Arxiv:2007.02943V2 [Astro-Ph.CO] 26 Jan 2021

Accurate precision Cosmology with redshift unknown gravitational wave sources

Suvodip Mukherjee,1, 2, 3, ∗ Benjamin D. Wandelt,4, 5, 6, † Samaya M. Nissanke,1, ‡ and Alessandra Silvestri2, § 1Gravitation Astroparticle Physics Amsterdam (GRAPPA), Anton Pannekoek Institute for Astronomy and Institute for High-Energy Physics, University of Amsterdam, Science Park 904, 1090 GL Amsterdam, The Netherlands 2Institute Lorentz, Leiden University, PO Box 9506, Leiden 2300 RA, The Netherlands 3Delta Institute for Theoretical Physics, Science Park 904, 1090 GL Amsterdam, The Netherlands 4Institut d’Astrophysique de Paris (IAP), UMR 7095, CNRS/UPMC UniversitéParis 6, Sorbonne Universités,98 bis boulevard Arago, F-75014 Paris, France 5Institut Lagrange de Paris (ILP), Sorbonne Universités,98 bis Boulevard Arago, 75014 Paris, France 6Center for Computational Astrophysics, Flatiron Institute, 162 5th Avenue, New York, NY 10010, USA (Dated: January 28, 2021) Gravitational waves can provide an accurate measurement of the luminosity distance to the source, but cannot provide the source redshift unless the degeneracy between mass and redshift can be broken. This makes it essential to infer the redshift of the source independently to measure the expansion history of the Universe. We show that by exploiting the clustering scale of the gravitational wave sources with galaxies of known redshift, we can infer the expansion history from redshift unknown gravitational wave sources. By using gravitational wave sources of unknown redshift that are detectable from the network of gravitational wave detectors with Advanced LIGO design sensitivity, we will be able to obtain accurate and precise measurements of the local Hubble constant, the expansion history of the universe, and the gravitational wave bias parameter, which captures the distribution of gravitational wave sources with respect to the redshift tracer distribution. While we showcase its application to low redshift gravitational waves, this technique will be applicable also to the high redshift gravitational wave sources detectable from Laser Interferometer Space Antenna (LISA), Cosmic Explorer (CE), and Einstein Telescope (ET). Moreover, this method will also be applicable to samples of supernovae and fast radio bursts with unknown or photometric redshifts.

I. INTRODUCTION H0LiCOW [16] also supports the mismatch. The discrepancy in the value of H0 between early-time and late- The measurement of the current expansion rate of the time probes is more than 4σ [17]. We shall note that independent, late-time measurements that use the Tip of Universe, known as Hubble constant (denoted by H0), as well as its value at different cosmological redshifts, is the Red Giant Branch (TRGB) to calibrate SNe have re- one of the key science goals in the field ofcosmology. cently given H0 = 69.8 ± 0.8 (stat)±1.7 (sys) km/s/Mpc This endeavour, which started with the first measure- [18], which reduces the discrepancy significantly. There are also studies that propose possible sources of system- ment of H0 by Edwin Hubble [1] has been typically performed via electromagnetic probes which can be classi- atics in the late-time measurements of H0 [19, 20]. As fied as standardized candles (e.g., supernovae (SNe)) [2– of yet, there is no conclusive evidence that settles this 4], standard rulers (e.g., cosmic microwave background mismatch by either any systematic, and/or invoking new (CMB), baryon acoustic oscillation (BAO)) [5–9], anda physics. Independent probes are required to settle this standard clock [10–13]. All these probes have become in- discrepancy. creasingly successful in making precise measurements of The direct detection of gravitational waves has recently H0, but have failed to converge to values which are con- offered a new independent probe of cosmic expansion. sistent with each other within their error-bars (including From the gravitational wave chirp generated by compact both statistical and known systematic uncertainties). In object binary mergers, one can infer the luminosity dis- fact, low redshift probes such as SNe [4] indicate a value tance of the source [21–27] leading to gravitational wave of H = 74 ± 1.4 km/s/Mpc , whereas the probes which sources being dubbed standard sirens. Interestingly the arXiv:2007.02943v2 [astro-ph.CO] 26 Jan 2021 0 depend on the high redshift Universe such as big bang intrinsic luminosity of the gravitational wave source de- nucleosynthesis (BBN), CMB, BAO indicate a value of pends on the chirp mass, and its evolution with the fre- H0 = 67.4 ± 0.5 km/s/Mpc [14, 15]. An independent quency is solely dictated by the general theory of rela- +1.7 tivity [21, 22], without the need of external calibration. measurement of H0 = 73.8−1.8 km/s/Mpc from the time delay of the strongly-lensed low redshift events by the The only limiting factors are any systematic uncertainties arising from the gravitational wave detector calibration [28, 29] and statistical uncertainty arising from the de- termination of the inclination angle, which is degenerate ∗ [email protected], [email protected] with the luminosity distance in setting the strain [25]. † [email protected] Though standard sirens are promising, using them for ‡ [email protected] the measurement of the expansion history requires an § [email protected] independent measurement of their redshift. The grav- 2

FIG. 1: We show the normalised posterior of the Hubble constant H0 = 100 h0 km/s/Mpc for different numbers of gravitational wave sources distributed up to redshift z = 0.5 for the sky localization error ∆ΩGW = 10 sq. deg. after marginalizing over cosmological parameter Ωm and nuisance parameters related to the gravitational wave bias parameters bGW (z). The constraints are also similar for ∆ΩGW = 25 sq. deg. In vertical dashed line we show the region between 16th and 84th percentile of the distribution for each cases. The vertical magenta dashed line denotes the injected value of h0 = 0.7 indicating reliable recovery in all cases. For comparison we also plot the measured value of h0 = 0.674 ± 0.005 by the Planck collaboration [8] and the value of h0 = 0.74 ± 0.014 by the SH0ES Team [4]. itational wave signal alone does not provide this infor- [35, 36]. Another way to reduce the error-bar on the value 1 mation in the absence of a known scale arising from ei- of H0 is by measurement of the inclination angle by either the tidal deformation [30], or the mass-gap in the ther measuring the two polarization states of the gravi- binary black hole (BBH) sources due to pair-instability tational wave signal using an expanded network of three supernova [31]. Another possibility to determine the or more gravitational wave detectors [25], or by using the redshift is by identifying the host galaxy using a co- higher-order multipole moments of the gravitational wave incident detection of an electromagnetic (EM) counter- signal [37]. Measurement of the inclination angle is also part from the gravitational wave source. The first-ever possible by accurately modeling the EM emission from multi-messenger detection happened for the binary neu- the jet of the gravitational wave source (e.g., [38–40]), tron star merger GW170817 detected by the Laser Inter- though this method may introduce astrophysical model- ferometer Gravitational-wave Observatory (LIGO) Sci- ing uncertainties. entific and Virgo Collaborations (LVC), for which the Consistent with the only definitive binary neutron star GRB170817A electromagnetic counterpart was observed, (BNS) detection with an EM counterpart so far [41, 42], leading to an independent measurement of the Hubble the expected number of gravitational wave sources with +12 constant H0 = 70−8 km/s/Mpc[14]. As shown in [32– an EM counterpart in the cosmic volume that can be 34], the joint estimation of the electromagnetic signal and explored by the Advanced LIGO/Virgo and KAGRA de- gravitational wave signal requires peculiar velocity cor- tectors is expected to be small since; in fact, only a frac- rection to the gravitational wave sources. In general, the tion of the total BNS and neutron star black hole (NS- error bar on H0 is more than 15% and is currently not BH) systems events are expected to have a detectable competitive with the measurements from CMB (< 1%) and SNe (∼ 1.5%). However in the future, with the measurement of a large number of sirens (∼ 50) with EM 1 The angle between the line of sight and the system’s orbital an- counterparts, one can achieve a 2% measurement of H0 gular momentum. 3

EM counterpart [43–47]. Successful detection of the EM find the redshift of the source is by exploring any mass counterpart requires its flux to be higher than the detec- scale associated with the compact objects originating due tion threshold of follow-up telescopes. It also requires the to the neutron star mass distribution [70, 71], the neu- sky localization area of the gravitational wave source to tron star tidal deformation [30], or using the mass-gap be small enough to do a fast search of the EM counterpart in the gravitational wave source population due to the before it fades away[48]. As a result, BNS and NS-BH pair-instability supernova [31]. systems, which are farther away with poor sky localiza- In this work, we explore a method that can be ap- tion, may not have a detectable EM counterpart, similar plied up to high redshift (up to which galaxy samples to the possible BNS (or NS-BH) event GW190425[46]. are going to be available) and can measure the value All these issues can be a serious bottleneck for measuring of H0 along with the density of dark energy, the equa- H0 using gravitational wave sources in the timescale of tion of state of dark energy, and also the spatial dis- ten years with a precision of ∼ 2% [35, 36, 49, 50]. tribution of black holes with respect to the dark mat- Gravitational wave sources, such as BBHs which have ter distribution. The galaxy catalogs from the com- higher intrinsic luminosity that can be detected at farther bination of several ongoing/upcoming surveys (such as distances in comparison to BNS systems, granting us ac- SDSS/BOSS, the Dark Energy Survey (DES) [72], the cess to a larger detectable cosmic volume. However, the Dark Energy Spectroscopic Instrument (DESI) [73], Eu- majority of BBHs which are detectable in the frequency clid [74], the Nancy Grace Roman Telescope 3 [75–77], band of the Advanced LIGO/Virgo detectors are not ex- the Rubin Observatory [78], Spectro-Photometer for the pected to have an EM counterpart by themselves, un- History of the Universe, Epoch of Reionization, and Ices less there is a presence of baryons surrounding the BBH, Explorer (SPHEREx) [79]) will be available up to redshift where a candidate was recently announced [51]. We refer z = 3. The combination of different missions will be able to the astrophysical systems without any EM counter- to cover nearly the full sky. We exploit the fact that both parts as dark standard sirens. Due to the absence of the the gravitational wave sources and galaxies are tracers of EM counterpart, identification of the host galaxy is not the matter density, and therefore, they are spatially cor- possible, and hence their redshift cannot be identified in related through the underlying matter field, to infer the the standard way. An alternative approach is required redshifts of dark standard sirens [21, 80, 81]. We build to exploit the expected large number of dark standard on previous work, where clustering with galaxies was ap- sirens for a measurement of the Hubble constant. plied to redshift unknown (or photometrically known) A possibility is to statistically obtain the host galaxy of SNe [81]. Our method does not identify the host galaxy of dark sirens from galaxy catalogs [35, 52–54]. An applica- the BBH source but finds its host redshift shell by explor- tion of this for the existing gravitational wave data was ing the three-dimensional spatial cross-correlation of the performed in previous studies [55–59]. These methods gravitational wave sources with redshift-known galaxies. can be promising but are not optimal, as we will discuss Host galaxy identification is, therefore, at the limit of our in the following section. Forecast studies of this method approach that only exploits very small, galaxy-scale cor- report the possibility of making H0 measurement at the relations [81]. The exploitation of the clustering aspect level of 5.8% in the future with 50 objects [35, 53, 57]2 is also implemented to identify the redshift distribution from only the low redshift sources and keeping the value of the galaxies [82–84]. of matter density of the Universe Ωm fixed. These meth- We detail the formalism of this method and the like- ods associate a probability to each galaxy as a possible lihood setup in Sec.II and Sec.III, respectively. Our host of the dark sirens [57], and is only effective up to low method does not require making any additional assump- redshift when the number of galaxies is limited. However, tion about the redshift dependence of the merger rate if the method is applied to the high redshift sources, then of gravitational waves sources but only requires that the the possible host along a particular direction of the sky is BBH mergers trace galaxies (incorporating the possibility going to be large in number, and as a result, the method is of natal birth kicks), so that there is a spatial correlation, not informative enough to choose the correct galaxy as a as discussed in Sec.IV. We show a forecast for the ac- host. As a result, it restricts the use of dark sirens to low curacy and precision of the measurements of H0 achiev- redshift even if accurate distance measurement is possi- able with our method in Fig.1 after marginalizing over ble for sources at high redshift from the LIGO/Virgo de- the matter density Ωm, and the redshift-dependent grav- α sign sensitivity [60, 61], and from the upcoming gravita- itational wave bias parameter bGW (z) = bGW (1 + z) . tional wave detectors such as the Kamioka Gravitational Details about this result are given in Sec.V. Moreover, Wave Detector (KAGRA) [62], LIGO-India [63], Laser since dark sirens can be detected up to high redshift, this Interferometer Space Antenna (LISA) [64], the Einstein method also makes it possible to explore the expansion Telescope (ET) [65], the Cosmic Explorer (CE) [66], and history of the Universe and provide an independent mea- the TianQin observatory [67–69]. An alternative way to surement of the cosmological parameters related to mat-

2 √ 3 Scaling the previous bounds from [35] and [57] as 1/ NGW in- Previously known as Wide-Field InfraRed Survey Telescope dicates similar error-bar. (WFIRST) 4 ter density Ωm, dark energy equation-of-state w0, and its formation of the real space galaxy density field δg(r). redshift dependence w(z) = wa(z/(1 + z)). This method The galaxy bias parameter encodes how galaxies trace can also explore the bias parameter of the gravitational the dark matter distribution [88]. A spectroscopic (or wave sources at different redshifts bGW (z), which will photometric) survey observes galaxies in the redshift capture its spatial distribution with respect to dark mat- space denoted by superscript s. The motion of the galax- ter. This method will also be applicable to the multi- ies along the line of sight induces a distortion in the messenger test of gravity proposed in [85, 86]. The position of the galaxies in redshift space, which causes breadth of the scientific returns possible from this avenue anisotropy in the statistical properties of the observed surpasses that of the statistical host identification meth- density field, known as redshift space distortion (RSD) ods [35, 53, 57]. For comparison, we apply our method [89, 90]. On large scales, the effect is captured by the d ln D to only low redshift sources with a fixed value of Ωm, where β ≡ f/bg(k, z) is defined in terms of f ≡ d ln a assuming a known value of the gravitational wave bias which is the logarithmic derivative of the growth func- parameter b . We find that the error-bar on H from ˆ GW 0 tion D with respect to the scale factor a, µkˆ = cosn. ˆ k is these methods [35, 53, 57] is more by only about 30% angle between the line of sight and the Fourier mode kˆ, than our method. This implies that in the limit of low and the superscript r denotes real space. The growth fac- 0 0 00 00 redshift sources, these methods [35, 53, 57] approach the H(z) R ∞ dz (1+z ) R ∞ dz (1+z ) −1 tor D = 0 3 [ 00 3 ] captures the optimal solution proposed in this work. We conclude in H0 z H(z ) 0 H(z ) Sec.VI. growth of the cosmological perturbations with redshift [91]. Astrophysical gravitational wave events are expected to occur in galaxies, and therefore, will follow the spatial II. FORMALISM: EXPLORING THE distribution of the galaxies with a bias parameter b CLUSTERING OF THE GRAVITATIONAL WAVE GW that is different from the bias parameters for galaxies b 5. SOURCES WITH GALAXIES g Following the definition Eq. (1), we can define the density r field for the gravitational wave sources in real space δGW The matter distribution in the Universe at large as scales (≥ 100 Mpc/h) is homogeneous and statistically r r isotropic, which agrees with the fundamental assumption δGW (k, z) = bGW (k, z)δm(k, z), (3) known as Copernican principle [87]. Under this setup, we can write the large scale distribution of galaxies in terms where bGW (k, z) is the gravitational wave bias parameter of a galaxy density field δg(r) = ng(r)/n¯g − 1, where [33, 81, 92–94]. The gravitational wave bias parameter ng(r) is the number density of galaxies at a position r captures how gravitational wave sources trace the large 4 andn ¯g is the mean number density of galaxies. Ac- scale structure in the Universe [92]. Since the gravita- cording to the standard model of cosmology, the spatial tional wave sources are tracers of luminosity distance and distribution of galaxies should trace the underlying dis- not redshift, they are not affected by RSD. tribution of matter in the Universe, and can be expressed The matter distribution also exhibits clustering prop- as a biased tracer of the matter density field δm(k) by erty, which can be statistically described by the correla- the relation, tion function ξ(r)6 [95–99]. The spatial clustering of the multiple tracers such as galaxies and gravitational wave δg(k) = bg(k)δm(k), (1) sources can be written in the Fourier space in terms of the three dimensional auto-power spectrum and cross power 7 where bg(k) is the galaxy bias, δg(k) is the Fourier trans- spectrum at different redshifts z as

s ss 0 −1 sr 0 δg(k, z) s 0 r 0 Pgg (k, z)δD(k − k ) +n ¯g(z) Pg GW (k, z)δD(k − k ) r δg(k , z) δGW (k , z) = sr 0 rr 0 −1 , δGW (k, z) Pg GW (k, z)δD(k − k ) PGW GW (k, z))δD(k − k ) +n ¯GW (z) (4)

Kaiser term [89, 90] 5 If primordial black holes (PBHs) are dark matter, then the distribution of PBHs is also going to be a biased tracer of the galaxy δs(k, z) = b (k, z)(1 + β µ2 )δr (k, z), (2) distribution. g g g kˆ m 6 Correlation function ξ(r), is related to the power spectrum P (k) by Fourier Transformation. 7 The angular bracket h.i denotes the ensemble average, which for a homogeneous and statistically isotropic Universe can be changed 4 P into an average over the spatial volume by ergodic theorem [91]. n¯g ≡ Ng/Vs = i ng(ri) 5

ij where Pxy(k, z) denotes the three dimensional power formation, stellar metallicity, and supernovae/AGN feed- spectrum at redshift z associated with the clustering be- back. The redshift dependence of the gravitational wave tween two tracers ({x y} ∈ {g, GW }) in redshift space or bias parameter is also unknown and we will discuss its 0 real space ({i j} ∈ {s, r}), δD(k − k ) denotes the Dirac implication in detail in the next section. −1 delta function, andn ¯x(z) is the shot noise contribu- One of the key aspects of Eq. (5) is that the underly- tion which is non-zero only when x and y are the same. ing cross power spectrum between galaxies and gravita- The shot noise contribution arises due to the discrete sr tional wave sources Pg GW (k, z) is related to the matter sampling of galaxies or gravitational waves sources. The power spectrum Pm(k, z), which is also measurable from redshift tomographic estimate of the auto power spec- the auto power spectrum of galaxies P ss(k, z). As a re- trum (x = y) and cross power spectrum (x 6= y) between gg sult, Pg GW (k, z) should follow similar statistical proper- galaxies and gravitational wave sources can be written in ss ties as Pgg (k, z). We exploit this very simple model to use terms of the matter power spectrum Pm(k, z) as the spatial cross-correlation of galaxies with gravitational wave sources to infer the luminosity-distance–redshift re- P ss(k, z) = b2(k, z)(1 + β µ2 )2P (k, z), gg g g kˆ m lation, and hence the cosmological parameters. P sr (k, z) = b (k, z)b (k, z)(1 + β µ2 )P (k, z), g GW g GW g kˆ m rr 2 PGW GW (k, z) = bGW (k, z)Pm(k, z). (5) III. LIKELIHOOD FOR INFERRING THE EXPANSION HISTORY USING DARK Astrophysical sources of gravitational waves are expected STANDARD SIRENS to form within galaxies. The latter, in turn, are expected to trace the underlying distribution of dark matter in the Let us consider a sample of N gravitational wave standard model of cosmology through the galaxy bias pa- GW sources (denoted by i) for which we have inferred the rameter b (k, z)[88, 91]. As a result, astrophysical gravi- g luminosity distance {di} to the source over a sky vol- tational wave sources are also expected to trace the dark l ume denoted by V . For each of these sources, there is matter distribution, but with a different bias parameter s also a measurement of the sky localization {θi , φi } b (k, z), which is yet to be measured. Even if a signif- GW GW GW with a 68% sky localization error ∆Ωi for each source. icant fraction of the gravitational wave sources would be GW Within the area of the sky localization, the exact posi- of primordial origin (such as primordial black holes), they tion of a gravitational wave source is not known. As a would still trace the underlying dark matter distribution result, any spatial information about the gravitational because under gravitational instability they would clus- wave source within that region is smoothed out. This ter in regions of higher dark matter density. However, results in smoothing the density field of the gravitational in such a scenario the gravitational wave bias parameter wave sources in Fourier space for the comoving modes would exhibit a different behavior from the astrophysical √ 1/2 gravitational wave bias parameter. The exploration of k > keff (z) ≡ 8 ln 2/(∆ΩGW dc(z)), where dc(z) is the 8 these differences is an interesting avenue to distinguish comoving distance to the source. Critically, assuming a between astrophysical and primordial black holes. This Gaussian distribution of the sky localization error, we can will be studied in details in a future work [100]. write the effect of sky localization on the density field as −k2/k2 (z) The presence of redshift space distortion (RSD) [89] in- δGW (k, ∆ΩGW , z) = δGW (k, z)e eff . Along with duces anisotropy in the observed auto (and cross) power the gravitational wave sources, we consider a number of spectrum with galaxies, as shown in Eq. (5). The bias galaxy samples Ng =n ¯gVs in the overlapping sky volume parameter for galaxies bg(k, z) and gravitational wave Vs with the known redshift zg and an error σz and the sources bGW (k, z) are redshift dependent and scale de- sky position denoted by {θg, φg} with an error on the 9 pendent. At large scales (k < 0.1 h/Mpc), the galaxy bias sky position ∆Ωg. Using galaxy samples with known is scale-independent, and behaves like a constant bg = 1.6 redshift, we can make tomographic bins of the galaxies [88, 101, 102]. We also expect similar scale-independent with Nz galaxies in each redshift bin. behavior of the bias parameter, bGW for gravitational At this point, we can combine the measurements to wave sources, at large scales (k < 0.1 h/Mpc), as it will be infer the underlying cosmology; in particular, the expan- mainly affected by the large scale spatial distribution of sion history of the Universe, which we model as H(z) = 3 R z 0.5 the galaxies. However, at smaller scales (k > 0.1 h/Mpc), H0(Ωm(1 + z) + Ωde exp (3 0 d ln(1 + z)(1 + w(z))) , the gravitational wave bias parameter is likely to be scale- and the corresponding cosmological parameters (Θc ∈ dependent as it will depend on the cluster scale (and {H0,Ωm, w(z) = w0 + wa(z/(1 + z))}) can be explored galaxy scale) astrophysical processes related to binary from dark standard sirens using the Bayes theorem [103]

8 Comoving distance dc(z) is related to the luminosity distance 9 For all practical purposes, sky localization error for galaxies can dl(z) by the relation dl(z) = (1 + z)dc(z). be considered to be zero. 6

ZZ NGW Y ss ss i i i P(Θc|ϑGW , dg) ∝ dΘn dz L(ϑGW |Pgg (k, z), Θn, dg(z))P(dg|Pgg (k, z))P({dl}GW |z, Θc, {θ , φ }GW )Π(z) i=1

× Π(Θn)Π(Θc), (6)

where, the gravitational wave data vector is composed of tional wave measurements within a luminosity distance i i i ϑGW ≡ {dl, θGW , φGW } and the galaxy data vector is bin. As a result, the combined posterior on the luminos- i i i composed of dg ≡ {δg(zg, θg, φg)}. Π(Θc) and Π(Θn) ity distance from NGW (dl) sources approaches a Gaus- denote the prior on the cosmological parameters Θc and sian distribution due to the central limit theorem. So the prior on the nuisance parameters Θn ∈ {bGW (k, z)}, re- assumption of Gaussian posteriors will not impact the respectively. Π(z) denotes the prior on the redshift range sults significantly when NGW (dl) is large. In any case, of the gravitational wave sources which can be taken to in the method proposed in this paper,a non-Gaussian be uniform over a wide range. In the presence of a red- posterior for individual gravitational wave sources can shift information about the gravitational wave sources, be trivially included in Eq. (6). an informative prior on the redshift can be considered. The posterior of the galaxy density field 10 In this analysis, we consider a uniform prior U(0, 1) on P(dg|Pgg(k, z)) given the galaxy power spectrum the redshift unknown gravitational wave sources. This is Pgg(k, z) can be written as sufficiently wide enough for the near-term and medium- s s∗ term gravitational wave surveys we are considering. ss δg(k, z)δg (k, z) i P(dg|Pgg (k, z)) ∝ exp − ss −1 , P({dl}GW |z, Θc) is the posterior on the luminosity dis- 2(Pgg (k, z) + ng(z) ) tance dl from the gravitational wave data ϑGW which, for (8) 11 s R 3 ik.r convenience, we model as a Gaussian distribution. where δg(k, z) = d r δg(r)e is the Fourier decom- position of the galaxy distribution. The first term in the denominator P ss(k, z) is the galaxy three dimensional i i i gg P({dl}GW |z,Θc, {θ , φ }GW ) power spectrum defined in Eq. (4), and ng(z) = Ng(z)/Vs (di({θi, φi} ) − d (z, Θ ))2 is the number density of galaxies in the redshift bin z. ∝ exp − l GW l c , 2σ2 Due to the non-linear structure formation, the Gaussian dl approximation of the field can break down at small scales. (7) In our analysis, we have taken the galaxy mock catalog (discussed in Sec.IV) which has the log-normal distri- where, σ is the error on the luminosity distance, and ¯ dl bution of the galaxy density field (δ(r) = ln(1 + δg(r))). R z c dz0 dl(z, Θc) = (1+z) 0 H(z0) is the model for the luminosity The statistics of the large scale structure density field can distance. The luminosity distance error is marginalised be be described well by the log normal distribution, as over the inclination angle. The posterior on the luminos- shown from cosmological simulations [104–106] as well as ity distance is expected to be non-Gaussian for individual from observations [107–109]. sources. In the cross-correlation technique, we combine The likelihood term L(ϑGW |Pgg(k, z), Θn, dg(z)) in the luminosity distance posteriors from multiple gravita- Eq. (6) is given by,

ss L(ϑGW |Pgg (k, z), Θn,dg(z)) ∝ 2 2 − k 2 k2 Pˆ(k, ∆Ω ) − b (k, z)b (k, z)(1 + β µ )P (k, z)e eff Z Z GW g GW g kˆ m Vs 2 exp − 2 k dk dµk ss −1 rr −1 , 4π 2(Pgg (k, z) + ng(z) )(PGW GW (k, z) + nGW (z) ) (9)

ˆ ∗ where P (k, z) = δg(k, z)δGW (k, ∆ΩGW ), nGW (z) = i NGW (dl(z))/Vs is the number density of gravitational 10 U(a, b) denotes the uniform function over the range (a,b). wave sources denoted in terms of the number of objects in 11 While this posterior is likely to be non-Gaussian in practice, we i the luminosity distance bin NGW (d (z)), and Vs denotes make this assumption purely to construct a forecast that can be l compared with other studies making similar assumptions. the total sky volume. The ﬁrst term in the numerator 7

Pˆ(k, z) denotes the observed cross-correlation signal be- sources bg and bGW respectively. As the cosmic density tween galaxies and gravitational wave sources, while the field evolves with redshift [91], we need to take this into second term is the model of the expected cross-correlation account in our cosmological simulations. To achieve this, power spectrum in redshift space in the presence of the we have generated several mock catalogs with box size anisotropic RSD, bias parameters, and limited sky reso- (in units of Mpc/h) [lx = 1350, ly = 1350, lz = 300] at lutionΩ GW of the gravitational wave sources. The bias each redshift bin, starting from z = 0 to z = 1.0 with parameters can be considered as nuisance parameters (or Planck-2015 cosmology [8]. Then all these mock cata- also cosmological parameters), and are marginalised over logs are combined to obtain a single mock catalog over in our setup. The term in the denominator denotes the the entire redshift range. The galaxy distribution and covariance matrix of the cross-correlation power spec- the gravitational wave sources are chosen from this dis- trum, in which the first term arises from the variance tribution which already includes cosmological evolution of the galaxy distribution, and the second term from as a function of redshift. The method is also repeated the variance of the gravitational wave distribution. The for finer/wider choices of the lz and the results obtained cross-correlation power spectrum is integrated over the from our method are robust. This mock catalog does not sky volume denoted by Vs and the Fourier wave num- take into account the contribution from weak lensing, ber k. The total number of Fourier modes which con- since it is going to have a marginal (≤ 1%) increase in tributes to the signal depends on the volume of the sky the variance of the inferred cosmological parameters for 2 2 survey given by Nm = k dkVs/4π . The integration the low redshift gravitational wave sources considered in in Eq. (9) takes into account the anisotropic shape of this analysis. the power spectrum by combining the contribution from Galaxy samples: The galaxy samples are produced for ˆ µk = cosn. ˆ k arising due to the RSD. The likelihood is a scale-independent bias parameter bg = 1.6 including maximized for the set of cosmological parameters, which the effect from RSD [112]. The galaxy mocks are ob- 4 transforms the galaxy density field from redshift space to tained for the number of galaxies Ng = 1.5 × 10 . The match or maximally correlate with the spatial distribu- redshift of these sources is assumed to be known spectro- tion of gravitational wave sources. scopically, which implies that the corresponding error in In the limit nx(z)Px(k, z) > 1, the likelihood is in the the redshift measurement is σz ≈ 0. cosmic variance limited regime, while nx(z)Px(k, z) < 1, Gravitational wave samples: For the same set of cos- it is in the shot noise dominated regime. For the grav- mological parameters and the same realization of the itational wave sources expected within 5 years (with an large scale structure density field from which we produced event rate R(z) = 100 Gpc−3 yr−1 [110, 111]), we will ex- the galaxy samples, we obtain the gravitational wave 12 plore the cross-correlation between the galaxies and grav- samples NGW with the gravitational wave bias param- α itational wave sources only for small values of k < keff eter bGW (z) = bGW (1 + z) with bGW = 2 and α = 0. ss For these samples we consider three different cases for in the shot noise regime nGW PGW GW (k, z) < 1. Galaxy 9 samples will have O(10 ) galaxies [73–79] and as a re- sky localization error ∆ΩGW = 10 sq. deg., ∆ΩGW = 25 sult, we will be in the cosmic variance limited regime for sq. deg., and ∆ΩGW = 100 sq. deg. [113, 114] which the values of k < keff . So the denominator of the ex- are possible to achieve from the network of five gravita- 4π2P ss(k,z) ponent in Eq. (9), is going to scale as gg . With tional wave detectors (LIGO-Hanford, LIGO-Livingston, nGW (z) Virgo, KAGRA, LIGO-India [60–63]). For each gravita- the availability of large numbers of gravitational wave tional wave source, the fractional error on the luminos- samples, the measurement will be in the cosmic variance ity distance depends inversely on the matched filtering limited regime n P rr (k, z) > 1, in which case GW GW GW signal-to-noise ratio (ρ) given by the relation[25, 115– the denominator of the exponent can be approximated 2 ss rr 118] as 4π Pgg (k, z)PGW GW (k, z). In this analysis, we have considered an analytical covariance matrix. This can also Z fmax |h(f)|2 be calculated from simulations for a specific mission of ρ2 ≡ 4 df , (10) large scale structure and gravitational waves experiment. 0 Sn(f)

where the value of fmax is considered as fmerg = 3 2 13 c (a1η + a2η + a3)/πGM [119] , Sn(f) is the detector IV. GENERATION OF MOCK CATALOG noise power spectrum, which we consider as the advanced

We implement our method on a mock catalog of large scale structure and gravitational wave sources which are 12 produced for the log-normal distribution of the density Different cases of NGW are considered in this analysis, and are discussed in the respective sections field using the publicly available package nbodykit [112]. 13 M = m1 + m2 is the total mass of the coalescing binary, η is The realization of the galaxies and gravitational wave 2 the symmetric mass ratio η = m1m2/M , c is the speed of light sources are obtained from the same random realization, and G denotes the gravitational constant. The values of the using a fixed matter power spectrum Pm(k, z) with differ- parameters are a1 = 0.29740, a2 = 0.044810, a3 = 0.095560 ent bias parameters for galaxies and gravitational wave [119]. 8

LIGO design sensitivity [61] 14. The template of the grav- V. RESULTS itational wave strain h(f) for f ≤ fmerg can be written in terms of the redshifted chirp mass Mz = (1 + z)Mc, Using the mock catalogs of galaxies and gravitational inclination angle with respect to the orbital angular mo- wave sources discussed in Sec.IV, we explore the cos- ˆ ˆ mentum L.nˆ (which is denoted by the function I±(L.nˆ)), mological parameters which affects the expansion his- and luminosity distance to the source d by the relation 17 L tory of the Universe (Hubble constant H0, matter den- [116, 119–122] sity Ωm, dark energy equation of state w(z)) using the formalism described in Sec.III. The precise and accurate inference of the cosmological parameters using this r 5 G5/6M2(f M )−7/6 h (f) = z z z I (L.ˆ nˆ). (11) method will rely on successfully mitigating the uncer- ± 3/2 2/3 ± 96 c π dL tainties associated with the unknown bias parameter and its redshift dependence associated with the gravitational In this analysis, we critically consider the posterior dis- wave sources. So, along with the cosmological parame- tribution of luminosity distance to be Gaussian with the ters, we also consider the gravitational wave bias parameter b (z) = b (1 + z)α to be unknown and jointly minimum matched filtering detection threshold ρth = 10 GW GW 15 infer the value of b and α (these are our nuisance for equal mass binaries with masses 30 M . The frac- GW parameters Θn ∈ {bGW , α}) in the analysis along with tional error in the luminosity distance σdl /dl can be about 10% for the bright sources having high detection the cosmological parameters. We consider three cases SNR ρ > 60 and as large as 70% for the objects at detec- in this analysis: (i) H0,Ωm, with fixed w0 = −1, and tion threshold ρ = 10. Sources with poor sky localization wa = 0; (ii) Ωm and ΩΛ, with fixed H0 = 70 km/s/Mpc, and large error on the luminosity distance will contribute w0 = −1, and wa = 0; (iii) w0 and wa with fixed onlya marginal improvement to the estimation of the cos- H0 = 70 km/s/Mpc and Ωm = 0.315. Uniform priors on the cosmological and nuisance parameters are consid- mological parameters. So only the fraction of events with better sky localization will effectively improve the preci- ered in the following range: Π H0 = U(20, 150), sion and accuracy of the estimation of the cosmological km/s/Mpc parameters. With the network of GW detectors such as Π(Ωm) = U(0.1, 1), Π(ΩΛ) = U(0, 1), Π(w0) = U(−2, 0), LIGO-Hanford, LIGO-Livingston, Virgo, KAGRA (and Π(wa) = U(−8, 8), Π(bGW ) = U(0, 6), Π(α) = U(−4, 4) in the future from LIGO-India), a sky localization error and Π(z) = U(0, 1). We show the results only for less than 100 sq. deg. is achievable [114]. The mean the ∆ΩGW = 10 sq. deg. However, the results for values of the luminosity distance are set to those of a ∆ΩGW = 25 sq. deg. only deteriorates marginally. For flat LCDM cosmological model with parameter values sky-localization error ∆ΩGW = 100 sq. deg., the impact [H0 = 70 km/s/Mpc, Ωm = 0.315, ΩΛ = 1 − Ωm, w0 = on the error-bars are about a factor of two on the inferred −1, wa = 0]. The chosen value of the Hubble param- parameters. Even with the increase in the sky localiza- eter, H0 is completely different from that considered tion errors, our method still gives unbiased results for all in the large scale structure mock catalog (H0 = 67.3 the cosmological parameters and the bias parameters. km/s/Mpc) to show that the inferred cosmological parameters are affected only by the luminosity distance and not by the parameters assumed in the mock catalog. For A. Measurement of H0, Ωm and bGW (z) gravitational wave sources, we do not assume any redshift information. The current estimate of the event rate The joint-estimation of the cosmological parameters of BBHs is R(z) = 102 Gpc−3 yr−1 [110]. With this event H0 and Ωm along with the nuisance parameters are rate, we expect a few thousand of events to be detected shown in Fig.2 for fixed value of w0 = −1 and wa = 0. every year with the advanced LIGO design sensitivity These results are obtained for the cases with Ng = [61]. In this analysis, we show the measurability of the 4 18 1.5 × 10 , NGW = 200 , and ∆ΩGW = 10 sq. deg expansion history by considering a few different cases of 19. Results show that we can make the measurement the number of gravitational wave sources N 16 and for GW of H0 = 70 km/s/Mpc with an accuracy of 1.9% with the sky localization which is expected to be achievable only NGW (z) = 40 BBHs in each redshift bin of width with a network of four/five gravitational detectors. ∆z = 0.1 up to redshift z = 0.5 detectable with the advanced LIGO design sensitivity [61]. The result shown in Fig.2 also indicates that the gravitational wave bias

14 The noise curves are available publicly on this website 17 https://dcc-lho.ligo.org/LIGO-T2000012/public Considering only the cosmological models with curvature ΩK = 15 30 M = 2 × 10 kg denotes the mass of the sun. 0. 16 18 We consider four cases of NGW = 50, 100, 200, 280 for this analy- The total number of gravitational wave sources NGW = sis in the LIGO design sensitivity, which is expected to be easily R N(z)dz. 19 available with the network of gravitational wave detectors. Results with ∆ΩGW = 25 sq. deg. changes only marginally. 9

FIG. 2: We show the joint posterior of the cosmological parameters H0 = 100h0 km/s/Mpc and Ωm along with the α nuisance parameters related to the gravitational wave bias parameter bGW (z) = bGW (1 + z) for number of gravitational wave sources NGW (z) = 40 extended up to redshift z = 0.5, and sky localization error ∆ΩGW = 10 sq. deg. The 68%, and 95% contours are shown in these plots along with the input values by the blue line. The mean value along with 1σ error-bar are mentioned in the title of the posterior distribution for all the parameters. Other cosmological parameters such as w0 = −1 and wa − 0 are kept ﬁxed for these results.

parameters bGW and α are uncorrelated with the cosmo- 200 BBHs at the advanced LIGO design sensitivity [61]. logical parameters H0 and Ωm. As a result, uncertainty The cross-correlation technique makes it possible to mea- associated with the gravitational wave bias parameter sure the bias parameter even with the currently ongoing does not aﬀect the inference of the cosmological parame- detector network and much before the operation of next- ters (for the parametric form of the bias considered in this generation gravitational wave detectors [65, 66] by us- analysis). This makes our method both precise and ac- ing the autocorrelation between the gravitational wave curate to infer the cosmological parameters. Using this sources. This is another additional gain which is not method we can measure the value of the gravitational possible from the other proposed methods [35, 53, 57]. wave bias parameter with σbGW /bGW ∼ 27%, with only The forecast posteriors on H0 (after marginalizing over 10

FIG. 3: We show the joint posterior of the cosmological parameters ΩΛ and Ωm along with the nuisance parameters α related to the gravitational wave bias parameter bGW (z) = bGW (1 + z) for number of gravitational wave sources NGW (z) = 40 extended up to redshift z = 0.7, and sky localization error ∆ΩGW = 10 sq. deg. The 68%, and 95% contours are shown in these plots along with the input values by the blue line. The mean value along with 1σ error-bar are mentioned in the title of the posterior distribution for all the parameters. Other cosmological parameters such as H0 = 70 km/s/Mpc, w0 = −1 and wa = 0 are kept ﬁxed for these results.

Ωm, bGW , α) for NGW = 50, 100 and 200 gravitational sirens (σH0 /H0 = 1.9%) compares favourably with that wave sources are shown in Fig.1 along with the measure- which could be obtained from 50 sources with EM coun- ment of Hubble constant H0 = 67.4 ± 0.5 km/s/Mpc and terparts (such as BNS and NS-BH, assuming σH0 /H0 = H0 = 74±1.4 km/s/Mpc from Planck [15] and SH0ES [4] 2%, [35, 36]). However, as the number of detected dark respectively. The uncertainty in the measurement of H0 sirens is expected to outnumber the sources with EM −1/2 counterparts (such as BNSs and NS-BHs), one can ex- decreases as the number of sources increases (∼ NGW ) and as the uncertainty in the luminosity distances de- pect the constraints on H0 from dark sirens to dominate those from BNSs and NS-BHs, with very conservative as- creases (∼ σdl /dl). sumptions about the availability of galaxy redshift survey Fig.1 shows that a measurement of H0 from 200 dark 11 covering a substantial fraction of the sky. In summary, with 3.4σ. However, the constraints on wa are going to our method will provide both accurate and precise mea- be weak with the modest number of gravitational wave surements of H0 from dark sirens along with Ωm, and red- sources. With more number of gravitational wave sources shift dependent gravitational wave bias parameter bGW (z) possible from the five years of observation with the Ad- from the network of the advanced (with or without op- vanced LIGO design sensitivity [61], we will be able to tical squeezing) gravitational wave detectors. Combin- infer the dark energy equation of state with higher ac- ing these two independent constraints, one would achieve −1/2 curacy (the error on the parameter reduces by NGW ) σH0 /H0 ∼ 1.4%, which is competitive with current con- for sources up to redshift z ∼ 1. This independent av- straints from standard candles [4]. Following our work, enue to measure w0 and wa will also be accessible from a recent work [123] has also obtained the constraints on the next-generation gravitational wave detectors such as H0 for a fixed value of Ωm, and with constant gravi- LISA [64], ET [65], CE [66] and TianQin observatory tational wave bias parameter using the cross-correlation [67] for sources which are beyond redshift z = 1. The technique. gravitational wave bias parameters bGW and α are also uncorrelated with the parameters describing the dark energy equation of state and can be measured with high B. Measurement of ΩΛ, Ωm and bGW (z) statistical significance as shown in Fig.4. This method can also be used for the joint estimation of all cosmolog- As our method can be applied to high redshift (up ical parameters and the gravitational wave bias parame- to which galaxy surveys will be available), we can also ter (H0, Ωm, Ωde, w0, wa, bGW (k, z)), provided one has a measure the energy budget in dark energy ΩΛ from dark large number of gravitational-wave sources up to high sirens. We make the joint estimation of the cosmological redshift (z > 1), so that the high redshift tomographic parameters ΩΛ–Ωm along with the two bias parameters bins can be used to constrain Ωm, and the low redshift bGW and α for the parametric form bGW (z) = bGW (1 + tomographic bins can be used to constrain the other cos- α z) for a fixed value of H0 = 70 km/s/Mpc, w0 = −1 and mological parameters. This will be possible only from wa = 0 with NGW (z) = 40 up to redshift z = 0.7. The the next generation of ground-based detectors (such as corresponding plot is shown in Fig.3. We show for the ET, and CE), and also from the space-based detectors first time that the energy budget of dark energy can be (such as LISA and TianQin observatory). measured from using dark sirens detectable within the modest timescale with the advanced LIGO design sensitivity [61] with only NGW = 280 BBHs. The Ωm and ΩΛ VI. CONCLUSIONS AND DISCUSSIONS are also uncorrelated with the bias parameters (bGW and α), and as a result will not affect the measurement of cos- Gravitational-wave sources are accurate luminosity mological parameters. The measurement of ΩΛ and Ωm distance tracers without requiring any external astro- gets less constrained fora limited number of gravitational physical calibration if instrument calibration can be wave sources if the value of H0 is not kept fixed. However, achieved [28, 29]. An accurate instrument calibration a joint estimation with H0 is possible with more number is essential for the measurement of the luminosity dis- of gravitational wave sources. This method is also useful tance [28, 29]. If this is possible, then gravitational wave for the future gravitational wave detectors such as LISA sources are an exquisite probe to measure the expansion [64], ET [65], CE [66] and TianQin observatory [67] to history of the Universe by exploiting the luminosity dis- measure ΩΛ,Ωm, and the gravitational wave bias param- tance and its redshift. However, the inference of the red- eter bGW (z). shift of the gravitational wave sources requires either an EM counterpart or a known mass scale (such as the mass scale associated with the tidal deformation [30] or pair- C. Measurement of w0, wa and bGW (z) instability of supernova [31]). For most of the gravitational wave sources, neither of these is available. In this The two-parameter phenomenological model of the paper, we apply the method introduced in [81], which dark energy equation of state wde = w0 + wa z/(1 + z) exploits the scale associated with the three-dimensional is usually considered to explore the redshift dependence clustering property of cosmic structure and the fact that of dark energy. Using our method, we show the joint both galaxies and gravitational wave sources, trace the estimation of w0 and wa along with the two bias pa- underlying dark matter field rameters bGW and α (for the parametric form bGW (z) = Using the detector sensitivity expected from the cur- α bGW (1 + z) ) in Fig.4 for NGW (z) = 40 extended up to rent generation gravitational wave detectors [60–63], we z = 0.7. We have kept the value of H0 = 70 km/s/Mpc show that with a modest number of gravitational wave and Ωm = 0.315 fixed for flat LCDM model. This plot sources (∼ 100) we will be able to infer the Hubble shows that this technique is capable to infer the dark en- constant with an accuracy ∼ 2.5% as shown in Fig.1 4 ergy equation of state with Ng = 1.5 × 10 , NGW = 280 for gravitational wave sources distributed up to redshift (up to redshift z = 0.7) and for ∆ΩGW = 10 sq. deg. z = 0.5. The exploration of clustering of gravitational The constraints on the values on w0 = −1 are possible wave sources with galaxies makes it a robust method to 12

FIG. 4: We show the joint posterior of the cosmological parameters w0 and wa along with the nuisance parameters α related to the gravitational wave bias parameter bGW (z) = bGW (1 + z) for number of gravitational wave sources NGW (z) = 40 extended up to redshift z = 0.7, and sky localization error ∆ΩGW = 10 sq. deg. The 68%, and 95% contours are shown in these plots along with the input values by the blue line. The mean value along with 1σ error-bar are mentioned in the title of the posterior distribution for all the parameters. Other cosmological parameters such as H0 = 70 km/s/Mpc and Ωm = 0.315 are kept ﬁxed for these results. infer the Hubble constant using the dark sirens. Besides Along with the measurement of the expansion history, the Hubble constant measurement, our method makes this method makes it possible to infer the gravitational it possible to measure the fraction of dark energy in wave bias parameter and its redshift dependence bGW (z). the Universe and its fundamental nature with the net- The gravitational wave bias parameter determines the work of current generation gravitational wave detectors, spatial distribution of the gravitational wave sources with as shown in Fig.3 and Fig.4. This is not possible respect to the dark matter distribution and provides an currently from the gravitational wave sources with EM avenue to measure this. Using our method, we can mea- counterparts (such as BNS and NS-BH) due to the low sure the bias parameter by more than 3σ precision with observable horizon (z < 0.5). only 200 BBHs distributed up to z = 0.5, as shown in 13

Fig.2. With the availability of more gravitational wave sq. deg. up to a redshift of z = 3. In a future work, we sources, the bias parameter can be measured with higher will explore the synergy between the EM telescopes and precision and accuracy. The cross-correlation with the the next generation GW detectors such as ET [65] and galaxies makes it possible to detect the bias parameters of CE [66] to map the cosmic history up to high redshift. gravitational wave sources sooner with higher statistical Finally, this method is not limited to gravitational significance than that possible from the auto-correlation wave sources but also applicable to any other distance [94]. The redshift dependent bias parameter is not de- tracers to infer the expansion history of the Universe generate with the cosmological parameters as shown in using the luminosity distance – redshift relation. Our Fig.2, Fig.3, and Fig.4, which makes it possible to method is readily applicable to SNe samples which will reliably detect the cosmological parameters even if the be detected with photometric redshift measurement from gravitational wave bias parameter is currently unknown. Rubin Observatory [78], as already pointed out by a pre- On longer timescales corresponding to the launch vious analysis [81]. In the future, this method can play of next-generation gravitational wave detectors such as a crucial role in studying cosmology with type-Ia SNe LISA [64], ET [65], and CE [66], we will be able to probe [125]. This method will be useful in exploring the syn- the expansion history of the Universe up to much higher ergy between the upcoming missions such as DES [72], redshift using the method proposed in this paper, with- DESI [73], Euclid [74], SPHEREx [79], Nancy Grace Ro- out inferring the EM counterparts of the gravitational man Telescope [75–77]. This method is also applicable wave sources. So, the method proposed in this paper will to Fast Radio Burst (FRBs) [126] to infer their redshifts help in building the observation strategy for the future for which host identification will be difficult. gravitational wave detectors. The success of the cross-correlation technique discussed in this paper depends on the availability of the ACKNOWLEDGMENTS EM telescopes with photometric/spectroscopic capabil- ities that can detect galaxies up to high redshift, with The authors would like to thank Archisman Ghosh for sufficiently low magnitude, and nearly full-sky cover- carefully reviewing the manuscript and providing use- age. Though such a rich data set is not possible from ful comments. S. M. also acknowledges useful discus- a single EM telescope, it is possible to achieve this by sion with Rahul Biswas, Neal Dalal, Will Farr, Archis- combining multiple EM telescopes operational in dif- man Ghosh, Salman Habib, Eiichiro Komatsu, Daniel ferent EM frequency bands covering different sky ar- M. Scolnic, Joseph Silk, and David Spergel. This anal- eas. Several ongoing or upcoming EM telescopes such ysis was carried out at the Horizon cluster hosted by as SDSS/BOSS [124], Dark Energy Survey (DES) [72], Institut d’Astrophysique de Paris. We thank Stephane Dark Energy Spectroscopic Instrument (DESI) [73], Eu- Rouberol for smoothly running the Horizon cluster. SM clid [74], Nancy Grace Roman Telescope [75–77], Ru- and SMN are also supported by the research program In- bin Observatory [78], Spectro-Photometer for the His- novational Research Incentives Scheme (Vernieuwingsim- tory of the Universe, Epoch of Reionization, and Ices puls), which is financed by the Netherlands Organization Explorer (SPHEREx) [79], will play a key role in build- for Scientific Research through the NWO VIDI Grant ing up nearly full-sky galaxy catalogs which can be cross- No. 639.042.612-Nissanke. This work is partly supported correlated with the sources detectable from the network by the Delta ITP consortium, a program of the Nether- of gravitational wave detectors such as LIGO/ Virgo/ lands Organisation for Scientific Research (NWO) that KAGRA. Ground-based ongoing surveys such as DES is funded by the Dutch Ministry of Education, Culture, will be measuring about 300 million galaxies over a sky and Science (OCW). The work of BDW is supported by area of approximately 5000 sq. deg up to redshift z ∼ 1.5. the Labex ILP (reference ANR-10-LABX-63) part of the In the future, SDSS-V will be measuring nearly full sky Idex SUPER, received financial state aid managed by and about 2500 sq. deg with the Multi-Object Spec- the Agence Nationale de la Recherche, as part of the troscopy and Integral Field Spectroscopy over the red- programme Investissements d’avenir under the reference shift range z = 0.3 to z = 2.5. Also, surveys such as ANR-11-IDEX-0004-02. The Center for Computational DESI and Rubin Observatory will be measuring galax- Astrophysics is supported by the Simons Foundation. AS ies up to z = 1.6 and z = 3 respectively with nearly acknowledges support from the NWO and the Dutch about one-third of the sky area. Upcoming space-based Ministry of Education, Culture, and Science (OCW) missions such as Euclid, SPHEREx, and Nancy Grace (through NWO VIDI Grant No. 2019/ENW/00678104 Roman telescope will also play a vital role in the syn- and from the D-ITP consortium). In this analysis, we ergy with the gravitational wave detectors. Euclid will have used the following packages: Corner [127], emcee: be measuring about 1.5 billion galaxies over a sky area The MCMC Hammer [128], IPython [129], Matplotlib of approximately 15000 sq. deg. in the redshift range [130], nbodykit [112], NumPy [131], and SciPy [132]. The z = 0.8 − 2. SPHEREx will be mapping about 500 mil- authors would like to thank the LIGO/Virgo scientific lion galaxies over the full sky up to redshift z ∼ 2.5. The collaboration for providing the noise curves. LIGO is Nancy Grace Roman telescope will be capable to mea- funded by the U.S. National Science Foundation. Virgo sure about a few billion galaxies over a sky area of 2500 is funded by the French Centre National de Recherche 14

Scientiﬁque (CNRS), the Italian Istituto Nazionale della Fisica Nucleare (INFN), and the Dutch Nikhef, with con- tributions by Polish and Hungarian institutes.

[1] E. Hubble, Proceedings of the National Academy of Sci- [31] W. M. Farr, M. Fishbach, J. Ye, and D. Holz, Astrophys. ence 15, 168 (1929). J. Lett. 883, L42 (2019), 1908.09084. [2] S. Perlmutter et al. (Supernova Cosmology Project), As- [32] C. Howlett and T. M. Davis (2019), 1909.00587. trophys. J. 517, 565 (1999), astro-ph/9812133. [33] S. Mukherjee, G. Lavaux, F. R. Bouchet, J. Jasche, [3] M. J. Reid, J. A. Braatz, J. J. Condon, L. J. Green- B. D. Wandelt, S. M. Nissanke, F. Leclercq, and K. Ho- hill, C. Henkel, and K. Y. Lo, ApJ 695, 287 (2009), tokezaka (2019), 1909.08627. 0811.4345. [34] C. Nicolaou, O. Lahav, P. Lemos, W. Hartley, and [4] A. G. Riess, S. Casertano, W. Yuan, L. M. Macri, and J. Braden (2019), 1909.09609. D. Scolnic, Astrophys. J. 876, 85 (2019), 1903.07603. [35] H.-Y. Chen, M. Fishbach, and D. E. Holz, Nature 562, [5] Planck collaboration (Planck), Astron. Astrophys. 571, 545 (2018), 1712.06531. A16 (2014), 1303.5076. [36] S. M. Feeney, H. V. Peiris, A. R. Williamson, S. M. Nis- [6] L. Anderson et al. (BOSS), Mon. Not. Roy. Astron. Soc. sanke, D. J. Mortlock, J. Alsing, and D. Scolnic, Phys. 441, 24 (2014), 1312.4877. Rev. Lett. 122, 061105 (2019), 1802.03404. [7] E. Aubourg et al., Phys. Rev. D92, 123516 (2015), [37] R. Abbott et al. (LIGO Scientific, Virgo) (2020), 1411.1074. 2004.08342. [8] P. Ade et al. (Planck), Astron. Astrophys. 594, A13 [38] G. Ghirlanda et al., Science 363, 968 (2019), (2016), 1502.01589. 1808.00469. [9] E. Macaulay et al. (DES), Mon. Not. Roy. Astron. Soc. [39] K. Mooley, A. Deller, O. Gottlieb, E. Nakar, G. Hal- 486, 2184 (2019), 1811.02376. linan, S. Bourke, D. Frail, A. Horesh, A. Corsi, and [10] R. Jimenez and A. Loeb, Astrophys. J. 573, 37 (2002), K. Hotokezaka, Nature 561, 355 (2018), 1806.09693. astro-ph/0106145. [40] K. Hotokezaka, E. Nakar, O. Gottlieb, S. Nissanke, [11] J. Simon, L. Verde, and R. Jimenez, Phys. Rev. D 71, K. Masuda, G. Hallinan, K. P. Mooley, and A. Deller, 123001 (2005), astro-ph/0412269. Nature Astron. (2019), 1806.10596. [12] D. Stern, R. Jimenez, L. Verde, M. Kamionkowski, and [41] B. P. Abbott et al. (LIGO Scientific, Virgo), Phys. Rev. S. Stanford, JCAP 02, 008 (2010), 0907.3149. Lett. 119, 161101 (2017), 1710.05832. [13] R. Jimenez, A. Cimatti, L. Verde, M. Moresco, and [42] B. P. Abbott et al. (LIGO Scientific, Virgo, 1M2H, Dark B. Wandelt, JCAP 03, 043 (2019), 1902.07081. Energy Camera GW-E, DES, DLT40, Las Cumbres Ob- [14] T. Abbott et al. (DES), Mon. Not. Roy. Astron. Soc. servatory, VINROUGE, MASTER), Nature 551, 85 480, 3879 (2018), 1711.00403. (2017), 1710.05835. [15] N. Aghanim et al. (Planck) (2018), 1807.06209. [43] F. Foucart, T. Hinderer, and S. Nissanke, Phys. Rev. D [16] K. C. Wong et al. (2019), 1907.04869. 98, 081501 (2018), 1807.00011. [17] L. Verde, T. Treu, and A. Riess, in Tensions between [44] M. W. Coughlin et al., Astrophys. J. Lett. 885, L19 the Early and the Late Universe (2019), 1907.10625. (2019), 1907.12645. [18] W. L. Freedman et al. (2019), 1907.05922. [45] I. Andreoni et al., Astrophys. J. 890, 131 (2020), [19] M. Rigault et al. (Nearby Supernova Factory) (2018), 1910.13409. 1806.03849. [46] B. Abbott et al. (LIGO Scientific, Virgo), Astrophys. J. [20] C. Kochanek, Mon. Not. Roy. Astron. Soc. 493, 1725 Lett. 892, L3 (2020), 2001.01761. (2020), 1911.05083. [47] R. Abbott et al., The Astrophysical Journal 896, L44 [21] B. F. Schutz, Nature 323, 310 (1986). (2020). [22] D. E. Holz and S. A. Hughes, ApJ 629, 15 (2005), astro- [48] M. M. Kasliwal et al. (2020), 2006.11306. ph/0504616. [49] S. Vitale and H.-Y. Chen, Phys. Rev. Lett. 121, 021303 [23] N. Dalal, D. E. Holz, S. A. Hughes, and B. Jain, Phys. (2018), 1804.07337. Rev. D 74, 063006 (2006), astro-ph/0601275. [50] D. J. Mortlock, S. M. Feeney, H. V. Peiris, A. R. [24] C. L. MacLeod and C. J. Hogan, Phys. Rev. D 77, Williamson, and S. M. Nissanke, Phys. Rev. D 100, 043512 (2008), . 103523 (2019), . [25] S. Nissanke, D. E. Holz, S. A. Hughes, N. Dalal, and [51] M. J. Graham, K. E. S. Ford, B. McKernan, N. P. Ross, J. L. Sievers, ApJ 725, 496 (2010), 0904.1017. D. Stern, K. Burdge, M. Coughlin, S. G. Djorgovski, [26] B. F. Schutz, Classical and Quantum Gravity 28, A. J. Drake, D. Duev, et al., Phys. Rev. Lett. 124, 125023 (2011), 1102.5421. 251102 (2020), . [27] S. Nissanke, D. E. Holz, N. Dalal, S. A. Hughes, J. L. [52] W. Del Pozzo, Phys. Rev. D 86, 043011 (2012), Sievers, and C. M. Hirata (2013), 1307.2638. 1108.1317. [28] L. Sun et al. (2020), 2005.02531. [53] R. Nair, S. Bose, and T. D. Saini, Phys. Rev. D 98, [29] D. Bhattacharjee, Y. Lecoeuche, S. Karki, J. Bet- 023502 (2018), 1804.06085. zwieser, V. Bossilkov, S. Kandhasamy, E. Payne, and [54] R. Gray, I. M. n. Hernandez, H. Qi, A. Sur, P. R. R. Savage (2020), 2006.00130. Brady, H.-Y. Chen, W. M. Farr, M. Fishbach, J. R. [30] C. Messenger and J. Read, Phys. Rev. Lett. 108, 091101 Gair, A. Ghosh, et al., Phys. Rev. D 101, 122001 (2020), (2012), 1107.5725. . 15

[55] M. Fishbach et al. (LIGO Scientific, Virgo), Astrophys. [82] J. A. Newman, Astrophys. J. 684, 88 (2008), 0805.1409. J. Lett. 871, L13 (2019), 1807.05667. [83] B. Menard, R. Scranton, S. Schmidt, C. Morrison, [56] B. Abbott et al. (LIGO Scientific, Virgo) (2019), D. Jeong, T. Budavari, and M. Rahman (2013), 1908.06060. 1303.4722. [57] M. Soares-Santos et al. (DES, LIGO Scientific, Virgo), [84] S. Schmidt, B. Menard, R. Scranton, C. Morrison, and Astrophys. J. Lett. 876, L7 (2019), 1901.01540. C. McBride, Mon. Not. Roy. Astron. Soc. 431, 3307 [58] R. Abbott et al. (LIGO Scientific, Virgo), Astrophys. J. (2013), 1303.0292. 896, L44 (2020), 2006.12611. [85] S. Mukherjee, B. D. Wandelt, and J. Silk (2019), [59] A. Palmese et al. (DES) (2020), 2006.14961. 1908.08951. [60] F. Acernese, M. Agathos, K. Agatsuma, D. Aisa, [86] S. Mukherjee, B. D. Wandelt, and J. Silk, Phys. Rev. D N. Allemandou, A. Allocca, J. Amarni, P. Astone, 101, 103509 (2020), 1908.08950. G. Balestri, G. Ballardin, et al., Classical and Quan- [87] J. Yadav, S. Bharadwaj, B. Pandey, and T. Seshadri, tum Gravity 32, 024001 (2014), . Mon. Not. Roy. Astron. Soc. 364, 601 (2005), astro- [61] B. P. Abbott et al., Phys. Rev. D 93, 112004 ph/0504315. (2016), [Addendum: Phys.Rev.D 97, 059901 (2018)], [88] V. Desjacques, D. Jeong, and F. Schmidt, Phys. Rept. 1604.00439. 733, 1 (2018), 1611.09787. [62] T. Akutsu et al. (KAGRA), Nat. Astron. 3, 35 (2019), [89] N. Kaiser, MNRAS 227, 1 (1987). 1811.08079. [90] A. Hamilton, in Ringberg Workshop on Large Scale [63] C. Unnikrishnan, Int. J. Mod. Phys. D 22, 1341010 Structure (1997), astro-ph/9708102. (2013), 1510.06059. [91] P. J. E. Peebles, The large-scale structure of the universe [64] P. Amaro-Seoane, H. Audley, S. Babak, J. Baker, E. Ba- (1980). rausse, P. Bender, E. Berti, P. Binetruy, M. Born, [92] S. Mukherjee and J. Silk, Mon. Not. Roy. Astron. Soc. D. Bortoluzzi, et al., arXiv e-prints arXiv:1702.00786 491, 4690 (2020), 1912.07657. (2017), 1702.00786. [93] F. Calore, A. Cuoco, T. Regimbau, S. Sachdev, and [65] M. Punturo et al., Class. Quant. Grav. 27, 194002 P. D. Serpico (2020), 2002.02466. (2010). [94] A. Vijaykumar, M. Saketh, S. Kumar, P. Ajith, and [66] D. Reitze et al., Bull. Am. Astron. Soc. 51, 035 (2019), T. R. Choudhury (2020), 2005.01111. 1907.04833. [95] P. J. E. Peebles and E. J. Groth, ApJ 196, 1 (1975). [67] J. Luo et al. (TianQin), Class. Quant. Grav. 33, 035010 [96] M. Davis and P. J. E. Peebles, ApJS 34, 425 (1977). (2016), 1512.02076. [97] M. Davis and P. J. E. Peebles, ApJ 267, 465 (1983). [68] H.-T. Wang et al., Phys. Rev. D 100, 043003 (2019), [98] A. J. S. Hamilton, ApJ 417, 19 (1993). 1902.04423. [99] S. D. Landy and A. S. Szalay, ApJ 412, 64 (1993). [69] S. Liu, Y.-M. Hu, J.-d. Zhang, and J. Mei, Phys. Rev. [100] S. Mukherjee, (Under preparation) (2020). D 101, 103027 (2020), 2004.14242. [101] L. Anderson, E. Aubourg, S. Bailey, D. Bizyaev, [70] L. S. Finn, Phys. Rev. Lett. 73, 1878 (1994), . M. Blanton, A. S. Bolton, J. Brinkmann, J. R. Brown- [71] S. R. Taylor, J. R. Gair, and I. Mandel, Phys. Rev. D stein, A. Burden, A. J. Cuesta, et al., MNRAS 427, 85, 023535 (2012), . 3435 (2012), 1203.6594. [72] D. E. S. Collaboration:, T. Abbott, F. B. Abdalla, [102] S. Alam, M. Ata, S. Bailey, F. Beutler, D. Bizyaev, J. Aleksic, S. Allam, A. Amara, D. Bacon, E. Balbinot, J. A. Blazek, A. S. Bolton, J. R. Brownstein, A. Bur- M. Banerji, K. Bechtol, et al., Monthly Notices of the den, C.-H. Chuang, et al., MNRAS 470, 2617 (2017), Royal Astronomical Society 460, 1270 (2016), ISSN 1607.03155. 0035-8711, https://academic.oup.com/mnras/article- [103] T. B. Price, LII. An essay towards solving a problem pdf/460/2/1270/8117541/stw641.pdf, . in the doctrine of chances. By the late Rev. Mr. Bayes [73] A. Aghamousa et al. (DESI) (2016), 1611.00036. (cited January 1763). [74] A. Refregier, A. Amara, T. D. Kitching, A. Rassat, [104] P. Coles and B. Jones, Mon. Not. Roy. Astron. Soc. 248, R. Scaramella, J. Weller, and f. t. Euclid Imaging Con- 1 (1991). sortium, ArXiv e-prints (2010), 1001.0061. [105] F. Bernardeau and L. Kofman, ApJ 443, 479 (1995), [75] J. Green, P. Schechter, C. Baltay, R. Bean, D. Bennett, astro-ph/9403028. R. Brown, C. Conselice, M. Donahue, X. Fan, B. S. [106] S. Colombi, Astrophys. J. Lett. 435, L536 (1994), astro- Gaudi, et al., arXiv e-prints arXiv:1208.4012 (2012), ph/9402071. 1208.4012. [107] E. Hubble, ApJ 79, 8 (1934). [76] D. Spergel, N. Gehrels, J. Breckinridge, M. Donahue, [108] V. Wild et al. (2dFGRS), Mon. Not. Roy. Astron. Soc. A. Dressler, B. S. Gaudi, T. Greene, O. Guyon, C. Hi- 356, 247 (2005), astro-ph/0404275. rata, J. Kalirai, et al., arXiv e-prints arXiv:1305.5425 [109] L. Clerkin et al. (DES), Mon. Not. Roy. Astron. Soc. (2013), 1305.5425. 466, 1444 (2017), 1605.02036. [77] O. Dore et al. (WFIRST), ArXiv e-prints (2018), [110] B. Abbott et al. (LIGO Scientific, Virgo), Phys. Rev. X 1804.03628. 9, 031040 (2019), 1811.12907. [78] LSST Science Collaboration, P. A. Abell, J. Allison, [111] B. Abbott et al. (LIGO Scientific, Virgo), Astrophys. J. S. F. Anderson, J. R. Andrew, J. R. P. Angel, L. Armus, Lett. 882, L24 (2019), 1811.12940. D. Arnett, S. J. Asztalos, T. S. Axelrod, et al., ArXiv [112] N. Hand, Y. Feng, F. Beutler, Y. Li, C. Modi, U. Seljak, e-prints (2009), 0912.0201. and Z. Slepian, Astron. J. 156, 160 (2018), 1712.05834. [79] O. Dore et al., arXiv (2018), 1805.05489. [113] S. Fairhurst, Class. Quant. Grav. 28, 105021 (2011), [80] M. Oguri, Phys. Rev. D 93, 083511 (2016), . 1010.6192. [81] S. Mukherjee and B. D. Wandelt (2018), 1808.06615. [114] M. L. Chan, Ph.D. thesis, Glasgow U. (2018). 16

[115] B. Sathyaprakash and S. Dhurandhar, Phys. Rev. D 44, [124] A. J. Ross et al., Mon. Not. Roy. Astron. Soc. 498, 2354 3819 (1991). (2020), 2007.09000. [116] C. Cutler and E. E. Flanagan, Phys. Rev. D49, 2658 [125] D. Scolnic et al. (2019), 1903.05128. (1994), gr-qc/9402014. [126] E. Petroff, J. Hessels, and D. Lorimer, Astron. Astro- [117] R. Balasubramanian, B. Sathyaprakash, and S. Dhu- phys. Rev. 27, 4 (2019), 1904.07947. randhar, Phys. Rev. D 53, 3033 (1996), [Erratum: [127] D. Foreman-Mackey, The Journal of Open Source Soft- Phys.Rev.D 54, 1860 (1996)], gr-qc/9508011. ware 24 (2016), . [118] A. Ghosh, W. Del Pozzo, and P. Ajith, Phys. Rev. D [128] D. Foreman-Mackey, D. W. Hogg, D. Lang, and 94, 104070 (2016), 1505.05607. J. Goodman, PASP 125, 306 (2013), 1202.3665. [119] P. Ajith et al., Phys. Rev. D77, 104017 (2008), [Erra- [129] F. Pérezand B. E. Granger, Computing in Science and tum: Phys. Rev.D79,129901(2009)], 0710.2335. Engineering 9, 21 (2007), ISSN 1521-9615, . [120] S. W. Hawking and W. Israel, Three hundred years of [130] J. D. Hunter, Computing In Science & Engineering 9, gravitation (Cambridge University Press, 1987). 90 (2007). [121] E. Poisson and C. M. Will, Phys. Rev. D52, 848 (1995), [131] S. van der Walt, S. C. Colbert, and G. Varoquaux, gr-qc/9502040. Computing in Science and Engineering 13, 22 (2011), [122] M. Maggiore, Gravitational Waves: Volume 1: Theory 1102.1523. and Experiments, Gravitational Waves (OUP Oxford, [132] E. Jones, T. Oliphant, P. Peterson, et al., SciPy: Open 2008), ISBN 9780198570745, . source scientific tools for Python (2001–), [Online; ac- [123] S. Bera, D. Rana, S. More, and S. Bose, Astrophys. J. cessed ¡today¿], . 902, 79 (2020), 2007.04271.