ProbingwithTMT the (more 時代のFirst than) Stars,GRBsGRB a forUVlittleで切り拓く超遠方宇宙� UNravelling Radiation, help from the B my DarkFields, friends Ages Dust… Mission� GeV Fermi ガンマ線バーストと遠方宇宙・元素の起源� 井上 進�（理研） ＋共同研究者の皆様 GRBs, blazars
R. Salvaterra, T. R. Choudhury, A. Ferrara, B. Ciardi, R. Schneider 分子 ALMA (MNRAS in press, arXiv:0906.2495) Y. Inoue, M. Kobayashi, T. Totani, Y. Niinou (in prep.)
z~30-6 ダスト ASTRO-H
blazar HiZ-G GRB
SPICA
再電離 金属 TMT submm R.ALMA Salvaterra,5@5 T.epoch R. Choudhury, of first starA. Ferrara, formation Inoue, （磁場）� Omukai & Ciardi, Benedetta MNRAS, Ciardi Firstsubmitted (MPA)Radiation (astro-ph/0502218) Fields, FirstB. MagneticCiardi, R. Schneider Fields,lots of First theory Star Formation SKA K. Ichiki, K. KiyotomoTakahashi Ichiki (CfCP/NAOJ) but almost no observations!!! K. Omukai Keitaro Takahashi (Princeton) IceCube
SKA CTA CTA “in an interstella burst i am back to reveal the universe” radio see - after T. Yorke also かなた outline�
I. GRBs and the high-z Universe - overview: early excitement vs recent calming down - quasar contribution to reionization - gamma-ray probe of UV background *star formation rate, metal evolution, damping wings… -> later talks recent reviews: Space Sci. Rev. (2016) 202� II. neutron star mergers and r-process nucleosynthesis - potential of X-ray diagnostics cosmic dark ages -> cosmic dawn� - big bang� z~3600� - matter-rad. eq.� z~1000� - recombination�
dark ages�
z>~30� - first stars, galaxies, black holes… -> reionization z~8-10�
z~6� - reion. ends
- star formation � z~1-3 peak�
z~0� adapted from Djorgovski� GRBs and high-z Universe� 2000’s: excitement, hope GRBs� high-z� 1967 discovery (670702) … 1997 afterglow, cosmological z (970508)
2000 potential recognition ~2000 first star simulations Lamb & Reichart 00 2001 quasar GP troughs Bromm & Loeb 01…� 2003 massive star (030329) 2003 CMB optical depth zreion~17?? 2005 z=6.3 (050904)
2009 z=8.2 (090423) high-z GRB mania!� record redshifts (spectroscopic): -2009
(GRB 090429 z~9.4)� GRB 090423 z=8.2�
©山崎� record redshifts (spectroscopic): -2017
(GRB 090429 z~9.4)� GRB 090423 z=8.2�
©山崎�
2015� 2020� record redshifts (spectroscopic): -2017 quasars strike back?�
(GRB 090429 z~9.4)� GRB 090423 z=8.2� ULAS J1120+0641 z=7.085�
©山崎�
2015� 2020� record redshifts (spectroscopic): -2017 galaxies strike back!�
GN-z11 z=11.09 EGSY8p7 z=8.68� (GRB 090429 z~9.4)� GRB 090423 z=8.2� ULAS J1120+0641 z=7.085�
©山崎�
2015� 2020� The Astrophysical Journal, 819:129current(11pp), 2016 March 10highest-z object: galaxy atOesch z=11.09 et al. Oesch+ 16�
The Astrophysical Journal, 819:129 (11pp), 2016 March 10 Oesch et al.
Table 3 Assumed LFs for z ∼ 10–11 Number Density Estimates
-5 Reference f * 10 M* α Nexp (Mpc−3)(mag)(<−22.1) Figure 5. Grism Spectrum of GN-z11. The top panel shows the (negative) 2D spectrum from the stack of our cycle 22 data (12 orbits) with the trace outlined by the Bouwensdark et red al. lines.(2015b For clarity) the 2D spectrum1.65 was smoothed by− a20.97 Gaussian indicated− by2.38 the ellipse in the 0.06 lower right corner. The bottom panel is the un-smoothed 1D flux density using an optimal extraction rebinned to one resolution element of the G141 grism (93Å). The black dots show the same further binned to 560Å, while the Finkelsteinblue line et shows al. ( the2015 contamination) level0.96 that was subtracted− from20.55 the original object−2.90 spectrum. We identify 0.002 a continuum break in the spectrum at λ=1.47±0.01 μm. The continuum flux at λ>1.47 μm is detected at ∼1–1.5σ per resolution element and at 3.8σ per 560Å bin. After excluding lower redshift solutions (see the text and Mashian et al. (2015fi ) 0.25 +0.08 −21.20 fl −2.20 0.03 fl Figure 4), the best- t grism redshift is zgrism = 11.09-0.12. The red line re ects the Lyα break at this redshift, normalized to the measured H-band ux of GN-z11. The Mason etagreement al. (2015 is excellent.) The fact that we0.30 only detect significant−21.05flux along the trace−2.61 of our target source, 0.01 which is also consistent with the measured H-band magnitude, is strong evidence that we have indeed detected the continuum of GN-z11 rather than any residual contamination. Trac et al. (2015) 5.00 −20.18 −2.22 0.002 this flux detection is ∼1–1.5σ per resolution element longward 3.1. The Best-fit Solution: A z∼11 Galaxy of 1.47 μm and consistent with zero flux shortward of that. The Note. The parameters f*, M*, and α represent the three parametersBased onour of previous the photometric redshift measurement for total spectral flux averaged over 1.47–1.65 μm represents a GN-z11 (z = 10.2), we expected to detect a continuum Schechterclear UV 5.5σ LFdetection. taken This from is fullythe different consistent with papers. the prediction phot break at 1.36±0.05 μm. This can clearly be ruled out. from the exposure time calculator for an H=26 mag source in However, the grism data are consistent with an even higher a 12-orbit exposure. The extracted 1D spectrum along the trace of GN-z11 is shown redshift solution. Interpreting the observed break as the 1216Å in Figures 4 and 5.Thesehighlightthedetectionofacontinuum break, which is expected for high-redshift galaxies based on absorption from the neutral inter-galactic hydrogen along the when assumingbreak with a flux a ratio more of ff()()ll realistic<><1.47 case 1.47 where 0.32 at the emission line line of sight, we obtain a very good fit to the spectrum with a fl 2σ when averaged over 560 Å wide spectral bins. 2 Figure 7. Redshift and UV luminosities of known high-redshift galaxies from ux isfl distributed over a combination ofreduced linesχ =(e.g.,1.2 (see Figures 4 and 5). The best-fit redshift is A ux decrement is seen around ∼1.6 μm, which is caused +0.08 blank field surveys. Dark filled squares correspond to spectroscopically fi zgrism = 11.09 , corresponding to a cosmic time of only Hβ+by[O negativeIII]),fl weux values can in conone of ourdently two visits invalidate slightly above such a solution.-0.12 confirmed sources, while small gray dots are photometric redshifts (Bouwens the peak of the trace of GN-z11. However, this dip is consistent ∼400 Myr after the Big Bang. The grism redshift and its fi The lowerwith Gaussian left panel deviates in from Figure the noise4 model.compares We also tested the measureduncertainty grism are derived from anet MCMC al. 2015bt to). the GN-z11 2D spectrum clearly stands out as one of the most luminous currently spectrumthat the with detected that continuum expectedflux is still for present the when best- adoptingfit lowerwhich also redshift includes the morphologicalknown galaxies information at all of redshifts the z>6 and is by far the most distantmeasured different stacking and extraction procedures (see Appendix A). source as well as the photometry,galaxy adopting with identical spectroscopy techniques(black squares; see Oesch et al. 2015b, for a full list solution we had previouslyfi identified (Oesch etas al. used2014 for) the.A 3D-HST survey redshifts (see Momcheva In particular, we con rmed the spectral break in a simple of references). Wider area surveys with future near-infrared telescopes (such as strongmedian line stack emitter of the individual SED is continuum-subtracted clearly inconsistent 2D grism withet al. 2015 the). data. observations of our six individual visits. Additionally, we This high-redshift solutionWFIRST also reproduces) will be the required expected to determine how common such luminous sources Apartcon fromfirmed the that the emission continuum break lines, is seen which only along we the trace do notcount detect, rate based this on an H=26really mag continuum are at z source,>10. as well modelof also our spectrum predicts by creating weak simple continuum 1D-extractions aboveflux and acrossas the the overall whole morphology of the 2D grism spectrum. This is below the trace of our target source. Finally, we confirmed that demonstrated in Figure 3, where we show the original data, the wavelength range. At <1.47 μm, this is higherneighbor than subtracted the 2D spectrum as well as the residual after short long a break is seen in both epochs separately (see Figure 10). the observed spectrum has a break of (1->ff) 0.68 observedIn mean, the following while sections, at > we1.47 discussμm the thepossible expected redshift flsubtractingux is too out low the z=11.09 model with the correct H-band nn solutions for GN-z11 based on the combined constraints of the magnitude. Note the drop of theatfl 2uxσ longward, thus of indicating 1.65 μm due again that we can marginally rule out a comparednew grism to the continuum observations. detection as wellOverall as the the pre-existing likelihoodto the of reduced a z∼ sensitivity2 of4000 the grism.Å Thisbreak is an important based on the spectrum alone without even extremephotometry. emission line SED based on our grism dataconstraint, is less because than it shows that the detected flux originates −6 including the photometric constraints. 10 and can be ruled out. 5 Note that in very similar grism observations for a source triply imaged by a CLASH foreground cluster, emission line 4. DISCUSSION contamination could also be excluded (Pirzkal et al. 2015). We 4.1. Physical Properties of GN-z11 thus have no indication currently that any of the recent z∼9–11 galaxy candidates identified with HST is a lower Despite being the most distant known galaxy, GN-z11 is redshift strong emission line contaminant (but see, e.g., relatively bright and reliably detected in both IRAC 3.6 and Brammer et al. 2013, for a possible z∼12 candidate). 4.5 μm bands from the S-CANDELS survey (Ashby et al. 2015). This provides a sampling of its rest-frame UV SED and even partially covers the rest-frame optical wave- 3.3. Excluding a Lower-redshift Dusty or Quiescent Galaxy lengths in the IRAC 4.5 μm band (see Figure 6). The photometry of GN-z11 is consistent with a SED of Another potential source of contamination for very high logMM ~ 9 using standard templates (Bruzual & Char- redshift galaxy samples are dusty z∼2–3 sources with strong lot 2003, see appendix). The UV continuum is relatively blue 4000 Å or Balmer breaks (Hayes et al. 2012; Oesch et al. with a UV spectral slope β=−2.5±0.2 as derived from a 2012). However, the fact that the IRAC data for GN-z11 show powerlaw fit to the H160, K, and [3.6] fluxes only, indicating that it has a very blue continuum longward of 1.6 μm, together very little dust extinction (see also Wilkins et al. 2016). with the very red color in the WFC3/IR photometry, already Together with the absence of a strong Balmer break, this is rules out such a solution (see SED plot in Figure 4). consistent with a young stellar age of this galaxy. The best fit Nevertheless, we additionally explore what constraints the age is only 40 Myr (<110 Myr at 1σ). GN-z11 thus formed its grism spectrum alone can set on such a solution. stars relatively rapidly. The inferred star formation rate is 24 ± The expected flux for such a red galaxy increases gradually −1 10 Me yr . All the inferred physical parameters for GN-z11 across the wavelength range covered by the G141 grism, unlike are summarized in Table 2. Overall, our results show that what we observe in the data (lower right in Figure 4). galaxy build-up was well underway at 400 Myr after the fi ∼ Compared to our best- t solution (see next section) we measure Big Bang. a Δχ2=15 when comparing the data with the expected grism flux. Apart from the extremely large discrepancy with the IRAC photometry, we can thus exclude this solution at 98.9% 4.2. The Number Density of Very Bright z > 10 Galaxies confidence based on the spectrum alone. The spectrum of GN-z11 indicates that its continuum break Similar conclusions can be drawn from the break strength lieswithin the H160 filter (which covers ∼1.4–1.7 μm; see Figure 6). alone (see e.g., Spinrad et al. 1998). Assuming that the The rest-frame UV continuum flux of this galaxy is therefore observed break at 1.47 μm corresponds to 4000 Å at z 2.7, a = ∼0.4 mag brighter than inferred from the H160 magnitude. The galaxy with a maximally old SED (single burst at z=15) estimated absolute magnitude is M =−22.1±0.2, which is fl short long UV would show a ux ratio of (1- 7 No. 1, 2003 WMAP FIRST-YEAR RESULTS: PARAMETERS 183 reionization z from CMB polarization anisotropy Planck Collaboration: Planck constraints on reionization history Planck 16 zr=8.8+-0.9 Spergel+ 03 zr=17+-5� Fig. 18. Constraints on ionization fraction during reionization. The allowed models, in terms of zre and z, translate into an allowed Fig. 7.—Left:Fractionofmassinboundobjectsasafunctionofredshift.Theblacklinesshowthemassincollapsedobjectswithmassgreaterthan2003 zr=17+-5 WMAP1 Spergel+ 03 MHRL z ,theeffective Jeans mass in the absence of H cooling for our best-fit PL ÃCDM model (thin lines are for the fit to WMAP only, andregion thick in linesxe( arez) (68 % and 95 % in dark blue and light blue, respectively), including the zend > 6 prior here. Left: Constraints from ð Þ 2 2006 z =10.9+2.7-2.3 WMAP3 Page+ 07 for the fit to all data sets). The heavy line uses the best-fit parameters based on all data (whichr has a lower 8), and the light line uses the best-fitCMB parameters data using a redshift-symmetric function (xe(z) as a hyperbolic tangent with z = 0.5). Centre: Constraints from CMB data based on fitting to the WMAP data only. The dashed lines show the mass in collapsed objects with masses greater than the Jeans mass assuming that the minimum mass is 106 M .Moreobjectsformiftheminimummassislower.Right:Ionizationfractionasafunctionofredshift.Thesolidlineshowsionization2008 zr=11.0+-1.4 WMAP5 Dunkley+using a redshift-asymmetric 09 parameterization (xe(z) as a power law). Right: Constraints from CMB data using a redshift-symmetric fraction for the best-fit PL ÃCDM model if we assume that H2 cooling is suppressed2010 by photodestruction z =10.6+-1.2 of H2.ThisfiguresuggeststhatH WMAP7 Komatsu+parameterization2 cooling may be11 with additional constraints from the kSZ e↵ect. necessary for enough objects to form early enough to be consistent with the WMAP detection. Ther heavy line is for the best-fit parameters for all data sets, and the light line is for the best-fit parameters for the WMAP only fit. The dashed lines assume that the objects with masses greater than 106 M can form stars. 2012 zr=10.3+-1.1 WMAP9 Hinshaw + 13 The gray band shows the 68% likelihood region for zr based on the assumption of instantaneous complete reionization (Kogut et al. 2003). 2014 zr=11.1+-1.1 Planck 14 function xe(z) with z = 0.5), a measurement of the Thomson for ⌧, and provide reionization redshifts which di↵er by less than 2016 z =8.8+-0.9 Planck 16 optical depth 0.4 . Constraints on the limits of possible early reionization are r = halo mass for star formation, Mmin, is controversial and 24%. We assume fburnd25%, fescd50%, fUVd50%, similar, leading to 10 % reionization levels at around z 10. depends on whether molecular hydrogen (H ) is available as f 90%, and 1 C 100, thus f 5:6 10 3. ⌧ = 0.058 0.012 (lollipop+PlanckTT), (24) 2 iond d clumpd netd  À ± To derive constraints on the duration of the reionization a coolant. If the gas temperature is fixed to the CMB tem- Figure 7 shows the fraction of collapsed objects and the epoch, we combined CMB data with measurements of the ampli- perature, then the Jeans mass, Mj 106 M .Ifmolecular maximum ionization fraction as a functionwhich of redshift is significantly more accurate than previous measure- ↵ hydrogen is available, then the Jeans¼ mass before reioniza- for our best-fit WMAP ÃCDM model. Thements. solid Thanks lines to the relatively high signal-to-noise ratio of the tude of the kSZ e ect. In the case of a redshift-symmetric model, 1:5 j0 3 HRL we found tion is M 2:2 10 wb=h wm 1 z =10 for z < 150 correspond to Mmin Mmin z ,whilethedashedlineslow-` polarization signal, the combination with lensing or data (Venkatesan et al. 2001).½ At zð > Þ150,ð theþ electronsÞ are ther- correspond to M ¼ Mj.Theð Þ WMAP detection of z < 2.8 (95 % CL), (27) min ¼ from high resolution CMB anisotropy experiments (ACT and mally coupled to the CMB photons. However, as Haiman reionization at high redshift suggests that H2 cooling likely et al. (1997) point out, a small UV background generated by played an important role in early star formation.SPT) does not bring much additional constraining power. The using the additional constraint that the Universe is entirely reion- ⇤ the first sources will dissociate H2, thus making the mini- Because early reionization requires the existenceimpact of on small- other CDM parameters is only significant for the ized at redshift 6 (i.e., zend > 6). mum mass much larger than the Jeans mass. They suggest scale fluctuations, the WMAP TE detectionamplitude has important of the initial scalar power spectrum As and (to a lesser Our final constraints on the reionization history are plot- HRL using a minimum mass that is much higher: Mmin z implications for our understanding of the natureextent) of the on dark its tilt ns. Other parameters are very stable compared 8 3=2 ð Þ¼ ted in Fig. 18 for each of the aforementioned cases, i.e., the 10 1 z =10 À .Ontheotherhand,ifthefirststars matter. to the Planck 2015 results. generated½ð þ Þ a significant flux of X-rays (Oh 2001), then this Barkana et al. (2001) note that the detection of reioniza- redshift-symmetric and redshift-asymmetric models, using only would have promoted molecular hydrogen formation tion at z > 10 rules out warm dark matter as a viableUsing candi-Planck data, we have derived constraints on two mod- the CMB, and the redshift-symmetric case using CMB+kSZ (all (Haiman et al. 2000; Venkatesan et al. 2001; Cen 2002), thus date for the missing mass as structure formsels very for the late reionization in history xe(z) that are commonly used in with prior zend > 6). Plotted this way, the constraints are not lowering the minimum mass back to Mj. these models. Warm dark matter cannot clusterthe literature: on scales a redshift-symmetric form using a hyperbolic tan- very tight and are still fairly model dependent. Given the low Following Tegmark & Silk (1995) we estimate the rate of smaller than the dark matter Jeans mass. Thus,gent transition this limit function; and a redshift-asymmetric form param- HRL value of ⌧ as measured now by Planck, the CMB is not able reionization by multiplying the collapse factor by an effi- applies regardless of whether the minimum masseterized is M by a power law. We have also investigated the e↵ect ciency factor. A fraction of baryons in the universe, f , falls or Mj. to give tight constraints on details of the reionization history. b of imposing the condition that the reionization is completed by into the nonlinear structures. We assume fb fDM (i.e., con- However, the Planck data suggest that an early onset of reion- stant baryon/dark matter ratio). A certain fraction¼ of these z = 6. ization is disfavoured. In particular, in all cases, we found that 5. COMBINING DATA SETS baryons form stars or quasars, fburn,whichemitUVradia- Allowing the ionization fraction shape and duration to vary, the Universe was less than 10 % ionized for redshift z > 10. tion with some efficiency, fUV. Some of this radiation In this section, we combine the WMAP datawe withhave other found very compatible best-fit estimates for the opti- Furthermore, comparisons with other tracers of the ionization escapes into the intergalactic medium photoionizing it; CMB experiments that probe smaller angularcal depth scales (0.059 and 0.060 for the symmetric and asymmetric however, the net number of ionizations per UV photons, 15 history show that our new result on the optical depth elimi- (ACBAR and CBI) and with astronomicalmodel, measurements respectively), showing that the CMB is indeed more sen- nates most of the tension between CMB-based analyses and fion, is expected to be less than unity (due to cooling and of the power spectrum (the 2dFGRS and Ly forest). We recombinations). Finally, the intergalactic medium might begin by exploring how including these data setssitive aff toects the our value of the optical depth than to the exact shape of constraints from other astrophysical data. Additional sources of be clumpy, making the photoionization process less effi- best-fit power-law ÃCDM model parametersthe ( reionization5.1). The history. However, the value of the reionization reionization, non-standard early galaxies, or significantly evolv- cient. This effect is counted for by the clumping factor addition of data sets that probe smaller scalesredshift systematicallyx does slightly depend on the model considered. In the ing escape fractions or clumping factors, are thus not needed. Cclump.Thusinthisapproximationtheionizationfractionis pulls down the amplitude of the fluctuations in the best-fit 5 case of a symmetric parameterization, we have found slightly Ongoing and future experiments like LOFAR, MWA, and given by xe 3:8 10 fnet fb, where fnet fburn fUVfesc fion= C .Thefactor3¼ Â:8 105 arises because¼ 7:3 10 3 of the larger estimates of zre than in the case of instantaneous reioniza- SKA, aimed at measuring the redshifted 21-cm signal from neu- clump   À rest mass is released in the burning of hydrogen to helium 15 In the following sections, we refer to the combined WMAPtion. This, ACBAR. can be understood through the shape of the degeneracy tral hydrogen during the EoR, should be able to probe reioniza- and we assume the primordial helium mass fraction to be and CBI data sets as WMAPext. surface between the reionization parameters. For an asymmetric tion directly and measure its redshift and duration to high ac- parameterization, zre is smaller, due to the fact that xe(z) changes curacy. Moreover, since reionization appears to happen at red- more rapidly at the end of reionization than the beginning. We shifts below 10, experiments measuring the global emission of specifically find: the 21-m line over the sky (e.g., EDGES, Bowman & Rogers 2010, LEDA, Greenhill & Bernardi 2012, DARE, Burns et al. z = 8.8 0.9 (redshift-symmetric) ; (25) re ± 2012), NenuFAR, Zarka et al. 2012, SARAS, Patra et al. 2013, z = 8.5 0.9 (redshift-asymmetric) . (26) re ± SCI-HI, Voytek et al. 2014, ZEBRA, Mahesh et al. 2014, and BIGHORNS, Sokolowski et al. 2015) will also be able to derive Assuming two di↵erent parameterizations of the reionization very competitive constraints on the models (e.g., Liu et al. 2015; history shows how much results on e↵ective parameters (like Fialkov & Loeb 2016). the redshift of reionization or its duration) are sensitive to the assumption of the reionization history shape. The best models of Acknowledgements. The Planck Collaboration acknowledges the support of: symmetric and asymmetric parameterization give similar values ESA; CNES, and CNRS/INSU-IN2P3-INP (France); ASI, CNR, and INAF 13 record redshifts (spectroscopic): -2017 + CMB zreion GN-z11 z=11.09 EGSY8p7 z=8.68� (GRB 090429 z~9.4)� GRB 090423 z=8.2� ULAS J1120+0641 z=7.085� Planck 2016 zreion=8.8+-0.9 reionization epoch: not as exciting? ©山崎� just around the corner!� 2015� 2020� z 10 UV LF from All Legacy HST Fields 11 ⇠ 8low abundance of z~10 galaxy Oesch et al. candidates? Oesch+ 1710.1113� significantly increases the search volume and the robust- Time [Gyr] ness of the SFRD10-2 measurement. 10-2 2 1.4 1 0.8 0.6 0.5 0.4 Comparison to Models and Simulations ] -1 4.4. The z>8 SFRD Trend and Dark Matter Halo 3 SFRD Model Comparison 10-3 10-3 ] Buildup ] -1.5 When compared-3 to the SFRD at z 8 our measured -3 -4 -4 - -2 z 10 valueMpc 10 lies almost exactly an order⇠ of magnitude Mpc 10 ⇠ -1 -1 > 0.3 M- /yr lower, given that the Schechter function normalization is -2.5 8 8 found to drop by-5 this amount.z~ Such a fast evolution of -5 z~ [mag 10 [mag 10 φ Mashian φ GALFORM -3 Mashian+16 the SFRD in only 170 Myr from z 8toz 10Mason may Meraxes ⇠z~10 ⇠ Mason+15 sound extreme.log However, it is important to note thatTrac the log Dayal -6 -6 -3.5 Liu+15 halo mass function10 also evolves extremely fast overBlueTides this 10 CoDa Data Sun Sun+16 Croc same redshift range. To illustrate this,Best-Fit we compute the Cai log SFR Density [M /yr/Mpc Behroozi+17 Behroozi cumulative number density of dark matter (DM) halos -4 10-7 9 10-7 as a function of redshift-22 using -21 the -20 code HMFcalc -19 -18(Mur- -17 -223 -214 5 -206 -197 8 -189 10 -17 11 12 13 M M Redshift ray et al. 2013). Our integration limitUV of the SFRD (0.3 UV M /yr) roughly corresponds to a dark matter halo mass Fig. 6.— The measured SFRD evolution compared model predic- Fig.10 3.— The z 10 UV LF results compared to models and simulations from the literature. The red line and points show our observed of LF,10 whileM theat bluez⇠ to green6 10 lines according are model LFs to (seeseveral text formod- references).tions. All modelsThe red predict dots are that the the same UV LFmeasured evolves values significantly as shown from in the els⇠ (e.g.,z 8(lightgrayline)toMason et al.⇠ 2015 z ;10.Mashian They typically et al. do2016 not; diSun↵er by & more thanprevious a factor figure of 3-4 corresponding over the luminosity to the range results of fromthe data.Bouwens While et al. ⇠ ⇠ (2016a)atz 4 8, Oesch et al. (2014)atz 9andthiswork Furlanettoseveral simulated2016). UV To LFs allow are in for good some agreement variation, with the we observations thus and generally within⇠ the 1 range, our observed UV⇠ LF lies at the lower end of essentially all theoretically predicted LFs. at z 10. The model SFRDs (integrated to the same limiting compute the cumulative number densities of halos down SFR)⇠ are shown in blue. They correspond to a subset of the same to logcomparedMh/M to=9 the.5 lowest and 10.5 of these as a earlierfunction estimates. of redshift. In (basedmodels on shown equation in the UV3). LF This plot. calculation Similar to results the UV in LF 28 result, We then normalize these densities to the measured SFRD we find that models generally overpredict the SFRD at z 10 a large part, this is due to the fact that the GOODS- galaxies,compared i.e., to the a factor observations.3 larger However, than mostour actual models sample do find⇠ the at z North8 to field compare appears the to evolutionary contain a significant trends, overdensity which is also of of only 9 sources, and⇠ can⇥ robustly be ruled out. In ⇠ SFRD to evolve very rapidly at these epochs and the higher values shownluminous in Fig. (M5.UV < 20 mag) galaxies, with three galax- particular,at z 10 could the partiallyMcLeod be et explained al. (2016 by) a LF di↵ predictserent normalization2 at z ⇠ 4 8. The open gray circles at lower redshift indicate⇠ previ- Theies evolution within a small of the projected cumulative area. halo Even mass though functions the re- galaxy⇠ candidates per HFF cluster field and 4.6 galaxies is inmaining excellent CANDELS-Wide agreement with fields the would fast have build-up reached of faint the inous the measurements six HFF parallel that used fields, di↵ meaningerent dust that correction in the factors HFF (see Bouwens et al. 2016a), which were used by most models to tune SFRDenough between to findz such10 sources, and z no reliable8, as is candidate evident could from dataset alone we should have found 16.7 z 10 galaxy their parameters. Nevertheless, the predicted SFRDs⇠ start to di- thebe figure. identified This in shows⇠ these that datasets, the⇠ observedresulting in fast the evolution lower in- candidates,verge significantly four times at z> as8suchthatitshouldbeveryeasytotest many as are present in the data. of theferred SFRD number should density. not have come as a surprise. It is andUsing rule our out somebest-fit of these LF, we models find with much early better JWST agreement observations. consistentOur measured with a model UV LF in which is in excellent the star-formation agreement with e - between the observed and predicted numbers. In partic- ciencythe is previous not varying analysis at by highBouwens redshift et (see al. (2015 also).Mashian This is ular, this LF predicts a total of 3.3 and 1.3 galaxies to very encouraging. Our analyses are completely indepen- beimaging found in data the analyzed six HFF clusters here span and more parallel than fields 800 com- arcmin2, et al.dent,2016 including; Stefanon the et simulations al. 2017a). of Orthe said selection another functions way, bined,and include respectively the HUDF/XDF, – in good agreement the HUDF09/12, with the thefour HFF a shallowerand the candidate evolutionary searches. trend However, would require this result a signifi- is also candidatedata, as well images as we all actually the five identified. CANDELS fields. In partic- cantnot change surprising in the given star-formation the fact that e weciency use a verywithin similar DM ular, we present a small sample of new z 10 galaxy halosapproach at high and redshift the fact – to that zeroth the datasets order. largelyRecent overlap, results candidates4.1.3. Comparison in the HFF to Predicted dataset UV (see LFs Section from Models⇠3.2), which fromwith clustering the exception measurements of the HFF also dataset provide which no compelling we newly areGiven self-consistently the small number folded of intoz the10 analysis galaxies while that we taking evidenceanalyze of here.such a As change can be (see seen,Harikane the HFF et candidates al. 2017). now identifiedspecial care in the of the combined position- HST and dataset,⇠ magnification-dependent it is interesting Inprovide Figure a6 measurement, we also compare of the UV the LF observed at MUV = SFRD19.25. to tocompleteness ask whether in this the might cluster be an fields indication due to of lensing. a reduced predictionsThis magnitude by some bin of previously the same did models not contain that were any galax- used star-formation e ciency in early halos at z>8. To this ies and corresponded to an upper limit. The main goal of this paper was to exploit all the for the UV LF comparison in section 4.1.3, and for which end,available we compareHST Legacy our observational fields in order results to with derive several the best- the SFRDsEven though at z> the8 have Schechter been function published. parameters In agreement of our theoretical models and simulations of the UV LF evo- best-fit model are very di↵erent from the ones quoted in possible UV LF measurement at z 10 before the advent with the UV LF results, most models predict a signifi- lution that have been published in⇠ the literature over Bouwens et al. (2015, see also Table 5), the UV LFs are in theof JWST last few. While years. it These is clear include that semi-empiricalthe UV LF and mod- the cos- cantlyvery higher good agreement SFRD at withz each10 than other we over observe. the luminosity Con- mic SFRD continue to decline with increasing redshift, sistent with our findings that⇠ the SFRD evolution can els tuned to the lower redshift LBG LFs or MFs (e.g., range we probe, as is shown in Figure 2. Similarly good Mashianthere was et al.some2016 debate; Mason in et the al. literature2015; Trac on et al. exactly2015; how be reproducedagreement is simply found with by the the DM Schechter halo MF function build-up, from Ishi- the fast this decline is between z 8 and z 10 – accel- model by Mashian et al. (2016), which is based on a Cai et al. 2014; Sun & Furlanetto 2016; Behroozi & Silk gaki et al. (2017). 2015erated), or or semi-analytical continuous with models respect (e.g.⇠ Dayal to z et⇠ al.4 20148?; The fixed SFR-MThe onlyhalo previousrelation UV at LF all redshiftsthat is clearly reproduces discrepant the CowleyHFF dataset et al. 2017 provide), and an full excellent, hydrodynamical independent⇠ simula- test of observedfrom our SFRD new trendsmeasurement at z> is8. the Other one from modelsMcLeod predict et al. tions.this, as The a power latter include law extrapolation UV LFs from of the theCoDa UV LFsim- trends (2016). This is based on only one UV LF point, shown higher SFRDs at z>8, and the discrepancy between ulationto z> (8Ocvirk was predicted et al. 2016 to), reveal the Croc moresimulation than 25 suite galaxies thesein predictions Fig 2,whichis are increasing3 higher with than redshift. our measurement. As is ev- ⇠ ⇥ (inGnedin the six2016 HFF), BlueTides clusters+parallel(Liu et al. fields.2016; Wilkins However, et al. after a identBased from on the this Figure, one point, by zMcLeod12 the et predictedal. (2016) argued SFRD 2017), and DRAGONS/Meraxes (Liu et al. 2016). for a much more gradual decline⇠ in the cosmic SFRD careful search for z 10 galaxies based on the Lyman range spans a factor 30 already. This clearly highlights break,The comparison our analysis to only⇠ these revealed theoretical four reliablepredictions candidate is at z>8. While we discuss the evolution of the cosmic shown in Figure 3. It is evident that the modeled UV LFs the great power that the JWST will have in testing these sources/images in these HFF fields, clearly demonstrat- modelSFRD predictions in detail in in the section near4.3 future., it is worth mentioning decrease in a similar way at z>8 compared to what we here that the combination of all the HST legacy fields areing finding that the observationally. evolution of Given the UVthe vastly LF is di faster↵erent thanna- at is clearly inconsistent with that UV LF (and hence the lower redshift, confirming our previous findings of an ac- 5. SUMMARY ture of these models, it seems clear that the main driver SFRD) from McLeod et al. (2016). In particular, we forcelerated this strong evolution. evolution is the underlying dark matter Thiscan paper compute presented how many az complete10 galaxies and self-consistentwe would have ⇠ haloWhen mass combining function, which all the is known HFF candidates to evolve rapidly with the at five analysisexpected of all in the our primecombined extragalactic dataset assumingHST theirdatasets UV LF to thesesources early previously times (see identifiedalso discussion in the in Section HUDF+CANDELS4.4). The constrain the evolution of the galaxy UV LF and cosmic data, we indeed find a best-fit UV LF that is almost star-formation rate densities to z 10. In particular, the exactly a factor 10 lower in normalization compared to ⇠ z 8 (see Fig. 2 and⇥ Table 5). This is consistent with 9 https://github.com/steven-murray/hmf several⇠ models of galaxy evolution at the 1 1.5 level, z>6 現在のフロンティア � cosmic reionization epoch status circa 2003 When? early? late? two-epoch? How? topology? What? Pop III? Pop II? mini-QSOs? dark matter decay? So what? regulation of dwarf galaxy formation … Madau 07 z>6 現在のフロンティア � cosmic reionization epoch status circa 2017 When? relatively late How? topology? What? Pop II? quasars? So what? regulation of dwarf galaxy formation baryon distribution in halos vs IGM CDM small-scale power spectrum Madau 07 2 D’Aloisio et al. is remarkablycosmic flat over reionization the redshift range 2 A83, page 9 of 14 cosmic reionization: quasars strike back? HeIIThe Astrophysical Ly Journal,α opacity825:144 (32pp), 2016 -> July He 10 reionization more extended?Worseck et al. Worseck+ 16� ionization energy: Figure 3. He II effective optical depth eff,He II vs. redshift for 17 He II sightlines in identical redshift bins of z 0.04 ( 10 proper Mpc at z 3), discovered in our Cycle17 survey (Figure 1) or reanalyzed from the HST archive (Figure 2). The measured eff,He II values are plotted as black circles with error bars distinguishing statisticalHeI errors– 24.6 due to Poisson eV count: near-simultaneous statistics (black, double-sided 1 errors corresponding with toH a conreionizationfidence level of 68.26%) and (massive additional systematic stars?) errors from background uncertainties (gray). For clarity, the data are plotted slightly offset with respect to the identical bin centers and total error bars smaller than the symbol size haveHeII been omitted. – 54.4 For every eV measurement: quasars we also plot only the 1 instrumental sensitivity limit (red horizontal dashes), which we adopt as measured values (arrow symbols) if the upper confidence level includes infinite eff,He II or if the signal is formally negative (P 0.1587). Overplotted are predictions from a semianalytic model of a reionized IGM matching low-redshift observations with two representative nHeII n H I ratios of 60 and 200 (green lines), and results evaluated in z 0.04 bins from a numerical simulation by McQuinn et al. (2009) in which He II reionization finishes at z 2.7 (blue; solid: median , dashed: 1 deviation). large fluctuations in HI Ly α opacityreion Becker+eff,He II 15� ->effective more optical depthspatchy in two contiguousreionization redshift bins bydecreasing rare, instrument bright sensitivity sources? at the corresponding� wave- (3.10 z 3.18), possibly continuing to lower redshifts that lengths λ>1350 Å combined with the faintness of the targets were excised due to geocoronal residuals. Other sightlines (e.g., results in low sensitivity to high values, some of which GRB damping wing observations eff,He II SDSSJ2346−0016) show complete Gunn–Peterson troughs at can be seen already at z 3.4. Statistically robust constraints theprobe same redshifts. of SMBH This indicates evolution that part of the� spread in the on the redshift evolution of the He II effective optical depth at data is due to large-scale variance between the sightlines. z 3.5 will require a larger sample of He II sightlines observed Figure 5 shows the 4 sightlines covering 3.34 z 3.5. at high S/N. Analysis of our recently obtained sample of three Half of the redshift bins have eff,He II 4, although all z 3.6 sightlines is forthcoming (Program 13875). sightlines are sensitive to 5. Again we see a strong eff,He II We may compare the eff,He II distributions with redshift to sightline-to-sightline variance, with the highest effective optical test statistically for evolution in the He II opacity. The median depths measured toward SDSSJ2346−0016, whereas the fi value of eff,He II is not well de ned at z 3 due to the frequent absorption in two sightlines remains low (SDSSJ1319+5202 sensitivity limits and limited statistics. In an attempt to better and SDSSJ1711+6052). Our Lyα effective optical depths are sample the underlying distribution of eff,He II at a given redshift, in good agreement with inferences from He II Lyβ absorption at we assumed that contiguous z 0.04 redshift bins of the these redshifts (Syphers et al. 2011). The lowest He II effective same sightline are independent, a strong approximation given optical depth at is robustly measured in a flux spike in z 3.3 the significant correlation between neighboring redshift bins, the SDSSJ1319+5202 sightline at z 3.44. Again we see especially at z 3. The median eff,He II increases gradually that the He II transmission occurs on smaller scales than our from 1.94 at z 2.70 (19 measurements at 2.66 z 2.74) z 0.04 redshift windows ( z 0.02 corresponding to 4 to at (10 measurements at ), proper Mpc at z 3.44). The other four detections occurring in 5.17 z 3.4 3.34z 3.50 the sightlines to SDSSJ1319+5202 and SDSSJ1711+6052 although the latter is poorly constrained to the highest robustly are closer to the sensitivity limit, meaning that some of them measured eff,He II value (50% of the data are sensitivity limits). may be Poisson background fluctuations (0.001 P 0.043). Nevertheless, this result highlights the trend described above: Large-scale underestimates of the background are unlikely, as the effective He II Lyα opacity increases monotonically from strong background oversubtractions would occur in other z 2.4 to z 3.4 by a factor of 2–3. regions. Consistency with a Poisson background fluctuation Armed with our statistical formalism to estimate the signal (i.e., P 0.01 for all four values) would require local significance, we combined the sightlines to estimate the overall increases of the mean background by more than its estimated significance of any residual flux. While this dilutes the 1.6%–3.2% uncertainty (Appendix B.4). We conclude that significance of individual detections, it also averages out Poisson background fluctuations cannot entirely account for individual background errors. For both high-redshift intervals these measurements. 3.06 z 3.26 and 3.34 z 3.50, the probability that all Only two sightlines in our sample probe z 3.5. Given the measured counts above the background are caused by Poisson fl 7 large observed variance in eff,He II at z 3, it is extremely background uctuations is very small (P 10 and difficult to draw firm conclusions on the redshift evolution of P 2 10 6, respectively). At 3.34 z 3.50 the flux the He II absorption at the highest redshifts. Moreover, the spike in SDSSJ1319+5202 dominates the signal. Discarding 11 裏技 � probing ionized IGM with radio dispersion SI 04 also Ioka 03 GRB H II H I 5 4 z= 0 zreion zGRB 3 (a) 2 2 e cdt xe(z)nIGM(z) 4 dz 10 IGM Δt= 2 2πm c ν dz 1+z 6 e 5 4 ] dispersion measure -3 3 2 3 3 nISM=(1+z) 10 3 nMC=10 DM [pc cm 6 5 4 3 2 nMC=10 2 102 Galactic 6 5 0 5 10 15 20 25 30 z e2 cdt x (z)n (z) -3 Inoue ‘04 MNRAS 348,dz 999 e IGM dispersion measure [pc cm ] Δt= 2 dz 1+z see also Ioka2πme c‘ 03ν Haddock & Sciama ‘65, Weinberg ‘72 Ginzburg ‘73, Palmer ‘93 dispersion measure mean effect assuming homogeneous IGM e2 Δt= 2 ne dl 2πmec ν dispersion measure delay time [hour] IGM dispersion measure with reionization by stars+quasars Mitra+16 model SI, Mitra, Choudhury, Ferrara, in prep. DM of mean IGM� difference with respect to best fit� 7000 100 ~MH15 6000 QSO-dom.� 50 5000 ~G15 new QSO� 4000 DM 0 DM 3000 delta ~HM12 old QSO� 2000 -50 solid: Mitra16 without HeII 1000 ~P16 new CMB� dotted: inst. zHI=8.3, zHeII=4� 0 -100 0 2 4 6 8 10 2 3 4 5 6 z z - model differences not large but relevant for obs. (δDM~+-100 at z~2-6) unique info on H+He reionization, evolution of low-L quasars - variance due to LSS averaged out with large enough sample - local DM main uncertainty -> can it be sufficiently constrained? - uncertainties in reionization important for interpreting DM at z>~3� xe=[1-YHe(xHII)+(YHe/4)(xHeII+2xHeIII)] gamma-ray “absorption”: probe of diffuse radiation fields gamma-ray absorption: probe of diffuse radiation fields γγ 吸収を通じた宇宙背景放射の精査 � 1. gamma-ray absorption: probe of diffuse background radiation (integrated starlight) γ + γ → e+ + e- E 2 4 ε σ peak ,, =4 me c 2 4 threshold: E ε (1-cos θ)>2 me c Costamante+ 03 e.g. 100 GeV + 10 eV (UV) TeV + 1eV (IR) γ absorption in blazars (z~<1) -> probe of IR/opt. background pair production cross section diffuse background radiation γ-ray source γ-ray telescope γ γVHE EBL e+ e- CM energy adapted from D. Mazin constraints on EBL at z=4.35 high-z手嶋さんのスライドより UV radiation field (“background”) from Fermi/LAT detection crucial for understanding early universe of GRB 080916C • reionizes the IGM (feedback on later galaxy formation) • dissociates H2 molecules (feedback on later star formation) • Ly α pumping (determines HI hyperfine 21cm level population) but direct detection impossible!!! Ciardi, Stoehr & White 03 -> suppresses formation of small galaxies –6– probe of high-z UV 1 This work (GRBs) 0.51.01.52.02.5 2FHL (Blazars) Finke+ 10 Model C background via ⌧, Dom´ınguez+ 11 Dom´ınguez+ 11 GRB gamma-rays Gilmore+ 12 Fiducial 0.1 Helgason+ 12 Desai+ 1710.02535 –13– redshift measurements exist. The low number ofGRB090926A photons detected from each single GRB at GRB090902B highEnergy [TeV] 0 energy,.01 predominantly due to the steep decline (with energy) of the LAT e↵ective area, 22 Fermi-LAT GRBs with z renders the detection of the EBL attenuation in the spectrum of a single source challenging. To overcome this, we analyze the combined set of GRB spectra (stacking) which allows us 2.8σ evidence of EBL attenuationto reject the null hypothesis of no EBL attenuation at 2.8 confidence. ⇠ at z~1.8 0.001 01234 Redshift LAT best fit -- 1 sigma Franceschini et al. 2008 Finke et al. 2010 -- model C Stecker et al. 2012 -- Low Opacity Fig. 2.— Prediction ofKneiske the cosmic et al. 2004 -ray -- horizonhighUV (i.e. the redshift and energy at which 10 Kneiske et al. 2004 -- best fit ⌧ =1)fromdi↵erentKneiske models & (see Dole legend) 2010 along with the highest energy photons from AGNs and GRBs at di↵Dominguezerent redshifts. et al. 2011 The GRBs from our sample are denoted by stars, Gilmore et al. 2012 -- fiducial AGNs by dots while theHelgason estimates & Kashlinsky from EBL 2012 models are denoted by lines. The two most γ constrainingγ GRBs in our analysis are labelled in the plot for reference. τ interesting prospects for CTAet al. 2016). We1 use the P 8R2 T RANSIENT 020 instrument response function. - more robust constraints The Minuit5 optimizer is used to determine the best-fit spectral parameters and the error estimate for the unbinned likelihood maximization analysis. GRB spectra are generally - quasar contribution to EBL?described using the “Band function” (Band et al. 1993), which consistsz of≈1.8 two power laws joined by a exponential cut-o↵, or a Comptonized model, which consists of a power law with - probe H+He reionization?� 10−1 10 102 103 4http : //F ermi.gsfc.nasa.gov/ssc/data/access/lat/BackgroundModels.htmlEnergy [GeV] 5http : //lcgapp.cern.ch/project/cls/work packages/mathlibs/minuit/doc/doc.html Fig. 4.— Constraint on the optical depth at a redshift of z 1.8, at 1 confidence level ⇡ (68%), derived for our GRB sample, compared with model estimates. The models of Frances- chini et al. (2008) and (Stecker et al. 2012, high and low opacity), not included in the numerical analysis (mentioned in Sec. 3), are included in the figure for completeness. The constraint on the -ray optical depth as derived from this analysis is reported in Figure 4. We report this constraint for an e↵ective redshift of 1.8. This value is derived ⇠ by separating the source sample into two redshift bins and finding the value of the redshift 542 O. Just et al. Moreover, a small fraction of the NS matter can be expelled during and after the coalescence of the binary components. Be- 560 O. Just et al. cause of its extremely neutron-rich initial state, this matter has long been speculated to be a possible site for the formation of r-process nuclei (Lattimer & Schramm 1974, 1976;Eichleretal.1989). In- deed, Newtonian (e.g. Ruffert et al. 1997;Freiburghaus,Rosswog Nucleosynthesis from compact binary mergers 559 & Thielemann 1999;Jankaetal.1999; Rosswog et al. 1999;Ruf- fert & Janka 1999, 2001;Rosswog2005; Korobkin et al. 2012)and conformally flat general relativistic (e.g. Oechslin, Janka & Marek 2007; Goriely, Bauswein & Janka 2011b;Bauswein,Goriely& Janka 2013b)aswellasfullyrelativistic(e.g.Kyutokuetal.2011; Downloaded from Nucleosynthesis from compact binary mergers 557 Deaton et al. 2013;Hotokezakaetal.2013a,b;Kyutoku,Ioka&Shi- Downloaded from bata 2013; Foucart et al. 2014;Wanajoetal.2014)hydrodynamical simulations of NS–NS and NS–BH mergers with microphysical r-process nucleosynthesis Downloaded from http://mnras.oxfordjournals.org/ equations of state (EOSs) have demonstrated that typically some 3 in NS mergers 10− M up to more than 0.1 M can become gravitationally un- ⊙ ⊙ Figure 14. Abundance distributions as functions of the atomic mass for Figure 16. Abundance distributions as functions of the atomic mass for four bound on roughly dynamical time-scales due to shock acceleration three systems with torus masses of Mtorus 0.03, 0.1 and 0.3 M and identical BH–torus systems (MBH 3M ; ABH 0.8; Mtorus 0.3M ) = = = = http://mnras.oxfordjournals.org/ a3M BH. Allhttp://mnras.oxfordjournals.org/ distributions are normalized to the same solar A ⊙ 130 computed with two different values of the⊙ viscosity parameter αvis, namely⊙ = and tidal stripping. Also the relic object, either a hot, transiently abundance.⊙ Calculations correspond to the M3A8m03a5, M3A8m1a5 and αvis 0.02 and 0.05, and two different prescriptions of the viscosity ten- = M3A8m3a5 models. The dotted circles show the solar r-abundance distri- sor (cf. Section 3.1). All distributions are normalized so that X 1. stable hypermassive NS (HMNS; stabilized by differential rotation = Just+ 15 bution (Goriely 1999). dynamical ejecta� Calculations correspond to the M3A8m3a2, M3A8m3a5, M3A8m3a2-v2 Downloaded from ! and M3A8m3a5-v2 models. The dotted circles show the solar r-abundance and thermal pressure; Baumgarte, Shapiro & Shibata 2000)andlater at Kokusai Hoken Keikakugaku (UNIV OF TOKYO) on December 21, 2016 distribution (Goriely 1999). supermassive NS (SMNS; stabilized by rigid rotation and thermal pressure) or a BH–torus system (Fig. 1, upper panel), can lose mass Figure 10. Abundance distributions as functions of the atomic mass for Figure 11. Abundance distributions as functions of the atomic mass for observe that for a higher dynamic viscosity coefficient ηvis ahigher the dynamical ejecta of three NS merger cases with prompt collapse of the the dynamical ejecta of three NS merger cases with delayed collapse of the relative amount of A > 130 elements is obtained (remembering also that the η for the two viscosity types are related by equation 2). along with its secular evolution in outflows that are driven by vis- remnants. Each binary system is characterized in the legend by the EOS used remnants. Each binary system is characterized in the legend by the EOS used http://mnras.oxfordjournals.org/ vis at Kokusai Hoken Keikakugaku (UNIV OF TOKYO) on December 21, 2016 in the simulation and the mass (in M )ofthetwoNSs.Alldistributions in the simulation and the mass (in M )ofthetwoNSs.Alldistributions This result can be explained by the fact that a higher dynamic Figure 18. Abundance distributions as functions of the atomic mass for two cous energy dissipation and turbulent angular momentum transport, are normalized to the same A 196 abundance.⊙ The dotted circles show the are normalized to the same A 196 abundance.⊙ The dotted circles show the viscosity leads to a smaller mean electron fraction Y¯ of the ejecta Figure 17. Time evolution of= the average radioactive heating rate per unit BH–torus systems with M = 3M , A 0.8, M 0.1 M (top) e solar r-abundance distribution (Goriely 1999). solar r-abundance distributionBH (Goriely 1999BH ). torus mass, Q , for two BH–torus systems with MBH 3M , ABH 0.8, = = = in our models (cf. Table 2). Sensitivity to r-process heating feedback magnetic pressure, nucleon-recombination heating and neutrino- and MBH 4M , ABH 0.8, Mtorus⊙ 0.3 M (bottom), when the⊙ heating ⟨ ⟩ = = = = = Mtorus 0.1 M (top) and MBH 4M , ABH 0.8, M⊙torus 0.3 M feedback due to⊙ the r-process β-decays and fission⊙ are included or not. All In Fig. 17, we compare the average temperatures as well as the While= the ejecta masses seem= to be in reasonable= agreement,= the energy deposition (Fig. 1, lower panel; e.g. Popham, Woosley & (bottom), when⊙ the heating feedback due⊙ to the r-process β-decays and⊙ distributions are normalized to the same solar A 130 abundance. Calcu- average heating rates for the models with and without radioactive similar pattern characteristics of the fission recycling= nucleosynthe- fissionbulk of is the included outflow or not. in the Calculations model of correspond Fernandez´ to & models Metzger M3A8m1a2, (2013) is lations correspond to models M3A8m1a2, M3A8m1a2-rh, M4A8m3a5 and heating as implemented by the approximate method described in Fryer 1999; Ruffert & Janka 1999; Dessart et al. 2009;Fernandez´ sis, as described at Kokusai Hoken Keikakugaku (UNIV OF TOKYO) on December 21, 2016 in Goriely et al. (2013, 2014). In the relativistic M3A8m1a2-rh,slightly more M4A8m3a5 neutron rich, andY M4A8m3a5-rh.e 0.1–0.21, than in our simulation, M4A8m3a5-rh. The dotted circles show the solar r-abundance distribution Section 3.1. Correspondingly, in Fig. 18,wecomparetheabundance ∼ simulations, inwind contrast to theejecta Newtonian� approximation (where &Metzger2013;Metzger&Fernandez´ 2014; Perego et al. 2014; Ye 0.18–0.30. The reason of this difference cannot be unam- (Goriely 1999). distributions for these models. The variations introduced by such a dynamically∼ (both components being composed of about 80– ejecta originate mostly from cold extended spiral arms; Korobkin biguously identified. The lower values of Ye could simply also be lowest order correction for radioactive heating are only marginal. 98 per cent of r-process material). For the three cases shown in et al. 2012; Rosswog et al. 2014), not all the mass elements lead at Kokusai Hoken Keikakugaku (UNIV OF TOKYO) on December 21, 2016 Siegel, Ciolfi & Rezzolla 2014). Such ejecta could be interesting a consequence of the error in the α-recombination heating, lead- Besides minor differences in the abundance distributions, also the Fig.ing to19 higher, this ratio escape amounts velocities to 1.4, and 5.3 therefore and 1.7 faster for the freeze-out 0.03, 0.1 from and to the same composition after ejection. Mass elements are ejected environments for r-process, p-process, and nickel nucleosynthesis total ejecta masses are hardly affected. With heating Mout increases 0.3weak-reaction M torus models, equilibrium. respectively. However, For Metzger this reason, & Fern whenandez´ normaliz- (2014) Figurewith considerably 15. Abundance different distributions velocities as functions so that ofthe the density atomic evolution mass for ⊙ from 22.1 to 22.4 per cent and from 22.7 to 22.8 per cent of the origi- ing to the A 196 abundance, the lighter elements, and in particular threemay BH–torusvary significantly systems with from BH one masses trajectory of MBH to the3, next4 and (cf. 6 M Bausweinand the (e.g. Surman et al. 2008, 2014;Caballero,McLaughlin&Surman mention that= the changes of nucleosynthesis-relevant properties as- = nal torus mass for models M4A8m3a5 and M3A8m1a2, respectively the A 130 peak, are found to typically vary within a factor of 3. sameet al. 0.32013b M ).tori. Fast All expanding distributions mass are normalized elements mightto the same not havesolar⊙A time130 to sociated≃ with their mistake are irrelevant. Another possible reason = (see Table 2). This indicates that including the radioactive heating 2012;Wanajo&Janka2012;Malkus,Friedland&McLaughlin Note that it is well known that calculations of the r-process abundance.capture all⊙ Calculations available free correspond neutrons to leading the M3A8m3a5, to no or only M4A8m3a5 one fission and for the Ye difference could be connected to the different treatments in a fully consistent manner is not necessary or at least does not abundances are still affected by large nuclear physics uncertain- M6A8m3a5cycle while models. more slowly The dotted expanding circles massshow elementsthe solar r-abundance allow for all distri- free 2014,andreferencestherein). of neutrino reactions and transport. lead to any significant changes of our modelling predictions (see ties (Arnould, Goriely & Takahashi 2007). Such uncertainties have butionneutrons (Goriely to be1999 captured). and typically three fission cycles to take also Rosswog et al. 2014, where a similar conclusion was drawn been extensively studied in the past, but each site provides its spe- place (Goriely et al. 2014). Delayed collapse cases The presence of significant mass fractions of radioactive material concerning the dynamical ejecta). cific3.4 Nucleosynthesisconditions and behaves in its own special manner so that an composedIn the delayed of 80 to collapse 94 per cent cases, of r-process the integrated material, mass the remainingassociated (nickel, r-nuclei,. . . ) was pointed out to lead to long-term decay assessment of the sensitivity to theoretical nuclear physics input 6with to 20 rapidly per cent expanding being made material essentially compared of 4He. to theSensitivity total ejecta to global mass 3.4.1 Nucleosynthesis in the dynamical ejecta parameters heating of the material ejected by compact binary mergers and requires careful and dedicated exploration. While we already partly remains relatively smallertotal than� in the prompt collapse cases. For 3.4.3 Combined nucleosynthesis in the dynamical and disc ejecta investigatedThe nucleosynthesis the sensitivity resulting of the from nucleosynthesis the dynamical in ejecta the dynamical has been thisThe reason, abundance similar distributions to Newtonian are modelsfound to of be the only NS-merger weakly sensitive hydro- thus to thermal radiation that can potentially be observed as elec- ejectastudied to previously masses, β-decay either based rates, on and parametrized fission probabilities ejecta trajectories (Goriely todynamics, the torus which mass, tend as shown to predict in Fig. rather14 for slow cases expansion with the (Korobkin same BH Assuming that the dynamical and disc nucleosynthesis components tromagnetic transient (Li & Paczynski´ 1998) termed ‘macronova’ et(Freiburghaus al. 2013), we et defer al. 1999 such;Gorielyetal. a sensitivity analysis2005) for or the more composition ‘realistic’ masset al. 2012 and spin), the and abundance the same distributions viscosity. The of relativisticdifferences models result from with are both ejected isotropically, we can combine the yields by sum- ofhydrodynamical the disc ejecta toNewtonian a future study. (Korobkin et al. 2012; Rosswog et al. thedelayed two subtle remnant trends collapse that exhibit the fraction a trough of material around A with200Ye (Fig.< 0.211 as). ming up the mass fractions for both components, weighted by their Figure 19. Abundance distributions as functions of the atomic mass for other scenarios: all from dynamical ejecta, 1stwell peak as the mean from entropy slightly other increase sources… for lower≃ torus masses.� corresponding total ejected masses. We only combine systems cor- (Kulkarni 2005)or‘kilonova’(Metzgeretal.2010;Metzger& Figure 1. Top panel: evolution2014) and (CFC-) paths relativistic of NS–NS (Goriely and et al. NS–BH2011b, 2013 mergers.;Bauswein De- threeThis combined deep trough systems is absent (merger in model the prompt plus remnant collapse model) cases correspond- (Fig. 10), As can be seen in Fig. 15, the abundance distribution is also found responding to the same BH mass and spin and the same torus mass et al. 2013b) simulations of NS–NS binary systems, recently also in- ingwhere to models the relative with torus contributions masses Mtorus of quickly0.03, expanding 0.1 and 0.3 trajectories M .Alldis- to Berger 2012). Indeed, a near-infrared data point at the position pending on the binary4COMPARISONWITHOBSERVATIONS parameters and the properties of the nuclear EOS, = cluding neutrino effects in relativistic merger models (Wanajo et al. tributionstothe be final only areabundance moderately normalized distribution sensitive to the same to are the solar found BHA mass. to196 be The appreciable abundance. observed⊙ Calcu- and slight to in a roughly consistent manner. For models with 0.03, 0.1 and = binary NS mergers canThe2014 lead striking). In to all similaritythese the simulations, formation between the the of number solar a stabledistribution of free neutrons NS, of r-element a per transient seed lationstrendfill the towards correspond trough relatively in tothe theA model heavier200 combinations region. elementsComparison for TMA_1616–M3A8m03a5, lower between BH masses NS–NS can 0.3 M tori, the abundance distributions for such combined mod- of the short-hard GRB 130603B about 7 d (rest-frame time) after ≃ HMNS (stabilized by differentialabundancesnucleus reaches in the rotation) a few 56 hundred.Z 76 and With range SMNS such and a the neutron corresponding(stabilized richness, fission abun- by rigidSFHO_13518–M3A8m1a5beand ascribed NS–BH to nucleosynthesis the lower and mean DD2_14529–M4A8m3a5. electron fractions of The the dottedejecta. cir- In els are⊙ given in Fig. 19.Thedynamicalejectacontributemainlyto the burst was interpreted as possible first detection of such an r- danceplays patterna fundamental observed role≤ in byultrametal-poor≤ recycling the starsmatter like during CS 22892 the neutron052 clesFig.Fig. show16,12 we theillustrates compare solar r-abundance the the abundance abundance distribution distributions distributions (Goriely 1999 obtained representative). with two of the production of the A > 140 nuclei, whereas the disc ejecta pro- − rotation), or a BH plusirradiation accretion-torus and by shaping system. the final r-abundance The last distribution scenario in is the also valuesthree NS–BH of the viscosity binary systems. parameter Inα thevis, NS–BH both for cases, the type the 1 expansion and type duce mostly the 90 A 140 nuclei. The combined distribution is process powered transient (Berger, Fong & Chornock 2013; Tanvir ≤ ≤ the outcome of a NS–BH110 ! A merger! 170 mass ifregion the BH/NS at the end of mass the neutron ratio irradiation. is not too 2velocities prescriptions, are rather as described low (see in Table Section1), comparable 2.2. In general to the terms, delayed the in surprisingly good agreement with the Solar system r-abundance et al. 2013), because its temporal delay, colour, and brightness are MNRASThanks to448, this541–567 property, (2015) the final composition of the ejecta is rather abundancecollapse cases distribution of the NS–NS is quite systems robust with and lowerrespect than to the those viscosity in the distribution (Fig. 19). However, the relative strength of the second large. Transitions betweeninsensitive the to different details of the evolution initial abundances stages and the can astrophysical be accom- treatmentprompt collapse for the cases. intermediate-mass The corresponding elements abundance 80 A distributions130, while to the third r-process peak depends sensitively on the ratio be- compatible with theoretical light-curve calculations for an assumed ≤ ≤ panied by mass-loss. Inconditions, this work, in particular we focus the mass exclusively ratio of the twoon NSs, the the quantity evolution itare is therefore rather sensitive rather to similar the viscosity to those for found the A in> the130 NS–NS elements. delayed We tween the amount of mass ejected from the disc and the one ejected 2 ejecta mass of some 10− M , taking into account the fact that of matter ejected, and the EOS (Goriely et al. 2011b; Korobkin et al. collapse with an underproduction of the A 202 nuclei. Impact of tracks and stages highlighted2012;Bausweinetal. by thick, solid2013b). red The lines. mentioned Bottom calculations panel: essen- mass- neutrinos ≃ ⊙ MNRAS 448, 541–567 (2015) r-process nuclei, in particular the abundant lanthanides, increase loss phases during thetially dynamical correspond interaction to the delayed-collapse of NS–NS scenario and and their NS–BH typical bi- During the binary merging phase, we disregard the effects of the opacity by roughly a factor of 100 compared to iron (Barnes naries and the subsequentyields secular are displayed evolution in Fig. 11. In of the afollowing, relic BH–toruswe present our firstsystem neutrinos in our hydrodynamic simulations as well as in the nucle- nucleosynthesis results obtained for cases with prompt collapse of osynthesis studies. In a few recent simulations with nucleosynthesis &Kasen2013; Hotokezaka et al. 2013b; Kasen, Badnell & Barnes (corresponding to the evolutionthe NS–NS binary paths merger indicated remnantby as well red as lines for NS–BH in the binary upper discussion neutrino effects were explored. While Korobkin et al. 2013;Tanaka&Hotokezaka2013; Grossman et al. 2014;Tanaka panel). The dynamical masssystems ejection (cf. Table 1 takes). Prompt place collapse within cases a few milliseconds (2012) did not find a significant impact in their Newtonian studies, Fig. 10 shows the abundance distributions resulting from the the relativistic NS–NS merger calculations of Wanajo et al. (2014) et al. 2014). when the two binary componentsprompt collapse of merge NS–NS withbinary systems. each other. The three Typical cases exhibit ejecta for a ‘soft’ nuclear EOS (i.e. NSs with small radii) point towards Nucleosynthesis calculations agree (e.g. Freiburghaus et al. 1999; masses are around 0.01 M and the average entropies of the ejecta are ⊙ Goriely et al. 2005, 2011b; Roberts et al. 2011; Korobkin et al. low. BH–torus systems eject matter mainly in viscously driven outflows MNRAS 448, 541–567 (2015) and to a smaller extent also in neutrino-driven winds. Baryon-poor polar 2012;Bausweinetal.2013b; Rosswog et al. 2014;Wanajoetal. funnels may provide suitable conditions for neutrino or magnetohydrody- 2014) that the dynamical ejecta of the merging phase provide suf- namically powered, ultrarelativistic, collimated outflows, which are likely to ficiently neutron-rich conditions for a robust r-processing up to the produce short GRBs. The image is adapted from Ruffert & Janka (1999)and third abundance peak and the actinides, although the theoretical Janka & Ruffert (2002). predictions show considerable differences in important details, de- pending on the exact conditions in the ejecta (electron fraction, Ye, expansion velocity, and specific entropy) and the employed nuclear physics, in particular the theoretical assumptions about the fission fragment distribution (cf. Goriely et al. 2013). Relativistic NS–NS MNRAS 448, 541–567 (2015) The AstrophysicalThe Astrophysical Journal, Journal,829:110829:110(20pp)(,20pp 2016), 2016October October 1 1 BarnesBarnes et et al. al. The Astrophysical Journal, 829:110 (20pp), 2016 October 1 Barnes et al. Figure 18. Select broadband light curves for our fiducial ejecta model for two treatments of thermalization: full (left panel), and the cruder treatment r-process nucleosynthesis in NS mergers:fttot ( ) Figureemployed 18. Select in earlier broadband kilonova light calculations curves for(right our panelfiducial). The ejecta curves model are dimmer for two fl treatmentsin the newer of thermalization: models, re ecting full fttot the( ) reduction(left panel in), thermalized and the cruder energy, treatment but employedrelationships in earlier between kilonova the calculations light curves(right in the panel various). The bands curves is are mostly dimmer in-situ probe? inunchanged, the newer so models, the kilonova reflecting colors the are reductionpreserved. in thermalized energy, but 42 K. Hotokezaka et al. relationships between the light curves in the various bands is mostly macro/kilonova Barnes+ 16 � unchanged,difficult so the kilonova colorsto obtain are preserved. detailed info� Figure 18. Select broadband light curves for our fiducial ejecta model for two treatments of thermalization: full fttot ( ) (left panel), and the cruder treatment employed in earlier kilonova calculations (right panel). The curves are dimmer in the newer models, reflecting the reduction in thermalized energy, but relationships between the light curves in the various bands is mostly unchanged, so the kilonova colors are preserved. Figure 17. Synthetic bolometric light curves calculated with Sedona for three different treatments of thermalization: full thermalization (blue curves); Sedonaʼs original thermalization scheme, which deposits charged particle energy but explicitly tracks the deposition of -ray energy (lime curves); and Decay products in NS merger ejecta Figure 19. Absolute (AB) J-band light curves for several ejecta models. As in 39 Figure 17.theSynthetic time-dependent bolometricft lightfrom curves our numerical calculated simulations with Sedona(red curvesfor three). Light tot ( ) earlier figures, the width of the curves is produced by differences in ftdue differentcurves treatments in the oftop thermalization:and bottom panels full adopt thermalizationftgiven by(blue randomHotokezaka curvesfields); and + 16 � very difficult totot ( ) detect� nuclear gamma linestot ( ) to different magnetic field configurations. The excess IR flux (gold star) Sedonaʼthes original FRDM mass thermalization model. To scheme, illustrate whichthe effect deposits of uncertainties charged in particlefton tot ( ) suggests an ejected mass between 5 10 2 and 10 1 M . energy but explicitly tracks the depositionfi of -ray energy (lime curves); and Downloaded from light curves, we plot for our ducial model (middle panel) Lbol for a range of Figure 19. Absolute (AB) J-band light curves for several ejecta models. As in the time-dependentmagnetic fieldsft(thickfrom red our curve numerical) and for simulations the DZ31 mass(red curvesmodel )(.red Light dashed tot ( ) earlier figures, the width of the curves is produced by differences in ftdue curves incurve the top). Accounting and bottom for panels time-dependent adopt ft thermalizationgiven by random efficienciesfields and has a tot ( ) tot ( ) toIf different the ejected magnetic materialfield con isfigurations. mainly con Thefined excess to IR theflux equatorial(gold star) the FRDMsigni massficant model. impact To on illustrate kilonova luminosity, the effect of particularly uncertainties for models in ft withon lower tot ( ) D=3 2 Mpc 1 � fi suggestsplane, an the ejected emission mass between will be5 brighter 10 whenand 10 theM system . is viewed light curves,masses we and plot higher for our velocities.fiducial Formodel our(middleducial model, panel) theLbol predictedfor a range luminosity of magnetic isfields lower(thick by a factor red curve of 2) andat peak, for and the byDZ31 10 days mass is model lower by(red an dashedfactor of 5. face-on (Roberts et al. 2011), which would reduce the inferred fi mass somewhat. In an oblate ejecta, thermalization will also be curve). Accounting for time-dependent thermalization ef ciencies has a If the ejected material is mainly confined to the equatorial significant impact on kilonova luminosity, particularly for models with lower more efficient, which could have a small impact on mass plane, the emission will be brighter when the system is viewed http://mnras.oxfordjournals.org/ masses and higher velocities. For our fiducial model, the predictedfl luminosity measurements. Finally, we note that different nuclear mass In Figure 19, we compare the detected ux to J-band light face-on (Roberts et al. 2011), which would reduce the inferred is lower by a factor of 2 at peak, and by 10 days isfi lower by an factor offl 5. models predict different rates of radioactive heating and curves for various ejecta models, and nd the observed ux is mass somewhat. In an oblate ejecta, thermalization will also be consistent with 5 10 2 MM 10 1 M. This mass is differing fttot (), which introduces additional uncertainties into Figure 17. Synthetic bolometric light curves calculatedej with Sedona for three moreour ef massficient, estimate. which Radiation could have transport a small simulations impact in on three mass differenthigher treatments than of what thermalization: is typically full predicted thermalization for the dynamical(blue curves ejecta); fl measurements.dimensions with Finally, time-dependent we note that thermalization different nuclearmodels will mass SedonaIn Figureʼsfrom original19 a, binary thermalization we compare NS merger, scheme, the detected suggesting which depositsux that to chargedJ if-band the particle kilonova light energy but explicitly tracks the deposition of fi-ray energy (lime curvesfl ); and modelsbetter constrain predict differentMej. rates of radioactive heating and curves forinterpretation various ejecta is correct, models, the and progenitornd the of observed GRB130603Bux is was Figure 19. Absolute (AB) J-band light curves for several ejecta models. As in the time-dependent ftfrom 2 our numerical simulations1 (red curves). Light differing fttot , which introduces additional uncertainties into consistentperhaps with an5tot ( NSBH) 10MM merger, ej or that 10 the M mass. This ejected mass is was earlier figures,() the width of the curves is produced by differences in ftdue curves in the top and bottom panels adopt ftgiven by random fields and our mass estimate. Radiation transport simulations intot three( ) higher thansignifi whatcantly is enhancedtypically bypredicted post-mergertot ( for) the disk dynamical winds. ejecta to different magnetic 6.4.field Late-time configurations. Light The Curve excess IR flux (gold star) the FRDM mass model. To illustrate the effect of uncertainties in fttot ( ) on dimensions with time-dependent thermalization 2 1 models will suggests an ejected mass between 5 10 and 10 M . lightfrom curves, a binaryOur we plot mass NS for estimate our merger,fiducial here suggestingmodel is an(middle improvement that panel if) Lbol theoverfor kilonova earlier a range work of Late-time kilonova light curves may probe the history of r- fi better constrain Mej. magneticinterpretationwhichelds (thick isneglected correct, red curve detailed the) and progenitor for thermalization, the DZ31 of GRBmass and model130603B gives(red dashed substan- was process nucleosynthesis in CO mergers. At ∼2 days after curveFigure). Accounting5. Light curves for time-dependent of nuclear γ -rays thermalization for NSM-solar efficiencies in the ranges has a of 10 keV Eγ < 30 keV (top left), 30 keV Eγ < 100 keV (top right), 100 keV at University of Tokyo Library on April 24, 2016 perhapstially an NSBH different merger, results. or For that example, the mass Piran ejected et al. was(2014) Ifmerger, the≤ ejectedfission material ceases to is be mainly important,≤ confi andned - to and the -decay equatorial signiEficant< 300 impact keV on (bottom kilonova left), luminosity, and 300 particularly keV E for models< 1 MeV with (bottom lower right) at a distance of 3 Mpc. Here, we show the four different ejecta models: Downloaded from γ fi suggested MMej 0.02 , less than halfγ our new value. dominate the kilonova6.4. Late-time’s energy Light supply. Curve Energy from -decay masses≤signi andcantly higher enhanced velocities. by For post-merger our fiducial model, disk≤ the winds. predicted luminosity plane, the emission will be brighter when the system is viewed (Mej,v) However,(0.01 M we, 0.3 havec), (0.01 not M accounted,0.05c), for (0.1 viewing M ,0.3 anglec), and effects. (0.1 M ,0.05c). Also shown are the sensitivity with exposure at 100 ks of current and Our mass= estimate here is an improvement over earlier work face-onis transferred(Roberts entirely et al. 2011 to fast), which-particles, would which reduce thermalize the inferred isfuture lower X-ray by a factor missions: of⊙ 2NuSTARat peak,(Harrison and by⊙ 10 et days al. is2013 lower), ⊙ASTRO-H by an factor(Takahashi of 5. ⊙ et al.Late-time2012), and kilonovaCAST (Nakazawa light curves et al. may2014). probe the history of r- which neglected detailed thermalization, and gives substan- mass somewhat. In an oblate ejecta, thermalization will also be 16process nucleosynthesis in CO mergers. At ∼2 days after tially different results. For example, Piran et al. (2014) merger,more effifissioncient, ceases which to could be important, have a small and impact- and on-decay mass suggested MM 0.02 , less than half our new value. measurements. Finally,’ we note that different nuclear mass macronovaIn Figure mechanisms19ej, we compare and the the sites detected of r-processflux to J nucleosynthe--band light dominateBrennecka the kilonova G. A., Weyers energy S., Wadhwa supply. M., Energy Janney from P. E., Zipfel-decay J., Anbar curvesHowever, for various we have ejecta not accounted models, and forfind viewing the observed angle effects.flux is ismodels transferred predict entirely different to fast rates -particles, of radioactive which heating thermalize and sis. However, such a nearby event is too rare to be detected with A. D., 2010, Science, 327, 449 http://mnras.oxfordjournals.org/ consistent with 5 10 2 MM 10 1 M. This mass is differing fttot (), which introduces additional uncertainties into current detectors. This detection withej a realistic rate will have to ourChurazov mass estimate. E. et al., 2014, Radiation Nature, transport 512, 406 simulations in three higher than what is typically predicted for the dynamical ejecta 16 Clayton D. D., Colgate S. A., Fishman G. J., 1969, ApJ, 155, 75 wait for a more sensitive generation that may be launched in the dimensions with time-dependent thermalization models will from a binary NS merger, suggesting that if the kilonova Chadwick M. B. et al., 2011, Nucl. Data Sheets, 112, 2887 future. better constrain M . interpretation is correct, the progenitor of GRB130603B was Cowan J. J., Thielemannej F.-K., Truran J. W., 1991, Phys. Rep., 208, 267 perhaps an NSBH merger, or that the mass ejected was Cowan J. J., Pfeiffer B., Kratz K.-L., Thielemann F.-K., Sneden C., Burles significantly enhanced by post-merger disk winds. S., Tytler D.,6.4. Beers Late-time T. C., 1999, Light ApJ, Curve521, 194 ACKNOWLEDGEMENTS Diehl R. et al., 2014, Science, 345, 1162 Our mass estimate here is an improvement over earlier work Late-time kilonova light curves may probe the history of r- East W. E., Paschalidis V., Pretorius F., Shapiro S. L., 2016, Phys. Rev. D, whichWe thank neglected Ben Margalit, detailed Satoshi thermalization, Chiba, and andRe’em gives Sari substan- for discus- process nucleosynthesis in CO mergers. At 2 days after 93, 024011 ∼ tiallysions. This different research results. was supported For example, by the PiranI-CORE et Program al. (2014 of) the merger, fission ceases to be important, and - and -decay Figure 2. Spectrum of γ -rays at 1, 3, 5, and 10 d after merger for NSM-solar. Black linesEichler depict D., Livio the γ -ray M., Piran spectrum T., Schramm produced D. by N., nuclei 1989, at Nature, rest. The 340, red 126 (blue) suggestedPlanning andMMej Budgeting 0.02 Committee, less than and half The Israel our new Science value. Foun- dominate the kilonova’s energy supply. Energy from -decay curve shows the spectrum with the Doppler broadening with an expansion velocity ofEichler 0.3c (0.05 M. etc). al., The 2015, normalization ApJ, 808, 30 is determined with the mass of ejected at University of Tokyo Library on April 24, 2016 However, we have not accounted for viewing angle effects. is transferred entirely to fast -particles, which thermalize r-processdation elements (grant of No 0.01 1829/12), M and the the RIKEN observed iTHES distance Project, of 3 Mpc. and Here, Grants- we do notFan take Y.-Z., any absorption Yu Y.-W., Xu and D., scattering Jin Z.-P., processes Wu X.-F., intoWei D.-M.,account. Zhang B., 2013, in-Aid for Scientific⊙ Research of JSPS (26400232, 26400237, ApJ, 779, L25 15H02075) and MEXT (15H00788, 15H00773, 15K05107). 16 Fernandez´ R., Metzger B. D., 2013, MNRAS, 435, 502 tdiff,o: efficientlyFernandez´ heat R., Kasen up the D., ejecta Metzger on B. the D., peak Quataert time-scale E., 2015, of MNRAS, supernovae 446, (Lucy7502005). 2 κγ c tdiff,o Here,Fong W. we et approximatelyal., 2014, ApJ, 780, evaluate 118 ϵγ (t) as the followings. In op- τγ (t) REFERENCES, (7) ≈ κo v t ticallyFoucart thick F. et regimes al., 2013,τ Phys.γ Rev.1, almost D., 87, 084006 all the γ -rays’ energy is de- Abadie J.! et al., 2010," Class. Quantum Gravity, 27, 173001 positedFoucart in F. the et al., ejecta. 2015, Phys. On≫ the Rev. contrary, D., 91, 124021 in optically thin regimes Argast D., Samland M., Thielemann F.-K., Qian Y.-Z., 2004, A&A, 416, Freiburghaus C., Rosswog S., Thielemann F.-K., 1999, ApJ, 525, L121 τ γ < 1, only a fraction τ of the photons are scattered and for each 997 Gao H., Ding X., Wu X.-F., Dai Z.-G., Zhang B., 2015, ApJ, 807, 163 t 2 κ scattered photon roughly half of γ -ray’s energy is transferred to an Arnould0.02 M.,diff Goriely,o S., Takahashiγ K., 2007, Phys. Rep., 450, 97 Goriely S., Sida J.-L., Lemaˆıtre J.-F., Panebianco S., Dubray N., Hilaire S., 2 1 electron via a single scattering process at energies of 1MeV. ϵγ ≈Barbuy B. et al.,t 2011, A&A,0.05 cm 534,/ A60g− Bauswein A., Janka H.-T., 2013, Phys. Rev. Lett., 111, 242502∼ Barnes J.,! Kasen D.," 2013,! ApJ, 775, 18 " is approximatelyGoriely S., Bauswein given A., by Just O., Pllumbi E., Janka H.-T., 2015, MNRAS, 1 − 1 Bauswein A., Gorielyκo S., Janka H.-T.,v − 2013, ApJ, 773, 78 452, 3894 N , (8) 1 1 (τ 1), Berger× E.,10 2014, cm2 ARA&A,/g 1 52,0 43.3c Grossman D., Korobkin2 O.,γ Rosswog S., Piran T., 2014, MNRAS, 439, ! − " ϵγ (t) − ≥ (9) Berger E., Fong W., Chornock# R., 2013,$ ApJ, 774, L23 ≈757 1 % 2 N & ' (τγ < 1), where κo is the opacity of r-process elements to photons in the optical bands. It is dominated by bound–bound transitions of lan- where N max(τ , τ 2) is the number of the scatterings that a MNRAS 459, 35–43 (2016) γ γ thanides (Kasen, Badnell & Barnes 2013;Tanaka&Hotokezaka photon undergoes= before escaping. Fig. 3 shows the heating rate 2013). For the dynamical ejecta, on the time-scale of tdiff,o, the opti- (left) and the thermalization efficiency (right). Here, the heat- cal depth to γ -rays is much smaller than unity, thereby only a small ing rate is normalized by a simple power-law heating Q˙ 1.3(t) 10 1.3 1 1 = fraction of the γ -rays’ energy is deposited in the ejecta on the peak 10 tday− erg s− g− (Korobkin et al. 2012), where tday is time in time-scale of macronovae. unit of day. The thick (thin) lines in the left-hand panel show the For the slowly expanding wind ejecta, in particular lanthanide heating rate taking the neutrino and γ -ray escape (only the neutrino free cases, the γ -ray heating efficiency is significantly different. The escape) into account. To calculate ϵγ (t), an ejecta mass of 0.01 M opacity to thermal photons and expansion velocity of the wind ejecta is assumed. The velocities of the ejecta is set to be 0.3c for NSM-⊙ 2 1 are κo 1cm g− and v 0.05c (see e.g. Tanaka & Hotokezaka solar and NSM-fission and 0.05c for NSM-wind. For NSM-solar and 2013 for∼ the opacity of the∼ wind case). The estimated optical depth NSM-fission, at 0.5 d, γ -rays start to escape from the ejecta so that to γ -rays is τ γ 3 on the time-scale of tdiff,o. Therefore, in the the thermalization efficiency drops from 0.7 to 0.2–0.3. The heating case of the lanthanide∼ free wind ejecta, γ -rays are weakly coupled rate and thermalization efficiency of NSM-wind are larger than the with the ejected material and heat up the ejecta until a few times other cases since γ -rays are well scattered. For NSM-fission, the tdiff,o. This situation is somewhat similar to those of supernovae, in thermalization efficiency turns to increase due to the spontaneous which γ -rays released from the radioactive decay of 56Ni and 56Co fission of transuranic nuclei and it reaches 0.5 around 10 d. ≈ MNRAS 459, 35–43 (2016) LETTER RESEARCH a 43 b 43 42 42 )] )] –1 –1 s s rg rg 41 41 (e (e L L Mej (M ) 0.017 0.034 Mej (M ) 0.017 0.018 log[ log[ vej 0.2c 0.2c vej 0.2c 0.2c 40 2 –1 40 (cm g ) 0.1 0.1 (cm2 g–1) 10 10 –1.3 –1.0 –2.1 –2.0 26 2 red 2.6 2.1 red 6.0 0.4 ej ej 0.04 0.27 c and .However,thismodelagain and scatter).Withinthiscontextwe rst assume an Fe-peak MMv » » fi 39 39 0.82 opacity as tted parameters (as well as the temperature oor cm g ,andanassociated nding a best- tvalueof k » fi fl fi fi − 1 –1 0 2 1 –1 0 1 This model includes the ejecta mass, ejecta velocity, and 10 Finally, we allow the10 opacity to vary as a free parameter,10 10 10 10 in Metzger (2017) and implemented in Villar et al. (2017). early times, especially in the UV and blue optical bands. We next turn to r-process heating, using the model outlined 4 days, in contrast to the observed rapid decline at rise for » Days since explosionDays since explosion Ni. not powered by the radioactive decay of as well. In particular, the model light curves exhibit an initial 56 ej ej (Nicholl et al. 2017). We therefore conclude that the transient is and ,producesapoor ttothedata 0.03 0.27 c MMv » » fi − Ni fraction is inconsistent with the optical spectra high 2 1 However, such a model (our red model),withbest- tvalues “ ” fi Figure 356 | Model bolometric light curve fits using the Arnett formalism. when opacity is forced to κ = 10 cm g , to all data (blue solid line) and model severely underestimates the NIR light curves, while the g ,leadingtoa red KN (e.g., Barnes & Kasen 2013). “ ” − 1 previous section, but the overall t is poor. In particular, this 10 cm are expected to have a much higher opacity of k = fi 2 ts to the ux and SEDs in the inferred from blackbody κ More recent calculations indicate that lanthanide-rich ejecta fl Mass (Mej), velocityfi (vej), opacity ( ) and a power-law slope for radioactive excluding the first three data points (green dashed line). In all models the 0.75 f . The parameters are comparable to those we » ( 3days),butagainisapoor ttotheNIRlightcurves. fi Ni ej ej ej 0.18 c ,adequatelydescribestheearlylightcurves v » , , and The best- t model has 0.01 0.26 c MMv » β » powering (fi ) are freely variable. Each of these parameters was allowed to maximum allowed velocity is 0.2c, which is also the preferred fit value. ej and ejecta mass and velocity. This model, with 0.03 MM the ejecta (as well as the temperature oor and scatter). » fl 2 2 ń assumed historically in Li & Paczy ski 1998) and tforthe Ni mass fraction in are the ejecta mass and velocity, and the χ χ fi σ vary to give the best fit (reduced56 ( ) are quoted). a, The blue solid line The errors are 1 uncertainties on the data, while the later points after = 0.1 cm g (see Villar et al. 2017). The model parameters opacity of κ 0.1 cm g (our blue model; e.g., as k = ” red “ − − 1 2 1 shows the best2 fit. The green dashed model also includes a thermalization 10 days are uncertain due to systematic effects. The full Markov chain 19 1 simultaneously. Error bars are given at the level in all panels, but may be smaller than the points. efficiency . The recovereds power law (β = − 1.0 to − 1.3) is close to the Monte Carlo analysis and uncertainties are discussed in Methods. parameter. Again, this model fails to capture the late-time NIR behavior. This is suggestive of the fact that we need to model multiple ejecta components model. This model clearly fails to capture any of the observed behavior. Bottom right: tting the data with a single-component KN model with the opacity as a free fi Like the SN model, this t is unable to capture the late-time NIR behavior and overall spectra shape. Bottom left: tting the data with a single-component red KN “ ” fi one predicted infi kilonova radioactivity models (β = − 1.2). b, Best fits ∼ and requires an unphysically large fraction of the ejecta to be synthesized into nickel ( 75%). Top right: tting the data with a single-component blue KN model. “ ” fi / Figure 3. Top left: tting the data with a Type I b c SN model powered by the radioactive decay of Ni. This model clearly fails to capture the late-time NIR behavior fi 56 opacitiesThe of Astrophysical around κ Journal = 10 Letters, cm2 g848:L17−1 would(10pp), 2017be likely October 204,20. In Fig. 3 we produced by high-opacity Cowperthwaite ejecta (dynamic et al. ejecta) could also be possible. show the best fits forcing κ = 10 cm2 g−1. No model with such a high The early blue flux is unlikely to be from a relativistic jet26 and an after- opacity is able to fit all of the data points well, but it can fit the later data glow from the weak gamma ray signal that was detected7,8, owing to the points. In these high-opacity models all observations are still within the rapid reddening and cooling and the X-ray non-detections. diffusion phase, but a steeper power law for energy input (β ≈ − 2) is The optical and near-infrared spectra support the ejecta being dom- favoured to produce the right emergent luminosity, no longer consistent inated by the light r-process elements at least at early stages. We used with t−1.3. If our reconstructed bolometric light curve is accurate at all the TARDIS code27 to construct simple models to guide interpretation epochs, there is not much room for a second component at later times of our spectra. The earliest spectrum (epoch + 1.4 d) we obtained from because the blue one cannot drop faster than the power source term. the New Technology Telescope (NTT, at La Silla, Chile) is fairly well However, it is possible that two-component spectral energy distribu- parameterized by a blackbody of Teff = 5,200 K, and does not show the tion (SED) fitting would give different late-time bolometric estimates. prominent spectral features (Ca, Mg or Si) usually detected in normal Then a two-componentmacro/ kilonovamodel where the early in light GW170817 curve is produced supernova spectra (see Extended Data Fig. 3). There are two broad and by low-opacity ejecta (a wind component), and the later light curve is blended structures at 7,400 Å and 8,300 Å, respectively, which become Thegood Astrophysical indication Journal Letters, 848:L17 but(10pp) ,not 2017 October definitive 20 proof of r-process Cowperthwaite� et al. abSpectral data (+1.4 d) 848:L17 (10pp), 2017 October 20 Cowperthwaite et al. The Astrophysical Journal Letters, 8 6.0 Model continuum Model Cs + Te ) 7 –1 Å 4.0 –1 s g 6 er 37 +1.4 d 0 fset 2.0 (1 λ L Cs I Te I ) + of 5 –1 +2.4 d Å 0 –2 0.4 0.6 0.8 4 cm Rest wavelength (μm) –1 +3.4 d s c g Spectral data (+2.4 d) er 3 Model continuum –16 2.0 Chornock+ 17� +4.4 d ) Model Cs + Te –1 Å Figure 3. Top left: fitting the data with a Type I b/c SN model powered by the radioactive decay of 56Ni. This model clearly fails to capture the late-time NIR behavior Cowperthwaite+ 17� –1 Smartt+ 17� and requires an unphysically large2 fraction of the ejecta to be synthesized into nickel (∼75%). Top right: fis tting1.5 the data with a single-component “blue” KN model. Flux (10 Like the SN model, this fit is unable to capture the late-time NIR behavior and overall+7.3 spectra d shape. Bottomg left: fitting the data with a single-component “red” KN fi model.blue This model kilonova clearly fails to capture any of the observed behavior. Bottom right: tting the data with er a single-component KN model with the opacity as a free parameter. Again, this model fails to capture the late-time NIR behavior. This is suggestive of the37 fact1.0 that we need to model multiple ejecta components simultaneously. Error bars are given1 at the 1s level in all panels,56 but may be smaller than the points. 0 indistinguishable from Ni � (1 +8.3 d λ L 0.5 Cs I Te I k = 0.1 cm2 g−1 (see Villar0 et al. 2017). The model parameters opacity of κ=0.1cm2g−1 (our “blue” model; e.g., as 56 are the ejecta mass and velocity, and the Ni mass fraction in assumed historically0 in Li & Paczyński 1998) and fitforthe fl the ejecta (as well as the temperature0.5 oor1.0 and scatterfeatures1.5). ejecta2.0 from mass and velocity.0.5 This model,1.0 with MMej » 1.50.03 and2.0 fi The best- t model has MMej » 0.01 Rest, vej wavelength» 0.26 c, and (μm) vej » 0.18 c,adequatelydescribestheearlylightcurvesRest wavelength (μm) f Ni » 0.75. The parameters are comparable to thoseindividual we (3days elements??),butagainisapoorfittotheNIRlightcurves. Figure 4inferred | Spectroscopic from blackbody data andfits to model the fl uxfits. and a, Spectroscopic SEDs in the dataMore (black recent calculationsadopted indicateblackbody that continuum lanthanide-rich model ejecta is also shown for reference). c, 2 curves) fromprevious + 1.4 section, d to but+ 4.4 the d overall after discovery,fit is poor. In showing particular, the this fast evolutionare expected toThe have Xshooter a much higher spectrum opacity at of +k 2.4= 10d alsocm shows Cs i and Te i lines that −1 “ ” of the SED.model The severely points underestimates are coeval UgrizJHK the NIR light photometry. curves, while b the, Comparisong ,leadingtoa of are redconsistentKN (e.g., with Barnes the broad & Kasen features2013) .observed in the optical and near- high 56Ni fraction is inconsistent with the optical spectra However, such a model (our “red” model),withbest-fitvalues the + 1.4 d spectrum fiwith a TARDIS spectral model that includes Cs i and56 infrared (here, the lines are indicated at velocities of 0.13c and we include Figure(Nicholl 3. Top et al. left:2017tting). the We data therefore with a Type conclude I b/c SN that model the powered transient by the is radioactiveMMej decay» 0.03 of Ni. Thisand modelvej » clearly0.27 fails c,producesapoor to capture the late-timefittothedata NIR behavior Te i (see andnottext). requires powered Thin an unphysically bylines the indicate radioactive large fraction the decay ofpositions the of ejecta56Ni. to beof synthesizedspectral into lines nickel blueshifted(as∼75% well.). Top In right: particular, fiadditional,tting the the data model with longer-wavelength a light single-component curves exhibit“blue” transitions anKN initial model. to supplement those in b). Like the SN model, this fit is unable to capture the late-time NIR behavior and overall spectra shape. Bottom left: fitting the data with a single-component “red” KN by 0.2c, correspondingmodel.We This next model turn clearly to tor-process the fails tophotospheric capture heating, any ofusing the velocity observed the model behavior. of outlined the Bottom model right: (therisefitting for the» data4 days, with a in single-component contrast to the KN observed model withrapid the opacity decline as a free at parameter.in Metzger Again,(2017 this) modeland implemented fails to capture thein Villarlate-time et NIR al. behavior.(2017). This isearly suggestive times, of the especially fact that we in need the to UV model and multiple blue ejectaoptical componen bands.ts simultaneously.This model Error includes bars are the given ejecta at the mass,1s level ejecta in all panels, velocity, but may and be smallerFinally, than the points. we allow the opacity to vary as a free parameter, 2 NOVEMBER 2017 | VOL 551 | N ATURE | 77 opacity as fitted parameters (as well as the temperature floor finding a best-fitvalueofk » 0.82 cm2 g−1,andanassociated © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. and scatter).Withinthiscontextwefirst assume an Fe-peak MMej » 0.04 and vej » 0.27 c.However,thismodelagain k = 0.1 cm2 g−1 (see Villar et al. 2017). The model parameters opacity of κ=0.1cm2g−1 (our “blue” model; e.g., as are the ejecta mass and velocity, and the 56Ni mass fraction in assumed historically in Li & Paczyński 1998) and fitforthe the ejecta (as well as the temperature floor and scatter). 6 ejecta mass and velocity. This model, with MMej » 0.03 and The best-fit model has MMej » 0.01 , vej » 0.26 c, and vej » 0.18 c,adequatelydescribestheearlylightcurves f Ni » 0.75. The parameters are comparable to those we (3days),butagainisapoorfittotheNIRlightcurves. inferred from blackbody fits to the flux and SEDs in the More recent calculations indicate that lanthanide-rich ejecta previous section, but the overall fit is poor. In particular, this are expected to have a much higher opacity of k = 10 cm2 model severely underestimates the NIR light curves, while the g−1,leadingtoa“red” KN (e.g., Barnes & Kasen 2013). high 56Ni fraction is inconsistent with the optical spectra However, such a model (our “red” model),withbest-fitvalues (Nicholl et al. 2017). We therefore conclude that the transient is MMej » 0.03 and vej » 0.27 c,producesapoorfittothedata not powered by the radioactive decay of 56Ni. as well. In particular, the model light curves exhibit an initial We next turn to r-process heating, using the model outlined rise for »4 days, in contrast to the observed rapid decline at in Metzger (2017) and implemented in Villar et al. (2017). early times, especially in the UV and blue optical bands. This model includes the ejecta mass, ejecta velocity, and Finally, we allow the opacity to vary as a free parameter, opacity as fitted parameters (as well as the temperature floor finding a best-fitvalueofk » 0.82 cm2 g−1,andanassociated and scatter).Withinthiscontextwefirst assume an Fe-peak MMej » 0.04 and vej » 0.27 c.However,thismodelagain 6 Evolution Channels, Multi-Messenger Astronom y short GRBs: X-rays from extended engine activity X-rays from X-ray powered transient magnetar spindown The Astrophysicalmacronova Journal, 818:104 (8pp model), 2016 February 20 Kisaka, Ioka, & Nakar GW than the dynamical timescale t, NS tt .5 τ ∼ 10^8 years Inspiral diff () NS/BH In the case that the central engine is active longer than the photon diffusion timescale, the light curve of the thermal emission follows the evolution of the central engine, i.e., the NS-BH NS-NS peak time is not necessarily when tt diff in the X-ray- powered model. This is the reason why the X-ray-powered model requires a smaller amount of ejecta than the previous models (see Section 3). Finally, the temperature of the ejecta T τ ∼ few ms Merger should be lower than the observed upper limit Tmax 4000 K (Berger et al. 2013; Tanvir et al. 2013), TT 4000 K, 6 prompt max () collapse which is obtained from the detected infrared flux and the optical upper limit (blue arrow in Figure 1). In Sections 2.1– 2.4, we consider these four conditions to give constraints on the ejecta properties. We also consider the luminosity ratio of X-ray and infrared excesses in Section 2.5. delayed collapse 2.1. X-Ray Absorption Figure 2. Schematic picture of theKisaka X-ray-powered model.+ 16 In this subsection, we consider condition (3) for the X-ray photons to be absorbed in the ejecta. The absorption optical where t is the time after the merger. Since the velocity structure depth is given by (Hyper-) Massive becomes homologous and the ejecta spreads to the radial Mej Black Hole – - some lines of sight ,7 direction, the typical mass density of the ejecta is given by X,abs bf 2 () Torus System Neutron Star must have ~1 4 Rej 3Mτej .2where bf is the bound-free opacity. The opacity bf depends quasi-isotropic? ej 3 () τ ∼ few s Post-Merger 4 Rej on the ionization state of the ejecta (Metzger & Piro 2014), broadly beamed?� - Doppler broadening Remnant fn For the composition of the ejecta, we consider both the iron- bf bf,8() adapted from Just rich ejectaless and thesevere heavy r-process ejecta. Since the tidally Am¯ p July. 27th, 2016 RIKEN-RESCEUejected Seminar matter is neutron-rich, the heavy r-process elements where A¯2is the average mass number of the composed elements may be synthesized (e.g., Lattimer & Schramm 1974). On the of the ejecta, mp is the proton mass, bf is the photoionization other hand, the iron-rich ejecta may be dominant if the shocks cross-section, and fn is the neutral fraction of the ejecta. The and/or neutrino irradiation make the electron fraction high ionization state is determined by the balance of the photo- (e.g., Metzger et al. 2009; Martin et al. 2015; Richers et al. ionization due to the X-ray emission from the central engine 2015; Sekiguchi et al. 2015). and the recombination. The neutral fraction fn is described by We consider four conditions to reproduce the infrared excess the ratio of the absorption rate of the ionizing photons and in our X-ray-powered model. First, the X-ray photons should ion the recombination rate as be absorbed by the ejecta, i.e., the optical depth for the X-ray rec absorption X, abs should be larger than unity, 1 ⎛ ion ⎞ fn ⎜1.⎟ () 9 ⎝ rec ⎠ X,abs 1.() 3 The recombination rate is approximately described by Second, the absorbed X-ray photons should be thermalized in the ejecta to produce infrared photons. Since the opacities for rec n e rec,10() the iron and r-process elements are a decreasing function of where ne is the number density of the electrons and rec is the wavelength (Kasen et al. 2013; Tanaka & Hotokezaka 2013), recombination coefficient. The ionization rate is we require the optical depth for infrared photons IR to be larger than unity, ion nc ph bf ,11() where the number density of photons whose energy is larger 1. 4 IR () than the ionization threshold energy is, Third, the infrared photons should escape from the ejecta to be LX nph ,12 detected by the observers. Since the optical depth IR satisfies 2 () 4 hRci condition (4), we consider a random walk for the propagation ej of infrared photons in the ejecta. The infrared photons can where h i is the ionization threshold energy and ò is the escape from the ejecta if the diffusion timescale tdiff is smaller fraction of the luminosity above the ionization energy. 3 6 Murase et al. 43 10-12 radiation luminosity disk emission, baseline (M=0.02Msun, V=0.3c, modest K) radioactive decay disk emission, baseline (M=0.02Msun, V=0.2c, high K) disk emission, optimistic (M=0.02M , V=0.3c, modest K) fall-back disk -13 sun 10 disk emission, optimistic (M=0.02Msun, V=0.2c, high K) 42 ] 10-14 -1 ]) s -1 -2 41 10-15 [erg cm E log(L [erg s -16 E F 10 40 10-17 39 10-18 5 10 15 20 25 30 105 106 107 108 t [d] t [s] 8 Murase etFigure al. 1.ThermalbolometricluminositiesfromaNS- Figure 3.X-raylightcurvesfromaBHwitharemnantdisk NS merger. We show the case where thermal radiation in the disk emission model, for E =3keV(thickcurves)and is mainly47 powered by radioactive decay of r-process ele- E =30keV(thincurves). opacity would be more favorable for long-lived NS rem- B =5x1016G, P =10ms (M=0.02M , V=0.2c, high K) ments and modified* 15 byi X-ray emissionsun from the disk with 40 B*=10−1G, Pi=3ms (M=0.02Msun, V=0.2c, modest K) nants. L 46 B =1013G, P =1ms (M=0.02M , V=0.2c, modest K) disk =10 erg* s 10fori demonstrationsun purposes. Studies of Galactic pulsar wind nebulae suggest that B*=10 G, Pi=1ms (M=0.02Msun, V=0.2c, high K) 6 45 Murase et al. 0 almost all the spin-down power is extracted in the form 10 dissipative outflow (M=0.02Msun, V=0.2c, high K) X-rays from remnant engine dissipativeMurase outflow (M=0.02M+ 1710.10757, V=0.2c, modest K) � ]) sun of a Poynting flux, and eventually released as non- 44 -12 -12 43-1 10 10 disk-1 emission, baseline (M=0.02M , V=0.3c, modest K) disk emission, baseline (M=0.02Mradiationsun, V=0.3c, luminosity modest K) 10 sun thermal nebular emission rather than thermal radiation. disk emission, baseline (M=0.02M , V=0.2c, high K) disk emission, baseline (M=0.02M , V=0.2c, high K) 43 radioactivesun decay sun [erg s disk emission, optimistic (M=0.02M fall-back, V=0.3c, disk modest K) BH disk emission, optimistic (M=0.02Msun, V=0.3c, modest K) th -13 sun -13 The Poynting-dominated pulsar wind is accelerated to 10 disk emission, optimistic (M=0.02M , V=0.2c, high K) 10 disk emission, optimistic (M=0.02Msun, V=0.2c, high K) sun 10-2 42 relativistic speed although its evolution beyond the light 42 log(L disk� ] -14 cylinder has been under debate. In this work, follow- ] -14 10 10 -1 -1 ]) 41 -3 s 10 s [mJy] -1 ν -2 ing Murase et al. (2015)andOmand et al. (2017), we -2 F 40 -15 -15 make the ansatz that microphysical parameters of em- 41 10 10 10-4 [erg cm [erg cm E bryonic pulsar wind nebulae are the same as those of E 39 log(L [erg s -16 5 10 15 20 25 30 -16 E F young pulsar wind nebulae, in which it is assumed that E F 10 10 -5 t [s] 10 electrons and positrons are accelerated with a broken 40 Figure10-17 5.Thermalbolometriclightcurvesofthermalradi- 10-17 power law with s1 =1.5ands2 =2.5 with a break 10-6 5 ation from a NS-NS merger leading to a long-lived NS. A 105 106 107 108 Lorentz factor γb =10 . The magnetic energy fraction is possible contribution from the spin-down energy is included. t [s] 39 10-18 10-18 − 3 4 5 6 5 6 7 8 set to ϵB =0.003 and the rest of the energy (ϵe =1 ϵB) 105 10 10 15 20 10 25 30 10 10 10 10 10 Figure 4.High-frequencyradiolightcurvesfromaBHwith is used for the acceleration of electrons and positrons. t [d] E [eV] Late-Time High-Energyt Signatures[s] from Compact Remnants of Double Neutron Star Mergers 9 aremnantdiskinthedissipativeoutflowmodel,for6 7 ν = In the very early stages of the nebular evolution, non- thick E=3 keV thin E=30 keV 10 thick t=10 ν s thin11 t=10 s FigureFigure 1.ThermalbolometricluminositiesfromaNS- 2. X-ray spectra from a BH with a remnant disk Figure10 3Hz.X-raylightcurvesfromaBHwitharemnantdisk (thin curves) and =10 Hz (thick curves). 106 B =5x1016G, P =10ms (M=0.02M ,6 V=0.2c, high K) 16 16 * i t sun B*=5x10 G, Pi=10ms (M=0.02ME sun, V=0.2c, high K) B*=5x10 G, Pi=10ms (M=0.02Msun, V=0.2c, high K) thermal emission can be thermalized even inside the neb- NS merger.in the disk-8 We emission show15 the model, case at where=10 thermals(thincurves)and radiation in the disk-8 emission15 model, for =3keV(thickcurves)and 15 107 B*=10 G, Pi=3ms (M=0.02Msun, V=0.2c, modest K) 10 B =10 G, P =3ms (M=0.02M , V=0.2c, modest K) 5 B =10 G, P =3ms (M=0.02M , V=0.2c, modest K) t B =1013G, P =1ms (M=0.02M , V=0.2c,r modestd K) E * 13 i sun 10 * 13 i sun ula (Metzger & Piro 2014; Fang & Metzger 2017). At is mainly=10 powereds(thickcurves).Thedistanceissetto* by radioactivei decaysun of -process=40Mpc. ele- =30keV(thincurves).B*=10 G, Pi=1ms (M=0.02Msun, V=0.2c, modest K) B*=10 G, Pi=1ms (M=0.02Msun, V=0.2c, modest K) mentsNote and that modified the baseline by X-ray and emission optimistic from cases the are disk degenerate with 40 6 −1 104 late times, given that the pulsar wind is highly relativis- L t -10 pulsar� diskat=10=1010 ergs. s for demonstration purposes. or10 its-10 corona may be reasonable when the mass accretion ] tic beyond the light cylinder, the pair density quickly ] 3 -1 10 -1 rate is not far from that of ultra-luminous X-ray sources. s 0 s 10 -2 drops, and subsequently the thermalization of non- -2 -12 dissipative outflow (M=0.02M , V=0.2c, high K) 2 10 At10 earlier-12 times, the disk radiationsun luminosity should not 10 dissipative outflow (M=0.02Msun, V=0.2c, modest K) thermal photons starts to occur instead in the merger -12 sion,10 Athena+ (Nandra et al. 2013), is planned to reach largely exceed the Eddington luminosity, but the outflow [mJy] -1 ν 1 F [erg cm disk emission, baseline (M=0.02M , V=0.3c, modest K) 10 10 sun − − − [erg cm ejecta. When intra-source electromagnetic cascades do E 17 2 1 a limiting-14disk sensitivityemission, baseline of (M=0.02MEFE sun∼, 10V=0.2c, higherg K) cm s in E kinetic-14 luminosity can be larger. A significant fraction 10 10 0 E F disk emission, optimistic (M=0.02M , V=0.3c, modest K) not occur in the so-called saturated cascade regime, it -13 sun E F 10 the10 0.2−disk2 emission, keV range, optimistic by (M=0.02M whichsun the, V=0.2c, X-ray high K) emission from of the-2 disk mass, ηw ∼ 0.1 − 1, may be ejected back 10 is relevant to solve kinetic equations to discuss high- -1 the accretion-16 onto the remnant BH is detectable up to into space as an ultrafast outflow, perhaps by viscous 10 10 10-16 energy signatures that can escape from the nebula and ] 10-14 -1 ∼ − -2 300 500 Mpc. Effects of the mass composition on heating10-3 (e.g., Dessart et al. 2009; Fern´andez & Metzger 10 s [mJy] ν ejecta (Murase et al. 2015). -2 the opacity can affect early-time fluxes by a factor of F 2013) and/or magnetohydrodynamic turbulence (e.g., 10-18 -18 -3 In Figure 5, we show thermal bolometric light curves -15 3 4 5 6 7 8 9 10 11 12 10 10 ∼10 -4 5 6 7 8 5 6 7 8 10, which10 can10 be10 regarded10 10 as one10 of10 the10 uncertainties.10 10 Price10 10 & Rosswog 200610 ; Kiuchi et al.10 2015) as well10 as ra- 10 10 10 10 of the thermal luminosity, Lth.Theejectamassandve- [erg cm E [eV] t [s] t [s] E For X-ray observations, deep measurements at ∼ diation pressure (e.g., Ohsuga et al. 2005; Jiang et al. -16 locity are assumed to be M =0.02 M⊙ and V =0.2 c, E F Figuret 10 − 6.High-energyphotonspectraofalong-livedpulsarare important. If the observational limits reach Figure2014 7-5; Narayan et al. 2017). The velocity of such ultra- Figure 8 HX thin 6 10 .X-raylightcurvesfromalong-livedpulsarasa .High-frequencyradiolightcurvesfromalong- respectively. If a long-lived NS exists with Pi ∼ 1ms, left as a compact remnant of the NS-NS merger, at t =10 s E E ν Ldisk tection prospects are quite sensitive to the spin-down ] to c, which is at least different from the case of GW+EM Notenent, that making the baseline the broadband and optimistic spectrum cases are as flat degenerate as EFE ∝ -1 t 6 s 16 at =10 s. 2 orparameters. its corona may We be find reasonable that radio when and the sub-mm mass accretion obser- -2 170817. The case of B∗ =5× 10 GandPi =10ms, const. even beyond E ∼ mec . Hard X-ray emis- 10-12 which is motivated by SGRB extended emission, keeps sion with a very hard spectrum is also expected above ratevations is not are far more from promising, that of ultra-luminous which is consistent X-ray sources. with [erg cm ∼ − AtMurase earlier et times, al. (2016 the), disk who radiation proposed luminosity synchrotron should nebu- not E the ejecta speed comparable to the original ejecta veloc- 10 100 keV. Note that the rotation energy for Pi ∼< 10-14 ity and could put an energy injection at early times. sion, Athena+E (Nandra× et51 al. 2013), is planned to reach∼ E largelylar emission exceed as the a probe Eddington of the luminosity, connection but between the outflow fast E F 10 ms is rot ∼> 4 10 erg, which−17 can exceed−2 −1 5 ej = a limiting sensitivity2 of EF51 E ∼ 10 erg cm s in kineticradio bursts luminosity and pulsar-drivencan be larger. supernovae, A significant including fraction Even if the long-lived pulsar model does not explain (5/2)MV ∼ 4 × 10 erg (Suzuki & Maeda 2017), so -16 − ∼ − 10 GW+EM 170817, such a long-lived NS remnant could the 0a.2 wind2 keV bubble range, breakout by which occurs the through X-ray emission Rayleigh-Taylor from ofsuper-luminous the disk mass, supernovae.ηw 0.1 High-frequency1, may be ejected radio emis- back ∼ 6 − 7 be born in low-mass NS-NS binaries. Our study here theinstabilities, accretion onto which the allowsremnant more BH X-rays is detectable to escape, up would to intosion space can escape as an around ultrafast outflow,10 10 perhapssthankstothe by viscous 10-18 ∼ 300 − 500 Mpc. Effects of the mass composition on heatingsmall ejecta (e.g., mass,Dessart the et fast al. velocity, 2009; Fern´andez and the expectation & Metzger 5 6 7 8 allows for the possible variety of NS remnants. occur. Here Eej is the ejecta kinetic energy. Note that 10 10 10 10 In Figure 6, we show spectra of embryonic pulsar thethe opacity gamma-ray can affect attenuation early-time by fluxes the extragalactic by a factor of back- 2013that) the and/or ejecta magnetohydrodynamic are largely neutral. In turbulence the case where (e.g., t [s] ∼ 15 wind nebulae embedded in the merger ejecta. The spec- 10,ground which light can be is not regarded included as one in this of the work. uncertainties. PriceB∗ =10 & RosswogGand 2006Pi; Kiuchi= 3 ms, et al.the 2015 sub-mm) as well emission as ra- Figure 9.Gamma-raylightcurvesfromalong-livedpulsar ∼ ∼ E tra consist of synchrotron and inverse-Compton com- ForWe X-ray show observations, the X-ray light deep curves measurements in Figure 7 atand the diationcan be detected pressure up(e.g., to Ohsugaz 1.5 et by al. ALMA 2005; Jiang with sensi- et al. as a merger remnant, for =1GeV(thickcurves)and ∼ E =100GeV(thincurves). ponents, and the latter is dominant in the GeV-TeV tHX−thin are important. If the observationalNuSTAR limits reach 2014tivity; Narayan of 0.01 et mJy. al. 2017 Note). thatThe velocity the radio of synchrotron such ultra- radio light curves in Figure 8. Using , with sen- −0.5 Ldisk ∼ Finally, accounting for the potentially dominant MS contri- ˙ −1 6 Inoue, Uchiyama, Arakawa, Renaud &bution, Wada Richter (2012, 2016) give Macc,HVC 0.7 M yr in ∼ ˙ ⊙ HI alone, and a total including ionized gas of Macc,HVC ! Finally, accounting for−1 the potentially dominant MS contri- 5 M yr , more than sufficient for M˙ acc,SF. However, note ⊙ ˙ −1 bution, Richterthat (2012, the MS 2016) contribution give Macc,HVC may0 be.7 transient,M yr in lasting only for ∼ ˙ ⊙ HI alone, and0.5-1 a total Gyr including (Fox et al. ionized 2014). gas of Macc,HVC ! −1 5 M yr , moreHVCs than of sufficient the Leading for M˙ Armacc,SF. is However, known to note be interacting with that⊙ the MS contributionthe disk in the may outer be Galaxy transient, (McClure-Griffiths lasting only for et al. 2008) 0.5-1 Gyr (Fox etdiagnostic al. 2014). to distinguish from more conventional possibili- HVCs of theties Leading like PWNe Arm oris known SNRs to be interacting with the disk in the outerradio Galaxy X-ray (McClure-Griffiths electron acceleration et al. measure 2008) magnetic fields diagnostic toto distinguish confirm sufficient from more to reach conventional maximum possibili- energy ties like PWNe orneutrinos SNRs FIG. 1.— This is a great figure. radio X-ray electrondistinguish acceleration from more measure conventional magnetic possibilities fields 1. energet- to confirm sufficientics 2. non-shell to reach maximum morphology energy 3. sometimes soft X emission, assume HVC properties generally similar to the CHVCneutrinos of IR lines? 4. locations uncorrelated with SF region interarm, FIG. 1.— ThisPark is a et great al. figure. (2016) distinguish from more conventional possibilities 1. energet- −1 5 −3 even outside stellar disk 5. correlated with extraplanar HI? vs = 300 km s MHVC = 10 M nHVC =0.15 cm ,ics so 2.that non-shellno morphology star formation 3. sometimes however soft induced X emission, (Franco et al. 1988; assume HVC properties−3 generally similar to⊙ the CHVC of IR lines? 4. locations uncorrelated with SF region interarm, nt =0.6 cm Izumi et al. 2014) Park et al. (2016) even outside stellar disk 5. correlated with extraplanar HI? v = 300 kmrHVC s−1 M 150pc= 105 M n =0.15 cm−3, so that outer galaxy Note that the 3 FVWs found to be associated s ∼ HVC HVC −3 no star formation however induced (Franco et al. 1988; −3D = 20 kpc R = 15 kpc⊙ ndisk 0.15 cm hdisk 300 pc nt =0.6 cm Izumi et al. 2014)with HVCs by Kang & Koo (2007) all lie at sky positions τ . ∼6 τ . ∼ X-ray6 opacity of NS merger ejecta SI, Hotokezaka, rHVC 150pcThis implies HVC 1 0 10 yr, disk 2 0 10 yrouter and galaxytoward Note that the outer the 3 Galaxy. FVWs found to be associated ∼ 6 ≃ ×−3 ≃ × Murase, Bamba+� D = 20 kpcτrad,Rs = 151.2 kpc10ndiskyr, 0.15 cm hdisk 300 pc with HVCs byGeV Kang Fermi & Koo TeV (2007) HAWC all lie at sky positions ≃ × ∼ ∼ 80 34 angularτ diameter. θ6 2τrHVC/D. 0.866 deg This is some- external galaxies? 1st peak e.g. Se This implies HVC 1 0 10 ∼yr, disk 2 0≃ 10 yr and toward the outer Galaxy. τ 1.2what106 largeryr, ≃ than× the RMS angular≃ diameter× of 0.52 degGeV ob- Fermi TeVouter HAWC galaxy CRs Fermi-LAT diffuseK: 12.7keV, emission L: 1.66keV rad,s ≃ × angularserved diameter byθ HESS,2rHVC/D 0.86 deg This is some- external galaxies?inhomogeneous, transient CR enhancement? low-level dif- ∼ ≃ fuse outer disk emission? what larger than the RMS angular diameter of 0.52 deg ob- outer galaxy CRs Fermi-LAT diffuse emission 2nd peak e.g. 130Te52 served by HESS, 3.2. Results inhomogeneous,HI emission transient CR faint enhancement? diffuse not easy low-level confusion dif- K: 31.8keV, L: 4.94keV outer galaxy fuse outer disk emission?T n R v, M distribution 3.2. Results HI emission faintilluminating diffuse not the easy interface confusion of cold accretion 3FGL point source 195 78 κ 3rd peak e.g. Pt outer galaxyradio T n R v, M distributionintergalactic filament cosmic web satellite gas 3FGL pointX-ray source illuminating the interface of cold accretion K: 78.4keV, L: 13.9keV intergalactic filament cosmic web satellite gas radio infrared? 5. CONCLUSIONS AND OUTLOOK X-ray The parent CHVC for this system may no longer be easily early epochs, for example productionhard of lightX-ray elementsrange infrared? used5. in CONCLUSIONSHotokezaka+ 16 AND� OUTLOOK perceivable due to deceleration and disruption after colliding (Suzuki & Inoue 2002) -> FORCE� The parentwith CHVC the disk. for this system may no longer be easily perceivable due to deceleration and disruption after colliding early epochs, for example production of light elements with the disk. (Suzuki & Inoue 2002) 4. DISCUSSION R v t homologous expansionej = ej low velocity shocks magnetic fields effects of neutral parti- −2 4. DISCUSSION κMej κ Mej vej t cles via charge exchange reactions (Ohira 2012; MorlinoR et al.= v t τ = =2 103 low velocity shocks magnetic fields effects of neutral parti- ej ej avg 4πR2 × 10cm2/g 0.01M 0.3c 104s 2013; Bykov et al. 2014) ej −2 cles via charge exchange reactions (Ohira 2012; Morlino et al. κMej κ Mej vej t ⊙ ! " ˙ τ−1= =2 103 (14) (14) 2013; BykovPutman et al. 2014) et al. (2012) give Macc,HVC 0.1 − 0.4 M yr , in- 2 2 4 ∼ ⊙ 4πRej × 10cm /g 0.01M 0.3c 10 s cluding the ionized gas seen in Hα, but excluding−1 the MS ⊙ Putman et al. (2012) give M˙ acc,HVC 0.1 − 0.4 M yr , in- ! " on the grounds that it may∼ not reach⊙ the disk within 1 cluding the ionized gas seen in Hα, but excluding the MS ∼ on the groundsGyr. that Estimating it may not the reach contribution the disk of within ionized1 gas detected We acknowledge valuable discussions with Changhyun ∼ ˙ Gyr. Estimatingvia metal the lines, contribution Lehner of& Howk ionized (2011) gas detected derive Macc,HVCWe acknowledgeBaek, valuable Takahiro discussions Kudoh, Yutaka with Changhyun Ohira, Ryo Yamazaki, −1 ˙ ˙ ∼ via metal0 lines,.45 − Lehner1.40 M &yr Howk, which (2011) would derive meetMacc the,HVC required MBaek,acc,SF. TakahiroMasahiro Kudoh, Nagashima, Yutaka Ohira, Ryo Yamazaki, −1 ⊙ ∼ 0.45 − 1.40 M yr , which would meet the required M˙ acc,SF. Masahiro Nagashima, ⊙ REFERENCES Abdalla, H., Abramowski, A., Aharonian, F., et al. 2017,REFERENCES ArXiv e-prints Beck, R. 2016, A&A Rev., 24, 4 Abdalla, H., Abramowski,Abdo, A. A., A., Allen, Aharonian, B., Berley, F., et D., al. et 2017, al. 2007, ArXiv ApJ, e-prints 664, L91 Beck, R. 2016, A&ABell, Rev., A. R. 24, 2004, 4 MNRAS, 353, 550 Abdo, A. A.,Abeysekara, Allen, B., Berley, A. U., D., Albert, et al. 2007, A., Alfaro, ApJ, 664, R., et L91 al. 2017, ArXiv e-printsBell, A. R. 2004,Binney, MNRAS, J. 353, 1977, 550 ApJ, 215, 483 Abeysekara,Acero, A. U., Albert, F., Ackermann, A., Alfaro, M., R., Ajello, et al. 2017, M., et ArXiv al. 2016, e-prints ApJS, 223, 26 Binney, J. 1977,Birnboim, ApJ, 215, 483 Y. & Dekel, A. 2003, MNRAS, 345, 349 Acero, F., Ackermann,Ackermann, M., M., Ajello, Ajello, M., M., et al. Atwood, 2016, ApJS, W. B., 223, et al.26 2012, ApJ, 750, 3Birnboim, Y. & Dekel,Bland-Hawthorn, A. 2003, MNRAS, J. & Gerhard, 345, 349 O. 2016, ARA&A, 54, 529 Ackermann, M.,—. 2016,Ajello, ApJS, M., Atwood, 222, 5 W. B., et al. 2012, ApJ, 750, 3 Bland-Hawthorn,Blandford, J. & Gerhard, R. & O. Eichler, 2016, ARA&A, D. 1987, 54, Phys. 529 Rep., 154, 1 —. 2016, ApJS,Ackermann, 222, 5 M., Ajello, M., Baldini, L., et al. 2017, ArXiv e-prints Blandford, R. &Blasi, Eichler, P. D. 2013, 1987, A&A Phys. Rev., Rep., 21, 154, 70 1 Ackermann, M.,Adam, Ajello, R., Ade, M., Baldini, P. A. R., L., Alves, et al. 2017, M. I. ArXiv R., et al. e-prints 2016, A&A, 596, A103Blasi, P. 2013, A&ABrooks, Rev., A. 21, M., 70 Governato, F., Quinn, T., Brook, C. B., & Wadsley, J. 2009, Adam, R., Ade,Aharonian, P. A. R., Alves,F., Akhperjanian, M. I. R., et al. A. 2016, G., Aye, A&A, K.-M., 596, et A103 al. 2005, Science,Brooks, 307, A. M., Governato,ApJ, 694, F., 396 Quinn, T., Brook, C. B., & Wadsley, J. 2009, Aharonian, F., Akhperjanian,1938 A. G., Aye, K.-M., et al. 2005, Science, 307, ApJ, 694, 396Bykov, A. M., Malkov, M. A., Raymond, J. C., Krassilchtchikov, A. M., & 1938 Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R., et al. 2006,Bykov, ApJ, A. M., Malkov,Vladimirov, M. A., Raymond, A. E. 2014, J. C., Collisionless Krassilchtchikov, Shocks A. in M., Partly & Ionized Plasma Aharonian, F., Akhperjanian,636, 777 A. G., Bazer-Bachi, A. R., et al. 2006, ApJ, Vladimirov, A. E.with 2014, Cosmic Collisionless Rays: Microphysics Shocks in Partly of Non-thermalIonized Plasma Components, ed. 636, 777 Armillotta, L., Werk, J. K., Prochaska, J. X., Fraternali, F., & Marinacci,with F. Cosmic Rays:A. Microphysics Balogh, A. Bykov, of Non-thermal P. Cargill, Components, R. Dendy, T. ed. Dudok de Wit, & Armillotta, L., Werk,2016, J. ArXiv K., Prochaska, e-prints J. X., Fraternali, F., & Marinacci, F. A. Balogh, A. Bykov,J. Raymond, P. Cargill, 523 R. Dendy, T. Dudok de Wit, & 2016, ArXivBaek, e-prints C. H., Kudoh, T., & Tomisaka, K. 2008, ApJ, 682, 434 J. Raymond, 523Caprioli, D. 2015, in International Cosmic Ray Conference, Vol. 34, 34th Baek, C. H.,Beck, Kudoh, M. T., C., & Tomisaka, Beck, A. M., K. 2008, Beck, ApJ, R., Dolag, 682, 434 K., Strong, A. W., & Nielaba,Caprioli, P. D. 2015, inInternational International Cosmic Cosmic Ray Ray Conference Conference, (ICRC2015), Vol. 34, 34th ed. A. S. Borisov, Beck, M. C., Beck,2016, A. JCAP, M., Beck, 5, 056 R., Dolag, K., Strong, A. W., & Nielaba, P. International CosmicV. G. Ray Denisova, Conference Z. M. (ICRC2015), Guseva, et al.,ed. A. 8 S. Borisov, 2016, JCAP, 5, 056 V. G. Denisova, Z. M. Guseva, et al., 8 6 Inoue, Uchiyama, Arakawa, Renaud & Wada Finally, accounting for the potentially dominant MS contri- ˙ −1 6 Inoue, Uchiyama, Arakawa, Renaud &bution, Wada Richter (2012, 2016) give Macc,HVC 0.7 M yr in ∼ ˙ ⊙ HI alone, and a total including ionized gas of Macc,HVC ! Finally, accounting for−1 the potentially dominant MS contri- 5 M yr , more than sufficient for M˙ acc,SF. However, note ⊙ ˙ −1 bution, Richterthat (2012, the MS 2016) contribution give Macc,HVC may0 be.7 transient,M yr in lasting only for ∼ ˙ ⊙ HI alone, and0.5-1 a total Gyr including (Fox et al. ionized 2014). gas of Macc,HVC ! −1 5 M yr , moreHVCs than of sufficient the Leading for M˙ Armacc,SF. is However, known to note be interacting with that⊙ the MS contributionthe disk in the may outer be Galaxy transient, (McClure-Griffiths lasting only for et al. 2008) 0.5-1 Gyr (Fox etdiagnostic al. 2014). to distinguish from more conventional possibili- HVCs of theties Leading like PWNe Arm oris known SNRs to be interacting with the disk in the outerradio Galaxy X-ray (McClure-Griffiths electron acceleration et al. measure 2008) magnetic fields diagnostic toto distinguish confirm sufficient from more to reach conventional maximum possibili- energy ties like PWNe orneutrinos SNRs FIG. 1.— This is a great figure. radio X-ray electrondistinguish acceleration from more measure conventional magnetic possibilities fields 1. energet- to confirm sufficientics 2. non-shell to reach maximum morphology energy 3. sometimes soft X emission, assume HVC properties generally similar to the CHVCneutrinos of IR lines? 4. locations uncorrelated with SF region interarm, FIG. 1.— ThisPark is a et great al. figure. (2016) distinguish from more conventional possibilities 1. energet- −1 5 −3 even outside stellar disk 5. correlated with extraplanar HI? vs = 300 km s MHVC = 10 M nHVC =0.15 cm ,ics so 2.that non-shellno morphology star formation 3. sometimes however soft induced X emission, (Franco et al. 1988; assume HVC properties−3 generally similar to⊙ the CHVC of IR lines? 4. locations uncorrelated with SF region interarm, nt =0.6 cm Izumi et al. 2014) Park et al. (2016) even outside stellar disk 5. correlated with extraplanar HI? v = 300 kmrHVC s−1 M 150pc= 105 M n =0.15 cm−3, so that outer galaxy Note that the 3 FVWs found to be associated s ∼ HVC HVC −3 no star formation however induced (Franco et al. 1988; −3D = 20 kpc R = 15 kpc⊙ ndisk 0.15 cm hdisk 300 pc nt =0.6 cm Izumi et al. 2014)with HVCs by Kang & Koo (2007) all lie at sky positions τ . ∼6 τ . ∼ X-ray6 opacity of NS merger ejecta SI, Hotokezaka, rHVC 150pcThis implies HVC 1 0 10 yr, disk 2 0 10 yrouter and galaxytoward Note that the outer the 3 Galaxy. FVWs found to be associated ∼ 6 ≃ ×−3 ≃ × Murase, Bamba+� D = 20 kpcτrad,Rs = 151.2 kpc10ndiskyr, 0.15 cm hdisk 300 pc with HVCs byGeV Kang Fermi & Koo TeV (2007) HAWC all lie at sky positions ≃ × ∼ ∼ 80 34 angularτ diameter. θ6 2τrHVC/D. 0.866 deg This is some- external galaxies? 1st peak e.g. Se This implies HVC 1 0 10 ∼yr, disk 2 0≃ 10 yr and toward the outer Galaxy. τ 1.2what106 largeryr, ≃ than× the RMS angular≃ diameter× of 0.52 degGeV ob- Fermi TeVouter HAWC galaxy CRs Fermi-LAT diffuseK: 12.7keV, emission L: 1.66keV rad,s ≃ × angularserved diameter byθ HESS,2rHVC/D 0.86 deg This is some- external galaxies?inhomogeneous, transient CR enhancement? low-level dif- ∼ ≃ fuse outer disk emission? what larger than the RMS angular diameter of 0.52 deg ob- outer galaxy CRs Fermi-LAT diffuse emission 2nd peak e.g. 130Te52 served by HESS, 3.2. Results inhomogeneous,HI emission transient CR faint enhancement? diffuse not easy low-level confusion dif- K: 31.8keV, L: 4.94keV outer galaxy fuse outer disk emission?T n R v, M distribution 3.2. Results HI emission faintilluminating diffuse not the easy interface confusion of cold accretion 3FGL point source 195 78 κ 3rd peak e.g. Pt outer galaxyradio T n R v, M distributionintergalactic filament cosmic web satellite gas 3FGL pointX-ray source illuminating the interface of cold accretion K: 78.4keV, L: 13.9keV intergalactic filament cosmic web satellite gas radio infrared? 5. CONCLUSIONS AND OUTLOOK X-ray The parent CHVC for this system may no longer be easily early epochs, for example productionhard of lightX-ray elementsrange infrared? used5. in CONCLUSIONSHotokezaka+ 16 AND� OUTLOOK perceivable due to deceleration and disruption after colliding (Suzuki & Inoue 2002) -> FORCE� The parentwith CHVC the disk. for this system may no longer be easily perceivable due to deceleration and disruption after colliding early epochs, for example production of light elements with the disk. (Suzuki & Inoue 2002) 4. DISCUSSION R v t homologous expansionej = ej low velocity shocks magnetic fields effects of neutral parti- −2 4. DISCUSSION κMej κ Mej vej t cles via charge exchange reactions (Ohira 2012; MorlinoR et al.= v t τ = =2 103 low velocity shocks magnetic fields effects of neutral parti- ej ej avg 4πR2 × 10cm2/g 0.01M 0.3c 104s 2013; Bykov et al. 2014) ej −2 cles via charge exchange reactions (Ohira 2012; Morlino et al. κMej κ Mej vej t ⊙ ! " ˙ τ−1= =2 103 (14) (14) 2013; BykovPutman et al. 2014) et al. (2012) give Macc,HVC 0.1 − 0.4 M yr , in- 2 2 4 ∼ ⊙ 4πRej × 10cm /g 0.01M 0.3c 10 s cluding the ionized gas seen in Hα, but excluding−1 the MS ⊙ Putman et al. (2012) give M˙ acc,HVC 0.1 − 0.4 M yr , in- ! " on the grounds that it may∼ not reach⊙ the disk within 1 cluding the ionized gas seen in Hα, but excluding the MS ∼ on the groundsGyr. that Estimating it may not the reach contribution the disk of within ionized1 gas detected We acknowledge valuable discussions with Changhyun ∼ ˙ Gyr. Estimatingvia metal the lines, contribution Lehner of& Howk ionized (2011) gas detected derive Macc,HVCWe acknowledgeBaek, valuable Takahiro discussions Kudoh, Yutaka with Changhyun Ohira, Ryo Yamazaki, −1 ˙ ˙ ∼ via metal0 lines,.45 − Lehner1.40 M &yr Howk, which (2011) would derive meetMacc the,HVC required MBaek,acc,SF. TakahiroMasahiro Kudoh, Nagashima, Yutaka Ohira, Ryo Yamazaki, −1 ⊙ ∼ 0.45 − 1.40 M yr , which would meet the required M˙ acc,SF. Masahiro Nagashima, ⊙ REFERENCES Abdalla, H., Abramowski, A., Aharonian, F., et al. 2017,REFERENCES ArXiv e-prints Beck, R. 2016, A&A Rev., 24, 4 Abdalla, H., Abramowski,Abdo, A. A., A., Allen, Aharonian, B., Berley, F., et D., al. et 2017, al. 2007, ArXiv ApJ, e-prints 664, L91 Beck, R. 2016, A&ABell, Rev., A. R. 24, 2004, 4 MNRAS, 353, 550 Abdo, A. A.,Abeysekara, Allen, B., Berley, A. U., D., Albert, et al. 2007, A., Alfaro, ApJ, 664, R., et L91 al. 2017, ArXiv e-printsBell, A. R. 2004,Binney, MNRAS, J. 353, 1977, 550 ApJ, 215, 483 Abeysekara,Acero, A. U., Albert, F., Ackermann, A., Alfaro, M., R., Ajello, et al. 2017, M., et ArXiv al. 2016, e-prints ApJS, 223, 26 Binney, J. 1977,Birnboim, ApJ, 215, 483 Y. & Dekel, A. 2003, MNRAS, 345, 349 Acero, F., Ackermann,Ackermann, M., M., Ajello, Ajello, M., M., et al. Atwood, 2016, ApJS, W. B., 223, et al.26 2012, ApJ, 750, 3Birnboim, Y. & Dekel,Bland-Hawthorn, A. 2003, MNRAS, J. & Gerhard, 345, 349 O. 2016, ARA&A, 54, 529 Ackermann, M.,—. 2016,Ajello, ApJS, M., Atwood, 222, 5 W. B., et al. 2012, ApJ, 750, 3 Bland-Hawthorn,Blandford, J. & Gerhard, R. & O. Eichler, 2016, ARA&A, D. 1987, 54, Phys. 529 Rep., 154, 1 —. 2016, ApJS,Ackermann, 222, 5 M., Ajello, M., Baldini, L., et al. 2017, ArXiv e-prints Blandford, R. &Blasi, Eichler, P. D. 2013, 1987, A&A Phys. Rev., Rep., 21, 154, 70 1 Ackermann, M.,Adam, Ajello, R., Ade, M., Baldini, P. A. R., L., Alves, et al. 2017, M. I. ArXiv R., et al. e-prints 2016, A&A, 596, A103Blasi, P. 2013, A&ABrooks, Rev., A. 21, M., 70 Governato, F., Quinn, T., Brook, C. B., & Wadsley, J. 2009, Adam, R., Ade,Aharonian, P. A. R., Alves,F., Akhperjanian, M. I. R., et al. A. 2016, G., Aye, A&A, K.-M., 596, et A103 al. 2005, Science,Brooks, 307, A. M., Governato,ApJ, 694, F., 396 Quinn, T., Brook, C. B., & Wadsley, J. 2009, Aharonian, F., Akhperjanian,1938 A. G., Aye, K.-M., et al. 2005, Science, 307, ApJ, 694, 396Bykov, A. M., Malkov, M. A., Raymond, J. C., Krassilchtchikov, A. M., & 1938 Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R., et al. 2006,Bykov, ApJ, A. M., Malkov,Vladimirov, M. A., Raymond, A. E. 2014, J. C., Collisionless Krassilchtchikov, Shocks A. in M., Partly & Ionized Plasma Aharonian, F., Akhperjanian,636, 777 A. G., Bazer-Bachi, A. R., et al. 2006, ApJ, Vladimirov, A. E.with 2014, Cosmic Collisionless Rays: Microphysics Shocks in Partly of Non-thermalIonized Plasma Components, ed. 636, 777 Armillotta, L., Werk, J. K., Prochaska, J. X., Fraternali, F., & Marinacci,with F. Cosmic Rays:A. Microphysics Balogh, A. Bykov, of Non-thermal P. Cargill, Components, R. Dendy, T. ed. Dudok de Wit, & Armillotta, L., Werk,2016, J. ArXiv K., Prochaska, e-prints J. X., Fraternali, F., & Marinacci, F. A. Balogh, A. Bykov,J. Raymond, P. Cargill, 523 R. Dendy, T. Dudok de Wit, & 2016, ArXivBaek, e-prints C. H., Kudoh, T., & Tomisaka, K. 2008, ApJ, 682, 434 J. Raymond, 523Caprioli, D. 2015, in International Cosmic Ray Conference, Vol. 34, 34th Baek, C. H.,Beck, Kudoh, M. T., C., & Tomisaka, Beck, A. M., K. 2008, Beck, ApJ, R., Dolag, 682, 434 K., Strong, A. W., & Nielaba,Caprioli, P. D. 2015, inInternational International Cosmic Cosmic Ray Ray Conference Conference, (ICRC2015), Vol. 34, 34th ed. A. S. Borisov, Beck, M. C., Beck,2016, A. JCAP, M., Beck, 5, 056 R., Dolag, K., Strong, A. W., & Nielaba, P. International CosmicV. G. Ray Denisova, Conference Z. M. (ICRC2015), Guseva, et al.,ed. A. 8 S. Borisov, 2016, JCAP, 5, 056 V. G. Denisova, Z. M. Guseva, et al., 8 X-ray “spectra” through NS merger ejecta assume τ~1 for 3rd peak K edge for line of sight at t=t0 τ ∝t-2 κ log t/t0=0,0.25,0.5,0.75� τ ) X-ray “spectra” through NS merger ejecta assume τ~1 for 3rd peak K edge for line of sight at t=t0 τ ∝t-2 κ log t/t0=0,0.25,0.5,0.75� τ ) X-ray “spectra” through NS merger ejecta assume τ~1 for 3rd peak K edge for line of sight at t=t0 τ ∝t-2 κ log t/t0=0,0.25,0.5,0.75� τ ) extended/plateau X-rays as GW counterparts -4 10 Extended Emission 100Mpc Kisaka+ 17 Plateau Emission CALET /HXM 10-6 INTEGRAL /SPI-ACS ] Fermi/GBM 2 -8 130603B 10 Swift/BAT MAXI /GSC -10 10 ISS-Lobster/WFI 130603B Einstein Probe/WXT Flux [erg/s/cm NB: Swift/XRT 10-12 “scattered component” eROSITA -14 quasi-isotropic 10 Scattered Plateau -> 1 2 3 4 5 6 fluorescent line 10 10 10 10 10 10 features? Duration [s] X-ray reflection spectra accretion disks in X-ray binaries, AGN Gilfanov 10� NSThe Astrophysical merger Journal Letters,ejecta809:L8 (5pp), 2015 August 10 Kisaka, Ioka, & Nakamura of the ejecta is described by 13 ⎛ v ⎞⎛ t ⎞ r 310⎜⎟⎜⎟cm.() 2 ⎝ 0.1c ⎠⎝ 104 s ⎠ On the other hand, the radius of thespectra? plateau emission region� is estimated as 2 212⎛ ⎞ ⎛ t ⎞ rctplateau 310 ⎜⎟⎜⎟cm,() 3 ⎝ 10⎠ ⎝ 1 s ⎠ where Γ is the bulk Lorentz factor of the emitter and t is the flux variability timescale. Both Γ and t have some range for each event, so that the emission continues over 11 14 rplateau 10– 10 cm in an approximately logarithmic way. KisakaSince the+ Lorentz 15� factor is low, 10, inside the jet due to the cocoon confinement (Nagakura et al. 2014) and thus the Figure 1. Schematic picture for the scattering of plateau emission and the relativistic beaming angle is larger than the jet opening angle engine-powered macronova. X-ray photons emitted from the inside of the jet 1 j, most emission from the jet can reach the boundary (light blue region) are scattered by the optically thick ejecta (thick arrow). The between the jet and the ejecta as shown in Figure 1. In addition, gray region is effectively thin, and the red region is effectively thick. the range of rplateau covers the sweet spot for the scattering, 12 rplateau r 3 10 cm. Therefore, the observers with a with the sensitivity of X-ray observations. In Section 3, we large viewing angle ( j) with respect to the jet axis would present the model of a macronova powered by the plateau detect the scattered X-ray photons of the plateau emission. activity, which explains the observations of GRB 130603B. We parameterize the scattered luminosity of the plateau Finally, we present discussions in Section 4. emission Lrf using a parameter ò as LLrf iso,pl,4() 2. SCATTERED X-RAY EMISSION where Liso,pl is the observed isotropic luminosity of the plateau Figure 1 shows a schematic picture for the scattering of the emission. We first consider the isotropically scattered compo- emission from the jet (the thick arrow). A significant fraction of nent whose energy is comparable to that before the scattering. photons that are emitted with angle j relative to the jet axis Then, the luminosity of the scattered component is 2 3 2 could be scattered at a large angle by the surrounding ejecta if Lrf [( j 25100.1 )]LLpl (j ) pl. Taking account the optical depth for the Thomson scattering is larger than of the widespread region of the emission, we use 10 3 as a unity, nrT 1, where n is the electron number density fiducial case. Some geometrical models suggest and T is the Thomson cross section. Using the assumption of 54 3 10 3 10 in the case of rplateau r 0.1 homologous expansion for the ejecta (Hotokezaka et al. 2013), (e.g., Equation (3) in Eichler & Levinson 1999), which is a the radius of the ejecta r is described by the velocity v and the bit smaller than the fiducial value. Note that the light crossing time since the merger t as r vt. The number density10 is time of the emission region, rc 100 s, is smaller than the described by n MAmvt¯ 33, where A¯ is the average mass j ej( p ) plateau duration so that the luminosity is not reduced by the number of the nuclei in the ejecta and m is the proton mass. If p time stretch due to the light crossing. the ejecta mainly consist of the r-process elements, we have In Figure 2, we plot the light curves of the plateau emission A 100 (e.g., Lattimer & Schramm 1974). Then a typical ¯ (the black dashed curve) and the scattered component with value of the optical depth is 10 3 (the red solid curve). A black dashed curve is the model light curve (Equation (12) in Kisaka & Ioka 2015) with 2 1 46 −1 2 ⎛ tA⎞ ⎛ ¯ ⎞ the luminosity Liso,pl 8 10 erg s and duration 10 ⎜⎟ 4 ⎜ 2 ⎟ 3 fl ⎝ 10 s⎠ ⎝ 10 ⎠ tinj 8.5 10 s. We also plot the ux (0.3–10 keV) of the 11 2 plateau emission assuming that a GRB 130603B–like event ⎛ Mej ⎞⎛ v ⎞ ⎜⎟,1occurs at the distance of 100 Mpc (red crosses). The photon ⎜ 2 ⎟ () ⎝ 10M ⎠⎝ 0.1 c ⎠ index of the plateau emission is about 2 so that the flux of the plateau emission does not strongly depend on the energy where c is the speed of the light. Therefore, the surrounding range. To see the detectability, we also plot the sensitivities of ejecta are optically thick to the Thomson scattering during the some soft X-ray detectors (blue dotted lines). 4 As shown in Figure 2, the flux of the scattered component plateau activity timescale (∼10 s). −3 fl Another condition to scatter a significant fraction of the with ò ∼ 10 at 100 Mpc is comparable to the ux sensitivity plateau emission is that the radius of the plateau emission limit of the ISS-Lobster/WTI with integration time 450 s (Camp et al. 2013) and the Einstein Probe/WXT with region is smaller than that of the expanding ejecta (Figure 1). integration time 1000 s (Yuan et al. 2015). ISS-Lobster/WTI Since the typical velocity of the ejecta is v 0.1c, the radius and Einstein Probe/WXT can cover a wide field of 900 and 10 Typically, the nuclei are weakly ionized. 11 http://www.swift.ac.uk/index.php 2 まとめ� I. 遠方宇宙 ��high-z GRB：伸び悩み？（期待外れ？） ��宇宙再電離： �以前ほど派手なことは起こっていないかも �� �が、まもなく手が届くところに？ �� �後の銀河の進化 、宇宙のバリオン・ DM分布などの �������理解にとっても重要 �� �クエーサーの寄与再考の価値あり ��ガンマ線吸収も II. 元素の起源 ��NS mergerにおけるr-processのX線診断 ��macro/kilonovaより直接的検証 、モデル判定の可能性 ��より詳細な検討中