Masaomi Tanaka...Optical Transient Search Survey Diameter (m) FOV (deg2) Depth (R mag) Area/ day(deg2) LOSS 0.76 0.01 19 galaxy CHASE 0.4 x 4 0.03 19 galaxy ROTSE-III 0.45 3.42 18.5

Optical Transient Search:Prospects for EM Counterparts of GW Sources

Masaomi Tanaka(National Astronomical Observatory of Japan)

The most famous optical transient

1 deg

Before

After

Before

After

Optical transient

search

GW detection

EM follow up

“EM-triggered” search

“GW-triggered” search

EM follow up

• EM-triggered Search

• Counterparts of GW Sources

• GW-triggered Search


Magnitude

5 20 3010mag

eye 1m

25

8m

15

space

SN@ 200 Mpc

m = �2.5 log10(F�)� 48.6

= �2.5 log10

�F�

3631� 10�23 erg s�1 Hz�1 cm�2

�

2 - 3.5 m : many

>8-10 m : 13

3.5-6.5 m : 24

4m

0

SN 1987A@ 50 kpc

Optical Transient Search

Survey Diameter(m)

FOV(deg2)

Depth(R mag)

Area/day(deg2)

LOSS 0.76 0.01 19 galaxyCHASE 0.4 x 4 0.03 19 galaxy

ROTSE-III 0.45 3.42 18.5 450Kiso 1.05 4 21 100PTF 1.26 7.8 21 1000

Pan-STARRS 1.8 7 21.5 6000SDSS-II 2.5 1.5 22.6 150

GOODS 2.5 (HST) 0.003 26 0.04SNLS 3.6 1 24.3 2

SkyMapper 1.3 5.7 22 --Subaru/HSC 8.2 1.75 26.5 3.5

(partly taken from Rau et al. 2009, PASP, 121, 1334)

ROTSE

PS (wide)

PS (deep)

SNLS

GOODS

PTF

0.01

0.1

1

10

100

1000

10000

18 19 20 21 22 23 24 25 26 27

Surv

ey a

rea

(deg

2 )

Limiting magnitude

SDSSPTF

LOSS(@6000A)

SkyMapper

CHASE

La SillaCRTS

KISS

HSC

ROTSE PS (deep)

SNLS

GOODS

0.01

0.1

1

10

100

1000

10000

18 19 20 21 22 23 24 25 26 27

Surv

ey a

rea

(deg

2 )

Limiting magnitude

< 1 d1-5 d> 5 d

PTF

PTFSkyMapper

PS (wide)

SDSS

LOSSCHASE

La SillaCRTS

(@6000A)

KISS

HSC

High cadence

ROTSE

HST

HSC

SNLS

Pan-STARRSSDSS

PTFPTF

16

18

20

22

24

26

28 0.01 0.1 1 10 100

Mag

nitu

de

Cadence (days)

Deep

BEFORE 2012

4 S. R. KULKARNI CALTECH OPTICAL OBSERVATORIES PASADENA, CALIFORNIA 91125, USA

The sources of interest to these facilities are connected to spectacular explosions. How-ever, the horizon (radius of detectability), either for reasons of optical depth (GZK cuto↵;�� ! e±) or sensitivity, is limited to the Local Universe (say, distance . 100Mpc). Un-fortunately, these facilities provide relatively poor localization. The study of explosions inthe Local Universe is thus critical for two reasons: (1) sifting through the torrent of falsepositives (because the expected rates of sources of interest is a tiny fraction of the knowntransients) and (2) improving the localization via low energy observations (which usuallymeans optical). In Figure 2 we display the phase space informed by theoretical considera-tions and speculations. Based on the history of our subject we should not be surprised tofind, say a decade from now, that we were not su�ciently imaginative.

Figure 2. Theoretical and physically plausible candidates are marked in theexplosive transient phase space. The original figure is from Rau et al. (2009).The updated figure (to show the unexplored sub-day phase space) is from theLSST Science Book (v2.0). Shock breakout is the one assured phenomenon on thesub-day timescales. Exotica include dirty fireballs, newly minted mini-blazars andorphan afterglows. With ZTF we aim to probe the sub-day phase space (see §5).

The clarity a↵orded by our singular focus – namely the exploration of the transientoptical sky – allowed us to optimize PTF for transient studies. Specifically, we undertakethe search for transients in a single band (R-band during most of the month and g bandduring the darkest period). As a result our target throughput is five times more relativeto multi-color surveys (e.g. PS-1, SkyMapper).

Given the ease with which transients (of all sorts) can be detected, in most instances, thetransient without any additional information for classification does not represent a useful,let alone a meaningful, advance. It is useful here to make the clear detection betweendetection7 and discovery.8 Thus the burden for discovery is considerable since for most

7 By which I mean that a transient has been identified with a reliable degree of certainty.8By which I mean that the astronomer has a useful idea of the nature of the transient. At the very

minimum we should know if the source is Galactic or extra-galactic. At the next level, it would be useful

LSST Science Book (after Rau+09, Kasliwal+,Kulkarni+)

High cadence

ROTSE

HST

HSC

SNLS

Pan-STARRSSDSS

PTFPTF

16

18

20

22

24

26

28 0.01 0.1 1 10 100

Mag

nitu

de

Cadence (days)

Deep

PTF 1m

KISS 1m

DECam 4m

Subaru 8m

KISS: KIso Supernova Survey

• Extremely high cadence

• 1-hr cadence

• ~100 deg2 /day

• ~ 21 mag in g-band (4000 A)

• High SFR field (within z=0.05, 30-100 Msun/yr)

2012/04: Dry run -2012/09: Main survey -

1.05m Schmidt Telescope

SN shock breakout7.5 Shock Breakout

-2

-1

0

1

2

3

4

5

-50 0 50 100

Flux

[10-3

2 W m

-2 H

z-1]

Days from the peak [Days]

NUVgri

0

1

2

3

0 5 10 15

22

24

26

280 0.5 1 1.5

Appa

rent

g’ m

agni

tude

[mag

]

Days since bolometric peak (observer frame) [Days]

z=0.2z=0.5z=1

z=1.5z=2

z=2.5z=3

Figure 7.6: (Left): Comparison between the GALEX and SNLS observations (points, SNLS-04D2dc,Schawinski et al. 2008) and a SN IIP model with Mms = 20M! and E = 1.2 × 1051 erg reddened for the hostgalaxy extinction with a color excess E(B − V ) = 0.14 mag (lines, Tominaga et al. 2009) (black and red: near UV,green: g-band, blue: r-band, magenta: i-band). The inset enlarges the phase when the SN emitted UV light. (Right):Apparent g′-band light curve of a shock breakout in AB magnitude system for a SN IIP model with Mms = 40M!

and E = 1051 erg. Limiting magnitudes for a 4σ detection in 3.3 min, 5 min, 10 min, and 20 min integrations arealso shown (dashed line), assuming 0.7” seeing, 1.6 arcsec aperture, and 3 days from New Moon.

identify the shock breakout with the blue g′ − r′ color as described below. Aims of this surveyare (1) detecting numerous high-z shock breakouts and obtaining their multicolor light curves, (2)observationally establishing the physics of shock breakouts and confirming that the shock breakoutstake place universally, (3) deriving a cosmic star formation history (CSFH) up to z ∼ 1.5 with theshock breakouts, and ultimately (4) developing the totally-new high-z study with shock breakouts.

Introduction

Core-collapse supernovae (CCSNe) have been so far observed only at z ! 0.9 (Dahlen et al. 2004;Poznanski et al. 2007), except for extraordinary events like Type IIn SNe (SNe IIn) (Cooke et al.2009) and gamma-ray bursts (GRBs; Salvaterra et al. 2009). The record will be broken by a shockbreakout. The shock breakout is the bolometrically-brightest phenomenon in SNe (> 1044 ergs s−1)lasting several seconds to several hours and emits dominantly in X-ray or ultraviolet (UV) with aquasi-blackbody spectrum (T > 105 K, Blinnikov et al. 2000). Although the shock breakouts areproposed to be a probe of the distant universe, its short duration and X-ray/UV-peaked spectramake it difficult to be observed. The first and currently last complete light curve of shock breakoutof normal CCSN was obtained for SN IIP SNLS-04D2dc (redshift z = 0.19, e.g. Schawinski et al.2008) by the GALEX satellite but the detection significance in near UV and far UV bands is only! 4σ and ! 2σ, respectively.

We adopted a multi-group radiation hydrodynamics code STELLA (Blinnikov et al. 2000) andpresented X-ray-to-infrared light curves (LCs), including the shock breakout, plateau, and tail, of

13

1 day !

Figure from N. Tominaga

progenitor star

> a few days

Breakout

Figure from T. Morokuma

SNshock

breakout

@ 200Mpc

17

18

19

20

21

22

23

24

25 0.01 0.1 1 10 100

Mag

nitu

de

Time (days)

17

18

19

20

21

22

23

24

25 0.01 0.1 1 10 100

Mag

nitu

de

Time (days)

Kyutokuet al.

NS merger

Metzger et al.Tominaga et al.

KISS

Depth

Area

CadenceOptical Transient

Search

Depth

Follow up

• Higher wavelength resolution

• Better image quality

• Polarization

• ...

Response

NS merger @ < 200 Mpc

~ 40 / yr / allsky

~ 1 / yr / 1000 deg2

(~ 500 / yr / 1000 deg2 for supernovae)

Blind optical search for NS merger

ROTSE

SNLS

GOODS

0.01

0.1

1

10

100

1000

10000

18 19 20 21 22 23 24 25 26 27

Surv

ey a

rea

(deg

2 )

Limiting magnitude

< 1 d1-5 d> 5 d

LOSSCHASE

CRTS

(@6000A)

KISS

HSC

PS (deep)

PTF

PTFSkyMapper

PS (wide)

SDSSLa Silla





EM signature from NS-NS merger

• On-axis short GRB

• Off-axis radio/optical afterglow

• Radioactive emission - kilonova- macronova- mini SN

The Astrophysical Journal, 746:48 (15pp), 2012 February 10 Metzger & Berger

with specific stellar populations). Because merger counterpartsare predicted to be faint, obtaining a spectroscopic redshiftis challenging (cf. Rowlinson et al. 2010), in which casespectroscopy of the host galaxy is the most promising meansof obtaining the event redshift.

It is important to distinguish two general strategies for con-necting EM and GW events. One approach is to search for aGW signal following an EM trigger, either in real time or ata post-processing stage (e.g., Finn et al. 1999; Mohanty et al.2004). This is particularly promising for counterparts predictedto occur in temporal coincidence with the GW chirp, such asshort-duration gamma-ray bursts (SGRBs). Unfortunately, mostother promising counterparts (none of which have yet beenindependently identified) occur hours to months after coales-cence.6 Thus, the predicted arrival time of the GW signal willremain uncertain, in which case the additional sensitivity gainedfrom this information is significantly reduced. For instance, ifthe time of merger is known only to within an uncertainty of∼ hours (weeks), as we will show is the case for optical (radio)counterparts, then the number of trial GW templates that mustbe searched is larger by a factor ∼104–106 than if the mergertime is known to within seconds, as in the case of SGRBs.

A second approach, which is the primary focus of this paper,is EM follow-up of GW triggers. A potential advantage in thiscase is that counterpart searches are restricted to the nearbyuniverse, as determined by the ALIGO/Virgo sensitivity range(redshift z ! 0.05–0.1). On the other hand, the large errorregions are a significant challenge, which are estimated to betens of square degrees even for optimistic configurations of GWdetectors (e.g., Gursel & Tinto 1989; Fairhurst 2009; Wen &Chen 2010; Nissanke et al. 2011). Although it has been arguedthat this difficulty may be alleviated if the search is restrictedto galaxies within 200 Mpc (Nuttall & Sutton 2010), we stressthat the number of galaxies with L " 0.1 L∗ (typical of SGRBhost galaxies; Berger 2009, 2011) within an expected GW errorregion is ∼400, large enough to negate this advantage for mostsearch strategies. In principle the number of candidate galaxiescould be reduced if the distance can be constrained from theGW signal; however, distance estimates for individual eventsare rather uncertain, especially at that low of S/Ns that willcharacterize most detections (Nissanke et al. 2010). Moreover,current galaxy catalogs are incomplete within the ALIGO/Virgovolume, especially at lower luminosities. Finally, some mergersmay also occur outside of their host galaxies (Berger 2010;Kelley et al. 2010). Although restricting counterpart searches tonearby galaxies is unlikely to reduce the number of telescopepointings necessary in follow-up searches, it nevertheless cansubstantially reduce the effective sky region to be searched,thereby allowing for more effective vetoes of false positiveevents (Kulkarni & Kasliwal 2009).

At the present there are no optical or radio facilities that canprovide all-sky coverage at a cadence and depth matched tothe expected light curves of EM counterparts. As we show inthis paper, even the Large Synoptic Survey Telescope (LSST),with a planned all-sky cadence of four days and a depth ofr ≈ 24.7 mag, is unlikely to effectively capture the range ofexpected EM counterparts. Thus, targeted follow-up of GW

6 Predicted EM counterparts that may instead precede the GW signal includeemission powered by the magnetosphere of the NS (e.g., Hansen & Lyutikov2001; McWilliams & Levin 2011; Lyutikov 2011a, 2011b), or cracking of theNS crust due to tidal interactions (e.g., Troja et al. 2010; Tsang et al. 2011),during the final inspiral. However, given the current uncertainties in thesemodels, we do not discuss them further.

BH

obs

jTidal Tail & Disk Wind

Ejecta ISM Shock

Merger Ejecta

v ~ 0.1 0.3 c

Optical (hours days)

KilonovaOptical (t ~ 1 day)

Jet ISM Shock (Afterglow)

GRB(t ~ 0.1 1 s)

Radio (weeks years)

Radio (years)

Figure 1. Summary of potential electromagnetic counterparts of NS–NS/NS–BH mergers discussed in this paper, as a function of the observer angle,θobs. Following the merger a centrifugally supported disk (blue) remains aroundthe central compact object (usually a BH). Rapid accretion lasting !1 spowers a collimated relativistic jet, which produces a short-duration gamma-ray burst (Section 2). Due to relativistic beaming, the gamma-ray emissionis restricted to observers with θobs ! θj , the half-opening angle of the jet.Non-thermal afterglow emission results from the interaction of the jet withthe surrounding circumburst medium (pink). Optical afterglow emission isobservable on timescales up to ∼ days–weeks by observers with viewing anglesof θobs ! 2θj (Section 3.1). Radio afterglow emission is observable from allviewing angles (isotropic) once the jet decelerates to mildly relativistic speedson a timescale of weeks–months, and can also be produced on timescales ofyears from sub-relativistic ejecta (Section 3.2). Short-lived isotropic opticalemission lasting ∼few days (kilonova; yellow) can also accompany the merger,powered by the radioactive decay of heavy elements synthesized in the ejecta(Section 4).(A color version of this figure is available in the online journal.)

error regions is required, whether the aim is to detect opticalor radio counterparts. Even with this approach, the follow-up observations will still require large field-of-view (FOV)telescopes to cover tens of square degrees; targeted observationsof galaxies are unlikely to substantially reduce the large amountof time to scan the full error region.

Our investigation of EM counterparts is organized as follows.We begin by comparing various types of EM counterparts, eachillustrated by the schematic diagram in Figure 1. The first is anSGRB, powered by accretion following the merger (Section 2).Even if no SGRB is produced or detected, the merger may stillbe accompanied by relativistic ejecta, which will power non-thermal afterglow emission as it interacts with the surroundingmedium. In Section 3 we explore the properties of such “or-phan afterglows” from bursts with jets nearly aligned towardEarth (optical afterglows; Section 3.1) and for larger viewingangles (late radio afterglows; Section 3.2). We constrain ourmodels using the existing observations of SGRB afterglows,coupled with off-axis afterglow models. We also provide a re-alistic assessment of the required observing time and achiev-able depths in the optical and radio bands. In Section 4 weconsider isotropic optical transients powered by the radioac-tive decay of heavy elements synthesized in the ejecta (referredto here as “kilonovae,” since their peak luminosities are pre-dicted to be roughly one thousand times brighter than thoseof standard novae). In Section 5 we compare and contrast thepotential counterparts in the context of our four Cardinal Virtues.

2

Metzger & Berger 2012, ApJ, 746, 48

Expected emission(Li and Paczynski 98)

No. 1, 1998 TRANSIENT EVENTS FROM NEUTRON STAR MERGERS L61

Fig. 2.—Time variation of the bolometric luminosity of the expanding spheregenerated by a neutron star merger, shown for models with a large mix ofradioactive nuclides, which provide a heating rate inversely proportional totime from the beginning of expansion. The models have three values of thefraction of rest mass released as heat: , 1024, 1025, the mass23f 5 10 M 5

, and the surface expansion velocity cm s21. For the adopted22 1010 M V 5 10,

opacity cm2 g21, we have days, as indicated5k 5 0.2 t 5 0.975# 10 s 5 1.13c

with filled circles, separating the solid lines (corresponding to the opticallythick case) and the dashed lines (corresponding to the optically thin case).

Fig. 3.—Evolution of the model from Fig. 2, with the fraction of rest-massenergy released in radioactive decay . The evolution begins at a very23f 5 10small radius, with a very high Teff and a very low . The open circles correspondLto the time of 0.1, 0.2, . . ., 1.1 days, and the vertical bar corresponds to thetime days, when the model becomes optically thin.5t 5 0.975# 10 s 5 1.13c

days), and , 102, 10, 1, 1021, and 1022, re-3º 1.13 t /t 5 10c radspectively. The initial condition is taken to be T(t 5

K (but the results are insensitive to the101 ms) 5 2.8# 10initial conditions). The light curves are extended to the opticallythin region by .L º eM

3.2. Power-Law Decay

The exponential-decay model was useful in demonstratingthat very short and very long timescale radioactivity is inef-ficient for generating a large luminosity. The most efficientconversion of nuclear energy to the observable luminosity isprovided by the elements with a decay timescale trad comparableto tc, when the expanding sphere becomes optically thin. Inreality, there is likely to be a large number of nuclides with avery broad range of decay timescales. Therefore, it is morerealistic to adopt a power-law decay model, which automati-cally selects the most efficient radioactive timescales.In the case of power-law decay with e(t) given by equation

(5), the analytic solution of equation (9) is

C 8b 1 3U 5 U 1 Y t , (10)Œ Œ21 ( )[ ]4 3t /8b 4t e 3 t 8b

where , , , and the in-2 3 3U 5 fMc / [(4p/3) V t ] t 5 t/t b 5 V/c1 c c

tegration constant C is determined by the initial conditions.Dawson’s integral Y is . In practice, the2 2x2x sY(x) { e e ds∫0solution is insensitive to the initial conditions. With this ap-proximation, we have . The corresponding luminosity isC º 0

8b 3L º L Y t , (11)Œ Œ0 ( )3 8b

where .2L 5 3fMc /(4bt )0 c

The time from the beginning of expansion to the peak lu-

minosity is

1/2t º 1.5b tm c

1/2 21/2 1/2M 3V k5 0.98 days . (12)( ) ( ) ( )0.01 M c k, e

The peak luminosity is

1/2 44 21L º 0.88b L 5 2.1# 10 ergs sm 0

1/2 1/2 21/2f M 3V k# . (13)( ) ( ) ( ) ( )0.001 0.01 M c k, e

The effective temperature at the peak luminosity is

21/8 4T º 0.79b T 5 2.5# 10 Keff, m 1

1/4 21/8 21/8 23/8f M 3V k# , (14)( ) ( ) ( ) ( )0.001 0.01 M c k, e

where . From equation (11), the time at which1/4T 5 (U /a)1 1the luminosity is down a factor of 3 from the peak luminosityis , which is in the optically thin regime for the1/2t º 4.9b tcsubrelativistic case. For the optically thin case, the luminosityof the nonthermal radiation from the expanding shell is roughlygiven by , which is the asymptotic form21L º eM º (4/3) L bt0of equation (11) for . Thus, equation (11) can be for-1/2t k bmally extended to the optically thin region, where the radiationbecomes nonthermal. The time from the peak luminosity to

No. 1, 1998 TRANSIENT EVENTS FROM NEUTRON STAR MERGERS L61

Fig. 2.—Time variation of the bolometric luminosity of the expanding spheregenerated by a neutron star merger, shown for models with a large mix ofradioactive nuclides, which provide a heating rate inversely proportional totime from the beginning of expansion. The models have three values of thefraction of rest mass released as heat: , 1024, 1025, the mass23f 5 10 M 5

, and the surface expansion velocity cm s21. For the adopted22 1010 M V 5 10,

opacity cm2 g21, we have days, as indicated5k 5 0.2 t 5 0.975# 10 s 5 1.13c

with filled circles, separating the solid lines (corresponding to the opticallythick case) and the dashed lines (corresponding to the optically thin case).

Fig. 3.—Evolution of the model from Fig. 2, with the fraction of rest-massenergy released in radioactive decay . The evolution begins at a very23f 5 10small radius, with a very high Teff and a very low . The open circles correspondLto the time of 0.1, 0.2, . . ., 1.1 days, and the vertical bar corresponds to thetime days, when the model becomes optically thin.5t 5 0.975# 10 s 5 1.13c

days), and , 102, 10, 1, 1021, and 1022, re-3º 1.13 t /t 5 10c radspectively. The initial condition is taken to be T(t 5

K (but the results are insensitive to the101 ms) 5 2.8# 10initial conditions). The light curves are extended to the opticallythin region by .L º eM

3.2. Power-Law Decay

The exponential-decay model was useful in demonstratingthat very short and very long timescale radioactivity is inef-ficient for generating a large luminosity. The most efficientconversion of nuclear energy to the observable luminosity isprovided by the elements with a decay timescale trad comparableto tc, when the expanding sphere becomes optically thin. Inreality, there is likely to be a large number of nuclides with avery broad range of decay timescales. Therefore, it is morerealistic to adopt a power-law decay model, which automati-cally selects the most efficient radioactive timescales.In the case of power-law decay with e(t) given by equation

(5), the analytic solution of equation (9) is

C 8b 1 3U 5 U 1 Y t , (10)Œ Œ21 ( )[ ]4 3t /8b 4t e 3 t 8b

where , , , and the in-2 3 3U 5 fMc / [(4p/3) V t ] t 5 t/t b 5 V/c1 c c

tegration constant C is determined by the initial conditions.Dawson’s integral Y is . In practice, the2 2x2x sY(x) { e e ds∫0solution is insensitive to the initial conditions. With this ap-proximation, we have . The corresponding luminosity isC º 0

8b 3L º L Y t , (11)Œ Œ0 ( )3 8b

where .2L 5 3fMc /(4bt )0 c

The time from the beginning of expansion to the peak lu-

minosity is

1/2t º 1.5b tm c

1/2 21/2 1/2M 3V k5 0.98 days . (12)( ) ( ) ( )0.01 M c k, e

The peak luminosity is

1/2 44 21L º 0.88b L 5 2.1# 10 ergs sm 0

1/2 1/2 21/2f M 3V k# . (13)( ) ( ) ( ) ( )0.001 0.01 M c k, e

The effective temperature at the peak luminosity is

21/8 4T º 0.79b T 5 2.5# 10 Keff, m 1

1/4 21/8 21/8 23/8f M 3V k# , (14)( ) ( ) ( ) ( )0.001 0.01 M c k, e

where . From equation (11), the time at which1/4T 5 (U /a)1 1the luminosity is down a factor of 3 from the peak luminosityis , which is in the optically thin regime for the1/2t º 4.9b tcsubrelativistic case. For the optically thin case, the luminosityof the nonthermal radiation from the expanding shell is roughlygiven by , which is the asymptotic form21L º eM º (4/3) L bt0of equation (11) for . Thus, equation (11) can be for-1/2t k bmally extended to the optically thin region, where the radiationbecomes nonthermal. The time from the peak luminosity to

Timescale

Luminosity

(e.g., Metzger+10, Roberts+11, Goriely+11, Metzger+12, Korobkin+12, Piran+13)

κ ~ 0.1 - 0.2 cm2 g-1

Good for Fe-rich supernova (Pinto & Eastman 2000)

What about r-process elements in NS merger??

Hotokezaka et al. 2013

Numerical relativitysimulation

3D Monte-Carloradiative transfer

* No nucleosynthesis calculation(assume solar abundance ratio)

opacity κ ~ 10 cm2 g-1

Radiative Transfer Simulations for NS Meger Ejecta 7

0.001

0.01

0.1

1

10

100

1000

5000 10000 15000

κ (c

m2 g

-1)

Wavelength (A)

NSM-allNSM-Fe

Fig. 3.— Mass absorption coefficient κ at v = 0.1c in model NSM-all and NSM-Fe as a function of wavelength (t = 3 days after themerger). In r-process element-rich ejecta, the opacity is higher than Fe-rich ejecta by factor of about 100 around the center of opticalwavelength (∼ 5000 A).

1040

1041

1042

1 10

UVO

IR L

umin

osity

(erg

s-1

)

Days after the merger

NSM-allNSM-tidalNSM-windNSM-Fe

Fig. 4.— Bolometric light curves for simple models with differentelement abundances: NSM-all (31 ≤ Z ≤ 92), NSM-tidal (55 ≤Z ≤ 92), NSM-wind (31 ≤ Z ≤ 54), and NSM-Fe (only Fe).

Figure 7 shows the bolometric light curves of realisticmodels. The models with the soft EOS APR4 (red) isbrighter than the models with the stiff EOS H4 (blue).This is interpreted as follows. For a softer EOS (i.e.,smaller radius of NS), the tidal mass ejection occurs at amore compact orbit. As a result, the mass of the ejecta ishigher for a softer EOSs (Hotokezaka et al. 2013). Sincethe heating rate is likely to be higher for a higher ejectamass (Metzger et al. 2010), the NS merger with the softEOS APR4 is brighter.

For the soft EOS APR4, the brightness does not de-pend strongly on the mass ratio of the binary NSs (redsolid and dashed lines in Figure 7). This is because fora soft EOS, such as APR4, the mass ejection is efficientenough. In contrast, for the stiff H4, the mass ejection ismore efficient for a higher mass ratio (Hotokezaka et al.2013). As a result, model H4-1215 (mass ratio of 1.25)is brighter than model H4-1314 (mass ratio of 1.08).

-18

-16

-14

-12

-10

-8

-6 2 4 6 8 10 12 14 16 18 20

Abso

lute

mag

nitu

de


UBVRI J HK Bol NSM-all

Fig. 5.— Multi-color light curves of model NSM-all (in Vegamagnitude). Light curves in redder bands are brighter and slower.

These results open a new window to study the natureof the NS merger and EOSs. By the combined analysisof GW signals and EM radiation, we may be able to pindown the masses of two NSs that merged and/or stiff-ness of the EOSs. Note that, in the current simulations,the heating rate per mass is fixed. To fully understandthe connection between the initial condition of the NSmerger and expected emission, detailed nucleosynthesiscalculations are necessary.

6. IMPLICATIONS FOR OBSERVATIONS

6.1. Follow-up Observations of EM Counterparts

In this section, we discuss the detectability of UVOIRemission from NS merger ejecta. Figure 8 shows expectedobserved light curves for a NS merger event at 200 Mpc.Model NSM-all (black) and 4 realistic models (red andblue) are shown. Note that all the magnitudes in Figure8 are given in AB magnitude for the ease of comparison

r-processFe

(consistent with Kasen+13, Barnes+13)

MT+13 (to be submitted)

6 Tanaka et al.

TABLE 1Summary of Models

Model Mej EK vch Abundance1

(M!) (erg)

NSM-all 1.0 × 10−2 1.3 × 1050 0.12c 31 ≤ Z ≤ 92NSM-tidal 1.0 × 10−2 1.3 × 1050 0.12c 55 ≤ Z ≤ 92NSM-wind 1.0 × 10−2 1.3 × 1050 0.12c 31 ≤ Z ≤ 54NSM-Fe 1.0 × 10−2 1.3 × 1050 0.12c Z = 26 (only Fe)

APR4-12152 8.6 × 10−3 4.3 × 1050 0.24c 31 ≤ Z ≤ 92APR4-13142 8.1 × 10−3 3.6 × 1050 0.22c 31 ≤ Z ≤ 92

H4-12152 3.5 × 10−3 1.4 × 1050 0.21c 31 ≤ Z ≤ 92H4-13142 7.0 × 10−4 1.9 × 1049 0.17c 31 ≤ Z ≤ 92

Note. — 1 Solar abundance ratios (Simmerer et al. 2004) areassumed. 2 Models from Hotokezaka et al. 2013.

1040

1041

1042

1 10

UVO

IR L

umin

osity

(erg

s-1

)


NSM-allκ=0.1 cm2 g-1

κ=1.0 cm2 g-1

κ=10 cm2 g-1

Fig. 2.— Bolometric light curve of model NSM-all (black, multi-frequency simulations). It is compared with the light curves for thesame model but with the gray approximation of UVOIR transfer(κ =0.1, 1, and 10 cm2 g−1 for blue, purple, and red lines, respec-tively). The result of multi-frequency transfer are most similar tothat of gray transfer with κ = 10 cm2 g−1.

Figure 3 shows the mass absorption coefficient as afunction of wavelength at t = 3 days in model NSM-allat v = 0.1c. The mass absorption coefficient is as high as1-100 cm2 g−1 in the optical wavelengths. The resultingPlanck mean mass absorption coefficient is about κ =10 cm2 g−1 (Figure A4). This is the reason why thebolometric light curve of multi-frequency transfer mostclosely follows that with gray κ = 10 cm2 g−1 in Figure2.

The high opacity in r-process element-rich ejecta is alsoconfirmed by the comparison with other simple models.Figure 4 shows the comparison of the bolometric lightcurve with that of models NSM-tidal, NSM-wind, andNSM-Fe. Compared with NSM-Fe, the other modelsshow the fainter light curves. This indicates that theelements heavier than Fe contributes to the high opacity.The opacity in model NSM-Fe is also shown in Figure 3.It is nicely shown that opacity in NSM-all is higher thanthat in NSM-Fe by a factor of about 100 at the center ofoptical wavelengths (∼ 5000 A).

As inferred from Figure 4, NSM-tidal (55 ≤ Z ≤ 92)has a higher opacity than that of NSM-wind (31 ≤ Z ≤54). This is because lanthanoid elements (57 ≤ Z ≤71) have the largest contribution to the bound-boundopacity, as demonstrated by Kasen et al. (2013). Notethat, however, even with the elements at 31 ≤ Z ≤ 54,the opacity is higher than Fe.

Figure 5 shows the multi-color light curves of modelNSM-all. In general, the emission from NS merger ejectais red because of (1) a lower temperature than SNe and(2) a higher optical opacity than in SNe. Especially, theoptical light curves in the blue wavelengths drops dra-matically in the first 5 days. The light curves in theredder band evolves more slowly. This trend is also con-sistent with the results by Kasen et al. (2013); Barnes &Kasen (2013).

Since our simulations include all the r-process ele-ments, spectral features are of interest. Since the sim-ulations by Kasen et al. (2013); Barnes & Kasen (2013)include only a few lanthanoid elements, they do not dis-cuss the detailed spectral features. Figure 6 shows thespectra of model NSM-all at t = 1.5, 5.0 and 10.0 daysafter the merger. Our spectra are almost featureless atall the epochs. This is because of the overlap of manybound-bound transitions of different r-process elements.As a result, compared with the results by Kasen et al.(2013); Barnes & Kasen (2013), the spectral features aremore smeared out.

Note that we could identify possible broad absorptionfeatures around 1.4 µm (in the spectrum at t = 5 days)and around 1.2 µm and 1.5 µm (t = 10 days). In ourline list, these bumps are mostly made by a cluster ofthe transitions of Y I, Y II, and Lu I. However, we arecautious about such identifications because the bound-bound transitions in the VALD database are not likely tobe complete in the NIR wavelengths even for neutral andsingly ionized ions. In fact, Kasen et al. (2013) showedthat the opacity of Ce from the VALD database drops inthe NIR wavelengths, compared with the opacity basedon their atomic models. Although we cannot exclude apossibility that a cluster of bound-bound transitions ofsome ions can make a clear absorption line in NS merg-ers, our current simulations do not provide prediction forsuch features.

5. DEPENDENCE ON THE EOS AND MASS RATIO

Full simulation

gray opacity(cm2 g-1)κ = 0.1κ = 1κ = 10

Fainter and longer


Observing Strategy


20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


u band200 Mpc

NSM-allAPR4 (soft)H4 (stiff)

4m8m

1.2 + 1.51.3 + 1.4

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


g band200 Mpc

1m

4m

8m

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


r band200 Mpc

1m

4m

8m

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


i band200 Mpc

1m

4m

8m

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


z band200 Mpc1m

4m

8m

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


J band200 Mpc

4m

space

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


H band200 Mpc

4m

space

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


K band200 Mpc

4m

space

Fig. 8.— Expected observed ugrizJHK-band light curves (in AB magnitude) for model NSM-all and 4 realistic models. The distanceto the NS merger event is set to be 200 Mpc. K correction is taken into account with z = 0.05. Horizontal lines show typical limitingmagnitudes for wide-field telescopes (5σ with 10 min exposure). For optical wavelengths (ugriz bands), “1 m”, “4 m”, and “8 m” limitsare taken or deduced from those of PTF (Law et al. 2009), CFHT/Megacam, and Subaru/HSC (Miyazaki et al. 2006), respectively. ForNIR wavelengths (JHK bands), “4 m” and “space” limits are taken or deduced from those of Vista/VIRCAM and the planned limits ofWFIRST (Green et al. 2012) and WISH (Yamada et al. 2012), respectively.

1m

4m

8m

opticali band

At 200 Mpc

10 min5 sigma

5 days

soft EOS

stiff


ROTSE

SNLS

GOODS

0.01

0.1

1

10

100

1000

10000

18 19 20 21 22 23 24 25 26 27

Surv

ey a

rea

(deg

2 )

Limiting magnitude

< 1 d1-5 d> 5 d

LOSSCHASE

CRTS

(@6000A)

KISS

HSC

PS (deep)

PTF

PTFSkyMapper

PS (wide)

SDSSLa Silla

LSST

Optical transient

search

GW detection

EM follow up

“EM-triggered” search

“GW-triggered” search

EM follow up





Observing Strategy


20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


u band200 Mpc

NSM-allAPR4 (soft)H4 (stiff)

4m8m

1.2 + 1.51.3 + 1.4

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


g band200 Mpc

1m

4m

8m

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


r band200 Mpc

1m

4m

8m

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


i band200 Mpc

1m

4m

8m

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


z band200 Mpc1m

4m

8m

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


J band200 Mpc

4m

space

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


H band200 Mpc

4m

space

20

21

22

23

24

25

26

27 0 5 10 15 20

Obs

erve

d m

agni

tude


K band200 Mpc

4m

space

Fig. 8.— Expected observed ugrizJHK-band light curves (in AB magnitude) for model NSM-all and 4 realistic models. The distanceto the NS merger event is set to be 200 Mpc. K correction is taken into account with z = 0.05. Horizontal lines show typical limitingmagnitudes for wide-field telescopes (5σ with 10 min exposure). For optical wavelengths (ugriz bands), “1 m”, “4 m”, and “8 m” limitsare taken or deduced from those of PTF (Law et al. 2009), CFHT/Megacam, and Subaru/HSC (Miyazaki et al. 2006), respectively. ForNIR wavelengths (JHK bands), “4 m” and “space” limits are taken or deduced from those of Vista/VIRCAM and the planned limits ofWFIRST (Green et al. 2012) and WISH (Yamada et al. 2012), respectively.

1m

4m

8m

opticali band

At 200 Mpc

10 min5 sigma

5 days

soft EOS

stiff


179 186 193 201 208 215 223 230 237 244 251

2 deg

~ 200 galaxies / 4deg2 (< 200 Mpc)

Kiso 1m Schmidt telescope

2 deg

GW alert error boxe.g. 10 deg x 10 deg

Typical 8-10m telescope

0.3 deg179 186 193 201 208 215 223 230 237 244 251

~ 5000 galaxies (< 200 Mpc)

PTF 1m

Pan STARRS 1.8m

3.0 deg

Depth

“GW-triggered” Follow up

Area(field of view)

Need “survey” observations/telescopes for follow up

Response

0.001

0.01

0.1

1

10

100

18 19 20 21 22 23 24 25 26 27

Fiel

d of

vie

w (d

eg2 )

Limiting magnitude

~ 8-10m~ 4m

~ 1-2m

KISS

HST

La SillaCRTS

ZTF

typical8-10m

Subaru

LSST

CFHT

PTFPS DES

SDSS

typical4m

2 deg

GW alert error boxe.g. 10 deg x 10 deg

Typical 8-10m telescope

0.3 deg179 186 193 201 208 215 223 230 237 244 251

~ 5000 galaxies (< 200 Mpc)

8m SubaruHyper Suprime-Cam

1.5 degPTF 1m

8m LSST

3.5 deg

Pan STARRS 1.8m

3.0 deg



• Strategy: Area * Depth * Cadence

• Counterpart of GW sources

• Fainter/longer than previously expected


• Need “survey” for follow-up observations

• Wide field 4m-class telescopes

• Best with wide field 8-10m telescopes=> Subaru/Hyper Suprime-Cam and LSST

0.001

0.01

0.1

1

10

100

18 19 20 21 22 23 24 25 26 27

Fiel

d of

vie

w (d

eg2 )

Limiting magnitude

~ 8-10m~ 4m

~ 1-2m

KISS

HST

La SillaCRTS

ZTF

typical8-10m

Subaru

LSST

CFHT

PTFPS DES

SDSS

typical4m

Documents

Masaomi Tanaka...Optical Transient Search Survey Diameter (m) FOV (deg2) Depth (R mag) Area/ day(deg2) LOSS 0.76 0.01 19 galaxy CHASE 0.4 x 4 0.03 19 galaxy ROTSE-III 0.45 3.42 18.5