld model of f (T ) gravity with noncanonical scalar matter field (K-fields) was studied in Ref. [67]. More f (T )-brane related studies can be found in Refs. [68][69][70][71][72]. In this paper we investigate the effects of torsion on thick branes structure and the resonance spectrum of KK gravitons. Based on Ref. [66], we would like to study the gravitational resonances of f (T )-brane. The fluctuation equation of f (T ) gravity was given in Ref. [66]. Note that the dynamical variables are the vielbein fields, but we will focus on the gravitational resonances described by the metric, for which the relations between the perturbed vielbein and perturbed metric should be well defined. Firstly, the perturbed vielbein can be defined as follows Using the relation between the metric and the vielbein we can get Then the tensor perturbation of the background metric is [66] The equation of motion for the tensor perturbation can be gotten as follows where Obviously, the resonance spectrum of KK gravitons is closely related to the form of the function f (T ). The spacetime torsion will cause the thick brane to split, making the thick brane appear more abundant internal structures. Therefore, more abundant resonance spectrum of KK gravitons may appear. In addition, it is also possible to reflect the structure of the extra dimension by studying the resonance spectrum of KK gravitons.
In this paper, we will construct some new f (T )-brane solutions. It will be shown that these f (T )-brane solutions are stable under the transverse-traceless tensor perturbation. Based on these solutions, we will study the effects of different parameters on the effective potential and the gravitational resonances. We will also give the region where the resonance exists in the parameter space. Besides, the effects of the torsion will also be studied by comparing different kinds of f (T ). More importantly, we will investigate the possibility that the first resonance to be a candidate for dark matter. For the warp factor that we choose, the first resonance can not be a candidate for dark matter for f (T ) = T and f (T ) = T + αT 2 . For f (T ) = T , the long-lived resonance requires a very large thickness of the brane, which is inconsistent with the gravitational experiment. The form of f (T ) = T + αT 2 has the similar result with f (T ) = T . But for the case of f (T ) = −T 0 e − T T 0 − 1 , the effect of torsion on both the effective potential and the lifetime of the first resonance can be large enough with suitable choice of the parameter T 0 . In this case, we do not need a thick brane with a very large thickness to make the lifetime of the first resonance long enough to be a candidate of dark matter.
The organization of this paper is as follows. In Sec. II, we will review f (T )-brane and its tensor perturbations [66], and get the zero mode normalization condition. In Sec. III, we will construct some thick f (T )brane models. Then we will investigate the possibility of the first resonance in these models to be a candidate for dark matter. Finally, in Sec. IV, we come to the conclusions and discussions.

II. BRANE WORLD MODEL IN f (T ) GRAVITY
Firstly, we give a brief review of the teleparallel gravity. This gravity theory was proposed by Einstein as an attempt of a unified theory of electromagnetism and gravity on the mathematical structure of distant parallelism. We usually use the veilbein fields e A (x M ) instead of the metric field g MN to describe the dynamics and structure of the spacetime. This is done in the tangent space associated with a spacetime point in the manifold, instead of the coordinate basis. These vielbein fields form an orthonormal basis of the tangent space at each point in the manifold with spacetime coordinates x M . The relation between the spacetime metric and the veilbein fields is given by where η AB = diag(−1, 1, 1, 1, 1) is the Minkowski metric (in this paper, we focus on five dimensions). In this paper, capital Latin indices A, B, C, · · · = 0,1,2,3,5 label tangent space coordinates, while M, N, O, · · · = 0, 1, 2, 3, 5 label spacetime ones. In this gravity theory, the spacetime is characterized by a curvature-free linear connection, i.e., the Weitzenböck connectionΓ P MN which is defined in terms of the vielbein fields: We use the Weitzenböck connection rather than the Levi-Civita connection Γ P MN to define the associated tensors. The torsion tensor is constructed from the Weitzenböck connection as The difference between the Weitzenböck connection and the Levi-Civita connection is given by It is useful to define another tensor S MN So the torsion scalar T is given by The Lagrangian of the teleparallel gravity can be written as where e is the determination of the vielbein e A (x M ) and M 5 is the five-dimensional mass scale, which is set to M 5 = 1 in this paper. It is known that the teleparallel gravity is equivalent to general relativity and hence is also called as the teleparallel equivalent of general relativity since R = −T − 2∇ M T N MN . In f (T ) gravity the torsion scalar is replaced with f (T ), a function of T. When f (T ) = T , it goes back to teleparallel gravity, and hence is equivalent to GR. In a fivedimensional f (T ) gravity, the action is given by where L M denotes the Langrangian density of the matter. After varying the action with respect to the vielbein, we can get the field equation: where and T MN is the energymomentum tensor of the matter field. Now, we would like to consider thick brane models in f (T ) gravity theory. The metric of the flat brane with codimension one is given by where η µν = diag(−1, 1, 1, 1) is the four-dimensional Minkowski metric and e 2A(y) is the wrap factor. The bulk vielbein is and the torsion scalar is We choose the matter Lagrangian density as where φ is the background scalar field only depending on the extra dimension y, and V (φ) is the potential of the scalar field φ. For such setup, the explicit equations of motion are given by It can be shown that only two of the above three equations are independent. So we need to give two of the four variables to solve these equations. From Eq. (20) and Eq. (21), we can obtain that Next, we consider the linear transverse-traceless tensor perturbation of the metric corresponding to the vielbein, which was investigated firstly in Ref. [66]. The perturbed vielbein can be written as where the Latin letters a, b, · · · denote the tangent space coordinates on the brane, and the Greek letters µ, ν · · · denote the spacetime coordinates on the brane. And then, the tensor perturbation of the background metric can be written as where satisfies the transverse-traceless condition Considering the above conditions, we obtain the main equation of the tensor perturbation [66]: where ✷ (4) = η µν ∂ µ ∂ ν . With the coordinate transformation Eq. (29) becomes where Now, we introduce the KK decomposition where Substituting Eq. (33) into Eq. (31), we get two equations: the Klein-Gordon equation for the four-dimensional KK gravitons ǫ µν : and the Schrödinger-like equation for the extradimensional profile: where m is the mass of the KK graviton and the effective potential is given by [66] U The Schrödinger-like equation (37) can be factorized as which ensures that the eigenvalues m 2 are non-negative, so there is no tensor tachyon mode with m 2 < 0. That is to say any brane solution of f (T ) gravity theory is stable under the transverse-traceless tensor perturbation. The solution of the graviton zero mode (the four-dimensional massless graviton) is where N 0 is the normalization coefficient. Note that, in order to recover the four-dimensional Newtonian potential on the brane, the zero mode of graviton should satisfy the following normalization condition

III. GRAVITATIONAL RESONANCES IN VARIOUS f (T )-BRANE MODELS
In this section, we will give some solutions of braneworld and investigate the gravitational resonances in thick f (T )-braneworld models. Because f (T ) is an arbitrary function of the torsion scalar T , so different functional forms can give different solutions.
In this paper, we consider the following warp factor Here the parameter k has mass dimension one. The parameter b has length dimension one and denotes the distance of two sub-branes. For convenience, we define the dimensionless scaled distanceb = kb. The shape of the warp factor is shown in Fig. 1(a), from which we can see that there is a platform near y = 0 for largeb. When y → ±∞, A(y) → −2k|y|, so the spacetime is asymptotically AdS 5 .
A. Model 1: f (T ) = T We first consider f (T ) = T , which is equivalent to GR. From Eqs. (23) and (24) we get the solution where F(y; q) and Π(y; q; p) are the first and third kind elliptic integrals, respectively. Plots of the scalar field are shown in Fig. 1(b). It can be seen that the scalar field has the configuration of a single kink for smallb. With the increase ofb, the single kink becomes a double kink, and the value of |φ(±∞)| increases accordingly. In particular, In general, the appearance of the double kink means that the brane splits into two sub-branes. The distance between the two sub-branes is b. Because the solution of the scalar field is more complicated, we have not inversely solved the expression of V (φ). Numerically, we plot the shape of the scalar potential V (φ) in Fig. 1(c). It can be seen that asb increases, V (φ) splits at φ = 0.
The effective potential (38) in the coordinate y is given by We plot the shape of the above effective potential in Fig. 2(a). It can be seen that the depth of the potential well decreases with the parameterb, and the height of the potential barrier increases withb. Both changes become smaller withb. On the other hand, it can be seen that the width of the effective potential increases withb.
With the increase ofb, the potential splits from one well to two wells. We will see later that the appearance of the two-well structure will result in gravitational resonances. Next, we will investigate the gravitational resonances in this model. Inspired by the investigation of Ref. [73], Almeida et al firstly proposed a method to find the fermion resonances by using large peaks in the distribution of the normalized squared wavefunction [28]. But this method is only applicable to even functions, and it is no longer valid when the solutions are odd. In order to find all resonances, Liu et al proposed another method by defining the relative probability [43] P where ψ(z) is the solution of Eq. (37), z b is approximately the width of the brane, and z max = 10z b . Here |ψ(z)| 2 can be explained as the probability density. If the relative probability P (m 2 ) has a peak around m = m n and this peak has a full width at half maximum, we can say that there exists a resonant mode with mass m n . Since the potentials considered in this paper are symmetric, the wave functions are either even or odd. Hence, we can use the following boundary conditions to solve the differential equation (37) numerically: where ψ even and ψ odd denote the even and odd modes of ψ(z), respectively. Then, substituting the effective potential (45) into Eq. (37) we can obtain the solution of the extra-dimensional profile ψ(z) with mass m and hence the relative probability P (m 2 ). According to the supersymmetric quantum mechanics, the supersymmetric partner potentials will share the same spectrum of massive excited states. So, we can judge whether there are resonances by analyzing the shape of the supersymmetric partner potential. The dual potential corresponding to the effective potential (38) is If there is no well or quasi-well in the dual potential, then there is no resonances [49]. Thus, for f (T ) = T , only for b > 1, there might exist resonances. Just as said before, the width of the potential barrier increases withb, which indicates there are more resonances for largerb, which can be seen from Figs. 2(b), 2(c), 2(d). Furthermore, we can obtain the corresponding lifetime τ of the gravitational resonances by the width Γ at half maximum of the peak, i.e., τ = 1 Γ . For convenience, we define the dimensionless scaled massm n = m n /k and the dimensionless scaled lifetimeτ = kτ . The relations of the scaled mass m 1 and the scaled lifetimeτ 1 of the first resonance with the parameterb are shown in Fig. 3. It can be seen that, the scaled massm 1 of the first resonance decreases withb, while the scaled lifetimeτ 1 of the first resonance increases withb. This behaviour means that if the parameterb is large enough, the lifetime of the first resonance with very light mass can be long enough as the age of our universe. So, we can consider such first gravitational resonance as one of the candidates for dark matter [50].
The scaled massm 1 and the scaled lifetimeτ 1 can be fitted as two functions of the parameterb, and the fit functions are given bȳ log(τ 1 ) = −0.159 + 1.183 ln(b).
In the brane world theory considered in this paper, the relation between the effective four-dimensional Planck scale M Pl and the fundamental five-dimensional scale M 5 is given by: for f (T ) = T , Eq. (50) becomes According to the recent experiment of the Large Hadron Collider (LHC), the collision energy is 13 TeV and the result shows that the quantum effect of gravity can be ignored. Theoretically, the quantum effect of gravity will appear if the energy scale is larger than the fivedimensional fundamental scale M 5 . Thus, from the condition M 5 > 13 TeV and Eq. (51), we can give the constraint on the parameter k in the Natural System of Units: On the other hand, it is known that the age of our universe is of about 13.8 billion years, i.e., 4.35 × 10 17 s. So, if we consider the first resonance as one of candidates for dark matter, its lifetime should be larger than the age of universe, i.e., τ 1 4.35 × 10 17 s, or in the Natural System of Units, Thus, the restriction of the parameter k can be expressed as k 1.5 × 10 −33τ 1 eV ≃b 2.724 × 10 −33 eV.
By combining the fit function (48) The shadow regions of Fig. 4 show the available ranges of the parameters k and m 1 , respectively. From Fig. 4(a), we can see that only if b > 7.9 × 10 9 , the two restricted conditions (52) and (54) of k could be satisfied, which means that the parameterb has a lower bound. And the corresponding constraint of the parameter k is k 9.5×10 −7 eV. From Fig. 4(b), we can see that the first resonance mass m 1 has a lower bound, i.e., m 1 4 × 10 −16 eV. But there are some problems here: (1) In any realistic brane scenario, the matter fields on the brane will cover the entire thickness of the brane along the extra dimension. On the other hand, such a largeb value means that the thickness of the brane is also very large. Within our constraints, k is about 10 −6 eV andb is about 10 10 . The corresponding size of b =b k is about 10 11 cm, which will cause the effective four-dimensional gravitational potential deviates from the squared inverse law at a very large distance. According to the method of Ref. [42], the correction term to the Newtonian potential from all massive gravitons in this model is We can see that the four-dimensional effective Newtonian potential becomes So whenb = 10 10 , the correction scale r is about 10 7 cm. This is contradicted with the current experiments of gravitational inverse-square law. In fact, Kiritsis et al [74] showed that in their thick brane model, when the thickness of the brane is large enough, the Newtonian potential of two points close to each other on the brane becomes five-dimensional.
(2) Note that, we use the relative probability method (46) to obtain the resonance and give the lifetime of the first KK resonance state atb ≤ 10 2 . Although we cannot guarantee that the fit functions (48) and (49) are still valid whenb is very large, we do not need to care about this. Because, the number of the resonances will increase with the parameterb, and the height of the effective potential is almost unchanged. That is to say, the width at half maximum of the first resonance will decrease with b. It will result in that the lifetime of the first resonance increases withb. Based on the above discussion, maybe the value of the equation (57) is not accurate, but our result is still valid. That is to say, under TEGR and warp factor (42), the first resonance can not be the candidate of dark matter. Therefore, we consider f (T ) gravity and see the effect of torsion.
In order to ensure that the scalar field φ is real, the parameter α should satisfy α ≤ 1 288k 2 . We can solve Eq. (60) numerically and show the plots of the scalar field in Figs. 6(a) and 6(b). It can be seen that, the shape of the scalar field depends on the values ofᾱ andb. The scalar field has the configuration of a single kink for small |ᾱ| andb. With the increase of |ᾱ|, the single kink becomes a double one, and the asymptotic value of |φ(±∞)| increases accordingly. The influence of the parameterb on the scalar field is the same as the case of f (T ) = T .
Since the expression of the effective potential for f (T ) = T + αT 2 is complicated and tedious, we only show its plots in Figs. 6(c) and 6(d). From Fig. 6(c) we can see that the depth of the effective potential well and the height of the potential barrier decrease with the parameterᾱ. With the decrease ofᾱ, the poten- tial splits from one well to two wells. From Fig. 6(d) we can see that the effective potential has four peaks. As −0.046 <ᾱ ≤ 1/288 ≈ 0.00347 for the case ofb = 4, the four peaks will become two peaks. We label two of them as peak 1 and as peak 2 in Fig. 6(d). Asᾱ increasing, the height of peak 1 decreases, but the height of peak 2 slowly increases. Whenᾱ = −0.212, the heights of peak 1 and peak 2 are approximately the same. The relative probabilities P (m 2 ) of the gravitational resonances for b = 4,ᾱ = 0.00347, −0.212, and −1 are shown in Fig. 7.
In order to investigate the effects of the parameterb, we fixᾱ = −1. Plots of the relative probability for different values ofb are shown in Fig. 8. We find that the number of resonances increases withb, and the mass m 1 of the  first resonance decreases withb. This result is the same as f (T ) = T . The relation between the number of the resonances and the parametersᾱ andb is shown in Fig. 9.
We can see that whenᾱ < −0.9, even if the value ofb is very small, there is still a resonance. As mentioned earlier, when |ᾱ| is large, there are two wells for the effective potential, which can be seen from Fig. 6(c). It is because of the existence of these two potential wells that whenb is small, the massive KK gravitons can be quasilocalized on the brane. However, when the value ofᾱ increases from −0.9 to 0.00347, a larger value ofb is required for the existence of resonances. The emergence of these properties is due to the difference between f (T ) = T + αT 2 and GR.
We know that there are positive correlations between the lifetime of the resonance and width and height of the effective potential. For the case of f (T ) = T , we adjust the value ofb to change the width of the potential well, but keep the height of the potential unchanged. For f (T ) = T + αT 2 , the height of the potential and the lifetime of the first resonance increase with |ᾱ|. When |ᾱ| is large enough, the height of the potential increases very slowly, and whenᾱ → −∞, it approaches a finite value. In other words, the parameterb plays a key role in the lifetime of the first resonance. This is similar to the case of f (T ) = T . Thus, we also run into the dilemma as the GR case, i.e., the effective four-dimensional gravitational potential with such a largeb may deviate from the squared inverse law at a large distance. So, let us consider another form of f (T ), which hopefully resolves this contradiction.
Lastly, we consider the model with where the mass dimension of the parameter T 0 is 2. For convenience, we define the dimensionless parameterT 0 = T 0 /k 2 . Then f (T ) can be rewritten as The expression (61) can be expanded at T = 0 as It can be seen that the smaller |T 0 |, the more deviation from GR. The warp factor is also considered as (42), and whenT 0 → −∞ the model reverts back to general relativity. In this model, Eq. (24) becomes Same as before, from Eq. (64) we can find the numerical solution of the scalar field φ. In order to ensure that the scalar field φ is real, the parameterT 0 should satisfȳ T 0 > 0 orT 0 ≤ −96. To compare the deviation of these two branches of T 0 from GR, we rewrite expression (63) as The T 2T0 term in the above equation is the dominant term deviating from GR. From T = −12A ′ (y) 2 and (42) we get that the range of T k 2 is −48 ≤ T k 2 ≤ 0. ForT 0 ≤ −96, Obviously, the effect of torsion for the caseT 0 > 0 can be more significant than the caseT 0 ≤ −96, this is even more evident in the effective potential. Plots of the scalar field are shown in Figs. 10(a) and 10(b). It can be seen that the scalar field has the configuration of a double kink for a largeb. Similarly, we get the effective potential U (z(y)) through Eq. (38). Because the expression of the effective potential is complicated, we only show its plots. The width of the potential well increases with the parameterb, and the height of the potential barrier decreases with the param-eterT 0 , which can be seen from Figs. 10(c) and 10(d). The effective potential forb = 10,T 0 = −96, −500, and T 0 → −∞ (corresponding to GR) are shown in Fig. 11. We can see that all shapes of the effective potential for all values ofT 0 are almost the same. That is to say, when T 0 ≤ −96, the torsion has an insignificant effect on the effective potential. Therefore, we only consider the case ofT 0 > 0.
The relative probability P (m 2 ) of the gravitational resonances forb = 10,T 0 = 2, 3, and 4 are shown in Fig. 12. We find that the number of the resonances decreases with  T 0 . The relation between the number of the resonances and the parametersT 0 andb is shown in Fig. 13. We can see that, whenT 0 < 6, even if the value ofb is very small, there is still a resonance. This is similar to the case of f (T ) = T + αT 2 .
The relations of the scaled massm 1 and the scaled lifetimeτ 1 of the first resonance with the parameterT 0 are shown in Fig. 14. It can be seen that, both the scaled massm 1 and the scaled lifetimeτ 1 of the first resonance decrease withT 0 . Similarly, the KK graviton resonances with long enough lifetime could be considered as one of the candidates for dark matter. The scaled massm 1 and the scaled lifetimeτ 1 can be fitted as two functions of the parameterT 0 , and the fit functions can be expressed as log(τ 1 ) = −0.783 + 8.738 On the other hand, since we have no analytical formulation for the integral expression (50) for f (T ) = −T 0 e − T T 0 − 1 , we cannot obtain an analytical con- In this case, we consider the constraint imposed by the test of the gravitational inverse-square law. According to the method of Ref. [42], for the brane world model considered here, the four-dimensional gravitational potential is where C is a dimensionless constant determined by the structure of the brane. The correction term is G N CM k 2 r 3 . For the case of f (T ) = −T 0 e − T T 0 − 1 , the correction occurs at the scale of r ∼ 1/k. Experimentally, in the recent experiments of the test of the gravitational inversesquare law, the usual Newtonian potential still holds down to the scale of about 50 µm [75][76][77]. Therefore, 1/k should be less than 50 µm, this is equivalent to On the other hand, the restriction of the lifetime of the first resonance on the parameter k is given by By combining the fit function (66) and the two conditions (69), (70), the restricted expressions of the mass of the first resonance m 1 with the parameterT 0 can be obtained The available ranges of the parameters k and m 1 are shown in Fig. 15. From Fig. 15(a), we can see that only ifT 0 < 0.28, the two restricted conditions (69) and (70) of k could be satisfied, which means that the parameter T 0 has an upper bound. From Fig. 15(b), we can see that the first resonance mass m 1 has a lower bound, i.e., d e f 15: The limit range of the parameter k and the corresponding range of the first resonance mass m1. The shadow region of the left panel is the limit range of the parameter k and the shadow region of the right panel is the corresponding range of the first resonance mass m1. The black solid lines are the limits that the lifetime of the first resonance should be longer than the age of the universe and the blue dashed lines are limits from 1/k should be less than 50 µm. m 1 2.8 × 10 −3 eV. It can be found that for model 3, the first resonance as a candidate for dark matter does not have the previous confliction of model 1 and model 2. The main reason is that the key role that determines the lifetime of the first resonance is no longerb butT 0 . This reminds us that, if the effect of the torsion is large enough, other forms of f (T ) can also produce a long-lived resonance that satisfies the conditions to be a candidate of dark matter.

IV. CONCLUSION AND DISCUSSION
In this work, we investigated the gravitational resonances in various f (T )-brane models. The branes are generated by a canonical scalar field. We first reviewed the tensor perturbations of f (T )-branes. Then we discussed the gravitational resonances in three f (T )-brane models with the same warp factor (42).
In model 1, we considered f (T ) = T and obtained the thick brane solution, which is equivalent to GR. We found that with the increase of the parameterb, the double kink structure appears in the scalar field, which generally corresponds to the appearance of sub branes and resonances. The influence of the parameterb on the gravitational resonances was analyzed. The result shows that, only when b > 1, the gravitational resonances could exist. The mass m 1 of the first resonance decreases with the parameter b. The lifetime τ 1 of the first resonance and the number of resonances increase with the parameterb. Ifb is large enough, the lifetime of the first resonance will be long enough as the age of our universe. This indicates that the first gravitational resonance could be one of the candidates for dark matter, which would lead to some constraints on the parameters k andb, i.e., k 9.5×10 −7 eV, b > 7.9 × 10 9 . And the corresponding available range of the first resonance mass is m 1 4 × 10 −16 eV. But there is a problem, with such a largeb, the effective fourdimensional gravitational potential may deviate from the squared inverse law at a large distance.
In model 2, we considered f (T ) = T + αT 2 . When the parameterᾱ changes, the peak of the effective potential changes from two to four, suggesting that f (T )-brane has more abundant internal structure and properties than GR-brane. The parameterᾱ can affect the resonance spectrum and the lifetime of resonances. The influence of the parameterb is the same as f (T ) = T . But the parameterb plays a key role in the lifetime of the first resonance. The relation between the number of the resonances and the parametersᾱ andb was obtained, which can be seen from Fig. 9. This can reflect the characteristics of the extra dimension. We found that for large |ᾱ|, there is a gravitational resonance even for smallb, which is different from GR. Unfortunately, when considering the resonance as a candidate for dark matter, this model has the same problem as GR.
In model 3, f (T ) = −T 0 e − T T 0 − 1 was considered. We analyzed the effects of the parametersT 0 andb on the effective potential, and found that the parameterT 0 significantly changes the height of the potential barrier. Then, we got the relation between the number of the resonances and the parametersT 0 andb (see Fig. 13). The influence of the parameterT 0 on the gravitational resonances was analyzed by takingb = 10. The result shows that the lifetime τ 1 of the first resonance decreases with the parameterT 0 . If the lifetime of the first resonance exceeds the age of our universe, the parameters k,T 0 , and mass of the first resonance must satisfy k 4 × 10 −3 eV, T 0 < 0.28, and m 1 2.8 × 10 −3 eV. We can see that model 3 does not have the two problems in model 1 and model 2. This indicates that the first resonance could be a candidate for dark matter asT 0 < 0.28, for which the f (T ) theory is very different from GR. Note that, if we choose a larger value ofb, the constraint onT 0 will be further relaxed. In a word, if we want the first gravitational resonance to be a candidate of dark matter, the effect of torsion should be significant. That is, other forms of f (T ) gravity may also satisfy the conditions of first resonance as a candidate for dark matter, which is worth further study.
In this paper, we only considered the KK gravitons in f (T ) gravity as the candidate for dark matter and investigated the corresponding properties of such resonances. In fact, KK fermions and KK vector particles, etc., may also be candidates for dark matter, which deserves further study.