The Standard siren tests of viable f ⁡ ( R ) f(R) cosmologies Yi Zhang e1,addr1, Xuanjun Niu addr1, Xianfu Su addr2, Dong-Ze He addr1 Abstract We constrain the Hu-Sawicki and Starobinsky f ⁡ ( R ) f(R) gravity models utilizing current electromagnetic (PP+CC, Planck and DESI2) datasets and simulate standard siren catalogs based on the resulting best-fit parameters. We demonstrate that the simulated SS data provide complementary sensitivity to the modified gravitational wave propagation friction term, thereby enhancing the discriminating power between f ⁡ ( R ) f(R) gravity and the Λ \Lambda CDM model. However, we note that standard sirens do not offer a viable resolution to the Hubble tension in this analysis, as the inferred constraints are predominantly driven by the fiducial cosmologies adopted in the simulations. Regarding the specific models, we find that for the Hu-Sawicki scenario, several data combinations favor F R ​ R ​ 0 < 0 F_{RR0}<0 , implying potential theoretical instabilities. And, for the Starobinsky model, while EM-only constraints are nearly symmetric between the two parameter branches ( b < 0 b<0 and b > 0 b>0 ), the inclusion of SS constraints introduces mild asymmetries, revealing the sensitivity of SS observables to the curvature dependence of the theory. Future truly independent standard siren observations would be crucial for a definitive assessment of f ⁡ ( R ) f(R) gravity as an alternative to Λ \Lambda CDM. ∎ 1 1 1 e-mail: zhangyia@cqupt.edu.cn 1 Introduction Recent baryon acoustic oscillation (BAO) measurements from the Dark Energy Spectroscopic Instrument (DESI) DESI:2025zgx ; DESI:2025fii , especially when combined with cosmic microwave background (CMB) and Type Ia supernova (SNe Ia) observations, suggest a mild preference for departures from a pure cosmological constant and therefore motivate tests of a dynamical dark-energy sector. At the same time, despite extensive theoretical efforts, many cosmological models remain challenged by the Hubble tension in electromagnetic (EM) data, commonly defined as a >5\sigma discrepancy between local distance ladder and CMB inferred determinations of the Hubble constant H_{0} DiValentino:2020zio ; Abdalla:2022yfr ; Planck:2018vyg ; Riess:2021jrx ; Breuval:2024lsv ; Murakami:2023xuy . Such tensions may arise from unaccounted systematics or, more intriguingly, from physics beyond the standard \Lambda CDM framework. These considerations provide strong motivation to test modified gravity scenarios as alternatives to \Lambda CDM model. Among modified gravity theories, f(R) gravity offers a natural and minimal extension of general relativity (GR) that can account for late-time cosmic acceleration without invoking a fundamental cosmological constant, while still admitting a continuous \Lambda CDM limit. Phenomenologically, this can be interpreted as an effective dark energy equation of state w_{\rm eff}(z) that deviates from -1 and may evolve with redshift z . However, when constrained solely by EM probes of the background expansion, most viable f(R) models are notoriously difficult to distinguish from \Lambda CDM Nunes:2016drj ; Farrugia:2021zwx ; DAgostino:2020dhv ; Basilakos:2013nfa ; Matos:2021qne ; Leizerovich:2021ksf ; Bessa:2021lnr ; Jana:2018djs ; Amendola:2006we ; Capozziello:2007eu ; Fu:2010zza ; Hu:2016zrh ; Amendola:2006kh ; Odintsov:2020qzd ; Kumar:2025mzo ; He:2026yhe . This motivates the use of complementary observables with enhanced sensitivity to modified gravity effects. A key question is therefore whether f(R) gravity can provide a viable alternative to \Lambda CDM when confronted with forthcoming standard siren (SS) tests. Gravitational waves (GWs), a cornerstone prediction of GR, provide such an independent probe. Since the first detection of the binary black hole merger GW150914, hundreds of compact-binary coalescences have been observed LIGOScientific:2016aoc ; LIGOScientific:2017vwq ; LIGOScientific:2017zic ; LIGOScientific:2018mvr ; LIGOScientific:2020ibl ; LIGOScientific:2025slb . In particular, standard sirens yield a direct measurement of the luminosity distance D_{L}^{\rm SS} without relying on the cosmic distance ladder. In modified gravity, the SS luminosity distance can differ from its EM counterpart due to an effective friction term in GW propagation, providing a powerful discriminator between modified gravity and \Lambda CDM Belgacem:2017ihm ; Belgacem:2019pkk . Although the current number of SS events is still limited, future detectors such as the Einstein Telescope (ET) are expected to observe large samples of binary neutron star mergers (including lensed events), enabling high-precision tests of GR and \Lambda CDM Holz:2005df ; Schutz:1986gp . Following the simulation frameworks of Refs. gwbook ; Maggiore2007 ; Zhao:2010sz ; Cai:2016sby ; Su:2024avk ; Zhang:2021kqn , we will generate mock ET standard siren catalogs to assess whether SS observations can break the degeneracy between f(R) gravity and \Lambda CDM here. In this work we focus on two representative and widely studied f(R) models: the Hu-Sawicki and Starobinsky models Hu:2007nk ; Starobinsky:2007hu , which satisfy standard viability conditions and recover \Lambda CDM in an appropriate limit. Motivated by the bimodal posteriors discussed in Ref. Basilakos:2013nfa , we further split the Starobinsky model into two branches with b>0 and b<0 . Since EM data (SNe Ia, BAO, and CMB) alone do not provide decisive evidence for dynamical dark energy, we use the EM best fitted results as fiducial inputs for our SS simulations in order to isolate and quantify the additional impact of standard sirens. This paper is organized as follows. In Sec. 2 we introduce the Hu-Sawicki and Starobinsky models. In Sec. 3 we describe the EM and standard siren datasets used in the analysis. The cosmological constraints and discussion are presented in Sec. 4 . Finally, we summarize our conclusions in Sec. 5 . 2 The f(R) models In the metric formalism, the Einstein-Hilbert action is generalized by replacing the Ricci scalar R with a generic function f(R) . A general f(R) theory minimally coupled to matter is described by \displaystyle S=\int d^{4}x\,\sqrt{-g}\left[\frac{f(R)}{16\pi G}+\mathcal{L}_{m}\right], (1) where \mathcal{L}_{m} denotes the Lagrangian of the non-gravitational matter sector, including dark matter. 2 2 2 We neglect radiation, which is subdominant at late times. The function f(R) encodes deviations from general relativity (GR) and may account for the late-time cosmic acceleration. It is convenient to decompose \displaystyle f(R)=R+F(R), (2) so that F(R) measures the departure from GR. Varying the action with respect to the metric yields the modified Einstein equations. For a spatially flat Friedmann-Robertson-Walker (FRW) background, the Ricci scalar is \displaystyle R=12H^{2}+6\dot{H}, (3) and the modified Friedmann equations can be written as \displaystyle 3f_{R}H^{2}-\frac{Rf_{R}-f(R)}{2}+3Hf_{RR}\dot{R}=8\pi G\,\rho_{m}, (4) \displaystyle-2f_{R}\dot{H}=8\pi G\,\rho_{m}+\ddot{f}_{R}-H\dot{f}_{R}, (5) where overdots denote derivatives with respect to cosmic time t , \rho_{m} is the matter energy density, and f_{R}\equiv df/dR , f_{RR}\equiv d^{2}f/dR^{2} . The coupled system in Eqs. ( 4 ) and ( 5 ) typically does not admit closed-form solutions. In practice, viable models such as the Hu-Sawicki and Starobinsky forms are constructed to smoothly approach \Lambda CDM in an appropriate limit. Their functional form can be cast as \displaystyle f(R)=R-2\Lambda\,y(R,b), (6) where \Lambda is an effective cosmological constant and y(R,b) parametrizes the deviations from \Lambda CDM via a dimensionless parameter b . For b>0 , one has \lim_{b\to 0}f(R)=R-2\Lambda and \lim_{b\to\infty}f(R)=R . In particular, as F_{R}\to 0 and F_{RR}\to 0 the theory continuously reduces to \Lambda CDM. Although originally proposed as cosmological constant free models, both the Hu-Sawicki and Starobinsky constructions effectively contain a \Lambda term in the relevant limit Starobinsky:2007hu ; Hu:2007nk . Defining the normalized Hubble rate E(a)\equiv H/H_{0} , an effective dark-energy equation of state can be expressed as \displaystyle w_{\rm eff}(a)=\frac{-1-\frac{2a}{3}\frac{d\ln E}{da}}{1-\Omega_{m}(a)}, (7) where \Omega_{m}(a)=\Omega_{m0}a^{-3}/E^{2}(a) . In the \Lambda CDM limit, w_{\rm eff}=-1 . Following Basilakos et al. Basilakos:2013nfa , we adopt an efficient approximation scheme based on expanding the Hubble rate around the \Lambda CDM solution, which avoids expensive numerical integrations and has been widely used Nunes:2016drj ; Farrugia:2021zwx ; DAgostino:2019hvh ; DAgostino:2020dhv . Rewriting Eq. ( 4 ) in terms of the e-folding number N=\ln a , we obtain \displaystyle-f_{R}H^{2}(N)+\Omega_{m0}e^{-3N}+\frac{1}{6}\bigl(Rf_{R}-f\bigr)=f_{RR}H^{2}(N)\,R^{\prime}(N), (8) where the prime “ \prime ” denotes derivative with respect to N . To quantify deviations from \Lambda CDM we expand around b=0 ; keeping the first two nonzero terms reproduces the numerical solution to better than 0.001\% for b\in[0.001,0.5] . Specifically, \displaystyle H^{2}(N)=H_{\Lambda}^{2}(N)+\sum_{i=1}^{M}b^{i}\,\delta H_{i}^{2}(N), (9) \displaystyle H_{\Lambda}^{2}(N)=\Omega_{m0}e^{-3N}+\bigl(1-\Omega_{m0}\bigr), (10) where M denotes the truncation order. The dynamical properties of f(R) models can be conveniently characterized by two dimensionless variables Amendola:2006we ; Tsujikawa:2007xu , \displaystyle r\equiv-\frac{Rf_{R}}{f(R)}=-\frac{R\bigl(1+F_{R}\bigr)}{f(R)}, (11) \displaystyle m\equiv\frac{Rf_{RR}}{f_{R}}=\frac{R\,F_{RR}}{1+F_{R}}. (12) Here, f_{R}=1+F_{R} and f_{RR}=F_{RR} . The quantity f_{R} controls the effective gravitational coupling, G_{\rm eff}\propto f_{R}^{-1} , whereas f_{RR} governs the dynamics of the additional scalar degree of freedom (the scalaron) and thus sets how rapidly deviations from GR/ \Lambda CDM can develop. The variable r measures the proximity to the \Lambda CDM limit, while the variable m quantifies the strength of the deviation; in many dynamical analyses m<0 signals an instability Amendola:2006we . Viable cosmologies correspond to trajectories in the (r,m) plane that pass close to the GR limit (r,m)\simeq(-1,0^{+}) during the radiation and matter eras and then approach a stable de Sitter fixed point near r\simeq-2 with 00 for viable f(R) models Yang:2011cp . Expanding the Friedmann equation yields \displaystyle H_{\rm Star}^{2}(N)=H_{\Lambda}^{2}(N)+b^{2}\,\delta H_{2}^{2}(N)+b^{4}\,\delta H_{4}^{2}(N)+\cdots, (24) where the leading correction appears at {\cal O}(b^{2}) . The explicit forms of \delta H_{2}^{2}(N) and \delta H_{4}^{2}(N) are given in Ref. Basilakos:2013nfa as well. For the Starobinsky case, EM constraints on b may exhibit a bimodal structure Kumar:2025mzo . To capture this feature, we impose the prior b\neq 0 , which partitions the parameter space into two branches, b<0 and b>0 , denoted as StarobinskyI and StarobinskyII (StarI and StarII for short). 3 Data and methodology We constrain the parameters of the f(R) models using a Markov Chain Monte Carlo (MCMC) analysis together with a maximum-likelihood estimation, implemented with CosmoMC Lewis:2002ah . In this section, we shall describe the electromagnetic datasets and the simulated standard siren data used in our analysis. 3.1 Electromagnetic datasets We consider EM probes of the cosmic distance scale, including measurements of the Hubble parameter from cosmic chronometers (CC) Moresco:2020fbm ; Favale:2023lnp , Type Ia supernovae from the PantheonPlus (PP) compilation Scolnic:2021amr ; Brout:2022vxf , CMB distance priors from Planck Planck:2018vyg , and BAO measurements from the DESI Data Release 2 (hereafter DESI2) DESI:2025zgx ; DESI:2025fii . The PP and CC data mainly probe the late-time Universe, whereas CMB and BAO measurements are also sensitive to earlier epochs. 3.1.1 PantheonPlus + cosmic chronometers (PP+CC) The PantheonPlus compilation provides 1701 distance modulus measurements from 18 independent surveys Scolnic:2021amr ; Brout:2022vxf over the redshift range 0.001220 ) models. For the corresponding standard siren simulations, the EM best-fit values from PP+CC/Planck/DESI2 are adopted as fiducial (baseline) parameters. Specifically, the mock catalogs StarI: \mathrm{SS}_{\rm PP+CC} , StarI: \mathrm{SS}_{\rm DESI2} , and StarI: \mathrm{SS}_{\rm Planck} are generated using (\Omega_{m0}^{\rm fid},H_{0}^{\rm fid},b^{\rm fid})=(0.293,73.05,-1.05\times 10^{-3}) , (0.325,64.32,-0.99\times 10^{-3}) , and (0.301,67.97,-0.53\times 10^{-3}) , respectively. Likewise, the mock catalogs StarII: \mathrm{SS}_{\rm PP+CC} , StarII: \mathrm{SS}_{\rm DESI2} , and StarII: \mathrm{SS}_{\rm Planck} are generated using (\Omega_{m0}^{\rm fid},H_{0}^{\rm fid},b^{\rm fid})=(0.293,73.08,1.05\times 10^{-3}) , (0.325,64.32,0.99\times 10^{-3}) , and (0.301,67.97,0.53\times 10^{-3}) , respectively. Table 2: Constraints on the derived quantities (F_{R0},\,F_{RR0},\,r_{0},\,m_{0}) and the information criteria (\Delta{\rm AIC},\,\Delta{\rm BIC}) for the Hu-Sawicki and Starobinsky models. Since HS: \mathrm{SS}_{\Lambda{\rm CDM:PP+CC}} is generated using the \Lambda CDM best-fit fiducial parameters (\Omega_{m0},\,H_{0},\,b)=(0.348,\,72.99,\,0) , we also report the corresponding AIC and BIC comparison between HS: \mathrm{SS}_{\Lambda{\rm CDM:PP+CC}} and \Lambda CDM: \mathrm{SS}_{\rm PP+CC} . Data F_{R0}(10^{-3}) F_{RR0}(10^{-7}) r_{0} m_{0}(10^{-3}) \Delta AIC \Delta BIC HS:PP+CC -47.72^{+14.21+47.45}_{-26.10-41.00} 12.75^{+4.15+4.71}_{-0.42-11.49} -1.602^{+0.104+0.166}_{-0.055-0.182} 86.01^{+48.78+62.60}_{-16.09-80.55} -0.8 4.7 HS:SS ΛCDM:PP+CC -12.00^{+17.82+56.03}_{-37.58-47.58} 3.51^{+14.49+15.64}_{-59.29-23.45} -1.726^{+0.156+0.203}_{-0.077-0.244} 21.32^{+71.12+88.35}_{-10.89-67.87} 2.5 7.4 HS:PP+CC+SS ΛCDM:PP+CC -16.62^{+14.18+50.51}_{-29.22-39.39} 5.30^{+10.36+12.46}_{-2.94-20.89} -1.716^{+0.130+0.155}_{-0.046-0.218} 30.80^{+56.21+73.29}_{-38.72-38.24} 2.4 8.3 HS: Planck 10.46^{+1.30+215.33}_{-84.51-96.43} -13.53^{+41.79+45.24}_{-10.09-130.32} -1.955^{+0.422+67.129}_{-0.219-28.968} -32.53^{+163.31+174.86}_{-19.69-467.53} 3.9 3.0 HS:SS Planck 4.35^{+16.93+125.93}_{-60.57-77.75} -8.35^{+29.39+32.58}_{-0.36-98.91} -1.830^{+0.267+0.338}_{-0.065-0.570} -14.56^{+121.36+148.31}_{-23.75-265.68} - - HS: Planck+SS Planck -21.46^{+7.04+15.86}_{-8.49-15.38} 9.40^{+3.66+6.22}_{-2.73-9.04} -1.714^{+0.044+0.078}_{-0.035-0.080} 41.45^{+16.30+28.19}_{-12.73-39.76} - - HS:DESI2 -4.24^{+14.59+46.88}_{-27.24-39.00} 1.31^{+12.22+16.53}_{-5.32-21.51} -1.805^{+0.134+0.190}_{-0.068-0.228} 7.15^{+54.69+75.27}_{-26.13-94.49} 2.9 3.4 HS:SS DESI2 -43.90^{+14.23+20.44}_{-41.74-54.21} 10.95^{+8.24+9.28}_{-1.96-25.33} -1.628^{+0.175+0.235}_{-0.066-0.312} 76.78^{+77.62+81.20}_{-9.39-132.6} - - HS:DESI2+SS DESI2 -11.64^{+11.73+19.53}_{-10.90-18.28} 4.87^{+4.77+7.39}_{-3.02-8.21} -1.767^{+0.053+0.089}_{-0.039-0.094} 22.67^{+21.81+35.13}_{-22.80-38.08} - - StarI:PP+CC -0.92^{+0.28+0.57}_{-0.28-0.56} 0.57^{+0.18+0.40}_{-0.21-0.39} -1.824^{+0.031+0.032}_{-0.022-0.028} 2.77^{+0.84+1.67}_{-0.84-1.72} -2.4 3.1 StarII:PP+CC -0.92^{+0.28+0.60}_{-0.28-0.56} 0.57^{+0.19+0.40}_{-0.21-0.41} -1.825^{+0.032+0.055}_{-0.022-0.059} 2.76^{+0.85+1.68}_{-0.91-1.79} -2.4 3.1 StarI:SS PP+CC -1.01^{+0.35+0.43}_{-0.07-0.75} 0.67^{+0.02+0.73}_{-0.32-0.37} -1.816^{+0.010+0.063}_{-0.031-0.040} 3.04^{+0.20+2.24}_{-1.06-1.31} - - StarII:SS PP+CC -1.01^{+0.27+0.40}_{-0.11-0.52} 0.64^{+0.06+0.47}_{-0.23-0.32} -1.815^{+0.012+0.045}_{-0.023-0.037} 3.02^{+0.33+0.15}_{-0.79-1.19} - - StarI:PP+CC+SS PP+CC -0.92^{+0.12+0.22}_{-0.09-0.21} 0.56^{+0.07+1.6}_{-0.08-0.16} -1.823^{+0.010+0.021}_{-0.010-0.020} 2.75^{+0.29+0.64}_{-0.34-0.65} - - StarII:PP+CC+SS PP+CC -0.97^{+0.10+0.20}_{-0.10-0.22} 0.60^{+0.07+0.17}_{-0.08-0.15} -1.817^{+0.010+0.019}_{-0.010-0.020} 2.92^{+0.32+0.66}_{-0.32-0.62} - - StarI: Planck -0.95^{+0.02+0.03}_{-0.01-0.03} 0.66^{+0.01+0.03}_{-0.01-0.02} -1.819^{+0.001+0.002}_{-0.002-0.003} 2.84^{+0.04+0.08}_{-0.04-0.07} 1.7 0.8 StarII: Planck -0.95^{+0.02+0.03}_{-0.01-0.03} 0.66^{+0.01+0.03}_{-0.01-0.02} -1.819^{+0.001+0.002}_{-0.002-0.003} 2.84^{+0.04+0.08}_{-0.04-0.07} 1.7 0.8 StarI:SS Planck -1.26^{+0.33+0.47}_{-0.07-0.85} 1.01^{+0.02+1.04}_{-0.37-0.49} -1.794^{+0.008+0.063}_{-0.026-0.039} 3.79^{+0.20+2.53}_{-1.00-1.42} - - StarII:SS Planck -1.12^{+0.29+0.46}_{-0.01-1.31} 0.87^{+0.05+1.68}_{-0.32-0.46} -1.806^{+0.004+0.093}_{-0.024-0.041} 3.36^{+0.04+3.90}_{-0.87-1.37} - - StarI: Planck+SS Planck -0.95^{+0.02+0.02}_{-0.01-0.02} 0.66^{+0.01+0.03}_{-0.01-0.02} -1.820^{+0.001+0.003}_{-0.001-0.001} 2.84^{+0.03+0.07}_{-0.04-0.06} - - StarII: Planck+SS Planck -0.94^{+0.01+0.02}_{-0.02-0.03} 0.66^{+0.01+0.02}_{-0.01-0.02} -1.820^{+0.001+0.003}_{-0.001-0.002} 2.83^{+0.04+0.08}_{-0.04-0.07} - - StarI:DESI2 -1.15^{+0.32+0.32}_{-0.13-0.39} 0.94^{+0.15+0.51}_{-0.36-0.40} -1.802^{+0.013+0.032}_{-0.022-0.028} 3.45^{+0.40+1.16}_{-0.80-0.97} -0.4 0.1 StarII:DESI2 -1.15^{+0.27+0.32}_{-0.13-0.39} 0.94^{+0.15+0.51}_{-0.36-0.40} -1.802^{+0.013+0.032}_{-0.022-0.028} 3.45^{+0.40+1.16}_{-0.80-0.97} -0.4 0.1 StarI:SS DESI2 -1.24^{+0.10+0.25}_{-0.14-0.27} 0.99^{+0.12+0.26}_{-0.12-0.24} -1.794^{+0.011+0.022}_{-0.008-0.021} 3.72^{+0.40+0.82}_{-0.31-0.77} - - StarII:SS DESI2 -1.47^{+0.33+0.46}_{-0.19-0.58} 1.30^{+0.19+0.72}_{-0.42-0.55} -1.777^{+0.017+0.043}_{-0.024-0.035} 4.42^{+0.57+1.73}_{-1.01-1.39} - - StarI:DESI2+SS DESI2 -1.22^{+0.18+0.27}_{-0.12-0.32} 1.01^{+0.12+0.37}_{-0.21-0.31} -1.796^{+0.011+0.026}_{-0.014-0.023} 3.67^{+0.34+0.95}_{-0.54-0.83} - - StarII:DESI2+SS DESI2 -1.28^{+0.16+0.28}_{-0.13-0.32} 1.10^{+0.15+0.40}_{-0.21-0.30} -1.791^{+0.011+0.025}_{-0.013-0.023} 3.84^{+0.40+0.97}_{-0.50-0.83} - - Figure 3: The evolutions of w_{\rm eff} for the Hu-Sawicki (left), StarobinskyI (middle) and StarobinskyII (right) models, which are based on the the 1\sigma regimes of PP+CC+SS ΛCDM:PP+CC /PP+CC+SS PP+CC , DESI2+SS DESI2 and Planck+SS Planck constraining results. Figure 4: The evolutions of F_{R} , F_{RR} , r , and m for the Hu-Sawicki (upper), StarobinskyI (middle) and StarobinskyII (bottom) models, which are based on the the 1\sigma regimes of PP+CC+SS ΛCDM:PP+CC /PP+CC+SS PP+CC , DESI2+SS DESI2 and Planck+SS Planck constraining results. 4 Discussion The cosmological constraints are summarized in Tables 1 and 2 . Generally, both the Hu-Sawicki and Starobinsky f(R) models are consistent with the data and can account for the late-time accelerated expansion. For each dataset, the best-fit \chi^{2}_{\min} is comparable to the number of data points. In particular, the standard siren dataset yields \chi^{2}_{\min}\simeq 1000 , indicating good internal consistency of the simulations. As expected, the joint EM+SS analyses provide the tightest constraints for all models, with the EM data contributing the dominant statistical weight.The impact of SS data on the Hubble tension is intrinsically dependent on the adopted simulations and fiducial cosmology, since the H_{0} values inferred from SS analyses largely depend on the assumed fiducial model. Consequently, SS datasets constructed with similar fiducial assumptions tend to yield comparable levels of Hubble tension. For the reference \Lambda CDM model, among the EM datasets, the Planck provides the tightest constraints on \Omega_{m0} and H_{0} . The 1\sigma confidence regions from PP+CC, DESI2, and Planck are clearly separated in the \Omega_{m0}-H_{0} plane, in which the PP+CC constraint favors a substantially larger H_{0} that is lying outside the 2\sigma regions inferred from Planck and DESI2. Within \Lambda CDM, SS datasets based on Planck or DESI2 fiducials may mildly reduce the inferred tension, whereas \mathrm{SS}_{\mathrm{PP+CC}} favors a higher H_{0} because it assumes a larger fiducial Hubble constant. As a result, PP+CC+ \mathrm{SS}_{\mathrm{PP+CC}} yields a tension of \sim 0.34\sigma , compared with 3.82\sigma and 6.06\sigma for the Planck and DESI2 related combinations, respectively, within the adopted simulations. As shown in Fig. 2 , for all considered f(R) models Planck yields the overall tightest constraints, while PP+CC-related combinations prefer larger values of H_{0} as \Lambda CDM model. Figs. 1 and 2 further show that our SS simulations are in good agreement with the PantheonPlus sample. For the Hu-Sawicki model, the SS luminosity-distance evolution follows a similar ordering to that in \Lambda CDM, \displaystyle D_{L}^{\mathrm{SS}_{\mathrm{PP+CC}}}0 . Consistently, Fig. 3 shows that the effective equation-of-state parameter w_{\rm eff}(z) displays oscillations with redshift. Also, w_{\rm eff} can cross the phantom divide and evolve from w_{\rm eff}<-1 to w_{\rm eff}>-1 at the redshift range 00 , which we denote as StarobinskyI and StarobinskyII, respectively. As shown in Table 1 , the best-fit \chi^{2} values of the Starobinsky model are smaller than those of \Lambda CDM for all datasets. In terms of model selection, only the PP+CC dataset yields \Delta{\rm BIC}=3.1 , corresponding to weak positive evidence against \Lambda CDM. For the remaining datasets, the information criteria do not provide decisive discrimination, indicating that \Lambda CDM and the Starobinsky model are statistically comparable. The PP+CC-related combinations systematically favor higher values of H_{0} than the Planck and DESI2 related combinations, thereby leading to a reduced Hubble tension, similar to the other considered models. Relative to the 5.87\sigma tension obtained from Planck alone in the StarobinskyI and StarobinskyII branches, adding \mathrm{SS}_{\rm Planck} mildly alleviates the tension to 4.70\sigma/4.73\sigma . In contrast, for PP+CC and DESI2, the inclusion of SS data tends to increase the inferred tension in the Starobinsky model. For instance, in the StarobinskyI model, the \mathrm{SS}_{\rm PP+CC} and \mathrm{SS}_{\rm DESI2} datasets yield tensions of approximately 0.55\sigma and 9.07\sigma respectively, whereas the corresponding EM datasets PP+CC/DESI2 give \sim 0.11\sigma and 1.88\sigma tensions, respectively. Owing to the imposed prior b\neq 0 , the posterior distributions for the Planck-related combinations do not always exhibit Gaussian shapes (see Fig. 2 ). In particular, the \mathrm{SS}_{\rm DESI2} posterior in StarobinskyI shows a clear non-Gaussian feature as well. Nevertheless, for EM-only constraints the two branches remain approximately mirror symmetric, again reflecting the evenness of the model. The \Omega_{m0} - H_{0} contours inferred from PP+CC, Planck, and DESI2 are clearly separated, indicating substantial inconsistencies among the EM datasets. Meanwhile, the SS simulations produce nearly parallel confidence contours, although the sizes of the allowed regions vary. Among the SS cases, the overall constraining power (as inferred from contour areas) follows \displaystyle\mathrm{SS}_{\rm DESI2}<\mathrm{SS}_{\rm PP+CC}\leq\mathrm{SS}_{\rm Planck}. Compared with EM only constraints, adding SS data systematically shifts \Omega_{m0} to larger values and H_{0} to smaller values. Despite the fact that the SS simulations are generated using symmetric EM best-fit values, the resulting SS constraints are not perfectly symmetric where the StarobinskyI branch typically exhibits a larger deviation from \Lambda CDM than StarobinskyII for the same |b| . This suggests that standard siren observables are sensitive to the detailed curvature dependence of the f(R) function. Although the Starobinsky model is even in b , cosmological evolution probes only the R>0 branch. And the background dynamics depend nonlinearly on F_{R} and F_{RR} . For the same |b| , the b<0 branch typically approaches the late-time de Sitter attractor more slowly, which can lead to a larger accumulated deviation in the luminosity distance measured by standard sirens. The \Omega_{m0}-b correlation further illustrates the potential of the Starobinsky model to break parameter degeneracies in a data dependent manner. Specifically, Planck and DESI2 related constraints show a negative correlation, whereas PP+CC favors a positive correlation; the SS constraints also tend to prefer a negative correlation. Although PP+CC alone provides the loosest bounds, its combination with SS data becomes significantly tighter which demonstrates the degeneracy breaking power of standard sirens within this model. The parameter correlations in the \Omega_{m0}-b , H_{0}-b , and w_{\rm eff0}-b planes shift signs between StarobinskyI and StarobinskyII because of the mirror symmetry. In particular, w_{\rm eff0} and b are negatively correlated for b<0 , but positively correlated for b>0 , producing a “butterfly”-like structure analogous to that found in the Hu-Sawicki case. At the same time, w_{\rm eff}(z) evolves from the quintessence regime to the phantom regime which is a stable behavior. Notably, the DESI2-related combinations can exclude the \Lambda CDM limit b=0 (equivalently w_{\rm eff0}=-1 ) at the 2\sigma level, providing evidence for dynamical dark energy within the Starobinsky framework. For the Starobinsky model, we constrain F_{R0} at the level of \mathcal{O}(10^{-3}) , tighter than that in the Hu-Sawicki model (typically \mathcal{O}(10^{-2}) ), and we obtain F_{RR0}\sim\mathcal{O}(10^{-7}) , which is positive and also smaller than the Hu-Sawicki case (typically \mathcal{O}(10^{-6}) ). Moreover, all data combinations exclude F_{R0}=0 , F_{RR0}=0 , r_{0}=-2 , and m_{0}=0 at the 2\sigma level for the Starobinsky model. We find that the redshift evolution of w_{\rm eff} in the combined DESI2 data, together with the trajectories of (F_{R},F_{RR},r,m) , can clearly distinguish the Starobinsky scenario from \Lambda CDM model. 5 Conclusion In this work, we first constrained the Hu-Sawicki and Starobinsky f(R) models with electromagnetic datasets (PP+CC, Planck, and DESI2) and then simulated standard siren catalogs based on the constrained results. Both f(R) scenarios provide global fits comparable to \Lambda CDM, and the combined EM+SS analyses yield the tightest constraints. The simulated SS data offer complementary sensitivity via the modified GW friction term, thereby enhancing the ability to discriminate f(R) gravity from \Lambda CDM. However, in our simulation-based setup, the inferred Hubble tension is largely dominated by the assumed SS fiducial cosmology. Consequently, while standard sirens can help distinguish modified gravity from \Lambda CDM, they do not provide a viable solution to the Hubble tension in this context. For the Hu-Sawicki model, SS data can mildly reduce the tension, but several data combinations prefer F_{RR0}<0 , suggesting potential theoretical instabilities. We conclude that the Hu-Sawicki model may be disfavored and potentially ruled out by future high-precision standard siren data. For the Starobinsky model, EM-only constraints are nearly symmetric between the b<0 and b>0 branches, whereas SS constraints show mild asymmetries, indicating sensitivity to the curvature dependence of f(R) . The redshift evolution of w_{\rm eff} in the combined DESI2 dataset, together with the trajectories of (F_{R},F_{RR},r,m) , clearly distinguishes the Starobinsky scenario from the \Lambda CDM model. Given the current inconsistencies in the data, future standard siren observations will be essential for a definitive assessment of f(R) gravity as an alternative to the standard cosmological paradigm. 6