Prospective bounds on f(Q) gravity with pulsar timing arrays Alireza Allahyari Abstract Pulsar timing arrays (PTAs) have recently provided compelling evidence for a stochastic gravitational wave background (SGWB) in the nanohertz frequency band, offering a unique window into fundamental physics. Here, we explore implications for symmetric teleparallel f ⁡ ( Q ) f(Q) gravity, a theory in which deviations from General Relativity (GR) arise through the non-metricity scalar f ⁡ ( Q ) f(Q) . Crucially, tensor modes propagate at the speed of light in this framework. However, their amplitude undergoes a modified damping during their evolution. We adopt a model-independent parameterization and derive an analytic approximation to the tensor mode transfer function to obtain the spectral energy density of primordial inflationary gravitational waves. Comparison with the NANOGrav 15-year and IPTA second data releases show that the inferred damping parameter n n remains consistent with GR, yet allows small deviations that could be observable. We then conduct a Fisher information matrix forecasts which demonstrate that the Square Kilometre Array (SKA) observatory will improve these constraints by several orders of magnitude, offering the potential to distinguish f ⁡ ( Q ) f(Q) gravity from GR with high precision. These results highlight PTAs as powerful probes of non-metricity-based modifications to gravity. 1 Introduction Recently, strong indications of the possibility of stochastic gravitational waves background in nHz have been obtained by various pulsar timing arrays (PTAs) observations [ 3 , 108 , 11 , 122 ] . But there is still no precise description for the source of these waves [ 2 ] . Focusing for concreteness on the NANOGrav [ 3 ] signal, this stochastic GW background has a blue-tilted tensor spectrum, with the spectral index n_{T}=1.8\pm 0.3 [ 116 ] . Modified gravity theories predict different generation and propagation mechanisms for gravitational waves (GWs) and could generate a blue-tilted spectrum. Additionally, we need to consider the fact that the recent measurement of propagation speed of GWs from GW170817 relative to its electromagnetic counterpart GRB170817A [ 1 ] severely limits the deviations from the speed of light. The f(Q) gravity originates from the symmetric teleparallel framework [ 98 , 21 ] , where gravitation is described not by curvature or torsion but by the non-metricity scalar Q . The f(Q) gravity [ 59 ] has gained considerable popularity in the past couple of years and the bulk of the research efforts have been concentrated on cosmological applications [ 76 , 105 , 96 , 95 , 36 ] . This model has also been applied to large structure formation [ 111 ] , the development of relativistic versions of Modified Newtonian Dynamics (MOND) [ 86 , 40 ] , bouncing cosmologies [ 18 , 7 , 49 ] , and even quantum cosmology [ 35 , 17 ] . A lot of effort has also gone into constraining or testing f(Q) models [ 34 , 15 , 20 , 46 , 4 , 33 , 8 , 44 , 68 , 93 , 25 , 45 , 114 ] . Extensions that involve incorporating boundary terms [ 28 , 32 , 102 ] or non-minimally coupled scalar field [ 61 , 54 ] have also been explored. Other very active area of research are black holes within f(Q) gravity [ 19 , 90 , 104 , 57 , 56 , 55 , 117 , 60 , 52 , 62 , 87 , 31 ] , modified stellar solutions [ 119 , 81 , 82 , 43 , 112 , 26 , 23 , 37 , 80 ] , and wormholes [ 19 , 90 , 104 , 57 , 56 , 55 , 117 , 60 , 52 , 91 , 87 ] . Also in this regard, some thought has been given to how observational data could be used to constrain f(Q) gravity [ 63 , 12 , 115 , 5 , 94 , 38 , 84 , 77 , 109 , 103 , 51 , 24 , 120 , 85 , 42 , 118 , 47 , 101 , 70 , 83 , 53 , 92 , 89 , 110 , 14 , 48 , 67 , 41 , 50 , 97 , 39 , 107 , 82 , 69 , 76 , 65 , 73 , 72 , 71 , 79 , 13 , 113 , 9 , 78 , 20 , 74 ] . In this paper, we assume that GW background has a primordial origin from an inflationary era with a modified damping term. To constrain the parameters of f(Q) gravity using pulsar timing arrays data we apply a model-independent parameterization implemented in [ 29 ] . In section 2 , we introduce primordial GWs and their relation to PTAs. In section 3 , we derive the approximate transfer function for tensor perturbations in f(Q) gravity and find the spectral energy density. Then, in section 4 by using NANOGrav 15-year data set (NG15) [ 3 ] and International PTA second data release (IPTA2) [ 106 , 10 ] , we constrain the model. Finally using Fisher formalism, we find the bounds from future PTAs data by Square Kilometre Array (SKA) observatory [ 121 ] . Section 5 is summary and discussion. 2 Primordial GWs The evolution of linear, transverse-traceless perturbations for the tensor modes in GR are described by the following equation \ddot{h}_{ij}+3H\dot{h}_{ij}+(k^{2}/a^{2})h_{ij}=0\,, (2.1) where H is the Hubble parameter. The relevant quantity in PTA observations is the spectral energy density. The spectral energy density of GWs can be derived as [ 30 ] \Omega_{\text{GW}}(k,\eta)=\frac{1}{12H^{2}a^{2}}[T^{\prime}(k,\eta)]^{2}P_{t}(k)\,, (2.2) where T(k,\eta) is the transfer function. Transfer function describes the evolution of GW modes after the modes re-enter the horizon. The quantity P_{t}(k) is the primordial power spectrum of GWs at the end of the inflationary period. We may write this in terms of a tilt n_{t} and a tensor amplitude A as P_{t}(k)=A(\frac{k}{k_{\star}})^{n_{t}}\,, (2.3) where k_{\star} is a pivot scale set as k_{\star}=0.05\,{\rm Mpc}^{-1} . The transfer function in context of GR, is numerically computed in [ 123 ] T(k,\eta_{0})_{\rm GR}=\frac{3j_{1}(k\eta_{0})}{k\eta_{0}}\sqrt{1.0+1.36(\frac{k}{k_{\rm eq}})+2.50(\frac{k}{k_{\rm eq}})^{2}}\,, (2.4) where j_{1} is Bessel function and \eta_{0} is the present conformal time. For GWs in PTA scales, we can use the approximation (k\gg k_{\rm eq}) and [T^{\prime}(k,\eta)]^{2}=k^{2}[T(k,\eta)]^{2} [ 116 ] . In pulsar timing experiments, it is convenient to express wavenumbers k in terms of frequencies f as [ 116 ] \displaystyle f\simeq 1.54\times 10^{-15}\left(\frac{k}{{\rm Mpc}^{-1}}\right)\,{\rm Hz}\,. (2.5) We find that f_{\star}\simeq 7.7\times 10^{-17}\,{\rm Hz} . Also in PTAs, the GW spectral energy density is rather written in terms of the power spectrum of the GW strain h_{c} given by \displaystyle\Omega^{\rm PTA}_{\rm GW}(f)=\frac{2\pi^{2}}{3H_{0}^{2}}f^{2}h_{c}^{2}(f)\,. (2.6) h_{c}(f) is supposed to take a powerlaw form with respect to a reference frequency f_{\rm yr} and can be expressed as \displaystyle h_{c}(f)=A\left(\frac{f}{f_{\rm yr}}\right)^{\alpha}\,, (2.7) where f_{\rm yr}=1\,{\rm yr}^{-1}\approx 3.17\times 10^{-8}\,{\rm Hz} . Finally using \alpha=\frac{3-\gamma}{2} , the present GWs spectral energy density is given by [ 116 ] \displaystyle\Omega^{\rm PTA}_{\rm GW}(f)=A^{2}\frac{2\pi^{2}}{3H_{0}^{2}}\frac{f^{5-\gamma}}{{\rm yr}^{\gamma-3}}\,. (2.8) 3 GWs in f(Q) gravity The f(Q) gravity originates from the symmetric teleparallel framework [ 98 , 21 ] , where gravitation is described not by curvature or torsion but by the non-metricity scalar Q , whose action is given by S=\int d^{4}x\,\sqrt{-g}\left[\dfrac{1}{2}f(Q)+\mathcal{L}_{m}\right]\,, (3.1) where \mathcal{L}_{m} is the matter field Lagrangian, and g is the determinant of metric. Notice that, up to a total derivative, the above action and the Einstein-Hilbert one are equivalent for f(Q)=Q . The non-metricity scalar is given by [ 22 ] Q=-Q_{\alpha\mu\nu}P^{\alpha\mu\nu}\,, (3.2) where the non-metricity tensor Q_{\alpha\mu\nu} , and the non-metricity conjugate P^{\alpha}{}_{\mu\nu} , are given by Q_{\alpha\mu\nu}\equiv\nabla_{\alpha}g_{\mu\nu}\,. (3.3) P^{\alpha}{}_{\mu\nu}=-\frac{1}{2}L^{\alpha}{}_{\mu\nu}+\frac{1}{4}(Q^{\alpha}-\tilde{Q}^{\alpha})-\frac{1}{4}\delta^{\alpha}_{(\mu}Q_{\nu)}\,, (3.4) where L^{\alpha}{}_{\mu\nu} is the disformation tensor, which takes the form L^{\alpha}{}_{\mu\nu}=\frac{1}{2}Q^{\alpha}{}_{\mu\nu}-Q_{(\mu\nu)}{}^{\alpha}\,, (3.5) where Q_{\alpha} and \tilde{Q}_{\alpha} are two independent traces of the non-metricity, defined as Q_{\alpha}\equiv g^{\mu\nu}Q_{\alpha\mu\nu},\quad\tilde{Q}_{\alpha}\equiv g^{\mu\nu}Q_{\mu\alpha\nu}\,. (3.6) In the homogeneous and isotropic universe described by the FLRW metric, the non-metricity scalar is given by [ 29 ] Q=6H^{2}=6\frac{\mathcal{H}^{2}}{a^{2}}\,. (3.7) The evolution of linear, transverse-traceless perturbations for the tensor modes due to modifications of the gravity theory are generally described by the following equation [ 99 ] h^{\prime\prime}_{ij}+\left(2+\nu\right)\mathcal{H}h^{\prime}_{ij}+(c_{T}^{2}k^{2}+a^{2}\mu^{2})h_{ij}=a^{2}\Gamma\gamma_{ij}\,, (3.8) where prime denotes differentiation with respect to conformal time \eta . \nu measures deviations from the standard amplitude damping, c_{T} is the GWs propagation speed, \mu an effective mass term, and \Gamma\gamma_{ij} possible source contributions. The propagation equation in f(Q) gravity for each polarization \lambda reads as [ 63 ] h^{\prime\prime}_{(\lambda)}+2\mathcal{H}\left(1+\frac{d\log f_{Q}}{2\mathcal{H}d\eta}\right)h^{\prime}_{(\lambda)}+k^{2}h_{(\lambda)}=0\,, (3.9) with f_{Q}\equiv\partial f/\partial Q . f(Q) models predict c_{T}=1 , i.e. gravitational waves propagate at the speed of light, consistent with current observational constraints [ 1 ] . The modification adds the damping parameter as \displaystyle\nu \displaystyle= \displaystyle\frac{1}{\mathcal{H}}\frac{\mathrm{d}\log f_{Q}}{d\eta}\equiv\frac{1}{H}\frac{d\log f_{Q}}{dt}\,. (3.10) The transfer function in modified gravity can be approximated as a product of a correction factor and a relativity part [ 100 ] . For f(Q) gravity we have T(k,\eta)_{\text{MG}}=\mathrm{exp}\left(-\frac{1}{2}\int^{\eta}\mathcal{H}\nu\,\mathrm{d}\eta^{\prime}\right)T(k,\eta)_{\text{GR}}\,. (3.11) We use a model-independent parameterization considering gravity mediated by non-metricity, with vanishing curvature and torsion; for describing the accelerated expansion of the universe with the function [ 29 ] f(Q)=\alpha+\beta Q^{n}\,. (3.12) We can see that one recovers GR for \alpha=0 and \beta=1=n , but for the \Lambda CDM model, we need \alpha>0 and \beta=1=n . Finally we find the transfer function as T(k,\eta)_{\text{MG}}=\left(\frac{\mathcal{H}(a)a_{e}}{\mathcal{H}(a_{e})a}\right)^{1-n}T(k,\eta)_{\text{GR}}\,, (3.13) where a_{e} refers to the scale factor at horizon entry. We consider the evolution of background in \Lambda CDM model [ 66 , 44 ] ; So the Hubble parameter can be written as \mathcal{H}^{2}(a)=a^{2}\mathcal{H}_{0}^{2}\Big[\Omega_{m,0}\,a^{-3}+\Omega_{r,0}\,a^{-4}+\Omega_{\Lambda,0}\,\Big]\,, (3.14) with matter density parameter \Omega_{m,0}=0.315 , radiation density parameter \Omega_{r,0}=9\times 10^{-5} , dark energy density parameter \Omega_{\Lambda,0}=0.68491 and \mathcal{H}_{0}=67.36 km/s/Mpc from Planck 2018 [ 6 ] . Also the scale factor at horizon entry is related to frequency as a_{e}\approx\frac{3.33870198\times 10^{-21}}{f}\,, (3.15) Using the above equations we obtain the spectral energy density of gravitational waves at present time as \begin{split}\Omega_{\text{GW}}(f)=A^{2}\frac{2\pi^{2}}{3H_{0}^{2}}\left(\frac{f^{5-\gamma}}{\text{yr}^{\gamma-3}}\right)\left(\frac{\mathcal{H}_{0}a_{e}}{\mathcal{H}_{e}}\right)^{2-2n}\\ \times\left(1+\frac{\mathcal{H}_{0}(1-n)(\Omega_{m,0}+\Omega_{r,0})}{4\pi f}\right)^{2}\,.\end{split} (3.16) 4 Bounds by PTAs In this section, we use NANOGrav 15 year data (NG15) and IPTA second data release (IPTA2) to constrain the parameter n . The parameters \alpha and \beta do not enter the final expression for spectral energy density. In the following, we show that more future data with improved precision can enhance the accuracy of the results. 4.1 NANOGrav and IPTA We use the python package PTArcade [ 88 ] , that provides an accessible way to perform Bayesian analyses with PTAs data. We consider uniform priors on the parameters as (0<\gamma<6) and (01 . Also upper amplitude gives n<1 and lower amplitude gives n>1 . Predictions are consistent with GR given uncertainties. This shows that pulsar timing arrays data is a very powerful tool for studying and constraining modified gravity theories. In figure 2 , for concreteness, we show h^{2}\Omega_{\text{GW}} as a function of frequency in logarithmic scales using the best fit of values of the parameters obtained from NG15 and IPTA2. Figure 1 : The posterior plots and marginal posteriors of f(Q) gravity parameters with \text{log}_{10}A=-14.2 . The contours show at 68% and 95% confidence levels for International PTA second data release (IPTA2) in blue and NANOGrav 15-year data set (NG15) in green. Parameter NG15 IPTA2 \text{log}_{10}A (fixed) n 0.9857\pm 0.0039 0.9802^{+0.0055}_{-0.0037} -13.6 \gamma 3.14\pm 0.34 3.96\pm 0.36 -13.6 n 1.0019\pm 0.0039 0.9965^{+0.0055}_{-0.0037} -14.2 \gamma 3.24\pm 0.34 4.06\pm 0.36 -14.2 n 1.0181\pm 0.0039 1.0127^{+0.0055}_{-0.0037} -14.8 \gamma 3.34\pm 0.34 4.15\pm 0.36 -14.8 Table 1 : Constraints at 68% confidence level from NG15 and IPTA2. Figure 2 : The current spectral energy density of GWs as a function of frequency in logarithmic scales for f(Q) model. The violin plots show NANOGrav 15-year data set (NG15) and International PTA second data release (IPTA2). 4.2 SKA observatory In this section we show that improving the quantity and precision of PTAs data results in stronger parameter constraints. The predictions are obtained by constructing the Fisher information matrix from estimated SKA observations. We first compare the model-derived gravitational wave signal spectrum with the SKA noise spectrum using the python package SGWBProbe [ 27 ] . As shown in Figure 3 , the SKA is sensitive enough to detect the signal predicted by the model. The signal is written as [ 58 ] S_{h}(f)=\Omega_{\text{GW}}(f)\frac{3H_{0}^{2}}{2\pi^{2}f^{3}}\,. (4.1) Figure 3 : The gravitational wave signal spectrum from f(Q) gravity (green) compared to the SKA noise spectrum (pink). The Fisher Information Matrix (FIM) for two pairs of pulsars I,L and J,K is written as [ 16 ] F_{\alpha\beta}\equiv\sum_{f_{k}}C^{-1}_{IJ}C^{-1}_{KL}\frac{\partial(R_{JK}S_{h})}{\partial\theta^{\alpha}}\frac{\partial(R_{LI}S_{h})}{\partial\theta^{\beta}}\,, (4.2) where \theta^{\alpha} are model parameters, f_{k}=\frac{k}{T_{\text{obs}}} , C is signal covariance matrix includes gravitational waves term C_{h} and noise term C_{n} as C_{IJ}=C_{n,IJ}+C_{h,IJ}\,, (4.3) C_{n,IJ}=\delta_{IJ}P_{I}^{WN}\,, (4.4) C_{h,IJ}=R_{IJ}S_{h}(f)\,, (4.5) we only consider white noise P_{I}^{WN} and ignore other noises P_{I}^{\rm WN}=2\sigma^{2}\Delta t\,, (4.6) where \Delta t is the inverse of the observing cadence and \sigma is the rms timing uncertainty. Also R_{IJ} introduced as R_{IJ}=\chi_{IJ}\cdot{\cal R}(f)\left[{\cal T}_{I}(f){\cal T}_{J}(f){T_{IJ}}/{T_{\text{obs}}}\right]^{1/2}, (4.7) \chi_{IJ}=\frac{1}{2}+\frac{3}{2}\xi_{IJ}\left[\ln\xi_{IJ}-\frac{1}{6}\right]+\frac{1}{2}\,\delta_{IJ}\,, (4.8) \xi_{IJ}\equiv\left({1-\cos(\zeta_{IJ})}\right)/{2}\,, where the frequency dependent, sky-averaged, quadratic response function defined as {\cal R}(f)\equiv 1/12\pi^{2}f^{2} and the transfer function {\cal T}(f)\simeq\left[1+1/\left(fT_{\text{\tiny obs}}\right)\right]^{-6} . The first geometrical factor \chi_{IJ} is the well-known Hellings-Downs correlation pattern as a function of the angular separation \zeta_{IJ} between pulsars as required by symmetries [ 64 ] . And effective overlapping time defined as T_{IJ}={\rm min}[T_{I},T_{J}] . For additional information see [ 16 ] . Finally, the covariance matrix C_{\alpha\beta} , is obtained by inverting the FIM, from which one can estimate uncertainties as \sigma_{\alpha}\equiv\sqrt{F_{\alpha\alpha}^{-1}} . In this calculation, we consider the two scenarios performed in [ 75 ] , which are listed in table 2 . SKA N_{\text{p}} T_{\text{obs}} [yr] \Delta t [week] \sigma [ns] Normal 200 20 2 50 Optimistic 1000 30 1 10 Table 2 : PTA parameters considered for SKA. Here N_{\text{p}} is the number of pulsars, T_{\text{obs}} is the observation time, \Delta t is the cadence and \sigma is the root-mean-square timing residuals. We set the mean values of the parameters from the best fit of NG15. For the convenience of calculating the inverse of the covariance matrix, we assume that there are N groups of 100 pulsars ( N_{\text{p}}=100N ). So we have F_{\alpha\beta}\equiv N\sum_{f_{k}}C^{-1}_{IJ}C^{-1}_{KL}\frac{\partial(R_{JK}S_{h})}{\partial\theta^{\alpha}}\frac{\partial(R_{LI}S_{h})}{\partial\theta^{\beta}}\,. (4.9) It should be noted that we have simulated pulsars at uniform random angles in the sky. The results are listed in table 3 and illustrated in figure 4 . As we see, future PTAs data from SKA can determine the parameters with higher precision. Parameter SKA (normal) SKA (optimistic) n 1.00191^{+0.00014}_{-0.00016} 1.0019\pm 0.000014 \gamma 3.24\pm 0.014 3.24^{+0.0025}_{-0.0024} \text{log}_{10}A -14.2\pm 0.0022 -14.2^{+0.00041}_{-0.00040} Table 3 : Constraints at 68% confidence level from SKA. Figure 4 : The posterior plots and marginal posteriors of the parameters. The contours show at 68% and 95% confidence levels for two scenarios SKA (normal) in indigo and SKA (optimistic) in emerald. The Fisher matrix forecasts presented here assume an idealized PTA noise model and are intended to illustrate the sensitivity of SKA-era observations to the modified gravity parameter n . In our analysis, n affects the frequency dependence of the stochastic gravitational wave background through a modified damping of tensor modes, leading to constraints that are driven primarily by the spectral shape rather than the overall amplitude. We have only considered white timing noise. However, in realistic PTA analyses, additional noise contributions such as intrinsic red timing noise or common-spectrum processes may be present and could moderately broaden the inferred parameter uncertainties. As usual, Fisher matrix estimates correspond to lower bounds on parameter uncertainties under the assumption of a Gaussian likelihood. A full Bayesian analysis of future PTA data is therefore expected to yield somewhat larger credible intervals, while preserving the qualitative sensitivity to deviations from GR. 5 Summary and discussion In this work, we have explored the constraints of recent pulsar timing array (PTA) observations for symmetric teleparallel f(Q) gravity. Adopting a model-independent parameterization, we focused on the modified damping of tensor modes encoded in the parameter n while ensuring that gravitational waves propagate at the speed of light, consistent with multi messenger constraints from GW170817. Using an analytic approximation for the tensor mode transfer function, we derived the present-day spectral energy density of primordial inflationary gravitational waves in the context of f(Q) gravity. Our analysis with NANOGrav 15-year (NG15) and International PTA second data release (IPTA2) sets tight constraints on n , which remains consistent with the GR value of n=1 . Nonetheless, the results allow small deviations, especially for blue-tilted tensor spectra or lower amplitudes, indicating that future, more sensitive observations may reveal subtle non-metricity effects. The posterior distributions demonstrate the potential of PTAs to probe modifications of gravity beyond the standard \Lambda CDM framework. Our Fisher matrix forecasts for the Square Kilometre Array (SKA) show that future PTA data could improve constraints on f(Q) parameters. In the optimistic scenario, uncertainties on n shrink to \mathcal{O}(10^{-5}) , making it feasible to distinguish f(Q) gravity from GR with more precision. This highlights the transformative role of next-generation PTAs in testing fundamental physics and constraining non-metricity based deviations from Einstein’s theory. Overall, our results underline that current PTAs already provide meaningful bounds on f(Q) gravity, and future observations will enable a high precision exploration of non-metricity effects in the propagation of gravitational waves.