The Astrophysics of the Binary Neutron Star Merger GW170817/GRB170817A

Home , Glitch (astronomy), GW170104, GW170817, PyCBC

Black and Gold The Astrophysics of the Binary Neutron Star Merger GW170817/GRB170817A

Leo P. Singer, Research Astrophysicist NASA Goddard Space Flight Center

1st GammaSIG Teleconference Outline

1. LIGO/Virgo

2. Discovery, Observing Strategies

3. Optical and Near Infrared Observations

4. Interpretation, Radio

5. A Few Open Questions 1. LIGO / Virgo VIRGO CASCINA, ITALY Upgrades completed and Advanced Virgo ADVANCED VIRGO became operational in August 2017. Main commences observations enhancements over initial Virgo: • Increased ﬁnesse of arm cavities

20 Test masses heavier, lower absorption, 10 Virgo • higher surface quality Hanford Livingston • Size of beam doubled (vacuum system and 21 input/output optics modiﬁed accordingly) 10 Hz) p • More robust control of ﬁnal pendulum stage

22 Not in this talk: LIGO/Virgo sensitivity 10 improvements in software developed and

Strain (1/ deployed in O2

23 10 • Feed-forward noise subtraction • Glitch removal by masking or subtraction 100 1000 LVC 2017, PRL • Fully coherent rapid localization for Frequency (Hz) inhomogeneous networks of detectors Advanced GW Detectors LIGO Hanford Operational since 2015/08

Advanced GW Detectors LIGO Hanford Operational since 2015/08

LIGO Livingston Advanced GW Detectors Operational since 2015/08 LIGO Hanford Virgo Operational since 2015/08 Operational since 2017/08

LIGO Livingston Advanced GW Detectors Operational since 2015/08 AUGUST 14, 2017, 10:30:43 UTC: the ﬁrst BBH signal observed with Virgo

Hanford Livingston Virgo 14 12 LVC 2017, PRL • Signal arrived ﬁrst at Livingston, 10 then 8.4 ms later at Hanford, then 8

SNR 6 6.6 ms later at Virgo 4 2 • Clearly visible chirp trace in Hanford 0 5.0 and Livingston spectrograms; faint 4.5 256 4.0 telltale visible in Virgo 3.5 128 3.0 •2.5 Independently detected in low 64 2.0 1.5 latency within 30s of data

Frequency [Hz] 32 1.0 0.5 acquisition by two real-time Normalized Amplitude 16 0.0

] 1.0 searches for compact binary inspiral

-21 2 5 5 signals: GstLAL (see C. Hanna’s 0.5 colloquium at GSFC on October 31) 0.0 0 0 0

noise and PyCBC 0.5 5 Initial matched-ﬁlter signal to noise 5 2•

Whitened Strain [10 1.0 0.46 0.48 0.50 0.52 0.54 0.56 0.46 0.48 0.50 0.52 0.54 0.56 0.46 0.48 0.50 0.52 0.54 0.56 ratios in H/L/V: 7.3/13.7/4.4 Time [s] Time [s] Time [s] OCTOBER 14, 2017, 10:30:43 UTC: the ﬁrst GW signal observed with Virgo

• A 31 on 25 M⊙ binary black hole merger

• Thanks to additional detector baseline, localized to 60 deg2

• First measurement of gravitational-wave polarization: conﬁrmation of tensor nature of gravitational waves as predicted by GR

• More detailed tests of GR in progress

Credit: LIGO/Virgo/NASA/Leo Singer PHYSICAL REVIEW LETTERS week ending PRL 119, 161101 (2017) AUGUST20 OCTOBER 17, 2017 2017, 12∶41:04 UTC ∼100 s (calculated starting from 24 Hz) in the detectors’ sensitive band, the inspiral signal ended at 12 41:04.4 UTC. the ﬁrst gravitational wave signal from a binary neutron star merger In addition, a γ-ray burst was observed 1.7∶ s after the coalescence time [39–45]. The combination of data from Detected in low latency in Hanford data. Chirp track clearly the LIGO and Virgo detectors allowed a precise sky • position localization to an area of 28 deg2. This measure- visible, long duration immediately implied a low mass binary PHYSICAL REVIEW LETTERS week ending PRL 119, 161101ment enabled (2017) an electromagnetic follow-up campaign that 20 OCTOBERmerger 2017 ! identified a counterpart near the galaxy NGC 4993, con- PHYSICAL REVIEW LETTERS week ending sistent with the localization and distance inferred from PRL 119, 161101 (2017) 20 OCTOBER 2017 Additionally, a short instrumental noise transient gravitational-wave data [46–50]. Additionally, a short instrumental noise transient • Chirp visible in Livingston too, but did not trigger due to a appeared in theFrom LIGO-Livingston the gravitational-wave detector signal, 1.1 the best s before measured appeared in the LIGO-Livingston detector 1.1 s before photodiode saturation glitch. the coalescencecombination time of GW170817 of the masses as is shown the chirp in Fig. mass2.[51] the coalescence time of GW170817 as shown in Fig. 2. 0.004 This transient noise, or glitch [71], produced a very brief This transientM noise,1.188 or−þ glitch0.002 M [71]. From, produced the union a of very 90% brief credible (less than 5 ms) saturation in the digital-to-analog converter ¼ ⊙ • No chirp visible in Virgo. (less than 5 ms)intervals saturation obtained in the using digital-to-analog different waveform converter models (see of the feedback signal controlling the position of the test Sec. IV for details), the total mass of the system is between masses. Similar glitches are registered roughly once every of the feedback2.73 signal and 3.29 controllingM . The individual the position masses of are the in the test broad few hours in each of the LIGO detectors with no temporal • H/L/V signal to noise ratio after noise subtraction and glitch ⊙ correlation between the LIGO sites. Their cause remains masses. Similarrange glitches of 0.86 are to 2.26 registeredM , due to roughly correlations once between every their unknown. To mitigate the effect on the results presented in ⊙ removal: 18.8/26.4/2.0 few hours in eachuncertainties. of the LIGO This suggests detectors a BNS with as no the temporal source of the Sec. III, the search analyses applied a window function to gravitational-wave signal, as the total masses of known zero out the data around the glitch [72,73], following the correlation between the LIGO sites. Their cause remains treatment of other high-amplitude glitches used in the BNS systems are between 2.57 and 2.88 M with compo- • Component masses: 1.4–1.6 on 1.2—1.4 M⊙ unknown. To mitigate the effect on the results presented⊙ in O1 analysis [74]. To accurately determine the properties nents between 1.17 and ∼1.6 M [52]. Neutron stars in of GW170817 (as reported in Sec. IV) in addition to the Sec. III, the searchgeneral analyses have precisely applied measured a window masses⊙ as function large as 2 to.01 noise subtraction described above, the glitch was modeled 2 zero out the data0.04 aroundM [53], the whereas glitch stellar-mass[72,73], black following holes found the Æ in with a time-frequency wavelet reconstruction [75] and • Localized to only 30 deg and 26–48 Mpc despite weak/ binaries⊙ in our galaxy have masses substantially greater subtracted from the data, as shown in Fig. 2. unresolved signal in Virgo due to proximity to antenna treatment of other high-amplitude glitches used in the Following the procedures developed for prior gravita- than the components of GW170817 [54 56]. O1 analysis [74]. To accurately determine the– properties tional-waveLVC detections 2017,[29,78], we concludePRL there is no pattern null Gravitational-wave observations alone are able to mea- environmental disturbance observed by LIGO environmen- of GW170817sure (as the reported masses of in the Sec. two objectsIV) in and addition set a lower to limit the on tal sensors [79] that could account for the GW170817 FIG. 2. Mitigation of the glitch in LIGO-Livingston data. Times their compactness, but the results presented here do not signal. are shown relative to August 17, 2017 12 41:04 UTC. Top panel: noise subtraction described above, the glitch was modeled ∶ exclude objects more compact than neutron stars such as The Virgo data, used for sky localization and an A time-frequency representation [65] of the raw LIGO-Living- with a time-frequency wavelet reconstruction [75] and estimation of the source properties, are shown in the ston data used in the initial identification of GW170817 [76]. The FIG. 1. Time-frequency representations [65] of data containing subtracted fromquark the stars, data, black as shown holes, or in more Fig. exotic2. objects [57–61]. bottom panel of Fig. 1. The Virgo data are nonstationary coalescence time reported by the search is at time 0.4 s in this The detection of GRB 170817A and subsequent electro- the gravitational-waveabove 150 Hz due to scattered event light from the GW170817, output optics figure and observed the glitch occurs 1.1 by s before the this time. LIGO- The time- Following themagnetic procedures emission developed demonstrates for the prior presence gravita- of matter. Hanford (top),modulated by LIGO-Livingston alignment fluctuations and below 30 Hz due (middle),frequency track and of GW170817 Virgo is clearly visible (bottom) despite the detectors.to Times seismic noise are from anthropogenic shown activity. relative Occasional topresence August of the glitch. 17,Bottom 2017panel: The raw 12 LIGO-Livingston41:04 tional-wave detectionsMoreover, although[29,78] a, neutron we conclude star black hole there system is no is not strain data (orange curve) showing the glitch in the time domain. – UTC. Thenoise amplitude excess around the European scale power in mains each frequency detector is normalized∶ to that ruled out, the consistency of the mass estimates with the of 50 Hz is also present. No noise subtraction was applied To mitigate the glitch in the rapid reanalysis that produced the sky environmental disturbance observed by LIGO environmen- detector’s noise amplitude spectral density.map shown in Fig. 3 In [77], the the raw detector LIGO data were multiplied data, dynamically measured masses of known neutron stars in to the Virgo data prior to this analysis. The low signal tal sensors [79] that could account for the GW170817 independently observable noise sourcesby an and inverse Tukey a glitch window (gray that curve, right occurred axis) that zeroed binaries, and their inconsistency with the masses of known FIG. 2. Mitigationamplitude observed in Virgo of significantly the constrained glitch the out in the LIGO-Livingstondata around the glitch [73]. To mitigate the glitch in the data. Times in the LIGO-Livingstonsky position, but meant that the Virgo detector data did not have been subtracted, as signal. black holes in galactic binary systems, suggests the source aredescribed shown relativein the text. to This August noise 17, mitigation 2017measurement of 12 the is source the41:04’s properties, same a model UTC. as of the that glitch Top panel: contribute significantly to other parameters. As a result, based on a wavelet reconstruction∶ [75] (blue curve) was sub- The Virgowas data, composed used of for two sky neutron localization stars. and an Aused time-frequency for thethe estimation results of the presentedrepresentation source’s parameters reported in Sec. in tractedIV[65] from. the data.of The time-series the data raw visualized in this LIGO-Living- figure estimation of the source properties, are shown in the ston data usedSec. IV is notin impacted the by initial the nonstationarity identification of Virgo have been bandpassed between of GW170817 30 Hz and 2 kHz so that the [76]. The II. DATA data at the time of the event. Moreover, no unusual detector’s sensitive band is emphasized. The gravitational-wave bottom panel of Fig. 1. The Virgo data are nonstationary coalescencedisturbance time was observed reported by Virgo environmental by sensors. thestrain search amplitude of GW170817 is is of at the order time of 10−22 and so 0.4 is s in this above 150 Hz dueAt the to scattered time of GW170817, light from the the Advanced output LIGO optics detec- figurein the and VirgoData the used data in glitch this study due can be occurs found to in the[80]. lower 1.1not visible s BNS before in the bottom horizon panel. this and time. the The time- tors and the Advanced Virgo detector were in observing frequencydirection of track the source of GW170817 with respect to is the clearly detector visible’s antenna despite the modulated by alignment fluctuations and below 30 Hz due III. DETECTION mode. The maximum distances at which the LIGO- presencepattern. of the glitch. Bottom panelgenerated reporting: The a highly significantraw detection LIGO-Livingston of a binary to seismic noise from anthropogenic activity. Occasional GW170817 was initially identified as a single-detector neutron star signal [85] in coincidence with the independ- Livingston and LIGO-Hanford detectors could detect a strainFigure data (orange1 illustrates curve) the showing data as they the were glitch analyzed in the time to domain. noise excess aroundBNS system the European (SNR 8), power known mainsas the detector frequency horizon determineevent astrophysical with the LIGO-Hanford detector source by a low-latency properties.ently observed γ-ray burst After GRB 170817A data[39–41]. col- To mitigatebinary-coalescence the glitch search [81– in83] using the template rapid wave- reanalysisA rapid binary-coalescence that reanalysis [86,87] produced, with the the sky of 50 Hz is also[32,62,63] present., were No 218 noise¼ Mpc subtraction and 107 Mpc, was while applied for Virgo lection, several independently measured terrestrial contribu- map shownforms in computed Fig. in post-Newtonian3 [77] theory,[11,13,36,84] the raw. time series detector around the glitch suppressed data with were a window multiplied to the Virgo datathe horizon prior was to 58 this Mpc. analysis. The GEO600 The detector low signal[64] was tions to theThe two detector LIGO detectors and noise the Virgo detector were were all subtractedfunction [73], as shown from in Fig. 2, confirmed the the LIGO presence of also operating at the time, but its sensitivity was insufficient bydata an using inversetaking Wiener data Tukey at the time; filtering however, window the saturation[66] at the LIGO-, as(gray describeda significant curve, coincident in signal right[67 in the– LIGO70] axis) detectors.. This The that zeroed 2 amplitude observedto contribute in Virgo to the significantly analysis of the constrained inspiral. The configu- the outsubtraction the dataLivingston removedaround detector prevented the calibrationthe search glitch from registering[73] a linessource. was and To rapidly mitigate 60 localized Hz to a ac region the power of 31 deg glitch, in the sky position, but meant that the Virgo data did not simultaneous event in both LIGO detectors, and the low- shown in Fig. 3, using data from all three detectors [88]. ration of the detectors at the time of GW170817 is measurementmains harmonicslatency transfer of of Virgo the from data was source both delayed. LIGO’s properties,This data sky map streams. was issued toa observing model The partners, sensi- allowing of the glitch contribute significantlysummarized into[29] other. parameters. As a result, basedtivity on of aVisual the wavelet inspection LIGO-Hanford of the LIGO-Hanford reconstruction and LIGO- detectorthe identification[75] was of an(blue electromagnetic particularly curve) counterpart was sub- the estimationA of time-frequency the source’ representations parameters[65] reportedof the data in from improvedLivingston by the detector subtraction data showed the presence of of a clear, laser[46,48,50,77] pointing. noise; several tracted fromlong-duration the chirp data. signal in time-frequency The time-series representations The combined data SNR of visualized GW170817 is estimated to in be this figure all three detectors around the time of the signal is shown in havebroad been peaks bandpassed in the 150 between–800 Hz region 30 Hz were and effectively 2 kHz so that the Sec. IV is notFig impacted1. The signal by is the clearly nonstationarity visible in the LIGO-Hanford of Virgo removed,of the increasing detector strain data. As a theresult, an initial BNS alert was horizon32.4, with values 18.8, of 26.4, that and 2.0 in the detector LIGO-Hanford, data at the timeand LIGO-Livingston of the event. data. Moreover, The signal no is unusual not visible detectorby 26%.’s sensitive band is emphasized. The gravitational-wave disturbance was observed by Virgo environmental sensors. strain amplitude of GW170817161101-3 is of the order of 10−22 and so is not visible in the bottom panel. Data used in this study can be found in [80]. 161101-2

III. DETECTION generated reporting a highly significant detection of a binary GW170817 was initially identified as a single-detector neutron star signal [85] in coincidence with the independ- event with the LIGO-Hanford detector by a low-latency ently observed γ-ray burst GRB 170817A [39–41]. binary-coalescence search [81–83] using template wave- A rapid binary-coalescence reanalysis [86,87], with the forms computed in post-Newtonian theory [11,13,36,84]. time series around the glitch suppressed with a window The two LIGO detectors and the Virgo detector were all function [73], as shown in Fig. 2, confirmed the presence of taking data at the time; however, the saturation at the LIGO- a significant coincident signal in the LIGO detectors. The Livingston detector prevented the search from registering a source was rapidly localized to a region of 31 deg2, simultaneous event in both LIGO detectors, and the low- shown in Fig. 3, using data from all three detectors [88]. latency transfer of Virgo data was delayed. This sky map was issued to observing partners, allowing Visual inspection of the LIGO-Hanford and LIGO- the identification of an electromagnetic counterpart Livingston detector data showed the presence of a clear, [46,48,50,77]. long-duration chirp signal in time-frequency representations The combined SNR of GW170817 is estimated to be of the detector strain data. As a result, an initial alert was 32.4, with values 18.8, 26.4, and 2.0 in the LIGO-Hanford,

161101-3 PHYSICAL REVIEW LETTERS week ending PRL 119, 161101 (2017) AUGUST20 OCTOBER 17, 2017 2017, 12∶41:04 UTC ∼100 s (calculated starting from 24 Hz) in the detectors’ sensitive band, the inspiral signal ended at 12 41:04.4 UTC. the ﬁrst gravitational wave signal from a binary neutron star merger In addition, a γ-ray burst was observed 1.7∶ s after the coalescence time [39–45]. The combination of data from Detected in low latency in Hanford data. Chirp track clearly the LIGO and Virgo detectors allowed a precise sky • position localization to an area of 28 deg2. This measure- visible, long duration immediately implied a low mass binary ment enabled an electromagnetic follow-up campaign that merger! identified a counterpart near the galaxy NGC 4993, consistent with the localization and distance inferred from gravitational-wave data [46–50]. • Chirp visible in Livingston too, but did not trigger due to a From the gravitational-wave signal, the best measured photodiode saturation glitch. combination of the masses is the chirp mass [51] M 1.188 0.004M . From the union of 90% credible −þ0.002 No chirp visible in Virgo. intervals¼ obtained using⊙ different waveform models (see • Sec. IV for details), the total mass of the system is between 2.73 and 3.29 M . The individual masses are in the broad • H/L/V signal to noise ratio after noise subtraction and glitch ⊙ range of 0.86 to 2.26 M , due to correlations between their removal: 18.8/26.4/2.0 uncertainties. This suggests⊙ a BNS as the source of the gravitational-wave signal, as the total masses of known BNS systems are between 2.57 and 2.88 M with compo- • Component masses: 1.4–1.6 on 1.2—1.4 M⊙ nents between 1.17 and ∼1.6 M [52]. Neutron⊙ stars in general have precisely measured masses⊙ as large as 2.01 2 0.04 M [53], whereas stellar-mass black holes foundÆ in • Localized to only 30 deg and 26–48 Mpc despite weak/ binaries⊙ in our galaxy have masses substantially greater unresolved signal in Virgo due to proximity to antenna than the components of GW170817 [54–56]. LVC 2017, PRL pattern null Gravitational-wave observations alone are able to measure the masses of the two objects and set a lower limit on their compactness, but the results presented here do not exclude objects more compact than neutron stars such as quark stars, black holes, or more exotic objects [57–61]. FIG. 1. Time-frequency representations [65] of data containing The detection of GRB 170817A and subsequent electro- the gravitational-wave event GW170817, observed by the LIGO- magnetic emission demonstrates the presence of matter. Hanford (top), LIGO-Livingston (middle), and Virgo (bottom) Moreover, although a neutron star black hole system is not detectors. Times are shown relative to August 17, 2017 12 41:04 – UTC. The amplitude scale in each detector is normalized∶ to that ruled out, the consistency of the mass estimates with the detector’s noise amplitude spectral density. In the LIGO data, dynamically measured masses of known neutron stars in independently observable noise sources and a glitch that occurred binaries, and their inconsistency with the masses of known in the LIGO-Livingston detector have been subtracted, as black holes in galactic binary systems, suggests the source described in the text. This noise mitigation is the same as that was composed of two neutron stars. used for the results presented in Sec. IV.

II. DATA At the time of GW170817, the Advanced LIGO detec- in the Virgo data due to the lower BNS horizon and the tors and the Advanced Virgo detector were in observing direction of the source with respect to the detector’s antenna mode. The maximum distances at which the LIGO- pattern. Livingston and LIGO-Hanford detectors could detect a Figure 1 illustrates the data as they were analyzed to BNS system (SNR 8), known as the detector horizon determine astrophysical source properties. After data col- [32,62,63], were 218¼ Mpc and 107 Mpc, while for Virgo lection, several independently measured terrestrial contribu- the horizon was 58 Mpc. The GEO600 detector [64] was tions to the detector noise were subtracted from the LIGO also operating at the time, but its sensitivity was insufficient data using Wiener filtering [66], as described in [67–70]. This to contribute to the analysis of the inspiral. The configu- subtraction removed calibration lines and 60 Hz ac power ration of the detectors at the time of GW170817 is mains harmonics from both LIGO data streams. The sensi- summarized in [29]. tivity of the LIGO-Hanford detector was particularly A time-frequency representation [65] of the data from improved by the subtraction of laser pointing noise; several all three detectors around the time of the signal is shown in broad peaks in the 150–800 Hz region were effectively Fig 1. The signal is clearly visible in the LIGO-Hanford removed, increasing the BNS horizon of that detector and LIGO-Livingston data. The signal is not visible by 26%.

161101-2 GW170104 GW150914 GW151226

Credit: LIGO/Virgo/NASA/Leo Singer GW170817

GW170104 GW150914 GW151226

Credit: LIGO/Virgo/NASA/Leo Singer 10

LVC, Fermi GBM, & INTEGRAL 2017 (ApJL) Merger GRB start 2500 Lightcurve from Fermi/GBM (10 50 keV) 2250 2000 1750 1500

Event rate (counts/s) 1250

1750 Lightcurve from Fermi/GBM (50 300 keV) A coincident short 1500 1250 gamma-ray burst!!! 1000

750 Event rate (counts/s) • Short-duration gamma-ray transient Lightcurve from INTEGRAL/SPI-ACS detected by Fermi GBM and INTEGRAL 120000 (> 100 keV) SPI-ACS 117500

115000 • Spatially consistent with GW localization, 112500

Event rate (counts/s) but arrived 1.74 s after GW merger signal

400 Gravitational-wave time-frequency map 300 • Two components: short (~0.6 s), hard 200 (Ep~185 KeV) pulse that resembles standard short GRB and delayed, longer 100 Frequency (Hz) (~1.2 s), softer (Ep~10 keV) tail 50 10 8 6 4 2 0 2 4 6 Time from merger (s)

Figure 2. The joint, multimessenger detection of GW170817 and GRB 170817A. Top: The summed GBM lightcurve for sodium iodide (NaI) detectors 1, 2, and 5 for GRB 170817A between 10 and 50keV, matching the 100ms time bins of SPI-ACS data. The background estimate from Goldstein et al. (2016) is overlaid in red. Second: The same as the top panel but in the 50–300keV energy range. Third: The SPI-ACS lightcurve with the energy range starting approximately at 100keV and with a high energy limit of least 80MeV. Bottom: The time-frequency map of GW170817 was obtained by coherently combining GW LIGO-Hanford and LIGO-Livingston data. All times here are referenced to the GW170817 trigger time T0 . TYPICAL OBSERVER REACTION 2. Discovery, Observing Strategies Discovery of a rapidly fading optical counterpart

NGC 4993 A SSS17a B

April 28, 2017 Hubble Space Telescope August 17, 2017 Swope Telescope

Figure 4 30 30 images centered on NGC 4993 with North up and East left. Panel A: Hubble ⇥ Space• Rapidly Telescope F606W-band fading optical (broad V )transient image from 4 detected months before by the One GW trigger Meter, (25, 35). PanelTwo B: Swope Hemisphere image of SSS17a. (1M2H The i-band) team image using was obtained Swope on 2017 1m August 17 at 23:33 UT bytelescope the Swope telescope at Las at Campanas Las Campanas Observatory. SSS17a is marked with the red arrow. No object is present in the Hubble image at the position of SSS17a (25, 35). 1M2H Team Coulter et al. 2017 Announced as SSS17a, also designated AT2017gfo • (Science) • Position in projected coincidence with the galaxy NGC 4993 at a distance of 40 Mpc • Also found using synoptic optical/near infrared telescopes (DECam, VISTA, MASTER) Confirmed by many teams doing independent opticalFigure 2 Sky region covering the 90th-percentile confidence region for the location of • • But defied common wisdom that it was found by galaxy targeted counterpart searches and targeted follow-up of sourceGW170817. follow-upThe 50th, on 70th, small-FOV and 90th-percentile (and even small contours aperture) are telescopes shown, with contours extracted from the same probability map as Fig. 1. Grey circles represent the locations of galaxies in our 12 galaxy catalog and observed by the Swope telescope on 2017 August 17-18 to search for the EM counterpart to GW170817. The size of the circle indicates the probability of a particular galaxy being the host galaxy for GW170817. The square regions represent individual Swope pointings with the solid squares specifically chosen to contain multiple galaxies (and labeled in the order that they were observed) and the dotted squares being pointings which contained individual galaxies. The blue square labeled ’9’ contains NGC 4993, whose location is marked by the blue circle, and SSS17a.

10 Galaxy strategies for LIGO-Virgo follow-up

10), and (30, 30) M binary mergers to be [low, mean, 8 6 5 8 8 7 high]: [10 , 10 , 10 ], [1 10 , 5 10 , 2 10 ], 9 8 ⇥ 8 ⇥ 3 1 ⇥ and [6 10 , 2 10 , 6 10 ]Mpc yr ,respec- tively (Abadie⇥ et al.⇥ 2010; Abbott⇥ et al. 2016h,c,a). From these rates we are able to estimate the number of systems that will be detected with each GW network, and then use the results in Fig. 2 to infer the number of systems that will be localized suciently to identify the unique host galaxy. We summarize our results in Tables 1 and 2. There are a number of interesting aspects of our results: The localization volumes are larger for more mas- • sive binaries. This is because these binaries are generally detected to higher distances, and therefore a fractional error in distance or sky position corresponds to a larger volume. Massive binaries also merge at lower frequency and spend shorter Fig. 2.— Cumulative distribution of the 90% confidence level 3D time in the GW detectable band, leading to a larger localization comoving volumes for 10,000 simulated 1.4M –1.4M error in estimates of the time-of-arrival. We find Nissanke, Kasliwal, & Gehrels, Cannizzo, Singer, Chen, Holz, (blue), 10M –10M (green), and 30M –30M (red) binary merg- that BNS systems are the most likely to be well ers. We consider threeChen GW detector & Holz network 2016 configurations at dif- Georgieva 2013 Kanner, Kasliwal, et al. 2016 ferent sensitivities: Solid–HLV at O3 sensitivity, dotted–HLV at localized, and therefore are the most likely to allow design sensitivity, and dashed–HLVJI at design sensitivity. The for unique host galaxy association. Nissanke, & Singer 2016 top axis shows the expected number of galaxies within the volume assuming the number density of galaxies is 0.01/Mpc3 The localization areas are larger for more massive • binaries. Massive binaries are detected at higher the posterior distribution for all source parameters using distances, so their signals are significantly red- full models of the source waveform and all detector data, shifted and, on top of their intrinsic lower merging and find good agreement. frequency, spend even shorter time in band. This is why, for example, the 10-10M binary black hole 3. RESULTS (BBH) localization area is 2.5 times that of the Our results are summarized in Figs. 2 and 3 and Ta- BNSs at HLV O3 sensitivity,⇠ with the ratio increas- bles 1 and 2. Fig. 2 shows a plot of the cumulative dis- ing to 5 at design sensitivity. tribution of the 90% localization volumes, while Fig. 3 ⇠ presents the cumulative distribution of 90% localization The localization areas are smaller for improved • sky areas. For example, we find that 50% of the BNS gravitational-wave networks. This is because the sources will be localized to a volume of 104 Mpc3,re- localization is primarily dependent on the timing gardless of detector network. ⇠ measurements at each detector, and as the net- We are particularly interested in the number of galax- works improve the relative timing improves. Going ies that can be expected in these volumes. To esti- from three detectors to five detectors causes a large mate this, we assume that the number density of galax- improvement in the 2D localization. 3 ies is 0.01/Mpc . This estimate takes the Schechter In all cases there is a tail to very well localized function (Schechter 1976) parameters in B-band = • 2 3 3 10 2 ⇤ 1.6 10 h Mpc , ↵=-1.07, L =1.2 10 h LB, and⇥h =0.7 ( Norberg et al. (2002);⇤ Liske⇥ et al. (2003); González et al. (2006); Gehrels et al. (2016), LB, is the solar luminosity in B-band), integrating down to 0.12 L⇤ and comprising 86% of the total luminosity. The top axis in Fig. 2 shows the expected number of galaxies in the 90% localization volumes. For example, a localization volume of 1000 Mpc3 corresponds to an expecta- tion of 10 galaxies, while a localization smaller than 100 Mpc3 corresponds to on average a single galaxy within the localization volume. In other words, for GW sources localized to within 100 Mpc3, it may be possible to di- rectly identify the host galaxy without the need for an associated EM transient. We have argued that a small fraction of GW sources are suciently well localized to allow for improved constraints on their host galaxies. But how many of these systems will actually be detected? To estimate this we Fig. 3.— Cumulative distribution of the 90% confidence level need to know the expected number of systems that will 2D localization areas for 10,000 simulated 1.4M –1.4M (blue), 10M –10M (green), and 30M –30M (red) binary mergers. We be detected by the various networks. With only a few consider three GW detector network configurations at di↵erent sen- systems detected to date, the rate of binary coalescences sitivities: Solid–HLV at O3 sensitivity, dotted–HLV at design sen- remains uncertain; we assume the rate for (1.4, 1.4), (10, sitivity, and dashed–HLVJI at design sensitivity. Pinpointing the optical counterpart • Order of magnitude A B C 60° Right ascension x´ x´ improvement in localization due to 90% 10 x Mpc Virgo: 190 → 30 deg² 50% 90% 30° HL Declination • Of the ~50 galaxies HLV 5° in the LIGO/Virgo 3D x localization volume, 0° Luminosity distance D NGC 4993 is the 50% Gemini+F2 JHKs third most massive IPN GBM

-30° • A priori, the most Declination 16h 12h likely host assuming 8h 10´´ stellar mass as a -60° Right ascension tracer for BNS merger rate Kasliwal, Nakar, Singer+ 2017 Localization of a Singer et al., in prep. GW170817 with future detectors

HL as built, HLV as built, HLV at O3 sensitivity, HLVIK at design sensitivity

Caveat: this was an exceptionally well localized event because it was so nearby! 3. Optical and Near Infrared Observations Near-infrared imaging

• 8.1m Gemini-South + FLAMINGOS2 JHK imaging (PI: Singer)

• 5m Palomar Hale + WIRC K imaging

• 1.4m IRSF JHK imaging (PI: Barway)

• 1.3m @ CTIO + ANDICAM (PI Cobb)

• 3.5m @ APO K imaging (PI Chanover)

Kasliwal, Nakar, Singer+ 2017 12 160821B 8,)336W,u−0 i,i' −0 −20 130603B J−3 +−4 14 .,.V−5

−18

−16

18 050709

−14 060614

20 AbVRluWH maJniWuGH (AB) AppaUHnW maJniWuGH (AB) −12

−10

24 BaUnHV+16 WRllaHJHU+17 5RVVwRJ+17 −8 CRmpaUiVRn G5B 26 0 2 4 6 8 10 12 14 16 18 20 7imH VinFH GW170817 (GayV)

Kasliwal, Nakar, Singer+ 2017 Figure 4: Comparison of the UiJHK lightcurve of EM170817 and literature macronova model lightcurves, which show a good match in the infrared but fail to produce the observed blue emission. Detections are shown as circles, upper limits as triangles. The models have been scaled to a distance of 40 Mpc and set to a galactic extinction of AV =0.34. The model lightcurves are the following: mej =0.05M⊙,vej =0.1c from (39), model N4 with the DZ31 mass formula from (17) and γA2 at a viewing angle of 30◦ from (18). In addition the OIR excess points from previously observed short GRBs are shown as squares (scaled to 40 Mpc and corrected for time dilation). GRB080503 (40) is not shown since there is no measured redshift.

23 A 1042 )

1 Blackbody ﬁts − 3 (erg s

L 2 1041 1 • Bolometric luminosity, eﬀective temperature, photospheric radius, and expansion velocity 13 11 B 9 K) 3 7 • Red line: hydrodynamic (10

T 5 simulation of cocoon + kilonova 3 1

8 ⊙ • 1: Cocoon cooling component

R ) 6 4 4 (10

R 2 C • 2: Fast-moving (>0.4c) 0 radioactive ejecta component 0.3 D

0.2 v/c • 3: Slow-moving (<0.4c) 0.1 radioactive ejecta component 0 1 2 3 4 5 6 7 8 9 10 11 12 13 Time since GW170817 (days) Kasliwal, Nakar, Singer+ 2017 Optical spectroscopy

• Gemini-South + GMOS

• Keck + LRIS

• Featureless blackbody continuum

Figure S3: Optical spectral sequence of EM170817 including the Gemini-S/GMOS spectra fromKasliwal, 2.5 and 3.5Nakar, days andSinger the Keck/LRIS+ 2017 spectrum from 7.7 days. The measured spectra are shown in gray, and versions smoothed with a Savitsky-Golay ﬁlter are shown in black.

Table S2: Optical and near-IR spectroscopic observations of EM170817. For each observation we give the observation date, the time since GW170817, the telescope, instrument, exposure time, approximate wavelength range, and spectral resolving power.

Observation Date ∆t Telescope Instrument Exposure Wavelength Range λ/∆λ (UTC) (days) (s) (A)˚ 2017-08-20 01:08 2.52 Gemini-S GMOS 2 × 300 6000–9000 1900 2017-08-21 00:15 3.48 Gemini-S GMOS 4 × 360 3800–9200 1700 2017-08-22 00:21 4.49 Gemini-S FLAMINGOS-2 8 × 150 12980–25070 600 2017-08-22 00:47 4.50 Gemini-S FLAMINGOS-2 6 × 150 9840–18020 600 2017-08-25 05:45 7.71 Keck I LRIS 300 + 2 × 600 2000–10300 1000 2017-08-29 00:23 11.49 Gemini-S FLAMINGOS-2 16 × 150 12980–25070 600

77 Near-infrared spectroscopy

• Gemini-South + FLAMINGOS-2 JH + HK spectrum @ 4.5 days after merger (black)

• Barnes & Kasen (2013) model kilonova spectrum at 4.5 days post merger

Kasliwal, Nakar, Singer+ 2017 Looks nothing like known explosive transients

Kasliwal, Nakar, Singer+ 2017

Figure S4: Near-infraredComparison spectrum of EM170817 with at t =4.5 SNdays post IaFigure merger S5: is shown Near-infrared at theComparison spectrum of EM170817 with at t =4.5 CCdays post SN merger is shown at the bottom, along with the spectra of the type Ia SN 2014J at 16 and 94 daysbottom, post maximum along with (144 the), spectra of the type IIP SN 2013ej at 12 days (147), the type IIb SN and the type Ia SN 2005df at 198 days post maximum (145). Each spectrum2011dh is at normalized 13 days (148 to), the type IIn SN 2010jl at 21 days (149), and the broad-lined type Ic the flux between 10000–10500 A˚ and shifted up from the one below for(Ic-BL) clarity. The SN spectrum 1998bw at of 8 days post maximum (150). The spectra of EM170817 are presented EM170817 was corrected for Milky Way reddening assuming E(B − Vas)=0 in figure.1 and S4. a standard We label several transitions commonly identified in the spectra of core-collapse RV =3.1 extinction law (146), and smoothed using a Savitzky–Golay filterSNe with to clearly vertical, show dashed the lines. prominent, broad spectral features at ∼ 10600 and 15000 A.˚ The unfiltered data are shown in gray. Regions of low S/N due to the strong telluric absorption features between the J, H, and K spectral windows are indicated by the vertical, gray bars. We label several transitions of Fe- peak elements on the top spectrum of SN 2014J. While qualitatively similar, the broad Fe-peak features characteristic of SNe Ia are inconsistent with the features observed in the spectrum of 79 EM170817.

78 Remarkable agreement between optical/near-IR data and r-process kilonova/macronova modeling

Kasen+ 2017 (Nature)

Pian+ 2017 (Nature) 4. Interpretation, Radio The puzzle

• Short GRB, but DELAYED and SUB-LUMINOUS

• No evidence for on-axis X-ray/optical/radio afterglow (although no deep observations before ~0.5 days)

• Bright, rapidly fading UV/optical emission → FAST and HOT as inferred from blackbody ﬁts

• DELAYED X-ray and radio emission A On-axis Weak sGRB B Slightly Off-Axis Classical sGRB Weak Afterglow Afterglow (X-ray/Radio) (X-ray/Radio) A concordant m um iu di ed e M M r r la lla l 30° e te t Weak s rs r e te Weak γ-rays t n picture n sGRB I Macronova I Macronova sGRB (UVOIR) (UVOIR) A. Early data sparse, but probably ruled out by lack of early panchromatic afterglow (requires implausibly low circumburst medium density)

C Cocoon with Choked Jet D On-axis Cocoon with Off-Axis Jet B. Requires some ﬁne-tuning of observer Afterglow Afterglow geometry (just barely outside of cone (X-ray/Radio) (X-ray/Radio)

m m of jet) or structured jet iu iu ed ed M M r r lla 30° lla 30° te te rs rs te Weak γ-rays te Weak γ-rays In In C. Highly energetic cocoon, ultra- Macronova sGRB Macronova (UVOIR) (UVOIR) relativistic jet is choked and does not break out

D. Low energy cocoon, ultra-relativistic jet successfully breaks out and is seen oﬀ axis Kasliwal, Gottlieb, Singer, et al. 2017 Radio at ~100 days post merger has the power to discriminate between the models

Hallinan, Corsi, Mooley, ..., Singer 2017

Fig. 3. ORadioﬀ-axis light jet curve (cases is consistent B or D) with predict either that an off the-axis radio jet or already a cocoon. peaked, Light curves are shown for both proposed cocoon models, i.e. a high energy cocoon due to a choked jet (red solid curve) andwhereas a low energy choked cocoon cocoon with jet (casebreak-out C) (blackpredicts dashed peak curve) at. The~100 light days curve favors the low energy cocoon with jet break-out. For the off-axis jet, we use a jet isotropic equivalent 50 energy, Eiso =1.5 × 10 erg, a jet half-opening angle, θj ~12°, and an angle between our line of -3 -3 sight and the jet, θobs ~30°, together with an ISM density of 10 cm (blue dash-dot curve). However, the light curve is consistent with a range of parameter space in jet energy and ISM density. For each set of jet and observer angle there is a single solution (namely n and E) that fit the data, which cannot be distinguished from the light curve. The pink curve is a model that represents the expected radio emission due to the high velocity tail of the sub-relativistic ejecta, which is yet to be ruled out. All the models indicate an ISM density of ~10-3 – 10-2 cm-3, which in turn is consistent with constraints from H I upper limits (Methods ref). We predict that the radio light curves will distinguish between the models shown within the first 100 days of the merger.

VLBI can break tie by resolving angular diameter of blast wave

Hallinan, Corsi, Mooley, ..., Singer 2017

AngularFig. 4.diameter Predicted time of evolution blast of wavethe radio sourcefrom size cocoon/choked for different models. Different jet possibleshould be smaller modelsat all for times the radio thanemission o fromﬀ-axis EM170817 relativistic expand at different jet because velocities. The it yellow takes area longer shows the approximate parameter space accessible to VLBI. Model parameters for the off-axis jet and hasand tococoon travel models fartherare the same to as used sweep in Figure up 3. The su subﬃ-relativisticcient circumburstejecta is assumed to have material a bulk velocity of 0.2c, as discussed in the main text.

5. A Few Open Questions

1. Why was the prompt emission so faint? Did we merely observe the merger oﬀ-axis, or was the burst intrinsically diﬀerent from typical well- studied sGRBs (e.g. from Swift with X-ray afterglows)?

2. What was the cause of the GW-GRB time delay?

3. What was the initial Lorentz factor of the outﬂow, and what process launched it? Did it interact in any interesting way with the sub-relativistic outﬂows?

4. What can we learn from early-time observations of future GW-GRB NS binary mergers, particularly in X-ray/optical/ultraviolet?