arXiv:1707.03719v2 [astro-ph.HE] 18 Aug 2017 otda el h w eecpso-or the ex- on-board significant any telescopes detect two elec- not The re- an could promptly for satellite well. on- were search Fermi from as observations to obtained ported serendipitous facilities Results by of going out counterpart. number carried tromagnetic large were observations a follow-up far. extensive so discovered event GW confirmed remote most M 2 e.Teeetwsascae ihtemerging the 31 with over of masses associated spanning with was holes each event black two The arcs, The of elongated deg. GW170104 2017). two 120 of al. of region et (Abbott localization consisted years confidence than 90% 000 less 70 LIGO was over detection event alarm the false one with The associated UTC. 10:11:58.6 probability 2017-01-04 on dis- covered GW170104, event, (GW) gravitational-wave nificant rpittpstuigL using 2018 typeset 28, Preprint September version Draft olwn h noneetb h IOteam, LIGO the by announcement the Following h IOVroclaoainrpre hr sig- third a reported collaboration LIGO/Virgo The ⊙ 7 P,Atoatcl tCsooi,UiestePrsDide Universit´e Paris Cosmologie, et AstroParticule APC, tadsac f880 of distance a at .Savchenko V. .Diehl R. 6 noartobtentepop nryrlae in released energy prompt the between ratio a into pcfi oiinwti hsrgo,teuprlmt nerdfo h I the from inferred s limits particular upper the the on region, from Depending this within trigger. position LIGO exam specific the to of us allowed region INTEGRAL The localization on-board instruments collaboration. the LIGO/Virgo of the capability by discovered GW170104, event rvttoa aeeeg fE of energy wave gravitational the pe-iiso the on upper-limits faysgicn eeto fapop iheeg event. high-energy prompt a we headings: of as Subject far detection as significant a results. which, any INTEGRAL optimal from of AstroSAT, the not with and relatively Fermi/GBM was compatible a by be istruments independently only still t INTEGRAL would is with E2 There the for compatible of region. estimated not sensitivity localization is LIGO the 90% E2, where the event, of AGILE/MCAL most the within of flux reported uoenSaeRsac n ehooyCnr (ESA/ESTEC) Centre Technology and Research Space European 8 S/ruSriedAtohsqe a.79Om e Meris des Orme 709 Bat. d’Astrophysique, DSM/Irfu/Service eue aafo h NEntoa am-a srpyisLa Astrophysics Gamma-Ray INTErnational the from data used We 10 13 ocwIsiueo hsc n ehooy ntttkyp Institutskiy Technology, and Physics of Institute Moscow γ a lnkIsiuefrAtohsc,Karl-Schwarzschil Astrophysics, for Institute Planck Max rypooe onepr oG100 yteAIEta iht with team AGILE the by GW170104 to counterpart proposed -ray 1 .Martin-Carillo A. F 9 2 SC eateto srnm,Uiest fGnv,chemi Geneva, of University astronomy, of Department ISDC, pc eerhIsiueo usa cdm fSine,Pr Sciences, of Academy Russian of Institute Research Space NFIsiuefrSaeAtohsc n lntlg,Vi Planetology, and Astrophysics Space for INAF-Institute 3 γ 12 T pc ainlSaeIsiueEetoe ulig3 Building - Elektrovej Institute Space National - Space DTU 4 =1.9 .Hanlon L. , nvri´ olue P-M;CR;IA;9A.Rce P4 BP Roche, Av. 9 IRAP; CNRS; UPS-OMP; Universit´e Toulouse; 5 1. pc cec ru,Sho fPyis nvriyCollege University Physics, of School Group, Science Space 1 .Ferrigno C. , A × INTRODUCTION T 10 E − + tl mltajv 12/16/11 v. emulateapj style X 390 450 − 7 g r cm erg 4 p.G100 stu the thus is GW170104 Mpc. 0reAieDmn tLeneDqe,725PrsCdx1,F 13, Cedex Paris 75205 L´eonie Duquet, et Domont Alice rue 10 a-lnkIsiu f Max-Planck-Institut γ 5 Dtd cetdXX eevdYY noiia omZZZ) form original in YYY; Received XXX. Accepted (Dated: .vnKienlin von A. , ryadhr -a rmteiso soitdwt h gravitatio the with associated emission prompt X-ray hard and -ray 11 5 NF AFMln,vaEBsii1,I213Mln,Italy Milano, I-20133 15, E.Bassini via IASF-Milano, INAF, 1 .Mereghetti S. , − .Bozzo E. , NERLOSRAIN FGW170104 OF OBSERVATIONS INTEGRAL 2 − + to 6 8 . γ 4 / F E M γ GW =10 ⊙ rf eso etme 8 2018 28, September version Draft 1 < n 19 and .Bazzano A. , 4 − rEtaersrsh hsk acig Germany Garching, Physik, Extraterrestrische ur ¨ 2.6 .Kuulkers E. , 6 r cm erg × 11 10 ABSTRACT .P Roques P. J. , o,CR/NP,CAIf,Osraor ePrsSorbonne Paris de Observatoire CEA/Irfu, CNRS/IN2P3, rot, − + 5 5 − . . 9 3 5 sn h NERLrsls ecnntconfirm not can we results, INTEGRAL the Using . − 2 γ 7 e e nryrne.Ti translates This range). energy MeV 2 - keV (75 2 ry ln h ieto oteosre n the and observer the to direction the along -rays .Brandt S. , ein h pe ii eie rmteFermi-GBM spanning the fluence from 5.2 1-second derived from a to limit corresponds localization upper observations LIGO The the time of al. the region. 82.4% et at enclosing (Burns 69.5% GW170104, of of event coverage 2017). Collaborations tem- sky GW LAT provided and Fermi Fermi-GBM and the spatially GBM Fermi was with 2017; that compatible background porally the over cess 00kVeeg ag n suigatpclBn spec- short Band a typical of a assuming trum and range energy keV 1000 ra oniec rbblt sbten2.4 between post- ratio is corresponding the noise probability and to coincidence 4.4 signal trial is detection estimated the The of (SNR) event. GW background time the instrument in with coincident the roughly over was that excess (AGILE-GW170104) an revealed also lite 2017b). was al. et (c.l.) (Svinkin level Konus-Wind confidence 2017). by al. 95% reported et at in (Bhalerao AstroSAT sky non-detection by the A reported of region was restricted event more the a of fluence the on -t.1 acigb uce -54,Germany D-85741, Munchen b. Garching 1, d-Str. 6 r ,Dlordy ocwRgo,110,Russia 141700, Region, Moscow Dolgoprudny, 9, er. esCASca,911GfsrYet ee,France Cedex, Gif-sur-Yvette 91191 Saclay, CEA iers .Laurent P. , marybrt–gaiainlwaves gravitational – burst amma-ray n fteisrmnso-or h GL satel- AGILE the on-board instruments the of One elran1 21A orwj,TeNetherlands The Noordwijk, AZ 2201 1, Keplerlaan , os e aair 0,013Rm,Italy 00133-Rome, 100, Cavaliere del Fosso a founy 43,179 ocw Russia Moscow, 117997 84/32, ofsoyuznaya 12 d’ n 7D-80KnesLnb Denmark Lyngby Kongens DK-2800 27 .Sunyaev R. , × uln efil,Dbi ,Ireland 4, Dublin Belfield, Dublin, cga 6C-20Vrox Switzerland Versoix, CH-1290 16 Ecogia, ´ 10 3 − .Chenevez J. , 7 36 -12 olue France Toulouse, F-31028 4346, r cm erg 7 γ , nqeon-ietoa viewing omni-directional unique 8 n h ul9%cndnelevel confidence 90% full the ine eta oe sue n the and assumed model pectral rybrt R) ihe pe limit upper tighter A GRB). burst, -ray .Lebrun F. , TGA bevtosrange observations NTEGRAL r wr,teeaen reports no are there aware, are oaoy(NERL oset to (INTEGRAL) boratory dtelws loe fluence allowed lowest the nd hsrgo a loobserved also was region This eITGA pe limits upper INTEGRAL he eMA ntuet The instrument. MCAL he 9 , − 13 rance. iie oto ftesky the of portion limited 2 n .Ubertini P. and , o9.4 to 3 .J-.Courvoisier J.-L. T. , 8 .Lutovinov A. , × 10 − 7 r cm erg a wave nal 2 − σ 2 ai Cit´e,Paris 9 , n 2.7 and 10 i h 8- the (in , 1 , σ 2 V. Savchenko et al.

(Verrecchia et al. 2017). based on the INTEGRAL mass model (Sturner et al. In this letter, we make use of the available data 2003) and simulated the propagation of monochromatic collected by the instruments on-board INTEGRAL parallel beams of photons in the 50 keV-100 MeV en- (Winkler et al. 2003) to search for possible hard X-ray ergy range. For each energy, we simulated 3072 sky and γ-ray counterparts to GW170104. We summarize positions (16-side HEALPix 2 grid). This allows us to the most relevant capabilities of the INTEGRAL instru- generate an instrumental response function for any po- ments for these kinds of searches in Sect. 2 and describe sition in the sky, which can then be used to compute all the obtained results in Sect. 3. We discuss the non- the expected number of counts for a given source spec- detection of a counterpart to the GW event in the IN- tral energy distribution. As shown in our previous paper TEGRAL data with respect to the findings reported by (Savchenko et al. 2017), this response produces results the AGILE team in Section 3.1. Our conclusions are for the bursts detected simultaneously by the SPI-ACS reported in Section 4. and other detectors (Fermi/GBM and Konus-Wind) that are consistent to an accuracy better than 20%. 2. THE INTEGRAL INSTRUMENTS AND THE FOLLOW-UP OF GW EVENTS 3. INTEGRAL OBSERVATIONS OF GW170104 As extensively described by Savchenko et al. (2017), At the time of the GW170104 (2017-01-04 10:11:58.6 INTEGRAL provides unique instantaneous coverage of UTC, hereafter T0) INTEGRAL was fully operational the entire high-energy sky by taking advantage of the and executing the pointing ID. 176700040010 in the di- synergy between its four all-sky detectors: IBIS/ISGRI, rection of Cas A / Tycho SNR, far from the likely lo- IBIS/PICsIT, IBIS/Veto, and SPI-ACS. These provide calization region of the LIGO trigger. All instruments complementary capabilities for the detection of tran- were performing nominally, yielding a virtually constant sient events characterized by different durations, loca- and stable background count rate from at least T0 - tions on the sky, and spectral energy distributions. In 2500 to T0 + 2500 ks. The SPI-ACS background count the case of the first GW event, GW150914, the most rate was about 1.14×105 counts s−1, which is higher than stringent upper limit on the non-detection of an elec- that observed during LVT151012 or GW150914 and close tromagnetic counterpart in 75 keV to 2 MeV energy to the maximum value ever observed in SPI-ACS data range with INTEGRAL was obtained with the SPI-ACS during the INTEGRAL mission lifetime (excluding the (Savchenko et al. 2016), while the peculiar localization time intervals affected by Solar flares). There are two of LVT151012 (Abbott et al. 2016) and its orientation reasons for the high background recorded at the time with respect to the INTEGRAL satellite required the of GW170104: the 11-years Solar activity cycle, which combination of the results from all detectors (together is close to its minimum, and the day-scale variations with a careful analysis of each instrument’s response and of the instrumental background, which have been com- background) to achieve an optimized upper limit. As we monly observed since the early stages of the instrument discuss in Sect. 3, it is again the SPI-ACS that provides operations. The enhanced background rate decreases the most stringent upper limit on the high energy emis- the sensitivity of INTEGRAL instruments by as much sion from the non-detected electromagnetic counterpart as 30%, when compared to the most favorable condi- to GW170104. tions and much less, when compared to our reports on The SPI-ACS (von Kienlin et al. 2003) is made of 91 LVT151012 and GW150914. However, it should be no- BGO (Bismuth Germanate, Bi4Ge3O12) scintillator crys- ticed that the effects of background fluctuations on the tals and it is the anti-coincidence shield of the SPI instru- sensitivity are typically smaller than those due to sky ment (Vedrenne et al. 2003). Besides its main function location. At the time of GW170104, the Earth was rela- of shielding the SPI germanium detectors, the ACS is tively distant from INTEGRAL, casting a small shadow also used as a nearly onmidirectional detector of transient of 49.0 deg2 on the instrument field-of-view (equivalent events, providing a large effective area at energies above to 0.12% of the sky) and occulting only about 0.032% ∼75 keV. The ACS data consist of event rates integrated of LIGO event localization probability. In the remaining over all the scintillator crystals with a time resolution part of this region, the SPI-ACS sensitivity was close to of 50 ms. No spectral and/or directional information of optimal. Thus, this instrument allowed us to carry out the recorded events is available. The typical number of the most accurate search for any electromagnetic coun- counts per 50 ms time bin ranges typically from about terpart to GW170104. For a fraction of the 90% LIGO 3000 to 6000. SPI-ACS features a high duty cycle of localization region, the IBIS sensitivity, including both 1 ∼85% and comprises events from the nearly complete ISGRI and PICsIT, (Ubertini et al. 2003) approached high energy sky. that of the SPI-ACS, but we checked that adding these SPI is partially surrounded by the satellite structure data did not significantly improve our results. Therefore and by the other INTEGRAL instruments, which shield we do not extensively comment on the IBIS data but re- the incoming photons and thus also affect the response port for completeness in Fig. 1 a comparison between the of the ACS in different directions. For this reason, the contributions provided by the SPI-ACS, IBIS/Veto, and computation of the ACS response requires detailed simu- ISGRI in searching for an electromagnetic counterpart of lations which take into account the entire satellite struc- GW170104. In this figure, we estimated for each value of ture. We developed a GEANT3 Monte-Carlo model the upper limit the integrated fraction of the entire LIGO localization region of the GW event that is probed by the 1 The reduction of 15% is due to the fact that the INTEGRAL instruments are switched-off near the perigee of every satellite rev- data of the different INTEGRAL instruments. The SPI- olution. The INTEGRAL orbit was as long as three sidereal days ACS is clearly able to provide the deepest limits in the until January 2015, but was later shortened to 2.7 to allow for a safe satellite disposal in 2029. 2 http://healpix.sourceforge.net INTEGRAL observations of GW170104 3 entire portion of the sky where the LIGO localization considered time interval. The upper limit derived in probability is significantly larger than zero. this way corresponds also to the 3-σ detection thresh- The INTEGRAL Burst Alert System (IBAS) old, which is the generally recommended approach to (Mereghetti et al. 2003) routinely inspects the IN- compute upper limits corresponding to the non detec- TEGRAL SPI-ACS and IBIS/ISGRI lightcurves in real tion of astrophysical events (Kashyap et al. 2010). Our time, searching for significant deviations from the back- method complies to the commonly accepted upper limit ground and producing automatic triggers. The closest definitions, used for example by the Fermi/GBM team IBAS trigger to GW170104 occurred on 2017-01-04 (Fermi GBM and Fermi LAT Collaborations 2017). The 22:12:40 (T0 + 43241 s) and was classified as a cosmic results obtained in these two cases are shown in Fig. 3 ray event, thus unlikely to be related to the LIGO and 4. The estimated upper limits (75 keV - 2 MeV) trigger. within the LIGO 90% localization region range from −7 −2 −7 −2 The closest event identified as a possible GRB in INTE- Fγ =1.9×10 erg cm to 3.5×10 erg cm for a 1- −7 −2 GRAL data occured at 2017-01-05 06:14:06 with a SNR second short hard GRB and from Fγ =5.2×10 erg cm of 9.3 and a duration of 5 seconds. The astrophysical na- to 10−6 erg cm−2 for an 8-second event characterized by ture of this event was confirmed by simultaneous obser- a typical long GRB spectrum. vations of Konus-Wind (Svinkin et al. 2017a), AstroSAT Assuming the reference distance to the event of (Sharma et al. 2017), POLAR (Marcinkowski & Xiao D=880 Mpc (Abbott et al. 2017), we can derive an up- 2017), and a combined IPN analysis (Svinkin et al. per limit on the isotropic equivalent total energy released 2017a). This was classified as a regular long GRB in the 75 - 2000 keV energy band in one second as (GRB170105) with an optical afterglow that could also × 49 Fγ D 2 be independently found in the ATLAS follow-up obser- Eγ < 3.2 10 erg −7 −2 . The energy  3.5×10 erg cm  880Mpc  vations of GW170104 (ATLAS17aeu; Tonry et al. 2017; emitted in gravitational waves can be estimated as EGW Melandri et al. 2017; Stalder et al. 2017; Bhalerao et al. +1.1 × 54 = 3.6−1.3 10 erg. The SPI-ACS upper limits we re- 2017). INTEGRAL observations contributed to the tri- ported above can constrain the fraction of energy emit- angulation which allowed the establishing the association ted in hard X-rays and γ-rays towards the observer dur- between GRB170105 and ATLAS17aeu (Svinkin et al. − ing the GW event to be f < 9 × 10 6 in the case of the 2017a). In general, INTEGRAL data are particularly γ × −5 useful to retrospectively search for GRB events, owing to short-hard burst, and fγ < 2.6 10 in the case of the its competitive and consistent omnidirectional sensitiv- long-soft one (in the 75 - 2000 keV energy range). ity, stable background, and high duty cycle (see e.g. a re- While the limit on the fraction of the gamma-ray en- cent case studied by Whitesides et al. 2017). GRB170105 ergy emitted in the energy range covered by SPI-ACS was later found to be likely unrelated to GW170104 has the advantage of depending the least on the as- (Stalder et al. 2017; Bhalerao et al. 2017). sumed source spectrum, it is of a general interest to We also inspected the SPI-ACS and IBIS light curves, estimate a limit on the bolometric luminosity. In the 1 - 10000 keV energy range that is conventionally used focusing on a time interval of ±500 s around T0 and prob- ing 5 different time scales in the range 0.05-100 s. The (e.g. Rowlinson et al. 2014; Pescalli et al. 2016), we can latter were selected to be representative of the dynamical constrain the total released electromagnetic energy and × 49 time scale of the accretion occuring in a coalescing com- its ratio to the GW energy as E1−105keV < 3.5 10 × −6 pact binary (e.g. Lee & Ramirez-Ruiz 2007). We did not ( f1−105keV < 9.8 10 ) in the case of the short-hard × 50 × −5 find any obvious detection of a significant signal tem- burst, and E1−105keV < 1.3 10 ( f1−105keV < 3.7 10 ) porally coincident with the GW event. A zoom of the in the case of the long-soft one. SPI-ACS lightcurve around the time of the LIGO trigger is shown in Figure 2. 3.1. On the possible AGILE detection of an Following the approach in Savchenko et al. (2016); electromagnetic counterpart to GW170104 Savchenko et al. (2017) and the non-detection of any sig- AGILE is an X-ray and γ-ray astronomical satellite of nificant electromagnetic counterpart to GW170104 in the The Italian Space Agency, launched in 2007. AGILE’s INTEGRAL data, we derived the corresponding upper scientific payload comprises a pair-conversion telescope, limits assuming the cases of (i) a short-hard burst, i.e. capable of detecting photons in the 30 MeV - 100 GeV a 1 s-long event characterized by a cut-off power-law energy range (GRID), and a hard X-ray monitor sensitive spectral energy distribution with parameters α = −0.5, in the 18 - 60 keV energy range (SuperAGILE or SA). Epeak = 600 keV; (ii) a long-soft burst, i.e. an 8-s long Additionally, AGILE is able to observe bright impulsive event whose spectral energy distribution is described events from a large fraction of the unocculted sky with by the Band model (Band et al. 1993) with parameters its mini-Calorimeter (MCAL), operating in the energy α = −1, β = −2.5, and Epeak = 300 keV. While the refer- band 0.4-100 MeV (Tavani et al. 2008). ence short GRB duration of 1 s is close to the peak of Verrecchia et al. (2017) reported on observations car- the short GRB duration distribution, the 8 s time scale ried out with the MCAL at the time of GW170104. These for the long-soft GRB is motivated by the sampling rate observations covered only a fraction of the LIGO localiza- of IBIS/Veto, in analogy with the approach presented tion, due to the occultation of the AGILE FoV caused by previously by Savchenko et al. (2017). the Earth. Several weak bursts were identified in the AG- To calculate the 3-σ upper limits, we fold the spec- ILE/MCAL data around the time of GW170104. Among tral model through the instrument response for each them, the 32 ms-long burst E2 was identified as a pos- sky location and adjust the model normalization until sible γ-ray counterpart of the GW event. The reported the predicted number of counts is equal to three times trigger time is at 0.46 ± 0.05 s before T0. the standard deviation of the background counts in the Following the report by (Verrecchia et al. 2017), we in- 4 V. Savchenko et al.

Hard, 1 s Soft, 8 s 1.0 1.0 Combined/SPI-ACS ISGRI IBIS/Veto 0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2 Combined/SPI-ACS

probed by the INTEGRAL instruments theINTEGRAL by probed instruments theINTEGRAL by probed ISGRI IBIS/Veto Fraction of the GW170104 localization probability the GW170104 of localization Fraction probability the GW170104 of localization Fraction 1 1 ,2, 75 - 2000 keV limit, erg cm- 2000 keV

Fig. 1.— Plot of the fraction of the LIGO localization probability of GW170104 probed by the data of the different INTEGRAL instruments as a function of the upper-limit (3σ c.l.) on the non-detected electromagnetic counterpart to the GW event. The figure on the left is for the case of the short-hard burst, while the figure on the right shows the case of a long-soft burst (see text for details). The ”Combined/SPI-ACS” text in the label indicates that the results do not quantitatively change if only the SPI-ACS data are used to draw the blue solid line or if the independent contributions from the other instruments are also merged.

Fig. 3.— Estimated 3σ upper limits on the 75-2000 keV flux Fig. 2.— Zoom of the INTEGRAL/SPI-ACS lightcurve in the of the non detected electromagnetic counterpart to GW170104 as ±10 s time interval around the LIGO detection of GW170104. Light derived from the SPI-ACS data assuming the case of a short-hard blue symbols represent the measurements at the natural instrument burst. The black contours show the most accurate localization of time resolution of 50 ms, while dark blue points correspond to the GW event at 50% and 90% c.l., as provided by the LALInfer- the data rebinned at a resolution of 250 ms. The dashed black ence (Abbott et al. 2017). curve represents the average instrument background obtained from a much longer span of data. vestigated the INTEGRAL data to check for any confir- mation of this detection. We note that, unlike the upper limit presented in the previous Section (Figs 3 and 4), we need to compute the upper bound in the flux of any possible celestial event corresponding to the measured signal in SPI-ACS at the exact time of the AGILE puta- tive event. However, the INTEGRAL orbit is very elon- gated resulting in a sizable difference in a celestial sig- nal arrival time, which depends on the unknown source sky location, reaching up to ±0.32 s. First, we com- puted for each position in the sky the time at which the event AGILE-GW170104 should have been observed by INTEGRAL. For each position in the sky at the proper trigger time, we show with a colour map in Fig. 5 the corresponding 90% c.l. values of the upper bound on the 400 – 40000 keV fluence consistent with the SPI- Fig. 4.— ACS count rate3. The reported values are calculated Same as Fig. 3 but in the case of a long-soft burst.

3 Note that the 90% c.l. was preferred to the 3σ approach to compare more easily the INTEGRAL and AGILE findings. INTEGRAL observations of GW170104 5

magenta contour on Fig. 5). The ensemble of these po- sitions covers about 4.2% of the LIGO localization re- gion and extends for a total of 77.5 deg2. Note that a few small regions enclosed within red dashed lines are sparsely present in the color map of Fig. 5. These are positions in the sky for which the AGILE trigger time of AGILE-GW170104 corresponds to positive count rate fluctuations in the SPI-ACS lightcurve. We inspected each of these fluctuations, but none of them exceeded a S/N of 1.5. Taking together all these results, we cannot exclude that the event AGILE-GW170104 is associated with the GW trigger if it originated from a restricted number of positions in the sky within the 90% LIGO localization region. However, this detection is compatible with the INTEGRAL results only if a fluence that is a factor of Fig. 5.— Plot of the estimated lowest detectable fluence at 90% 1.2 lower than the best fit value obtained from the AGILE c.l. by the SPI-ACS for a 32 ms long burst going off at the time data is considered. of AGILE-E2 in different positions of the sky (a spectral energy We noticed that the limited positions in the sky distribution with a slope of -2 has been assumed). The large dashed circle corresponds to the location occulted for AGILE by within the 90% LIGO localization region for which the the Earth. The small dark circle represents the region occulted AGILE/MCAL detection is compatible with the INTE- by the Earth to INTEGRAL. Solid red lines enclose the regions GRAL results were also accessible to the Fermi/GBM where the lowest detectable SPI-ACS fluence is higher than the (Fermi GBM and Fermi LAT Collaborations 2017; best fit one (8.9×10−8 erg cm−2) obtained for the tentative AGILE counterpart of GW170104 (i.e., the event E2). Dashed red lines Burns et al. 2017) and, in an even more limited way, are used for the same comparison with the lower boundary of the by the AstroSAT/CZTI (Bhalerao et al. 2017). Further AGILE fluence (5.9×10−8 erg cm−2). The thick magenta lines en- analysis of the observations performed by these two circle the position of the sky within the 90% LIGO localization facilities could help to confirm or not the AGILE region of GW170104 in which the minimum fluence reported for detection. AGILE-GW170104 is compatible with the INTEGRAL results. The conclusions above depend significantly on the as- assuming a 32 ms-long event characterized by a power- sumed spectral energy distribution of the event. A de- law shaped spectral energy distribution with a slope of tailed description of the spectral parameters of AGILE- -2 (as done in Verrecchia et al. 2017). As SPI-ACS ob- GW170104 is not provided by Verrecchia et al. (2017), serves positive and negative count rate fluctuations in the and thus we followed their assumption of a power-law background, all positions corresponding to a certain time shaped energy distribution with a slope of -2. At the delay between the INTEGRAL and AGILE locations de- same time, the authors also indicated that AGILE- fine circularly-shaped regions in the sky within which the GW170104 features similar timing and spectral proper- upper bound on the event flux is constant. This is the ties to the precursor of GRB090510. This weak precursor reason why the map of the upper bound values in Fig. 5 was detected by both AGILE/MCAL and Fermi/GBM. comprises stripes of different colors. The source positions It was also detected by INTEGRAL/SPI-ACS with a in the sky coincident with the direction toward AGILE S/N of 6.1, even though the location in the sky was not as seen from INTEGRAL and the diametrically opposite covered with the optimal sensitivity of the SPI-ACS. By direction correspond to the maximum absolute time de- analyzing the response of this instrument in the direction lays. Since the altitude of AGILE’s orbit is much smaller of GRB090510 and using the results obtained from the than that of INTEGRAL’s orbit, the direction from IN- observation of the precursor of the GRB, we were able TEGRAL towards AGILE is very close to the direction to derive a nearly model-independent conclusion that a from INTEGRAL to Earth, and the circularly-shaped re- similar event occurring anywhere within the LIGO 90% gions are all approximately centered on the position of localization region of GW170104, excluding the area in- the Earth (a small dark circle in Fig. 5). The median visible to AGILE, should have been detected by the SPI- value of the fluence in sky locations compatible to the ACS with a median S/N of 13.0, and certainly no lower time delay between the spacecrafts is 1.7×10−8 erg cm−2 than 4.6. and it does not exceed 7.1×10−8 erg cm−2 in any sky Finally, we stress that it is entirely possible that the position enclosed within the LIGO 90% localization re- AGILE/MCAL event was a real weak short GRB going gion of GW170104. In Fig. 5, we highlighted with red off in a region of the sky covered with a low SPI-ACS contours the portions of the sky where the minimum de- sensitivity and completely unrelated to GW170104 (i.e. tectable fluence by INTEGRAL is consistent with the outside the 90% LIGO localization region). Combining best fit (solid) and the lowest allowed (dashed) fluence of the area of the sky with unfavorable orientations for the AGILE-GW170104 inferred from the AGILE data. SPI-ACS observations and not occulted by the Earth for We found that there are no sky positions within the AGILE, we inferred a remaining allowed region spanning 90% LIGO localization region for which the best fit flu- about 3533 deg2. ence of the AGILE event is compatible with the INTE- GRAL results. There are, however, positions within the 4. CONCLUSIONS 90% LIGO localization region for which the lowest al- All GW events reported so far by LIGO were found to lowed value of the fluence of AGILE-GW170104 would be most likely associated with binary back hole mergers. still be compatible with the INTEGRAL results (thick The extensive multi-wavelength follow-up campaigns car- 6 V. Savchenko et al. ried out after each of these discoveries led to the detection ACKNOWLEDGEMENTS of at least two possible electromagnetic counterparts to Based on observations with INTEGRAL, an ESA the GW events (Connaughton et al. 2016; Greiner et al. project with instruments and science data centre funded 2016; Verrecchia et al. 2017). Although none of these as- by ESA member states (especially the PI countries: Den- sociations was firmly confirmed, they led to discussion of mark, France, Germany, Italy, Switzerland, Spain), and exotic scenarios in explaining EM emission in these merg- with the participation of Russia and the USA. The SPI- ers (e.g. Perna et al. 2016; Loeb 2016; Woosley 2016; ACS detector system has been provided by MPE Garch- Lyutikov 2016). The INTEGRAL efforts to follow-up ing/Germany. We acknowledge the German INTEGRAL as much as possible all relevant LIGO triggers will even- support through DLR grant 50 OG 1101. The Ital- tually help to revealing which, if any, of these scenarios is ian INTEGRAL/IBIS team acknowledges the support of applicable. So far, the INTEGRAL results have provided ASI/INAF agreement n. 2013-025-R.0. AL and RS ac- the most stringent upper limits on any associated prompt knowledge the support from the Russian Science Founda- hard X-ray and γ-ray emission in the 75 keV to 2 MeV tion (grant 14-22-00271). Some of the results in this pa- energy range for each of the announced GW events when per have been derived using the HEALPix (G´orski et al. INTEGRAL observations were available, challenging the 2005) package. We are grateful the Fran¸cois Arago Cen- possible association of GW 150914 and GW 170104 with tre at APC for providing computing resources, and Vir- the tentatively reported electromagnetic counterparts. tualData from LABEX P2IO for enabling access to the StratusLab academic cloud. Finally, we thank the anony- mous referee for the insightful comments.

REFERENCES Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2016, Physical Rowlinson, A., Gompertz, B. P., Dainotti, M., et al. 2014, Review X, 6, 041015 MNRAS, 443, 1779 Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2017, Physical Savchenko, V., Ferrigno, C., Mereghetti, S., et al. 2016, ApJ, 820, Review Letters, 118, 221101 L36 Band, D., Matteson, J., Ford, L., et al. 1993, ApJ, 413, 281 Savchenko, V., Bazzano, A., Bozzo, E., et al. 2017, eprint Bhalerao, V., Kasliwal, M. M., Bhattacharya, D., et al. 2017, arXiv:1704.01633, arXiv:1704.01633 ArXiv e-prints, arXiv:1706.00024 Sharma, V., Bhalerao, V., Bhattacharya, D., & Rao, A. R. 2017, Burns, E., Blackburn, L., Briggs, M. S., et al. 2017, GRB GRB Coordinates Network, 20389 Coordinates Network, 20365 Stalder, B., Tonry, J., Smartt, S. J., et al. 2017, ArXiv e-prints, Connaughton, V., Burns, E., Goldstein, A., et al. 2016, ApJ, 826, arXiv:1706.00175 L6 Sturner, S., Shrader, C., Weidenspointner, G., et al. 2003, A&A, Fermi GBM and Fermi LAT Collaborations. 2017, ArXiv e-prints, 411, L81 arXiv:1706.00199 Svinkin, D., Golenetskii, S., Aptekar, R., et al. 2017a, GRB G´orski, K., Hivon, E., Banday, A., et al. 2005, ApJ, 622, 759 Coordinates Network, 20406 Greiner, J., Burgess, J. M., Savchenko, V., & Yu, H.-F. 2016, —. 2017b, GRB Coordinates Network, 21158 ApJ, 827, L38 Tavani, M., Collaboration, f. t. A., Argan, A., et al. 2008, Kashyap, V. L., Van Dyk, D. A., Connors, A., et al. 2010, Astronomy and Astrophysics, Volume 502, Issue 3, 2009, arXiv:arXiv:1006.4334v1 pp.995-1013, 502, 995 Lee, W., & Ramirez-Ruiz, E. 2007, New Journal of Physics, 9, 17 Tonry, J., Denneau, L., Heinze, A., et al. 2017, GRB Coordinates Loeb, A. 2016, ArXiv e-prints, arXiv:1602.04735 Network, 20377 Lyutikov, M. 2016, eprint arXiv:1602.07352, arXiv:1602.07352 Ubertini, P., Lebrun, F., Di Cocco, G., et al. 2003, A&A, 411, Marcinkowski, R., & Xiao, H. 2017, GRB Coordinates Network, Vedrenne, G., Roques, J.-P., Sch¨onfelder, V., et al. 2003, A&A, 20387 411, L63 Melandri, A., Piranomonte, S., Branchesi, M., et al. 2017, GRB Verrecchia, F., Tavani, M., Ursi, A., et al. 2017, ArXiv e-prints, Coordinates Network, 20735 arXiv:1706.00029 Mereghetti, S., G¨otz, D., Borkowski, J., Walter, R., & Pedersen, von Kienlin, A., Beckmann, V., Rau, A., et al. 2003, A&A, 411, H. 2003, Astronomy and Astrophysics, 411, L291 L299 Perna, R., Lazzati, D., & Giacomazzo, B. 2016, ArXiv e-prints, Whitesides, L., Lunnan, R., Kasliwal, M. M., et al. 2017, ArXiv arXiv:1602.05140 e-prints, arXiv:1706.05018 Pescalli, A., Ghirlanda, G., Salvaterra, R., et al. 2016, A&A, 587, Winkler, C., Courvoisier, T. J.-L., di Cocco, G., et al. 2003, A&A, doi:10.1051/0004-6361/201526760 411, L1 Woosley, S. E. 2016, ApJL, 824, 10