arXiv:1612.04095v2 [gr-qc] 12 Mar 2017 savne IOVROadtetidgnrto detectors (such generation detectors third the based and ground LIGO/VIRGO the advanced by as detected routinely be to equiva 8]. Einsteins using [7, the [6] test as principle GW well lence of as measurement, detection dispersion direct per- the first been already the has following test formed Inde wavelength therein). Compton references and the massiv [5] review of break- existence the the the (see or through principle light equivalence of weak GW speed of of ing the speed from the For different , be techniques. of could other theories to alternative unaccessible in way example, a as- in various Genera (GR), test of validity ativity to the able like physics, be gravitational also of pects would so one operatin these data detectors of gathering GW and With existence of systems. the merger binary proving speculative the far holes, from black came massive laboratory two first in the detected Moreover, Universe. signal the which GW on achievement window great new a a the up was opened by [4] (GW150914) detector waves LIGO advanced gravitational first of the detection of announcement direct Recent [3]. Weisberg follow-up and subsequent Taylor its by PSR and [2] system Taylor pulsar and binary Hulse by the 1913+16 of wa discovery the gravitational after of made o existence was first the The for [1]. speed mass in evidence the Einstein servational the Albert at by traveling of predicted and were variations light, source of time the of by moment generated quadrupole strain, spatial of iaynurnsas(SN)aeoeo rmsn sources promising of one are (NS-NS) stars neutron Binary waves transverse the are which (GWs), waves Gravitational h pe fgaiainlwvsfo toglne gravita lensed strong from waves gravitational of speed The iLn Fan Xi-Long eosre ahya.Tepwro hsmto smil limi mainly is around method is this kilonova of for power which The gravitational-wav year. each third-generation requ observed of It be Gener era as the acts. similar in emission lensing possible EM gravitational predicts and and GW metric between G delays and time ( details diff internal modeling the lens s against measuring any robust it (i.e. on makes setting rely differential not Its does evengravity. method GW-EM this lensed The Conceptually counterpar counterparts. EM EM counterpart. and and GWs GWs in of seen times images arrival electromagneti comparing their by and way: GWs fold lensed strongly on based (GWs) ASnmes 53.f 95.85.Sz 95.30.Sf, numbers: PACS rad fast and burst gamma-ray short include wit transients associated short shorter-duration much of transients lensed epooeanwmdlidpnetmaueetsrtg fo strategy measurement model-independent new a propose We INTRODUCTION 1 , 2 , ∗ 5 1 a Liao Kai 3 eateto srnm,BiigNra nvriy 10087 University, Normal Beijing Astronomy, of Department colo hsc n ehooy ua nvriy ua 4 Wuhan University, Wuhan Technology, and Physics of School colo cec,WhnUiest fTcnlg,Whn43 Wuhan Technology, of University Wuhan Science, of School nvriyo iei,Uiestca4 007Ktwc,P Katowice, 40-007 4, Uniwersytecka Silesia, of University 2 4 eateto hsc n ehncl&Eetia Engineer Electrical & Mechanical and Physics of Department eateto srpyisadCsooy nttt fPhy of Institute Cosmology, and Astrophysics of Department 3 ae Biesiada Marek , ue nvriyo dcto,Whn400,China 430205, Wuhan Education, of University Hubei 10 4 e.Ti netit a esprse yafco of factor a by suppressed be can uncertainty This sec. 4 Rel- l , 5 lkadaPi´orkowska-Kurpas Aleksandra , ves ed, b- g e - rnebtentm easi WadE oan)- domains) EM and GW in delays time between erence h Weetcnb dnie.Cniae o such for Candidates identified. be can event GW the h eetr,we bu 0lne Weet would events GW lensed 10 about when detectors, e ec h iigprecision timing the reach hri)wt ekrbn magnitude r-band peak referenc with by and therein) [10] accompanied review the be visib (see kilonovae/mergernovae be to as could which expected counterparts (EM) are electromagnetic more they the even that them makes is What interesting [9]. Telescope Einstein the like ihtepeito htEnti eecp hudb bet able be should Telescope Einstein that prediction [16–18] the in with studied been has BH-BH) fo NS-BH, (NS-NS, compact Perspectives objects double merging from magnitude. GWs of lensed strongly orders observing three Universe by the increase of volume will probed Ad that the means over This sensitivity LIGO. in vanced magnitude of order an improve will paper. this in proposed method the wil future of the power in the counterpart enhance EM the of accuracy bot timing improveme the from Any on astrophysics view. of current point observational in and topics theoretical research strat top identification the and signals are rates GW properties, of their counterparts on EM focused the of studies Detailed [15]). the be of also order These (of could [14]. transients NS-BH they shorter-duration and much but NS-NS from known, GW not of counterparts [13 is EM FRB attention of considerable origin attracted ra- The fast have the (FRB) Recently, bursts SGRB. observable dio an give to as aligned eprso Wfo SN n SB eg 2.Because 12]. [e.g. collimation NS-BH jet and the NS-NS of from GW of terparts SRs,wihhv ipeadsaptmoa etrs( features temporal sharp of order and of simple bursts have gamma-ray transie which Short days) (SGRBs), (SNe). of supernovae order to (of similar shorter-duration events are They [11]). lRltvt.W xetta uhts ilbecome will test such that expect We Relativity. al obursts. io stae ntesm esn oeta)adagainst and potential) lensing same the in travel Ws etgnrto Wdtcoslk h isenTelescope Einstein the like detectors GW generation Next E)cutrat.Ti a edn natwo- a in done be can This counterparts. (EM) c eicter fmsiegaioso modified or gravitons massive of theory pecific e ytmn cuayo h Mcounterpart, EM the of accuracy timing by ted h rpgto pe fgaiainlwaves gravitational of speed propagation the r sadb oprn h iedly between delays time the comparing by and ts rs oee,ta h hoyo rvt is gravity of theory the that however, ires, sprastebs a oietf nEM an identify to way best the perhaps is t 0 . inlwv-lcrmgei signals wave-electromagnetic tional 1 − 1 ,Biig China Beijing, 5, s ln and oland ,aeaohrvr rmsn Mcoun- EM promising very another are ), 07,China 30072, 00 China 0070, ∼ sics, 4 10% ing, n ogHn Zhu Zong-Hong and , ∼ ∼ 0 fteN/Ssseswl be will systems NS/NS the of . 01 10 10 ms fstrongly if , ∼ 22 eg R 368in 130628 FRB (e.g. − 25 ms BMg(e.g. Mag AB 1 losu to us allows ) , 5 † egies es nt nt le ]. o h r - l 2 detect several tens up to more than hundred of such events per certainty of the time of arrival method. Concerning speed of year. This statistics is however dominated by BH-BH systems. GW, it has been proposedin [34] that using the phase informa- Although a pure BH-BH merger is not expected to has an EM tion of the GWs from inspiralling compact binary estimated counterpart, yet several papers [e.g. 7, 19, 20], motivated by by matched filtering technique, a bound on the graviton mass the plausible GBM transient associated with GW150914 [21], (hence on the speed of GW) could by made using GW alone. have proposed the formation channels of EM counterpart in However, the expected bounds depend strongly on other phys- BH-BH merger system and discussed their applications. ical effects relevant for the particular inspiralling system de- With new generation of dedicated surveys (e.g. SLACS, tected such as spin-induced precessions, orbital eccentricity, CASSOWARY, BELLS, SL2S), strong gravitational lensing higher waveform harmonics, the merger-ringdown phase, etc. has developed into a serious technique in extragalactic astron- Here, we propose a method to directly constrain the speed omy (galactic structure studies) and in cosmology. In this of GW by using the strong lensing time-delays measured with phenomenon, a source (typically a or a distant galaxy) GW and their EM counterparts. The differential setting of lensed by a foreground massive galaxy or cluster appears in our method makes it free from the intrinsic time delays in the multiple images. Light rays of these images travel along paths source (i.e. different emission times of GW and EM signal). differing in length and probe of the lens General idea to use gravitationally lensed signals registered in at different depth experiencing different gravitational time de- GW and EM windows has independently been proposed by lays. These two effects: geometrical and the Shapiro effect [35]. Our formulation is slightly different from theirs and is combine to produce the time delay between images [22]. If supported with more rigorous calculations. Also worth not- the source is intrinsically variable (and most are), the ing is the paper by Takahashi [36] which claims that even light curves of its images can be used to extract out the time within it is possible for a lensed GW signal delay [23]. This technique requires high-quality monitoring to come earlier than EM one (emitted simultaneously) due to with sufficient cadence, season and campaign lengths so that wave effects in gravitational lensing (breakdown of geometric the microlensing effects caused by the stars can be eliminated. optics approximation). This result does not apply in our case Moreover, quite recently the first detection of the gravitation- where we consider galaxies acting as lenses. ally lensed supernova have been reported [24] and the original dream of Sjur Resfdal came true [25]. In the case of lensed transient sources, like SNe, the measurements of time-delays METHOD between images can be much more accurate. Besides its typi- cal use to determine the Hubble constant [26], measurements Our method is an extension of the idea proposed by [37] in of strong lensing time-delays has also been used to constrain the context of testing the Lorentz Invariance Violation by us- the amplitude of the background [27]. The ing dependence of time delays in gravitationally lensed forthcoming Large Synoptic Survey Telescope (LSST) will systems. Let us assume that we observed a strongly lensed find about 8000 lensed quasars, 3000 of which will be GW signal and identified its electromagnetic counterpart in monitored and∼ have the well-measured∼ time delays in six fre- the optical, radio waves or in gamma-rays. Then we would quency bands within ten years. The estimated number of ro- be able to measure time delays between the images indepen- bust time-delay measurements for these is around 400, each dently in GW detectors – ∆tGW and in the electromagnetic with precision < 3% and accuracy 1% [28]. The LSST should window – ∆tγ. They will be different if the speed of gravity also find some 130 lensed supernovae during its survey du- vGW is different from c. The difference (∆tγ ∆tGW ) will − ration, while the deep, space-based supernova survey done bear information about the speed of GW. The bound on the by Joint Dark Energy Mission (JDEM) is expected to find vGW will have the following general form valid for a broad 15 lensed SN [29]. Note that, similar to SNe, the pro- set of analytical lens models: ∼ posed isotropic counterparts to NS-NS mergers (e.g., kilono- 2 vGW δT vae) have signals a few days of duration with a limited timing 1 , (1) − c ≤ ∆t Flens(z ,z ) accuracy.   γ l s As already mentioned, one of the most important issues to where δT is timing accuracy with which time delays are de- be studied with GW detectors is testing the validity of GR, termined and Flens(zl,zs) O(1) is some factor (weakly) in particular the question whether GW travels with the speed dependent on the lens model∼ and background cosmology (see of light. Similar questions arise within the Lorentz Invariance below). Let us stress that our method is purely empirical Violating (LIV) theories where the dispersion relation for the one: we do not assume any (non-existing so far as a consis- could be modified making the energy tent theory) model of massive gravitons. We just refer to dif- dependent. The observed time of arrival delay between two ferences in time delays in GW and EM windows, assuming, events, such as emission of different energy photon-photon however, that the theory of gravity is metric and predicts grav- [30], photon-neutrino[31], and GW-EM signals [32, 33], have itational lensing similar as General Relativity. It means we re- been proposed to constraint the respective propagation speed. fer to purely classical, rather than quantum regime. However, The unknown intrinsic time delay in the emission time of such in order to be more specific in calculations we will assume two signals to be compared contributes considerablyto the un- below that gravitons are massive and travel along time-like 3 geodesics. In this sense the term “graviton” should be per- and with the notation: ceived as a useful jargon rather than reference to the quantum z2 dz′ nature of GWs. There is a wide diversity of possible alter- 1 2 In(z ,z ) := ′ n ′ , (12) native theories of gravity not all of which will be well con- Zz1 (1 + z ) h(z ) strained by the method we propose. For some more recent one has: reviews see, e.g. [38] or [39] where the constraining power of observed GW signals has been demonstrated and discussed. 2 4 1 c mGW c 2 Propagation of massive gravitons on the cosmological ∆rGW = 2 (1 + z) I2(0,z). (13) 2 H0 E background. The hypothesis that the speed vGW of GW could be differ- The above formulae should be understood in the following ent from c means that gravitons should be treated as massive way: if the emission time te and detection time t0 are fixed, particles (having the rest mass mGW ) moving along timelike i.e. the same for the GW and electromagneticsources then the geodesics. Therefore their dispersion relation would be GW source is by ∆rGW closer than electromagnetic source. 2 2 2 2 4 On the other hand if they are emitting from the same location E p c = m c (2) GW − GW GW the GW signal would come by ∆tGW = ∆rGW /c later than instead of electromagnetic counterpart. In other words, travel time for GW would be by ∆tGW longer, as if the source was located 2 2 2 Eγ pγ c =0. (3) by ∆rGW farther. − Strong lensing time delays For the purpose of illustrating as for the photons. Let us moreover assume that GW travel our ideas we shall restrict our attention to the singular isother- along radial geodesics in the flat Friedman-Robertson-Walker mal sphere (SIS) model which has been proved to be a useful (FRW) model with the metric and reliable phenomenological model of early type galaxies ds2 = c2dt2 a(t)2 dr2 + r2dθ2 + r2sin2θdφ2 . (4) which dominate the population of lenses. The generalization − to SIE (singular isothermal ellipsoids) and general power-law   Generalization to non-flat FRW would be straightforward. spherically symmetric mass distribution is rather straightfor- Covariant and contravariant radial components of GW four ward and would not change our conclusions. are related as: The Einstein ring radius for the SIS model is:

2 r 2 pr = a p (5) Dls σv ϑE =4π 2 (14) and obviously Ds c

r 2 2 where σv denotes one-dimensional velocity dispersion of stars dr p c prc = = . (6) in lensing galaxy. If the lensing is strong i.e. the misalignment dt E a2E angle β between the directions to the lens and to the source Then it is easy to see that velocity of gravitons is is β < ϑE then two co-linear images A and B form on the 2 2 2 opposite side of the lens, at radial distances ϑA = β + ϑE and dr c 1 mGW c a vGW = = 1 2 . (7) ϑB = ϑE β having time delays between the images: dt a  − 2 pr  − If the GW signal was emitted at the moment and detected 1+ zl DlDs 2 2 te ∆tSIS = (ϑA ϑB) (15) (observed) at t0, then the travel distance of GW is: 2c Dls − which according to the equations Eq.( 9) and Eq.( 14) can also rGW = rγ ∆rGW , (8) − be written as where:

t0 z 2 4 c dz 32π σ r(zl)r(zl,zs) rγ = dt = c , (9) ∆tSIS = y , (16) Zte a(t) Z0 H(z) H0 c r(zs)   e e is the usual comoving distance to the GW source, and where r˜(zl) denotes the dimensionlesse (i.e. with c/H0 fac- 2 3 t0 tored out) comoving distance to the lens and y = β/ϑE. 1 mGW c ∆rGW = 2 a(t)dt. (10) In the context of massive photons it was shown by [40] 2 p Z 2 4 r te m c that the bending angle is modified by a factor 1 + 2E2 . Besides the expansion rate H(z) we will also use its dimen- These considerations are valid in our case, so it means that sion less form h(z) defined as H(z) = H0h(z) where H0 is impact parameters of photons and GW from the same im- the Hubble constant. Using age are different and the Einstein angle gets modified to: 2 4 mGW c E ϑE,GW = ϑE(1 + 2E2 ). Therefore while calculating the p = a(t ) , (11) r e c time delay between images seen in GW, one has to consider 4 this effect, which affects together with ge- like kilonova or better yet, SGRB associated with GW signal. ometrical terms using corrections Eq.(10) in the distances Dls Then, according to the Eq.(10), the expected time delay (in and Dl. each image) would be: Now, we can see that the difference between image time delays observed in GW detectors and in the electromagnetic 1 2 domain is ∆tγ,GW = (1 + zs) I2(0,zs) (21) 2H0 2 4 mGW c lens For the source at the redshift zs = 2 one has the following ∆tSIS,GW ∆tSIS,γ = ∆tSIS,γ 2 F (zl,zs) (17) 2 − E bound: 1 vGW 9.92 10−22 − c ≤ × where: It would be appropriate to compare the above bounds with the results published in [4] and [6] concerning the constraints (1 + zs)I2(0,zs) Flens(zl,zs)=1+ on violations of general relativity leading to massive gravi- 2r(zl,zs) tons. The bound obtained from the GW150914 event was for- (1 + zl)I2(0,zl) (1 + zl)I2(0,zl) mulated in terms of graviton Compton wavelength λGW > e (18) 13 − 2r(zl) − r(zl, rs) 10 km which turned out to be the strongest dynamical bound probing the propagation of gravitational interactions. Therefore we could specifye the speed ofe GW through Eq. 1 Translating this into a bound on the speed of gravity, one ob- using the information from the lensed GW-EM system in term 2 19 tains: 1 vGW < 10− . This constraint is much stronger of a “graviton” : − c than one can get from differences in time delays. Let us re- 2 4 2 mind, however, that the aforementioned bound was obtained m c vGW GW =1 , (19) as a result of sophisticated analysis using waveform models E2 − c   that allow for parameterized general-relativity violations dur- If one would be able to measure such a difference in time de- ing the inspiral and merger phases and using the dispersion lays this would also be a proof that gravitons are massive (i.e. measurement. On the other hand, the method dis- that GR needs to be modified). cussed by us - using the GW vs. EM arrival times in lensed The accuracy δT of time delay measurements sets con- images - is by three ordersof magnitude more restrictive. This straints on the vGW . Assuming the galaxy-galaxy strong lens- means that even a single instant of gravitationally lensed GW ing system with zl = 1 and zs = 2 one has the following signal accompanied by EM transient counterpart would be bound coming from the GW/EM difference in lensing time valuable. delays

4 v 2 δT σ − y −1 PERSPECTIVES 1 GW 4.26 10−10 − c  ≤ × 1 ms250 km/s 0.1 (20) One can expect that the next decades of observations car- where we also assumed ΛCDM cosmology with the Hubble ried together in the GW and EM windows will be sufficient to −1 −1 constant H0 = 68 km s Mpc , Ωm = 0.3. Numerical give a strong constraint on the GW speed and graviton mass. factor setting the scale corresponds to lens velocity dispersion Concerning our method, its main limitation is the accuracy σ = 250 km/s, timing accuracy of δT = 1 ms and source of EM time delay, while timing in the GW detectors is very - lens misalignment y = 0.1 Recent discovery of “Refsdal precise (< 10−4 ms). supernova” [24] and especially its reappearance [41] demon- The planed third generation gravitational wave detector, strated that we are starting discover transient events lensed by such as the Einstein Telescope, could observe the strongly a cluster. The cluster-scale images have much bigger time de- lensed GW. The rate of yearly detections of strongly lensed lays than in the case of galaxy scale lenses. This means that GW from NS-NS and NS-BH sources are in the range 2 in such case our method of differences in time delays would to 10/yr [16, 17], depending on different ET configurations,∼ be more restrictive. For example, taking the value of time de- stellar population synthesis models [42]. Cadenced wide-field lay for the Refsdal supernova image SX which reappeared as EM imaging surveys in the next decade will increase the cat- 2 predicted one would get a bound 1 vGW 3.2 10−11 alog of strongly lensed systems by two orders of magnitude. − c ≤ × assuming 1ms timing accuracy.  Besides, one can imagine a dedicated follow-up project based Apart from the limitation due to accuracy with which EM on the observed GW events. Short-duration EM counterparts lensing time delay can be measured, the method described transients (such as kilonova) have a strong and pronounced above is less restrictive than travel time techniques because feature on the light curve ( i.e. the maximum point), which it cannot take advantage of cumulative effect along the whole creates a unique opportunity for time-delay extracting algo- path. However, the strong lensing system seen both in EM rithms, resulting in a accurate estimate. For these objects, and GW offers additional possibility to compare the moments we expect to get the time delay precision 104 s through of arrival of the same image seen in the EM and GW respec- the dedicated photometry related to maximum∼ point. Much tively. This would be possible only for transient EM sources shorter-duration EM counterparts, like SGRBs, FRBs or any 5 new signal discovered in the future, will be measured with China and gained special fund support of foreign knowledge much better time delay accuracy. The constraint on GW speed introducing project. Z. Z. was supported by the NSFC No. can be enhanced by increasing the number N of lensed sys- 11633001. This research was also partly supported by the tems observed in GW and EM windows. In such a case, the Poland-China Scientific & Technological Cooperation Com- statistical uncertainty would be reduced by a factor of √N. mittee Project No. 35-4. Such a population of much shorter-duration EM counterparts, such as 10 FRBs with 0.01ms time delay accuracy, could suppress∼ the uncertainty of time delays by a factor of 1010 ∼ comparing with measured by a kilonova. The issue of ex- ∗ pected rates of joint EM/GW strongly lensed events is inter- [email protected] † [email protected] esting on its own and merits further studies. However, even a [1] A. Einstein, Sitzungsberichte der K¨oniglich Preußischen single such event – discovered either serendipitously or as a Akademie der Wissenschaften (Berlin), Seite 688-696. (1916). result of dedicated surveys – would be very important. [2] R. A. Hulse and J. H. Taylor, Astroph. J. Lett. 195, L51 (1975). Our approach has a number of advantages. First is its dif- [3] J. H. Taylor and J. M. Weisberg, Astrophys. J. 253, 908 (1982). ferential setting making it robust, as already mentioned. How- [4] B. P. Abbott, R. Abbott, T. D. Abbott, M. R. Abernathy, F. Ac- ever, the price paid for this is that it is much less restrictive. ernese, K. Ackley, C. Adams, T. Adams, P. Addesso, R. X. Ad- Second, the lensed EM/GW event is perhaps the best way to hikari, and et al., Physical Review Letters 116, 061102 (2016), arXiv:1602.03837 [gr-qc]. identify an EM counterpart: even with poor resolution of GW [5] C. M. Will, Living Reviews in Relativity 17 (2014), 10.12942/lrr-2014-4, detectors if we see a lensed GW (two strains of similar tem- arXiv:1403.7377 [gr-qc]. poral structure) coincident with lensed EM source we can be [6] The LIGO Scientific Collaboration and the Virgo Collaboration, almost sure about the source location. Extra bonus, then is ArXiv e-prints (2016), arXiv:1602.03841 [gr-qc]. that besides differential time delays we would be able to mea- [7] X.-F. Wu, H. Gao, J.-J. Wei, P. M´esz´aros, B. Zhang, Z.-G. Dai, sure time of flight differences GW vs. EM in each image. If S.-N. Zhang, and Z.-H. Zhu, Phys. Rev. D 94, 024061 (2016), there are more than two images - e.g quads which are typi- arXiv:1602.01566 [astro-ph.HE]. [8] E. O. Kahya and S. Desai, Physics Letters B 756, 265 (2016), cal in strong lensing systems discovered so far, we could have arXiv:1602.04779 [gr-qc]. several measurements from a single lensed source. One has to [9] B. Sathyaprakash, M. Abernathy, F. Acernese, remark, however, that as discussed in [43] the peak amplitude P. Ajith, B. Allen, P. Amaro-Seoane, N. Anders- of GW emission associated with the time of the merger could son, S. Aoudia, K. Arun, P. Astone, and et al., be registered long before the EM prompt SGRB signal. The Classical and 29, 124013 (2012), two signals would be separated by the lifetime of the supra- arXiv:1206.0331 [gr-qc]. massive NS, which can easily exceed 103 s. The intrinsic time [10] R. Fern´andez and B. D. Metzger, ArXiv e-prints (2015), arXiv:1512.05435 [astro-ph.HE]. delay between EM and GW signal is very hard to disentangle [11] P. S. Cowperthwaite and E. Berger, from the possible time delay due to hypothetical difference Astrophys. J. 814, 25 (2015), arXiv:1503.07869 [astro-ph.IM]. between the speed of light and speed of gravity. Therefore it [12] D. B. Fox, D. A. Frail, P. A. Price, S. R. Kulkarni, E. Berger, is a serious obstacle to the method of EM/GW time of flight T. Piran, A. M. Soderberg, S. B. Cenko, P. B. Cameron, A. Gal- differences in each image. Yam, M. M. Kasliwal, D.-S. , F. A. Harrison, E. Nakar, We can conclude that according to the anticipated devel- B. P. Schmidt, B. Penprase, R. A. Chevalier, P. Kumar, K. Roth, opment of GW astrophysics, massive EM surveys and the D. Watson, B. L. Lee, S. Shectman, M. M. Phillips, M. Roth, P. J. McCarthy, M. Rauch, L. Cowie, B. A. Peterson, J. Rich, synergy between them will create possibility to use strongly N. Kawai, K. Aoki, G. Kosugi, T. Totani, H.-S. Park, A. Mac- lensed GW-EM events as complementary tests of fundamen- Fadyen, and K. C. Hurley, Nature (London) 437, 845 (2005), tal physics and astrophysics.When this Letter was under re- astro-ph/0510110. view we became aware of the paper [44], in which the authors [13] D. Thornton, B. Stappers, M. Bailes, B. Barsdell, S. Bates, independently discussed the idea and applications of multi- N. D. R. Bhat, M. Burgay, S. Burke-Spolaor, D. J. messenger time delays from lensed GWs. Champion, P. Coster, N. D’Amico, A. Jameson, S. John- ston, M. Keith, M. Kramer, L. Levin, S. Milia, C. Ng, A. Possenti, and W. van Straten, Science 341, 53 (2013), arXiv:1307.1628 [astro-ph.HE]. ACKNOWLEDGMENTS [14] T. Totani, Publications of the Astronomical Society of Japan 65 (2013), 10.1093/pasj/65.5.L12, arXiv:1307.4985 [astro-ph.HE]. The authors would like to thank the referees for their valu- [15] D. J. Champion, E. Petroff, M. Kramer, M. J. Keith, able comments which allowed to improve considerably the M. Bailes, E. D. Barr, S. D. Bates, N. D. R. Bhat, original text. X. F. thanks M. Sereno and Y. Chen for valu- M. Burgay, S. Burke-Spolaor, C. M. L. Flynn, A. Jame- son, S. Johnston, C. Ng, L. Levin, A. Possenti, B. W. Stappers, able comments. X. F. was supported by the National Natural W. van Straten, D. Thornton, C. Tiburzi, and A. G. Lyne, Science Foundation of China (No. 11303009 and 11673008) Mon. Not. R. Astron. Soc. (2016), 10.1093/mnrasl/slw069, and Newton International Fellowship Alumni follow on fund- arXiv:1511.07746 [astro-ph.HE]. ing. K. L. was supported by the NSFC No. 11603015. M. [16] M. Biesiada, X. Ding, A. Pi´orkowska, and Z.-H. Zhu, B. obtained approval of foreign talent introducing project in Journal of Cosmology and Astroparticle Physics 10, 080 (2014), 6

arXiv:1409.8360 [astro-ph.HE]. der, K. Mandel, X.-L. Meng, G. Meylan, L. A. Moustakas, [17] X. Ding, M. Biesiada, and Z.-H. Zhu, T. P. Prabhu, A. Romero-Wolf, A. Shafieloo, A. Siemigi- Journal of Cosmology and Astroparticle Physics 12, 006 (2015), nowska, C. S. Stalin, H. Tak, M. Tewes, and D. van Dyk, arXiv:1508.05000 [astro-ph.HE]. Astrophys. J. 800, 11 (2015), arXiv:1409.1254 [astro-ph.IM]. [18] R. Perna, D. Lazzati, and B. Gia- [29] M. Oguri and P. J. Marshall, comazzo, Astroph. J. Lett. 821, L18 (2016), Mon. Not. R. Astron. Soc. 405, 2579 (2010), arXiv:1001.2037. arXiv:1602.05140 [astro-ph.HE]. [30] X.-F. Wu, H. Gao, J.-J. Wei, X.-L. Fan, P. M´esz´aros, B. Zhang, [19] B. Zhang, Astroph. J. Lett. 827, L31 (2016), Z.-G. Dai, S.-N. Zhang, and Z.-H. Zhu, ArXiv e-prints (2016), arXiv:1602.04542 [astro-ph.HE]. arXiv:1602.01566 [astro-ph.HE]. [20] A. Loeb, Astroph. J. Lett. 819, L21 (2016), [31] M. J. Longo, Phys. Rev. D 36, 3276 (1987). arXiv:1602.04735 [astro-ph.HE]. [32] A. Cooray and N. Seto, Phys. Rev. D 69, 103502 (2004), [21] V. Connaughton, E. Burns, A. Goldstein, L. Blackburn, astro-ph/0311054. M. S. Briggs, B.-B. Zhang, J. Camp, N. Christensen, C. M. [33] C. Cutler, W. A. Hiscock, and S. L. Larson, Hui, P. Jenke, T. Littenberg, J. E. McEnery, J. Racusin, Phys. Rev. D 67, 024015 (2003), gr-qc/0209101. P. Shawhan, L. Singer, J. Veitch, C. A. Wilson-Hodge, [34] C. M. Will, Phys. Rev. D 57, 2061 (1998), gr-qc/9709011. P. N. Bhat, E. Bissaldi, W. Cleveland, G. Fitzpatrick, [35] T. E. Collett and D. Bacon, Phys. Rev. Lett. 118, 091101 (2017), M. M. Giles, M. H. Gibby, A. von Kienlin, R. M. Kip- arXiv:1602.05882 [astro-ph.HE]. pen, S. McBreen, B. Mailyan, C. A. Meegan, W. S. Pa- [36] R. Takahashi, ArXiv e-prints (2016), arXiv:1606.00458. ciesas, R. D. Preece, O. J. Roberts, L. Sparke, M. Stan- [37] M. Biesiada and A. Pi´orkowska, bro, K. Toelge, and P. Veres, Astroph. J. Lett. 826, L6 (2016), Mon. Not. R. Astron. Soc. 396, 946 (2009), arXiv:0712.0941. arXiv:1602.03920 [astro-ph.HE]. [38] E. Berti, V. Cardoso, L. C. B. Crispino, [22] P. Schneider, J. Ehlers, and E. E. Falco, L. Gualtieri, C. Herdeiro, and U. Sperhake, Gravitational Lenses, XIV, 560 pp. 112 figs.. Springer-Verlag Berlin HeidelbergInternational New York. Journal Also of Astronomy Modern Physics and Astrophys D 25, 1641022ics Library (2016), (1992) p. 112. http://www.worldscientific.com/doi/pdf/10.1142/S0218271816410224. [23] M. Tewes, F. Courbin, and G. Mey- [39] N. Yunes, K. Yagi, and F. Pretorius, ArXiv e-prints (2016), lan, Astron. Astroph. 553, A120 (2013), arXiv:1603.08955 [gr-qc]. arXiv:1208.5598 [astro-ph.CO]. [40] D. D. Lowenthal, Phys. Rev. D 8, 2349 (1973). [24] P. L. Kelly, S. A. Rodney, T. Treu, R. J. Foley, G. Bram- [41] T. Treu, G. Brammer, J. M. Diego, C. Grillo, P. L. Kelly, mer, K. B. Schmidt, A. Zitrin, A. Sonnenfeld, L.-G. Strol- M. Oguri, S. A. Rodney, P. Rosati, K. Sharon, A. Zitrin, ger, O. Graur, A. V. Filippenko, S. W. Jha, A. G. Riess, I. Balestra, M. Bradaˇc, T. Broadhurst, G. B. Caminha, M. Bradac, B. J. Weiner, D. Scolnic, M. A. Malkan, A. von A. Halkola, A. Hoag, M. Ishigaki, T. L. Johnson, W. Karman, der Linden, M. Trenti, J. Hjorth, R. Gavazzi, A. Fontana, R. Kawamata, A. Mercurio, K. B. Schmidt, L.-G. Strolger, S. H. J. C. Merten, C. McCully, T. Jones, M. Postman, A. Dressler, Suyu, A. V. Filippenko, R. J. Foley, S. W. Jha, and B. Patel, B. Patel, S. B. Cenko, M. L. Graham, and B. E. Tucker, Astrophys. J. 817, 60 (2016), arXiv:1510.05750. Science 347, 1123 (2015), arXiv:1411.6009. [42] M. Dominik, K. Belczynski, C. Fryer, D. E. Holz, [25] S. Refsdal, Mon. Not. R. Astron. Soc. 128, 307 (1964). E. Berti, T. Bulik, I. Mandel, and R. O’Shaughnessy, [26] S. H. Suyu, P. J. Marshall, M. W. Auger, S. Hilbert, R. D. Astrophys. J. 779, 72 (2013), arXiv:1308.1546 [astro-ph.HE]. Blandford, L. V. E. Koopmans, C. D. Fassnacht, and T. Treu, [43] R. Ciolfi and D. M. Siegel, Astroph. J. Lett. 798, L36 (2015), Astrophys. J. 711, 201 (2010), arXiv:0910.2773 [astro-ph.CO]. arXiv:1411.2015 [astro-ph.HE]. [27] B. Allen, Physical Review Letters 63, 2017 (1989). [44] T. Baker and M. Trodden, ArXiv e-prints (2016), [28] K. Liao, T. Treu, P. Marshall, C. D. Fassnacht, N. Rum- arXiv:1612.02004. baugh, G. Dobler, A. Aghamousa, V. Bonvin, F. Courbin, A. Hojjati, N. Jackson, V. Kashyap, S. Rathna Kumar, E. Lin-