Astronomy & Astrophysics manuscript no. AO_cosmography ©ESO 2020 November 13, 2020

TDCOSMO VI: Distance Measurements in Time-delay Cosmography under the Mass-sheet transformation

Geoff C.-F. Chen ,1,★ Christopher D. Fassnacht,1 Sherry H. Suyu,2, 3, 4 Akın Yıldırım,2 Eiichiro Komatsu,2, 5 and Jose´ Luis Bernal6

1 Department of Physics and Astronomy, University of California, Davis, CA 95616, USA 2 Max Planck Institute for Astrophysics, Karl-Schwarzschild-Strasse 1, D-85740 Garching, Germany 3 Physik-Department, Technische Universität München, James-Franck-Straße 1, 85748 Garching, Germany 4 Academia Sinica Institute of Astronomy and Astrophysics (ASIAA), 11F of ASMAB, No.1, Section 4, Roosevelt Road, Taipei, 10617, Taiwan 5 Kavli IPMU (WPI), UTIAS, The University of Tokyo, Kashiwa, Chiba 277-8583, Japan 6 Department of Physics and Astronomy, Johns Hopkins University, 3400 North Charles Street, Baltimore, Maryland 21218, USA Received xxx, xxxx; accepted xxx, xxxx

ABSTRACT

Time-delay cosmography with gravitationally lensed quasars plays an important role in anchoring the absolute distance scale and hence measuring the Hubble constant, 퐻0, independent of traditional distance ladder methodology. A current potential limitation of time delay distance measurements is the “mass-sheet transformation” (MST) which leaves the lensed imaging unchanged but changes the distance measurements and the derived value of 퐻0. In this work we show that the standard method of addressing the MST in time delay cosmography, through a combination of high-resolution imaging and the measurement of the stellar velocity dispersion of the lensing galaxy, depends on the assumption that the ratio, 퐷s/퐷ds, of angular diameter distances to the background quasar and between the lensing galaxy and the quasar can be constrained. This is typically achieved through the assumption of a particular cosmological model. Previous work (TDCOSMO IV) addressed the mass-sheet degeneracy and derived 퐻0 under the assumption of ΛCDM model. In this paper we show that the mass sheet degeneracy can be broken without relying on a specific cosmological model by combining lensing with relative distance indicators such as supernovae type Ia and baryon acoustic oscillations, which constrain the shape of the expansion history and hence 퐷s/퐷ds. With this approach, we demonstrate that the mass-sheet degeneracy can be constrained in a cosmological-model-independent way, and hence model-independent distance measurements in time-delay cosmography under mass-sheet transformations can be obtained. Key words. method: gravitational lensing: strong – cosmology: distance scale

1. Introduction tathiou 2020) or not, clearly demonstrates that it is crucial to test any single methodology by different and independent datasets. The Hubble constant (퐻0) is one of the most important parame- ters in cosmology. Its value directly sets the age, the size, and the Time-delay cosmography (TDC; e.g., Treu & Marshall 2016; critical density of the Universe. Despite the great success of the Suyu et al. 2018) provides a technique to constrain 퐻0 at low ΛCDM model (Komatsu et al. 2011; Hinshaw et al. 2013; Planck redshift that is completely independent of the traditional distance Collaboration et al. 2018), a stringent challenge to the model ladder approach. When a quasar is strongly lensed by a galaxy, its multiple images have light-curves that are offset by a well- comes from a discrepancy between the extremely precise 퐻0 (= 67.4±0.5 km s−1 Mpc−1) value derived from Planck measure- defined time delay, which depends on the mass profile of the lens ments of the cosmic microwave background (CMB) anisotropies and cosmological distances to the galaxy and the quasar (Refsdal under the assumption of ΛCDM (Planck Collaboration et al. 1964). A critical aspect of this technique is a model that describes the mass distribution in the lensing galaxy and along the line 2018), and the 퐻0 value from direct measurements of the local Universe (Verde et al. 2019). of sight between the background object and the observer. This arXiv:2011.06002v1 [astro-ph.CO] 11 Nov 2020 model is constrained by the morphology of the lensed emission The recent direct 퐻0 measurements (퐻0 = 74.03 ± −1 −1 of the background object, the stellar velocity dispersion in the 1.42 km s Mpc ) from Type Ia supernovae (SN1a), calibrated lensing galaxy, and by deep imaging and spectroscopy of the by the traditional Cepheid distance ladder (SH0ES collabora- fields containing the lens system. This model is combined with the tion; Riess et al. 2019), show a 4.4휎 tension with the Planck re- time delays (e.g., Bonvin et al. 2018) to measure the characteristic = ± ( ) ± sults. However, a recent measurement of 퐻0 69.8 0.8 stat distances for the lens system: the angular diameter distance to the 1.7(sys) km s−1 Mpc−1 from SN1a calibrated by the Tip of the lens (퐷d) and the time-delay distance, which is a ratio of the Red Giant Branch (CCHP) agrees at the 1.2휎 level with Planck angular diameter distances in the system: and at the 1.7휎 with SH0ES results (Freedman et al. 2019). The spread in these results, whether due to systematic effects (Efs-

퐷d퐷s −1 ★ Current email address: [email protected] 퐷Δt ≡ (1 + 푧d) ∝ 퐻0 , (1) 퐷ds Article number, page 1 of 11 A&A proofs: manuscript no. AO_cosmography

+4.1 −1 −1 where 푧d is the redshift of the lens, 퐷s is the distance to the population, the inference on 퐻0 shifted to 67.4−3.2 km s Mpc , background source, and 퐷ds is the distance between the lens in agreement with the Planck value and results from distance lad- and the source. In turn, these distances are used to determine ders (Riess et al. 2019; Freedman et al. 2019). Comparison of cosmological parameters, primarily 퐻0 (e.g., Suyu et al. 2014; the galaxy population distributions show that several observed Bonvin et al. 2016; Birrer et al. 2019; Chen et al. 2019; Rusu properties, such as central stellar velocity dispersion, are similar. et al. 2019; Wong et al. 2019; Jee et al. 2019; Taubenberger et al. In addition, elliptical galaxies are a very homogenous population, 2019; Shajib et al. 2020a). as evidenced by the tightness of correlations like the fundamental A recent analysis with this technique, using a blind analy- plane (Auger et al. 2010, and references therein). However, some sis on data from six gravitational lens systems1, inferred 퐻0 = major differences between the samples are that (1) the SLACS +1.7 −1 −1 73.3−1.8 km s Mpc , a value that was 3.8휎 offset from the lensing galaxies are at lower redshifts than those in the time-delay Planck results (Wong et al. 2019; Millon et al. 2020). This analy- sample, and (2) the SLACS lensing galaxies have smaller ratio of sis used two commonly used descriptions of the mass distribution effective radius to Einstein radius than the time-delay sample (see of the lensing galaxy. The first description consists of a NFW halo Fig. 16 in Paper IV). Possible potential biases and limitations of (Navarro et al. 1996) plus a constant mass-to-light ratio stellar using the SLACS sample are discussed by Paper IV and Shajib distribution (the “composite model”). The second description et al.(2020b). models the three dimensional total mass density distribution, i.e., In this work, we take a more general approach to constrain the luminous plus dark matter, of the galaxy as a power law (Barkana internal MST by combining the time-delay lens system with rela- 1998), i.e., 휌(푟) ∝ 푟−훾 (the power-law model). These models tive distance indicators without assuming a specific parametriza- yield 퐻0 measurements that are consistent within the errors for tion of the cosmological model. We show that one can hence individual lens systems; the final uncertainties on 퐻0 incorporate constrain the internal MST in a cosmological independent way a marginalization over the choice of mass model (Millon et al. and obtain more broadly applicable distance posteriors. In Sec- 2020). tion2, we introduce the basics of the mass-sheet transformation. Although the power-law and composite models are well- In Section3 and Section4, we discuss the distance measurements motivated by both observations (e.g., Koopmans et al. 2006, under the effects of the internal and external MST. In Section5, 2009; Suyu et al. 2009; Auger et al. 2010; Barnabè et al. 2011; we discuss error propagation under MST. In Section6, we provide Sonnenfeld et al. 2013; Humphrey & Buote 2010; Cappellari a cosmological-model-independent way to constrain the internal 2016) and simulations (Navarro et al. 1996), there is a well- MST. We summarize our work in Section7. known degeneracy in gravitational lensing known as the mass- sheet transformation (MST), that leaves imaging observables in- 2. The Mass-sheet transformation variant but will bias the determination of 퐻0 (Falco et al. 1985; Gorenstein et al. 1988). The line-of-sight mass distribution con- The mass-sheet transformation is a degeneracy affecting grav- tributes to first order mass-sheet-like effect (Fassnacht et al. 2002; itational lens systems. One can transform any projected mass Suyu et al. 2013; Greene et al. 2013; Collett et al. 2013); we will distribution, 휅(휃), into infinite sets of 휅휆 (휃) via refer to this as an external MST. However, for the mass dis- tribution of the lensing galaxy, there are different models that 휅휆 (휃) = 휆휅(휃) + 1 − 휆, (2) can give the same lensing observables but would give different without degrading the fit to the lensed emission (Falco et al. time-delays. The most degenerate case is the one with spherical 1985), although MST does change the source size accordingly. symmetry in which the density profiles differ by a component Here 휅(휃) is a scaled two-dimensional projected mass density that is uniform in within the radial ranges probed by lensing. distribution, 휅(휃) = Σ(휃)/Σcrit, where Σ(휃) is the mass surface This component which could be described by a large-core mass density and Σ is the lensing critical density, distribution (Blum et al. 2020, see detail in Section 2), change the crit distribution of the mass density profile of the lensing galaxy. It 푐2 퐷 Σ = s . (3) fits with recent work that question whether that elliptical galaxies cr 4휋퐺 퐷 퐷 do not necessarily follow a power-law or composite model to the d ds desired precision (Schneider & Sluse 2013; Xu et al. 2016). The physical picture of MST comes from both the environment Birrer et al.(2020) (hereafter Paper IV) showed that allowing (a.k.a., an external MST, 휅ext) and the mass models of the lens- for an internal MST on the power-law model increased the un- ing galaxy (a.k.a., an internal MST, 휆int). We separate these two certainty of the 퐻0 measurement of a seven-lens sample from the components of the MST because we use different observables to 2.4% precision of Millon et al.(2020) to 8%, in a ΛCDM cos- assess their effects. For example, the estimation of the external mology. Interestingly, the central value of 퐻0 remained almost MST uses weighted number counts of galaxies and/or weak grav- +5.6 −1 −1 itational lensing, based on spectroscopy and deep imaging of the unchanged in this analysis (74.5−6.1 km s Mpc ). To improve the precision of the 퐻0 inference, Paper IV added data from the field containing the lens. This approach has been extensively used SLACS sample (Bolton et al. 2004, 2006). In this lens sample, the in TDC (e.g., Fassnacht et al. 2006; Suyu et al. 2010; Collett et al. background objects are galaxies, not quasars, so they cannot be 2013; Greene et al. 2013; Rusu et al. 2017; Tihhonova et al. 2018; used for time-delay cosmography. However, the combination of Buckley-Geer et al. 2020). Information about the internal MST high-resolution imaging and kinematic measurements allows the is derived from high-resolution imaging and the stellar velocity SLACS sample to improve the constraints on the mass profiles dispersion of the lensing galaxy. of massive elliptical galaxies. With the inclusion of the SLACS The theoretical version of the internal MST, i.e., a mass sheet information (Shajib et al. 2020b) and the assumption that the sam- with infinite extent, is clearly non-physical. Therefore, in as- ple of time-delay and SLACS lenses are drawn from the same sessing the internal MST we need to find a physical model that approximates the behavior of a mass sheet at small projected dis- 1 Except the first lens, B1608+656, which was not done blindly, the tances from the center of the lensing galaxy, but that vanishes at subsequent five lenses in H0LiCOW are analyzed blindly with respect large radii (see Fig. 7 in Schneider & Sluse 2013). One example to the cosmological quantities of interest. of this was proposed by Blum et al.(2020), who introduced a

Article number, page 2 of 11 G. C.-F. Chen et al.: Distance measurements in TDC under MST

3. The measurement of 푫횫t under the MST Once the time delays between multiple images are observed, one can measure the time-delay distance via

퐷 Δ푡 = Δt Δ휙(휃, 훽), (8) 푐 where 푐 is the speed of light and 휃, 훽, and 휙(휃) are the image coordinates, the source coordinates, and the Fermat potential respectively. The form of Equation (8) allows the inference of the cosmological information contained in 퐷Δt without any need for cosmological priors on the lens modeling. However, in the presence of a MST, given the same time delays and imaging data, the transformed projected mass profile produces a different time-delay distance via

퐷 퐷 = Δt . (9) Δ푡,휆 휆

Fig. 1: Illustration of the transformed power-law profile in the Thus, additional information is required to constrain both the mean dimensionless enclosed projected mass distribution under internal and external MST, and thus to obtain unbiased 퐷Δt mea- 00 the internal MST with 휃s = 10 . All the transformed mass profiles surements. share the same Einstein radius (red dashed line). All the mass distributions in this figure produce essentially the same model images but different unlensed size of the source, which is not 4. The measurement of 푫d under the MST directly observable. Once the velocity dispersion of the lensing galaxy is measured, one can use high-resolution imaging of the lens system to measure the ratio 퐷 /퐷 via mass profile that can satisfy these requirements. For this profile, s ds the physical internal MST, which redistributes any specific mass   p 2 퐷s 2 profile, 휅(휃), should be written as (휎푣 ) = 푐 퐽(휂lens, 휂light, 훽ani), (10) 퐷ds

휅mst,int (휃) = 휆int휅(휃) + (1 − 휆int)휅c (휃), (4) p where 휎푣 is the line-of-sight luminosity-weighted velocity dis- persion that is predicted by the mass distribution in the lensing where galaxy. Here, 퐽 contains the angular-dependent information in- cluding the parameters describing the 3D deprojected mass dis- 휃s 휅c (휃) = , (5) tribution, 휂lens, the surface-brightness distribution in the lensing √︁ 2 + 2 휃s 휃 galaxy, 휂light, and the stellar orbital anisotropy distribution, 훽ani. In a similar way to the time-delay distance, the separability in and 휃s is the scale radius. When we set 휃s to a large value, e.g. Equation (10) allows us to infer the cosmological distance ra- 00 ( ) 10 , 휅c 휃 approximates the theoretical internal MST very well tio 퐷s/퐷ds without the need of cosmological priors on 퐽. Since over the region of interest (Paper IV). 퐷Δt ∝ 퐷d (퐷s/퐷ds), we can use the combination of the 퐷Δt We illustrate the effects of adding such a mass-sheet profile measurement from the time delays and 퐷s/퐷ds from velocity to the lensing galaxy mass distribution in Figure1, by plotting dispersion to obtain 퐷d. the mean dimensionless enclosed projected mass distribution, We discuss the effect of 휅ext and 휆int on the 퐷d measurement in the following two sections. 2 ∫ 휃 휅¯(휃) ≡ 휃0휅 (휃0)푑휃0. (6) 2 mst,int 휃 0 4.1. External MST only

The Einstein radius of the lens system, 휃E, is defined as the Jee et al.(2015) found that 퐷d is an invariant quantity under an angular radius for which 휅¯(휃E) = 1. external MST. This is because 휅ext only contributes to the change Thus, the general MST which accounts for both 휅ext and 휆int of the normalization of 3D mass profile and does not affect its can be written as overall shape given any mass model (Suyu et al. 2013; Chen et al. 2019). That is, Equation (10) changes to 휅휆 (휃) = (1 − 휅ext)휅mst,int (휃) + 휅ext = 휅true (휃), (7)  퐷  (휎p)2 = (1 − 휅 ) s 푐2퐽(휂 , 휂 , 훽 ), (11) 푣 ext 퐷 lens light ani where 휅true represents the true projected mass profile. In this ds 휅ext paper we set the stage for future investigations by dissecting where the constraining power on the distance measurements in where the minus sign means that one needs to remove the mass TDC comes from, and exploring what assumptions have to be contributed from the environment (i.e., the mass along the line of made and data have to be used in order to break the internal sight) inside the Einstein radius to obtain the mass of the lensing MST. galaxy.

Article number, page 3 of 11 A&A proofs: manuscript no. AO_cosmography

defined in terms of the mean dimensionless enclosed projected mass distribution (휅¯), is invariant under an internal MST (i.e., 휅¯(휃E) = 휅¯휆 (휃E) = 1; see also Figure1), while the physical mass inside the Einstein radius is unconstrained without assum- ing a cosmological model. Secondly, from Equation (15) we show hereafter that if a cosmological model is not assumed, then the values of 휆int and 퐷s/퐷ds, which affect the shape and normaliza- tion respectively, are degenerate. Hence, even with a measured velocity dispersion, the mass inside the effective radius is also not constrained. Therefore, a single-aperture velocity dispersion is in- sufficient to break the degeneracy and constrain the internal MST if one does not assume a cosmological model. Spatially resolved kinematics of the lensing galaxy would be required (Yildirim et al. in prep.). In order to illustrate these dependencies, we use the power-law mass model, which was obtained by fitting to the real imaging data of a four-image gravitational lens system (J0924+0219) shown Fig. 2: The multiple lensed images and the extended arc around in Figure2 (see Chen et al. in prep for details), and analysed the lensing galaxy are from the background AGN and its recon- it in the context of an internal MST (i.e., added to our model a structed host galaxy. The foreground main lens is located in the MST component as described in Equation (5)). For the anisotropy center of the lens system. The solid horizontal line represents component, we assume 훽ani varies with radius, and parameterize 00 1 scale. The detailed lens modeling will be presented in Chen et this behaviour in the form of an anisotropy radius, 푟ani, in the al. in prep. Note that the study of this paper is not limited to any Osipkov-Merritt formulation (Osipkov 1979; Merritt 1985), specific configuration of the lens systems. 푟2 훽ani = , (17) 푟2 + 푟2 In order for the predicted velocity dispersion to match the ani as an example. In this formulation, 푟 = 0 indicates pure radial observed value, 퐷s/퐷ds must transform under an external MST ani via orbits and 푟ani → ∞ is isotropic with equal radial and tangential velocity dispersions. In our models, we use a scaled version of       퐷s −1 퐷s 퐷s the anisotropy parameter, 푎ani ≡ 푟ani/푟eff, where 푟eff = 퐷d휃eff, = (1 − 휅ext) = . (12) 퐷 퐷 퐷 and 휃eff is the effective radius. The redshift of the lens and source ds 휅ext ds ds true are 푧d = 0.393 and 푧s = 1.523, respectively. Note that the study Then the time-delay distance changes via of this work is not limited to any specific configuration of the lens systems. We set mock time delays (Δ푡AB = 10 ± 1.5 days, = ( − )−1 퐷Δ푡,휅ext 1 휅ext 퐷Δt. (13) Δ푡CB = 15 ± 1.5 days, Δ푡DB = 10 ± 1.5 days), and a mock veloc- ity dispersion measurement (= 279 ± 15 km s−1) for the analysis. Thus, by combining Equation (1), Equation (12), and Equa- The uncertainties on the time delays and velocity dispersion are tion (13), one can show that 퐷 is invariant under the external d typical of time-delay lens systems. Note that all the figures pro- MST, duced in this work are based on this single lens. Since 휅ext does not affect the 퐷d measurement and is well-understood, we set (퐷d)휅ext = 퐷d. (14) 휅ext = 0 throughout the paper for simplicity. We consider two situations, one with a flat ΛCDM cosmology, and another in which we only use the velocity dispersion, time 4.2. Internal MST+External MST delays, and imaging data without assuming any cosmological Since the velocity dispersion depends on the enclosed 3D mass of model. Note that throughout the paper, for flat ΛCDM cosmology, the lensing galaxy, whose shape is not conserved under an internal we assume Ωm = [0.05, 1.0], ΩΛ = 1 − Ωm, and 퐻0 uniform in −1 −1 MST, the constraint on the 퐷s/퐷ds ratio is not mathematically [0, 150] km s Mpc . The results are shown in Figure3, where −1 the ΛCDM results are shown as points that are color-coded by scaled by 휆int under an internal MST. We therefore must expand the Equation (11) by including 휆int in the 퐽 term, the velocity dispersion in order to demonstrate the link between 휎푣 and the other parameters. Figure3 clearly demonstrates that  퐷  휆 can be constrained only when ΛCDM is assumed, on the ( p)2 = ( − ) s ( ) int 휎푣 1 휅ext 퐽 휂lens, 휂light, 훽ani, 휆int , (15) contrary, when the cosmological model constraint is relaxed, 퐷 퐷ds true d changes very little within the physically-allowed values of 휆int. since 퐽 contains the 3D de-projected dimensionless mass model Thus the 퐽 term in Equation (15) can be very well approximated components, whose structure is affected by the value of 휆int. Thus, as 퐽(휆int) = 휆int퐽 with < 1% shift on the distance measurement 00 퐷d can be expressed as of 퐷d for 휃s = 10 and with single aperture averaged velocity dispersion. This approximation was also used by Paper IV. 1 퐷 푐2 퐷 = Δt 퐽(휂 , 휂 , 훽 , 휆 ). (16) d 2 lens light ani int 1 + 푧d 휆int 휎 푣 5. Error propagation in MST Many previous investigations (e.g. Suyu et al. 2014, Paper 5.1. Error propagation without assuming a cosmological model IV) show that the internal MST can be broken with a single- aperture velocity dispersion, given a cosmological model. This In the previous section, we showed that 퐷Δt is directly affected can be explained as follows: firstly, the Einstein ring radius, as by both external and internal MST, while 퐷d is not affected by

Article number, page 4 of 11 G. C.-F. Chen et al.: Distance measurements in TDC under MST

Fig. 3: The comparison of the 퐷d, 퐷Δt, 휆int, and 푎ani measurements with and without the assumption of ΛCDM model from single time-delay mock lens. ‘TD’ represents time delay information and ‘VD’ represents velocity dispersion information. When ΛCDM model is not assumed, the internal MST (휆int) is not constrained. When ΛCDM model is assumed, the degeneracy can be broken and hence 퐷Δt is constrained. The anisotropy parameter is not constrained in either case. Color-coded velocity dispersion shows that 퐷Δt is positively correlated with 퐷d but anti-correlated with 휎푣 . The contours represent the 68.3% (shaded region) and 95.4% quantiles.

휅ext and is nearly invariant. Thus, based on Equation (9) the error 5.2. Error propagation under ΛCDM model on 퐷Δt (훿퐷Δt) given 휆 and 휆int scales as However, if one assumes a ΛCDM model, the error on 휆int is 훿퐷 훿휆 훿휆 Δt ∼ − = − int , (18) 훿휆int 훿휎푣 퐷 휆 휆 ∼ 2 . (20) Δt int 휆int 휎푣 while based on Equation (16), the error of 퐷d scales as By combining Equations 18–20, the correlations between the errors on 퐷Δt, 퐷d, and 휎푣 are 훿퐷d 훿휎푣 ∼ −2 , (19) 훿퐷Δt 훿퐷d 훿휎푣 퐷d 휎푣 ∼ ∼ −2 . (21) 퐷Δt 퐷d 휎푣 where 휎 is the measured line-of-sight velocity dispersion. Thus, 푣 In Figure3, we indeed see that under the assumption of ΛCDM the uncertainties on 퐷 are dominated by the velocity dispersion d model, 퐷 is positively correlated with 퐷 but anti-correlated measurement errors, while the 퐷 uncertainties are dominated Δt d Δt with 휎 . by both internal and external MST. Therefore, 퐻 inferred solely 푣 0 Most importantly, while Equation (20) and Equation (21) tell from 퐷d is robust against the MST (Jee et al. 2019)2. us that both 퐷Δt and 퐷d are anti-correlated with 휆int under ΛCDM 2 Note that Jee et al.(2019) shows that different anisotropy model may model, 퐷d is nearly uncorrelated with 휆int without assuming a slightly shift the inferred 퐷d value. cosmological model inside the physically-allowed range of 휆int.

Article number, page 5 of 11 A&A proofs: manuscript no. AO_cosmography

Fig. 4: The decomposition of the constraining power from time delays (TD), velocity dispersion (VD), and imaging data under the assumption of ΛCDM model. When only ‘TD+imaging’ is used, the values of 퐷d and 퐷Δt are not constrained, and both of them are fully degenerate with 휆int. Since the velocity dispersion constrains the 휆int, the joint constraint (‘TD+VD+imaging data’) breaks the degeneracy and hence constrains 퐷Δt and 퐷d. The anisotropy parameters are not constrained in all cases.

6. Constraining the internal MST ments of time delays and luminosity-weighted stellar velocity dispersions with errors that are typical of those found in previous In the previous sections of this paper, we have shown that in works in this field. the presence of an internal MST, parameterized by 휆int, the time-delay distance measurement is poorly constrained unless 6.1. Method 1: Choosing a cosmological model a specific cosmological model is picked. This situation is clearly demonstrated in the 휆int vs. 퐷Δt panel in Figure3, where 휆int In Section 4.2, we demonstrated that 퐷s/퐷ds and 휆int are degener- is essentially unconstrained without the assumption of a cosmo- ate quantities. However, once a cosmological model is assumed, logical model. In turn, the large uncertainties in 휆int translate 퐷s/퐷ds can be determined up to a range depending on the other into imprecise inferences on 퐷Δt. Therefore, in this section we cosmological parameters such as Ω푚 and Ω푘 (Grillo et al. 2008), describe two approaches to improving the constraints on 휆int. since the unknown factor of 퐻0 cancels out in the ratio. This The first approach constrains 휆int in a fashion that depends on means that the measurement of the velocity dispersion constrains the cosmological model (Paper IV), while the second approach 휆int, hence the mass inside the effective radius of the lensing works even if we are agnostic about the cosmological model. In galaxy. Once 휆int is constrained, 퐷Δt is constrained and 퐷d can both cases, we assume that observations have provided measure- be inferred from 퐷Δt. Hence the mass inside the Einstein radius is

Article number, page 6 of 11 G. C.-F. Chen et al.: Distance measurements in TDC under MST

퐷푉 /푟s, the dilation scale normalized by the standard ruler length (or 퐷A/푟s and 퐻푟s in the anisotropic analysis), where

 푐푧 1/3 퐷 ≡ 퐷2 (푧) , (23) 푉 M 퐻(푧)

퐷M = (1 + 푧)퐷A, and 퐷A is the angular diameter distance. If we vary 푀 and 푟s freely, those data sets provide the information on the shape of the expansion history (e.g., Cuesta et al. 2015) and thus 퐷s/퐷ds. However, we do still require some model for the redshift dis- tance relationship to connect these data. In this work, we choose piece-wise natural cubic splines3 to describe 퐻(푧) that fit to the data. The spline method has been used in many studies to recon- struct the expansion history (Bernal et al. 2016; Poulin et al. 2018; Aylor et al. 2019; Bernal et al. 2019). The splines are set by the values they take at different redshifts. These values can uniquely Fig. 5: Results of the reconstruction of 퐸 (푧) (= 퐻(푧)/퐻0) using define the piece-wise cubic spline once we require continuity of the SN1a and BAO data, normalised by the values for our fiducial 퐻(푧) and its first and second derivatives at the knots, and set two cosmology given by the best-fitting parameters from the Planck boundary conditions. We also require the second derivative to analysis for a ΛCDM model. Each thin line comes from a random vanish at the exterior knots. We set the minimal assumptions of draw among the points in parameter space within 68% confidence this work: level. The thick line in the middle represents the best fit. The shape of expansion history is described by piece-wise natural 1. Cosmological principle of homogeneity and isotropy (i.e., cubic splines. The splines can be used to constrain 퐷s/퐷ds and Friedmann–Lemaître–Robertson–Walker metric), hence break the internal mass-sheet transformation. 2. Assumption of general relativity: the curvature density pa- 2 rameter is given by Ω푘 = 푘 [푐/(푅0퐻0)] where 푘 = {1, 0, −1}, 푅0 denotes the present value of the scale factor, assigned. Therefore, by assuming a cosmological model, the in- 3. Spline 퐻(푧) completely specifies the FLRW metric. 2 ternal MST can be broken to a level that depends on the precision 4. Cosmic distance duality relation: 퐷 퐿 = 퐷 퐴(1 + 푧) . of the velocity dispersion measurement. To further illustrate the effect of assuming a cosmological model, we show the constrain- To get a good constraint on the cosmological-model- ing power on 퐷 , 퐷 , 휆 , and the anisotropy parameter (푎 ) independent 퐷s/퐷ds at the redshift of the mock lens, we need the d Δt int ani 4 when setting the cosmological model to ΛCDM in Figure3. We data that cover the redshift up to the source redshift (푧s = 1.523 in this case). Current existing data show that one can constrain clearly see the correlation between 퐷Δt and inferred 퐷d under the assumption of ΛCDM. the shape of the expansion history up to ∼ 푧 = 2.5 (see Fig. 3 in For the case where no cosmological model is used in Fig- Bernal et al. 2019). Therefore, we update the likelihood used in Bernal et al.(2019) and use Pantheon data sets of SN1a (Scol- ure3, we see that 퐷Δt is degenerate with 휆int. In contrast, by assuming a cosmological model, one restricts the allowed range nic et al. 2018); BAO from galaxies (Kazin et al. 2014; Alam et al. 2017; Gil-Marín et al. 2020), quasars (Hou et al. 2020; of 휆int and thus places stronger constraints on the inferences on the cosmological distances. We can decompose the black con- Neveux et al. 2020) and the Lyman-훼 forest (du Mas des Bour- tour in Figure3 into separate cases to examine the constraining boux et al. 2020). For the eBOSS likelihoods, we use the Gaussian power from the velocity dispersion only (VD only), time-delay approximation and that BAO can be used to apply cosmological measurements only (TD only), and a joint constraint from both models beyond LCDM (Bernal et al. 2020; Carter et al. 2020). measurements (VD+TD). Figure4 clearly shows that the velocity We summarize the BAO measurements and the redshift informa- tion in Table1. We set 5 “knots” at different redshifts ( 푧0 = 0, dispersion constrains 휆int when assuming a cosmological model 푧1 = 0.25, 푧2 = 0.5, 푧3 = 1., 푧4 = 2.5). The complete set of pa- (e.g., ΛCDM). In other words, the value of 휆int depends on the rameters for the Spline model is {퐻0, 퐻1, 퐻2, 퐻3, 퐻4, 푟s, Ω푘 , 푀}. cosmological model, and the measurement of 퐷Δt in this case is not a cosmological-model-independent quantity. Uniform priors are assumed for all parameters. We show the re- constructed 퐸 (푧)(= 퐻(푧)/퐻0) normalised by the values for our fiducial cosmology given by the best-fitting parameters from the 6.2. Method 2: Using external data sets to constrain 퐷s/퐷ds Planck analysis for a ΛCDM model in Figure5. The posterior of 퐷 /퐷 can be obtained by integrating 퐸 (푧). To break the internal MST without assuming a particular cos- s ds In Figure6, we show that by combining the inference from mological model (e.g., ΛCDM model), we require additional the external datasets on 퐻(푧)/퐻 , which constrain 퐷 /퐷 , with information to constrain 퐷 /퐷 . This can be done by including 0 s ds s ds time-delay strong lensing systems, one can obtain cosmological- data on Type Ia supernovae (SN1a) and Baryon acoustic oscilla- model-independent 퐷 and 퐷 measurements that are compara- tions (BAO). The SN1a data are given as measurements of the d Δt ble to those obtained by assuming a ΛCDM or 푤CDM cosmol- distance modulus ogy. In addition, we also see that the values of 휆int under ΛCDM 휇(푧) ≡ 푚 − 푀 = 25 + 5log10퐷 퐿 (푧), (22) 3 Note that linear interpolation (e.g. Verde et al. 2017), Gaussian Pro- cesses (e.g. Joudaki et al. 2018; Liao et al. 2020), or smooth Taylor where 푚 is the apparent magnitude, 푀 is a fiducial absolute expansion (e.g. Macaulay et al. 2019; Wojtak & Agnello 2019; Arendse magnitude, and 퐷 퐿 is the luminosity distance. When 푀 is a free et al. 2019) are alternatives. parameter without calibration, SN1a only constrain the shape of 4 For the current 7 TDCOSMO lenses, the source redshifts are all below the expansion history. The BAO data provide measurements of 푧s = 2.5.

Article number, page 7 of 11 A&A proofs: manuscript no. AO_cosmography

Fig. 6: Comparison of the inferred 퐷d, 퐷s/퐷ds, 퐷Δt, and 휆int in different cases: (1; orange) The distance measurements directly from single time-delay mock lens without assuming any cosmological model. (2; blue) The distance measurements under the assumption −1 −1 of flat 푤CDM model with Ωm = [0.05, 1.0], Ωde = 1 − Ωm, 푤 = [−2.5, 0.5], and 퐻0 uniform in [0, 150] km s Mpc . (3; black) The distance measurements under the assumption of ΛCDM model with Ωm = [0.05, 1.0], ΩΛ = 1 − Ωm, and 퐻0 uniform in [0, 150] km s−1 Mpc−1. (4; green) The distance measurements from single time-delay mock lens, Pantheon dataset, and BAO dataset by using splines with free Ω푘 . (5; red) The distance measurements from single time-delay mock lens, Pantheon dataset, and BAO dataset by using splines with Ω푘 = 0. For the cases without assuming particular cosmological model, the constraining power on 휆int is comparable to the cases with an assumption of having a underlying cosmological model. and 푤CDM are slightly offset from the cases which include SN1a 6.3. Comparison with Paper IV and BAO data sets. This is because the flat priors on Ωm and 푤 in ΛCDM and 푤CDM models do not reflect the expansion his- In the previous section, we demonstrate that 퐷s/퐷ds is fully tory described by the SN1a and BAO datasets. Thus, it indicates degenerate with 휆int, which affects the time-delay distance mea- the importance of including the external datasets which directly surement. Therefore, we compare the redshift-dependent median constrain the expansion history to get distance measurements. value of 퐷s/퐷ds, constrained directly by the SN and BAO data at the redshift of the current 7 TDCOSMO lens samples, with In Figure7, we show the distance measurements from com- the 퐷s/퐷ds in Paper IV, which used the prior based on Pantheon bining a single time-delay lens with SN1a and BAO with free Ω푘 sample with Ωm = 0.298 ± 0.022 (Scolnic et al. 2018) under the (the green contour in Figure6). This distance measurements can assumption of ΛCDM model. These results shown in Figure8 be used to infer 퐻0 in generic cosmological models. We empha- demonstrate that in the case of these 7 lenses, using the prior infor- size that this approach does not require the absolute calibration mation from Pantheon data sets which constrain the low redshift of SN1a or BAO; thus, the derived constraint on 퐻0 remains expansion history, and then using ΛCDM model to extrapolate independent of the distance ladder and the sound horizon scale. the constraint on 퐷s/퐷ds to high redshift are valid approaches

Article number, page 8 of 11 G. C.-F. Chen et al.: Distance measurements in TDC under MST

Fig. 7: The cosmological-model-independent distance measurements from combining single time-delay mock lens with Pantheon and BAO datasets. Each thin line comes from a random draw among the points in parameter space within 68% confidence level. The thick line in the middle represents the best fit.

since the deviations from the median value of 퐷s/퐷ds are all 휆int. These distance measurements with SN1a and BAO shown below 1% demonstrating that the shape of the expansion history in Figure7 can then be used to infer 퐻0 in generic cosmological is well-described by the ΛCDM model. However, the time-delay models. It is important to stress that this approach does not re- distance measurements derived by the method developed in this quire the absolute calibration of SN1a or BAO; thus, the derived work are broadly applicable distance posteriors, which can be constraint on 퐻0 remains independent of the distance ladder and used to infer 퐻0 in various cosmological models. the sound horizon scale. Acknowledgements. GC-FC thanks James Chan, Tommaso Treu, Dominique Sluse, Matt Auger, and Elizabeth Buckley-Geer for many insightful comments and 7. Conclusions espeically Simon Birrer and Adriano Agnello for GC-FC thanks Alessandro Son- nenfeld for useful suggestions on using stellar kinematics code5 and Ken Wong for In this work, we use a mock gravitational lens system to study the the code review. GC-FC acknowledges support by NSF through grants NSF-AST- correlation between distance measurements under the mass-sheet 1906976 and NSF-AST-1836016. GC-FC acknowledges support from the Moore Foundation through grant 8548. GC-FC and CDF acknowledge support for this transformation (MST) with or without assuming a cosmological work from the National Science Foundation under Grant Nos. AST-171561 and model. We verify that although 퐷Δt is directly correlated with AST-1907396. SHS and AY thank the Max Planck Society for support through both the internal and external MST, 퐷d is not only invariant under the Max Planck Research Group for SHS. SHS and EK are supported in part by an external MST but is also insensitive to the internal MST. Thus, the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) un- der Germany’s Excellence Strategy - EXC-2094 - 390783311. The Kavli IPMU without assuming any particular cosmological model, the role of is supported by World Premier International Research Center Initiative (WPI), velocity dispersion is to obtain the angular diameter distance to MEXT, Japan. JLB is supported by the Allan C. and Dorothy H. Davis Fellow- the lens (퐷d) rather than break the internal MST (휆int). To break ship. 휆int, in addition to the velocity dispersion, we identify that con- straining 퐷s/퐷ds is the key, which is typically achieved through the assumption of a particular cosmological model, and hence References 휆int and 퐷Δt are both cosmological-model-dependent quantities. In this work, we show that cosmological-model-independent 퐷Δt Alam, S., Ata, M., Bailey, S., et al. 2017, MNRAS, 470, 2617 measurement can be achieved when one uses relative distance indicators (e.g., SN1a and BAO) to constrain 퐷s/퐷ds and hence 5 https://github.com/astrosonnen/spherical_jeans

Article number, page 9 of 11 A&A proofs: manuscript no. AO_cosmography

Fig. 8: The percent difference between the median value of 퐷s/퐷ds, directly constrained from the SN1a and BAO, and the 퐷s/퐷ds under the assumption of ΛCDM model with Ωm = 0.298 ± 0.022 (Scolnic et al. 2018), which was used in Paper IV. All 7 time-delay lenses show < 1% deviation validate the use of the Pantheon data sets which constrain the low redshift expansion history and the use of ΛCDM model which extrapolates the constraint on 퐷s/퐷ds at high redshift.

Measurement 푧eff Reference 푟s/퐷푉 0.106 Beutler et al.(2011) 퐷푉 /푟s 0.15 Ross et al.(2015) 퐷 푀 (푟s,fid/푟s) (Mpc) 0.38 Alam et al.(2017) −1 −1 퐻(푧)(푟s/푟s,fid) (km s Mpc ) 0.38 퐷 푀 (푟s,fid/푟s) (Mpc) 0.51 −1 −1 퐻(푧)(푟s/푟s,fid) (km s Mpc ) 0.51 퐷 푀 (푟s,fid/푟s) (Mpc) 0.61 −1 −1 퐻(푧)(푟s/푟s,fid) (km s Mpc ) 0.61 퐷푉 /푟s 0.44 Kazin et al.(2014) 퐷푉 /푟s 0.6 퐷푉 /푟s 0.73 퐷 푀 /푟s 0.698 Gil-Marín et al.(2020) 푐/(퐻(푧) ∗ 푟s) 0.698 퐷 푀 /푟s 1.48 Hou et al.(2020) and Neveux et al.(2020) 푐/(퐻(푧) ∗ 푟s) 1.48 퐷 푀 /푟s 2.33 du Mas des Bourboux et al.(2020) 푐/(퐻(푧) ∗ 푟s) 2.33

Table 1: Summary of the BAO measurements that are used in this work. In our fiducial cosmology, 푟s,fid = 147.78 Mpc.

Arendse, N., Wojtak, R. J., Agnello, A., et al. 2019, arXiv e-prints, Birrer, S., Treu, T., Rusu, C. E., et al. 2019, MNRAS, 484, 4726 arXiv:1909.07986 Blum, K., Castorina, E., & Simonović, M. 2020, ApJ, 892, L27 Auger, M. W., Treu, T., Bolton, A. S., et al. 2010, ApJ, 724, 511 Bolton, A. S., Burles, S., Koopmans, L. V. E., Treu, T., & Moustakas, L. A. 2006, Aylor, K., Joy, M., Knox, L., et al. 2019, ApJ, 874, 4 ApJ, 638, 703 Barkana, R. 1998, ApJ, 502, 531 Bolton, A. S., Burles, S., Schlegel, D. J., Eisenstein, D. J., & Brinkmann, J. 2004, Barnabè, M., Czoske, O., Koopmans, L. V. E., Treu, T., & Bolton, A. S. 2011, AJ, 127, 1860 MNRAS, 415, 2215 Bonvin, V., Chan, J. H. H., Millon, M., et al. 2018, A&A, 616, A183 Bernal, J. L., Breysse, P. C., & Kovetz, E. D. 2019, Phys. Rev. Lett., 123, 251301 Bonvin, V., Tewes, M., Courbin, F., et al. 2016, A&A, 585, A88 Bernal, J. L., Smith, T. L., Boddy, K. K., & Kamionkowski, M. 2020, arXiv Buckley-Geer, E. J., Lin, H., Rusu, C. E., et al. 2020, MNRAS, 498, 3241 e-prints, arXiv:2004.07263 Cappellari, M. 2016, ARA&A, 54, 597 Bernal, J. L., Verde, L., & Riess, A. G. 2016, J. Cosmology Astropart. Phys., 10, Carter, P., Beutler, F., Percival, W. J., et al. 2020, MNRAS, 494, 2076 019 Chen, G. C. F., Fassnacht, C. D., Suyu, S. H., et al. 2019, MNRAS, 2193 Beutler, F., Blake, C., Colless, M., et al. 2011, MNRAS, 416, 3017 Collett, T. E., Marshall, P. J., Auger, M. W., et al. 2013, MNRAS, 432, 679 Birrer, S., Shajib, A. J., Galan, A., et al. 2020, arXiv e-prints, arXiv:2007.02941 Cuesta, A. J., Verde, L., Riess, A., & Jimenez, R. 2015, MNRAS, 448, 3463

Article number, page 10 of 11 G. C.-F. Chen et al.: Distance measurements in TDC under MST du Mas des Bourboux, H., Rich, J., Font-Ribera, A., et al. 2020, ApJ, 901, 153 Efstathiou, G. 2020, arXiv e-prints, arXiv:2007.10716 Falco, E. E., Gorenstein, M. V., & Shapiro, I. I. 1985, ApJ, 289, L1 Fassnacht, C. D., Gal, R. R., Lubin, L. M., et al. 2006, ApJ, 642, 30 Fassnacht, C. D., Xanthopoulos, E., Koopmans, L. V. E., & Rusin, D. 2002, ApJ, 581, 823 Freedman, W. L., Madore, B. F., Hatt, D., et al. 2019, ApJ, 882, 34 Gil-Marín, H., Bautista, J. E., Paviot, R., et al. 2020, MNRAS, 498, 2492 Gorenstein, M. V., Falco, E. E., & Shapiro, I. I. 1988, ApJ, 327, 693 Greene, Z. S., Suyu, S. H., Treu, T., et al. 2013, ApJ, 768, 39 Grillo, C., Lombardi, M., & Bertin, G. 2008, A&A, 477, 397 Hinshaw, G., Larson, D., Komatsu, E., et al. 2013, ApJS, 208, 19 Hou, J., Sánchez, A. G., Ross, A. J., et al. 2020, arXiv e-prints, arXiv:2007.08998 Humphrey, P. J. & Buote, D. A. 2010, MNRAS, 403, 2143 Jee, I., Komatsu, E., & Suyu, S. H. 2015, J. Cosmology Astropart. Phys., 11, 033 Jee, I., Suyu, S. H., Komatsu, E., et al. 2019, Science, 365, 1134 Joudaki, S., Kaplinghat, M., Keeley, R., & Kirkby, D. 2018, Phys. Rev. D, 97, 123501 Kazin, E. A., Koda, J., Blake, C., et al. 2014, MNRAS, 441, 3524 Komatsu, E., Smith, K. M., Dunkley, J., et al. 2011, ApJS, 192, 18 Koopmans, L. V. E., Bolton, A., Treu, T., et al. 2009, ApJ, 703, L51 Koopmans, L. V. E., Treu, T., Bolton, A. S., Burles, S., & Moustakas, L. A. 2006, ApJ, 649, 599 Liao, K., Shafieloo, A., Keeley, R. E., & Linder, E. V. 2020, ApJ, 895, L29 Macaulay, E., Nichol, R. C., Bacon, D., et al. 2019, MNRAS, 966 Merritt, D. 1985, AJ, 90, 1027 Millon, M., Galan, A., Courbin, F., et al. 2020, A&A, 639, A101 Navarro, J. F., Frenk, C. S., & White, S. D. M. 1996, ApJ, 462, 563 Neveux, R., Burtin, E., de Mattia, A., et al. 2020, MNRAS, 499, 210 Osipkov, L. P. 1979, Pis ma Astronomicheskii Zhurnal, 5, 77 Planck Collaboration, Aghanim, N., Akrami, Y., et al. 2018, arXiv e-prints, arXiv:1807.06209 Poulin, V., Boddy, K. K., Bird, S., & Kamionkowski, M. 2018, Phys. Rev. D, 97, 123504 Refsdal, S. 1964, MNRAS, 128, 307 Riess, A. G., Casertano, S., Yuan, W., Macri, L. M., & Scolnic, D. 2019, arXiv e-prints, arXiv:1903.07603 Ross, A. J., Samushia, L., Howlett, C., et al. 2015, MNRAS, 449, 835 Rusu, C. E., Fassnacht, C. D., Sluse, D., et al. 2017, MNRAS, 467, 4220 Rusu, C. E., Wong, K. C., Bonvin, V., et al. 2019, MNRAS, 498, 1440 Schneider, P. & Sluse, D. 2013, A&A, 559, A37 Scolnic, D. M., Jones, D. O., Rest, A., et al. 2018, ApJ, 859, 101 Shajib, A. J., Birrer, S., Treu, T., et al. 2020a, MNRAS, 494, 6072 Shajib, A. J., Treu, T., Birrer, S., & Sonnenfeld, A. 2020b, arXiv e-prints, arXiv:2008.11724 Sonnenfeld, A., Treu, T., Gavazzi, R., et al. 2013, ApJ, 777, 98 Suyu, S. H., Auger, M. W., Hilbert, S., et al. 2013, ApJ, 766, 70 Suyu, S. H., Chang, T.-C., Courbin, F., & Okumura, T. 2018, Space Sci. Rev., 214, 91 Suyu, S. H., Marshall, P. J., Auger, M. W., et al. 2010, ApJ, 711, 201 Suyu, S. H., Marshall, P. J., Blandford, R. D., et al. 2009, ApJ, 691, 277 Suyu, S. H., Treu, T., Hilbert, S., et al. 2014, ApJ, 788, L35 Taubenberger, S., Suyu, S. H., Komatsu, E., et al. 2019, A&A, 628, L7 Tihhonova, O., Courbin, F., Harvey, D., et al. 2018, MNRAS, 477, 5657 Treu, T. & Marshall, P. J. 2016, A&A Rev., 24, 11 Verde, L., Bernal, J. L., Heavens, A. F., & Jimenez, R. 2017, MNRAS, 467, 731 Verde, L., Treu, T., & Riess, A. G. 2019, Nature Astronomy, 3, 891 Wojtak, R. & Agnello, A. 2019, MNRAS, 486, 5046 Wong, K. C., Suyu, S. H., Chen, G. C.-F., et al. 2019, Monthly Notices of the Royal Astronomical Society, 498, 1420 Xu, D., Sluse, D., Schneider, P., et al. 2016, MNRAS, 456, 739

Article number, page 11 of 11