INTEGRAL UPPER LIMITS ON GAMMA-RAY EMISSION ASSOCIATED WITH THE GRAVITATIONAL WAVE EVENT GW150914

The IBIS instrument (Ubertini et al. 2003) is composed of two detectors, ISGRI (20–1000 keV; Lebrun et al. 2003) and PICsIT (175 keV–10 MeV; Di Cocco et al. 2003), which, using a coded mask, provide images over a field of view of 30° × 30°. The ISGRI data are used to automatically search and localize in real time GRBs and other transients through the INTEGRAL Burst Alert System (IBAS; Mereghetti et al. 2003). IBAS did not reveal any new transients in the IBIS field of view at the time of the LIGO trigger, down to a peak flux sensitivity of ∼0.1 ph cm⁻² s⁻¹ (20–200 keV, 1 s integration time). We also carried out an offline search in the time interval 09:28–10:00 UT, again with negative results. Note, however, that the instruments of INTEGRAL were pointed at a position (R.A. = 271°, decl. = –31°) outside the high-probability region of the gravitational signal. This also prevented  the X-ray monitor instrument JEM-X (Lund et al. 2003) from collecting constraining data.

IBIS can also provide a response to photons outside the field of view, due to high-energy photons passing through the passive and active shields of the instrument, allowing the detection of transient events. Indeed, most of the shielding of IBIS is passive and relatively thin, becoming transparent to photons above ∼200 keV. For high-energy events, ratemeters for the PICsIT detector are available in 8 energy bands in the range 210–2600 keV. We investigated the count rate light curve in ±10 s around the GW150914 for possible excesses on timescales from 0.016 to 10 s, but found no positive signal. We set 3σ upper limits to fluences in the 570–1200 keV energy range of $2.5\;\times \;{10}^{-7}$ erg cm⁻² and $6.5\times {10}^{-7}$ erg cm⁻², assuming durations of 1 s and 10 s, respectively. These values apply to a fully exposed detector area. The detection efficiency is highly dependent on the source position and it is considerably reduced for sources located at large angles with respect to the instrument pointing direction, due to the lower exposed area and the presence of the 2 cm-thick BGO anticoincidence shield. The localization region of GW150914 is positioned at a large offset (∼80° to 140°) from the telescope axis. This implies that the sensitivity is decreased and is strongly dependent on the source position in the sky. Nevertheless, the PICsIT observation provides an important independent limit on gamma-ray emission above 500 keV associated with the GW150914.

The Fermi/GBM team reported a possible hard X-ray transient on 2015 September 14 at 09:50:45.8 UTC, about 0.4 s after the reported LIGO burst trigger time, and lasting for about one second (Blackburn et al. 2015; Connaughton et al. 2016). The light travel time can introduce a time difference between INTEGRAL and Fermi detections of up to ±0.5 s, depending on the source position within the LVC error region. We do not observe any excess within a −0.5 s to +0.5 s window around the Fermi/GBM trigger (Figure 1), and set a 3σ upper limit of $1.5\times {10}^{-7}$ erg cm⁻² for a one-second integration time, assuming a typical short hard GRB, characterized by a Band model with parameters $\alpha =-0.5$, $\beta =-2.5$, ${E}_{\mathrm{peak}}=1000$ keV. A substantial part of the candidate event in the GBM comes from the high-energy BGO detector, above 100 keV (Blackburn et al. 2015), where the Fermi/GBM effective area is about a factor of 30–40 smaller than that of the INTEGRAL/SPI-ACS.

Connaughton et al. (2016) find that preferable localization of their candidate is in the direction of the Earth, or close to it, limited to the southern, dominant, arc of the GW150914 localization. Assuming the preferred localization, they conclude that the spectrum can be best fit by a hard power law with a slope of −1.4 and a 10–1000 keV fluence of ${2.4}_{-1.0}^{+1.7}\times {10}^{-7}$ erg cm⁻². Extrapolating this spectrum to the full 75 keV–100 MeV energy range accessible to the SPI-ACS without a cutoff is clearly unphysical and incompatible with the SPI-ACS upper limit. However, no best-fit parameters for a model comprising a cutoff power law are reported. On the other hand, Connaughton et al. (2016) found a best fit to the Comptonized model in the northeastern tip of the the southern arc, with a power-law index ${\alpha }^{\mathrm{COMP}}=-0.16$ and ${E}_{\mathrm{peak}}^{\mathrm{COMP}}=3500\;\mathrm{keV}$, harder than a typical Fermi/GBM spectrum. We assume this spectral model to compute the expected signal in the SPI-ACS: for the southern (northern) arc, SPI-ACS would detect 4740 (1650) counts, with a signal significance of 15σ (5σ) sigma above the background. It should be noted that the northern arc is disfavored by both the GBM and the LIGO localizations.

We stress that to compare the GBM and SPI-ACS sensitivities, it is inappropriate to use a soft spectral model, as in the computation of our early fluence upper limits (Ferrigno et al. 2015), since the spectral properties of the GBM candidate are very different.

Considering the reported hardness of the GBM candidate, and the favorable orientation of the SPI-ACS with respect to the GW150914 localization, we are inclined to claim that the non-detection by SPI-ACS disfavors a cosmic origin for the Fermi/GBM excess. If the origin of the event was near the Earth, INTEGRAL would not detect it, due to the large INTEGRAL—Earth distance at the time of GW150914 (140,000 km). Connaughton et al. (2016) discussed a possible terrestrial origin for the GBM excess and came to the conclusion that such an origin is not compatible with the characteristics of a terrestrial gamma-ray flash. However, they do not exclude the possibility that the event had a magnetospheric origin. Eventually, considering that the false alarm probability of the GBM association is relatively high (0.2%; Connaughton et al. 2016) and that SPI-ACS does not detect it, it is likely that the GBM excess is a random background fluctuation.

INTEGRAL/SPI-ACS is the only instrument covering the whole GW150914 position error region at the time of the GW trigger. The limit depends on the position, burst duration, and the assumed spectral model, and ranges from ${F}_{\gamma }=2\times {10}^{-8}$ erg cm⁻² to ${F}_{\gamma }={10}^{-6}$ erg cm⁻² in the 75 keV–2 MeV energy range for a typical range of GRB models and sky positions (see Figure 3). Assuming a reference distance to the event of D = 410 Mpc (Abbott et al. 2016), this implies an upper limit on the isotropic equivalent luminosity of ${E}_{\gamma }\lt 2\times {10}^{48}\;\mathrm{erg}\left(\frac{{F}_{\gamma }}{{10}^{-7}\;\mathrm{erg}\;{\mathrm{cm}}^{-2}}\right){\left(\frac{D}{410\mathrm{Mpc}}\right)}^{2}$. The LIGO observation corresponds to the energy emitted in gravitational waves ${E}_{\mathrm{GW}}=1.8\pm 0.3\times {10}^{54}$ erg. Our SPI-ACS upper limit constrains the fraction of energy emitted in gamma-rays in the direction of the observer: ${f}_{\gamma }\;\lt \;{10}^{-6}\;\left(\frac{{F}_{\gamma }}{{10}^{-7}\;\mathrm{erg}\;{\mathrm{cm}}^{-2}}\right){\left(\frac{{E}_{\mathrm{GW}}}{1.8\times {10}^{54}\mathrm{erg}}\right)}^{-1}{\left(\frac{D}{410\mathrm{Mpc}}\right)}^{2}$.

Our analysis of the gravitational wave signal indicates that it was produced by the coalescence of two black holes (Abbott et al. 2016). If at least one of the merging black holes was charged, following the Reissner–Nordstrom formulation, up to 25% of the gravitational energy could have been converted into electromagnetic radiation (Zilhão et al. 2012). However, it is expected that the charge of the black hole is spontaneously dissipated and is not significant for astrophysical applications. To date, there is no theoretical work predicting electromagnetic emission from the coalescence of two non-charged back holes in a vacuum. Indeed, it is not possible to create photons in a system with no matter outside of the gravitational horizon and only gravitational interaction involved, without invoking effects of quantum gravity, a theory which has not yet been developed.

The coalescing black holes may be surrounded by matter, in the form of spherically symmetric inflow or/and an accretion disk, which can form if the inflow possesses sufficient angular momentum. The accretion disk can have high density and large potential energy. Rapid changes in the accretion dynamics during binary coalescence may lead to bright observational signatures (Farris et al. 2012). Magnetic fields, anchored in the accretion disk, can cause bright radio emission that is simultaneous with the gravitational waves (Mösta et al. 2010).

While supermassive black holes are often accompanied by substantial disks, black holes of stellar mass lose the disk created during the progenitor star collapse on a timescale of the order of ${\tau }_{\mathrm{disk}}\sim 100\;{\rm{s}}$ (Woosley 1993). Sustainable accretion disks can be expected when a constant inflow of matter is provided by a companion star: in these cases, the black hole—star binary can be a bright and variable X-ray and gamma-ray source. However, it remains to be established how likely it is to find a dynamically stable triple system composed of a binary black hole and an additional companion star.

Isolated stellar-mass black holes or binary black holes are bound to accrete from the interstellar medium (ISM). This process can be described as quasi-spherical Bondi–Hoyle accretion (Bondi & Hoyle 1944), characterized by very low accretion rates $\dot{M}\sim {10}^{15}$ g s⁻¹ ${\left(\frac{{M}_{H}}{65\times {M}_{\odot }}\right)}^{2}\quad \left(\frac{{\rho }_{\infty }}{{10}^{-24}\;{\rm{g}}\;{\mathrm{cm}}^{-3}}\right){\left(\frac{{c}_{s}}{10\mathrm{km}{{\rm{s}}}^{-1}}\right)}^{-3}$. In the case of a merger the accretion rate may be enhanced by up to two orders of magnitude on a one-second timescale (Farris et al. 2010), but in the case of a stellar black hole binary accreting from the ISM, the isotropic peak luminosity cannot exceed ${L}_{\mathrm{iso}}=100\times 0.3\times \dot{M}{c}^{2}\;=2.5\times {10}^{37}$ $\mathrm{erg}\;{{\rm{s}}}^{-1}{\left(\frac{{M}_{H}}{65{M}_{\odot }}\right)}^{2}\quad \left(\frac{{\rho }_{\infty }}{{10}^{-24}\;{\rm{g}}\;{\mathrm{cm}}^{-3}}\right){\left(\frac{{c}_{s}}{10\mathrm{km}{{\rm{s}}}^{-1}}\right)}^{-3}$. This luminosity is almost 17 orders of magnitude lower than the GW luminosity and more than ∼11 orders of magnitude below the current gamma-ray upper limits. Agol & Kamionkowski (2002) calculated a possible range of Bondi–Hoyle accretion rates in a Milky Way-like galaxy, yielding, in very rare cases, $\dot{M}\sim {10}^{17}$ g s⁻¹ or peak luminosity ${L}_{\mathrm{iso}}=3\times {10}^{39}$ erg s⁻¹, still a factor of 10⁹ below what is observable. The conditions necessary to produce observable emission may be reached in dense molecular clouds, where $\left(\frac{{\rho }_{\infty }}{{10}^{-16}\;{\rm{g}}\;{\mathrm{cm}}^{-3}}\right){\left(\frac{{c}_{s}}{1\mathrm{km}{{\rm{s}}}^{-1}}\right)}^{-3}\gt 1$. Therefore, our upper limit on the hard X-ray burst associated with the merger disfavors the possibility that the binary was embedded in such a cloud, unless the emission was very anisotropic.

Recently, different mechanisms for producing the gamma-ray emission in a black hole binary merger were suggested. For example, a binary black hole with a very small separation could be formed immediately after the collapse of a massive star, resulting in a gamma-ray burst produced nearly simultaneously with a gravitational wave signal (Loeb 2016). Alternatively, if an unusually long-lived disk is present around the black hole binary it could produce a bright gamma-ray signature at the time of the coalescence (Perna et al. 2016).

Abbott et al. (2016) were able to make use of the gravitational wave data to constrain the compactness of the merging objects, excluding the possibility that either of them is a neutron star.

Strange stars are more compact than neutron stars, and their coalescences can have different gravitational wave signatures (Moraes & Miranda 2014). Very exotic equations of state for strange quark stars would allow them to reach 6 ${M}_{\odot }$ (Kovács et al. 2009, not considering rotation). This is well below the 90% lower limit inferred for this event (25 ${M}_{\odot }$).

Boson Stars (see Schunck & Mielke 2003 for a review) might reach arbitrarily high masses, while only being slightly bigger than their gravitational radius. The existence of these objects requires an extension of the minimal standard model with a new fundamental scalar field, responsible for a stable particle. The properties of this field would determine the macroscopic properties of boson stars. This field has to be compatible with the non-detection established by particle physics experiments on Earth, cosmological simulations, and models of stellar evolution. Because of these limitations, the preferred model is generally a field with minimal coupling to standard model fields. A boson star consisting of a non-charged scalar field cannot be directly involved in any electromagnetic radiation, even in the case of an energetic coalescence event. On the other hand, the coalescence of boson stars might have distinct gravitational wave signatures (Palenzuela et al. 2007).

Another exotic star type, Q-stars (Bahcall et al. 1989, 1990; Miller et al. 1998; where Q here does not stand for quark) can reach 10 or even 100 solar masses. The existence of these objects was suggested based on finding the possibility of a peculiar baryonic state of matter, without introducing new matter fields. No predictions on their coalescence exist, to the best of our knowledge.