GECKO Optical Follow-up Observation of Three Binary Black Hole Merger Events: GW190408_181802, GW190412, and GW190503_185404

Since the first detection of the binary black hole (BBH) merger event by the Laser Interferometer Gravitational-Wave Observatory (LIGO) Collaboration and Virgo Collaboration (LVC) in 2015 September 14 (Abbott et al. 2016), the reported number of BBH events has been increasing fast, from two in the Advanced LIGO and Advanced Virgo (LIGO/Virgo) first (O1) observing run, and three in the second (O2) observing run, up to 20 in the first half (O3a) of the third (O3) observing run (Abbott et al. 2020b). Gravitational wave (GW) astronomy is now flourishing, with the GW signals providing valuable information on the compact star mergers, such as the masses of the binary compact objects, their approximate distance, and the location in the sky of the event.

The power of GW astronomy was much enhanced when the first electromagnetic-wave (EM) counterpart of a GW event was identified on 2017 August 17 during the LIGO/Virgo O2 run. The GW event, GW170817, was found to be due to a binary neutron star (BNS) merger with 90% localization of 31 deg² and a luminosity distance of 40 Mpc (Abbott et al. 2017a). The combined efforts of the GW and EM observations produced a series of important results, including the first identification (AT2017gfo) of the exact location and the distance to the GW event (Abbott et al. 2017b; Arcavi et al. 2017; Coulter et al. 2017; Lipunov et al. 2017; Soares-Santos et al. 2017; Tanvir et al. 2017; Valenti et al. 2017), convincing observational evidence for the link between a short gamma-ray bursts (GRB) and a BNS merger (Abbott et al. 2017c; Goldstein et al. 2017; Savchenko et al. 2017), undeniable discovery of a kilonova and the first detailed characterization of such an event (Andreoni et al. 2017; Arcavi et al. 2017; Chornock et al. 2017; Cowperthwaite et al. 2017; Drout et al. 2017a; Evans et al. 2017b; Haggard et al. 2017; Hallinan et al. 2017; Kasliwal et al. 2017; Margutti et al. 2017; Nicholl et al. 2017; Pian et al. 2017; Smartt et al. 2017; Troja et al. 2017; Utsumi et al. 2017), a glimpse of the BNS merger environment (Im et al. 2017; Tanvir et al. 2017), and the first application of the GW siren method to measure the Hubble constant (Abbott et al. 2017d). In short, this was a monumental event that marked the beginning of multi-messenger astronomy (MMA) using GW and EM observations. If a similar study can be performed for BBH GW events that currently occupy the majority of the detected GW events, it would greatly expand the power of MMA and enhance our knowledge of this previously rarely studied aspect of the universe.

As a result, many follow-up observations have been carried out to find EM counterparts of BBH GW events, yet no EM counterparts have been identified for the BBH events (Cowperthwaite et al. 2016; Morokuma et al. 2016; Smartt et al. 2016a, 2016b; Soares-Santos et al. 2016; Lipunov et al. 2017; Stalder et al. 2017; Turpin et al. 2017; Yoshida et al. 2017; Doctor et al. 2019; Smith et al. 2019; Yang et al. 2019; Gompertz et al. 2020; Grado et al. 2020; Herner et al. 2020). There are several reasons for the nondetections. First of all, the strong gravitational force of black holes do not allow debris from the merger to escape, thus prohibiting EM-producing physical processes from occurring (Berger 2014). A few models suggest that an EM counterpart might be produced if, e.g., the BBH merger preferentially happened in the active galactic nucleus (AGN) accretion disk, but the expected strength of EM radiation is weak or very uncertain (Loeb 2016; Perna et al. 2016, 2018; Bartos et al. 2017; de Mink et al. 2017; Stone et al. 2017; McKernan et al. 2018). For the O3 run, a candidate optical counterpart for a BBH merger, GW190521 (Abbott et al. 2020b), is reported as a possible AGN-related transient (Graham et al. 2020). Second, the early LIGO/Virgo runs have been hampered by limited detector sensitivities, and as a result, the locations of the events estimated from the GW signals were very uncertain at a level of several hundreds to thousands deg², and their distance estimates were uncertain as well (Abbott et al. 2019). In spite of these daunting challenges, identification of BBH-induced GW EM counterparts remains attractive due to its potential for providing unexpected physical processes during the extreme compact merger and the prospects for using BBH merger EM counterparts for a number of MMA applications.

The LIGO/Virgo O3 run provided a unique opportunity to improve the limits of the BBH EM counterpart search. With the improved sensitivity of the GW detectors, the number of GW events increased (36 BBH events) and the localization area was improved as well (514 deg², median of 90% localization area for BBH events ⁵ ). Hence, we organized an EM follow-up observation group in Korea, utilizing a network of telescopes worldwide that were already in use for monitoring nearby galaxies for new transients such as supernova (Intensive Monitoring Survey of Nearby Galaxies, IMSNG; Im et al. 2019). Our EM-follow-up effort was named the Gravitational-wave EM Counterpart Korean Observatory (GECKO; Im et al. 2020; G. S. H. Paek et al. 2021, in preparation). Several of the GECKO facilities have wide-field capabilities (field of view >1 deg²) and are suited for searching EM counterparts over a wide area.

In this paper, we present the GECKO optical follow-up observation for the first three BBH merger events in the O3a run, namely GW190408_181802 (hereafter GW190408, for brevity), GW190412, and GW190503_185404 (hereafter GW190503). We focus our attention on the observation made by the Korea Microlensing Telescope Network (KMTNet; Kim et al. 2016), which is the most sensitive wide-field telescope of GECKO. The description of the KMTNet observation and analysis results will serve as a reference for future EM follow-up attempts. However, observations made by other telescopes are also reported in order to show the capabilities of GECKO facilities. In Section 2, we summarize three GW events that are targeted for our optical follow-up observation. Section 3 describes the telescopes we used and the observation strategy. Data analysis of the observed images is given in Section 4. We report the observation results and transient candidates in Section 5. Finally, we discuss the results in Section 6, and conclude the paper in Section 7.

Here, we describe the characteristics of three BBH events observed by GECKO in the early phase of the O3 run. We note that our observations were performed using information from the initial alerts (Bayester.fits), which were updated later in GWTC-2 catalog (Abbott et al. 2020b).

2.1. GW190408

On 2019 April 8 18:18:02 UT, the first event of LIGO/Virgo O3 run, designated originally as S190408a (Ligo Scientific Collaboration & VIRGO Collaboration 2019a), and later as GW190408 (Abbott et al. 2020b), was detected by the Advanced LIGO and Advanced Virgo GW detectors. This event was classified as a BBH merger with >99% probability. The luminosity distance and the 90% sky localization area were initially reported as 1473 ± 358 Mpc and 387 deg² (Table 1). Later, further analysis of the GW data showed a luminosity distance of ${1580}_{-590}^{+400}$ Mpc with a redshift of ${0.30}_{-0.10}^{+0.06}$, a 90% credible sky area of 140 deg², and a false alarm rate (FAR) of <7.9 × 10⁻⁵ yr⁻¹ (Abbott et al. 2020b). The BH masses of the compact objects are found to be ${24.5}_{-3.4}^{+5.1}\,{M}_{\odot }$ and ${18.3}_{-3.5}^{+3.2}\,{M}_{\odot }$. The most recent GWTC-2 localization shows that the event happened in the Northern Hemisphere near R.A. = 351.0 deg and decl. = 53.9 deg, but the initial 90% localization area contained a localized region in the Southern Hemisphere (Figure 1). Hence, the Southern Hemisphere localization area was observed by the KMTNet.

Zoom In Zoom Out Reset image size

Table 1. Target GW Events

Event	Type	Event time	Initial	Initial	GWTC-2	GWTC-2	FAR
		(UT)	90% Area (deg2)	Distance (Mpc)	90% Area (deg2)	Distance (Mpc)	(yr−1)
GW190408	BBH	2019-04-08 18:18:02	387	1473 ± 358	140	${1580}_{-590}^{+400}$	<7.9 × 10−5
GW190412	BBH	2019-04-12 05:30:44	156	812 ± 194	21	${740}_{-170}^{+140}$	<7.9 × 10−5
GW190503	BBH	2019-05-03 18:54:04	448	412 ± 105	94	${1520}_{-660}^{+710}$	<7.9 × 10−5

Note. The FAR values are taken from the result of PyCBC BBH pipeline in the GWTC-2 catalog.

Download table as:  ASCIITypeset image

2.2. GW190412

2.3. GW190503

3.1. Telescopes

3.2. Basic Observation Strategy

3.3. Observations

All the observed images were processed with basic reduction procedures (bias subtraction, dark subtraction, and flat-fielding). Astrometric calibration is done using SCAMP (Bertin 2006) or Astrometry.net (Lang et al. 2010), using tens to hundreds of stars as astrometry references depending on the field of view and depth. This produces images with astrometric accuracy of rms scatter of 02–08 in both R.A. and decl. directions. For the KMTNet observation for a given epoch, only two frames were taken, which made it challenging to remove cosmic rays during co-addition. Therefore, we removed the cosmic rays from each single frame using LA-Cosmic software (van Dokkum 2001). The sequentially observed two images were combined to produce a deeper image with SWarp (Bertin et al. 2002).

We used Pan-STARRS (PS1; Chambers et al. 2016) images at the closest wavelength to the observed image filter as reference images (for example, we used a PS1 r-band image for the KMTNet R-band image), and the PS1 reference images were subtracted from observed images by convolving and rescaling them to match the spatial resolution and the flux scale of the observed images using High Order Transform of PSF ANd Template Subtraction (HOTPANTS; Becker 2015).

The declinations of the KMTNet images for GW190408 and GW190503 are lower than −30 deg, and out of the PS1 coverage. For these images, we observed the same area of the sky twice with a time separation of about one day, and we used the images observed ∼24 hr after the first epoch as reference images for the two events.

Transient candidates were detected by identifying positive signals having signal-to-noise ratios greater than five in the subtracted images, using SExtractor (Bertin & Arnouts 1996). Many of the detected sources in the residual images were false identifications found at over- or undersubtracted bright sources. This imperfect subtraction was often caused by a slight mismatch in astrometry and/or in the wavelength coverage of the reference and the observed images, and they appear as positive signals right next to negative pixels (see the upper part of the residual image of SN 2018imf in Figure 2). To remove these artifacts, we ran SExtractor on the inverse images of the residual images. The positive detections in the residual images were rejected if there were positive detections in the inverse residual image right next to their position (<3''). Additionally, we removed candidates matched with nontransient sources in the reference images, to exclude variable sources like variable stars or active galactic nuclei. We also removed candidates that were on bad pixels or whose central pixels were saturated. The remaining candidates were visually inspected and classified as new transient candidates if there were no other artifacts. Figure 2 shows the science, reference, and residual images of three transients found in the KMTNet images (see Section 5).

Zoom In Zoom Out Reset image size

For the images from the other telescopes, we subtracted images using PS1 images and visually inspected the residual images to find transient candidates.

After the identification of the credible transient candidates, we matched them with known solar system objects using SkyBoT (Berthier et al. 2006) version 3.0. SkyBot is a system that provides the location of known Solar System objects at a given epoch. Also, we matched our transient candidates with those that were already reported in the Transient Name Server (TNS ⁶ ).

For GW190408, we found two transient candidates. They are matched with known Solar System objects.

In the observed areas for the GW190412 event, we detected 275 transient candidates. As we describe below, the majority of them are found to be Solar System objects, since the localization area of GW190412 was near the ecliptic plane, with an ecliptic latitude of about 14 deg. Among these candidates, 258 objects were found to have matches with Solar System objects within 1' of their positions (most of the sources are matched within 5''). In addition, we found three moving objects that did not have matching Solar System objects. We consider these to be previously unknown Solar System objects or known Solar System objects with high positional errors. One transient candidate was found to be a fast proper motion star of ∼1'' yr⁻¹ (Gaia Collaboration et al. 2018). Three objects were found in the vicinity of possible host galaxies, and they are identified in the TNS as SN 2019aqz (Type Ia), SN 2018imf (Type IIP), and AT 2018jcu (Figure 2). The remaining 10 candidates do not have possible host galaxies in their vicinity, and were only observed in a single frame. Their nature cannot be judged from our data alone, but we suspect that most of them are Solar System objects, considering the large number of such objects in this field. The observing strategy at that time was geared toward minimizing the overhead by reducing the time required to move to another field. However, we realize that efficient identification of moving objects can be done more easily if subsequent frames are taken about an hour later. We are now considering such a strategy for future observations.

No transient candidates were found in the observed areas of event GW190503. Also, no transient candidates were found in WIT, DOAO, SAO, and LOAO images in any of the events.

Table 5 shows the three transients and 10 transient candidates with uncertain nature from the GECKO GW190412 observation. In Table 6, we also report the epochs and the positions of the Solar System objects found from our follow-up observations.

6.1. Limits on BBH EM Counterpart Brightness

6.2. Comparison with Kilonova Light-curve Models

We compare model light curves of kilonovae and the imaging depths to discuss the possibility of GW EM counterpart detection (Figure 3). One of the models is the AT2017gfo-like kilonova model of Kasen et al. (2017), hereafter K17. The model is for the AT2017gfo that combines the blue and red kilonova components with ejecta masses of 0.025 M_⊙ and 0.04 M_⊙, respectively. We plotted the two kilonova components in Figure 3, and it shows that the blue kilonova is more dominant than the red kilonova in the early phase, <2 days. Similarly, the AT2017gfo-like kilonova models of Tanaka et al. (2017) and Banerjee et al. (2020) are plotted in Figure 3 (with ejecta masses of 0.03–0.05 M_⊙). We also plot earlier kilonova light-curve predictions for NS–BH (NSBH) binary merger and BNS mergers with ejecta masses of 0.0007–0.07 M_⊙ (Hotokezaka et al. 2013; Tanaka & Hotokezaka 2013; Tanaka et al. 2014). These models assume a higher mass fraction of lanthanides (X_lan ∼ 10^−1.5) than the K17 model (X_lan = 10⁻⁴), and consequently predict fainter light curves than the AT2017gfo-like models. For the distances of these models for each event, we assumed the values from the GWTC-2 catalog. Overall, we confirm that these events are too far (700–1600 Mpc) for their EM counterparts to be detected with KMTNet at a nominal depth. But if AT2017gfo-like kilonovae appeared at about 400 Mpc or less, KMTNet could have detected them (Figure 3).

Zoom In Zoom Out Reset image size

6.3. Prospects for Rapid Follow-up Observation

6.4. Need for Deep Reference Images

In this paper, we presented the GECKO follow-up observation of three BBH events, GW190408, GW190412, and GW190503, that were reported early in the O3a run. In particular, we reported in detail the KMTNet observations that at 100 minutes to 6 hr from the GW detection with a sky coverage of 29–63 deg² and ∼22.5 mag depth in the R band.

From our observation, we identified 13 transient candidates for GW190412 and hundreds of moving objects such as asteroids. Among these transient candidates, three transients were already reported (two confirmed as a supernova), and the remaining 10 are likely to be moving objects, due to their proximity to the field near the ecliptic plane, the lack of a host galaxy near them, and having been observed only once. Therefore, no probable EM counterpart was found in our observation. The improved localization of these events in the recently published GWTC-2 catalog reveals that the initial position estimates were very broad and many of the optical follow-ups can point to regions of the sky that are not in the improved localization area. However, we find that our KMTNet imaging observation covered a 69% probability region of GWTC-2 localization map for GW190503, and our nondetection places an upper limit on the rest-frame g-band magnitude of the EM counterpart of this BBH merger event to be M_g > − 18.0 AB mag. Together with this, the comparison between the image depths of KMTNet data and the theoretical kilonova light-curves show that brighter AT2017gfo-like kilonova events can be detected when the luminosity distance is out to ∼400 Mpc.

Our follow-up observation started as quickly as about 100 minutes after the GW event alert, and could cover an area of about 50 deg² in one hour, demonstrating that the observation of the 90% localization area in the future O4 run can be done in about two hours. Work is ongoing to improve the response time. Overall, our results show promise that GECKO can identify kilonova events out to 400 Mpc within hours from the GW event alert in future GW observing runs that will improve the detector sensitivity event localization capability.

We thank the referee for useful comments that helped and improved the presentation of the paper. We appreciate the support and encouragement from colleagues of the Korean Gravitational Wave Group. This research has made use of the KMTNet system operated by the Korea Astronomy and Space Science Institute (KASI) and the data were obtained at three host sites of CTIO in Chile, SAAO in South Africa, and SSO in Australia. We also thank the operators at LOAO and KMTNet for performing the requested observations and the staff at the McDonald observatory for their support of WIT. This work was supported by the National Research Foundation of Korea (NRF) grant, No. 2020R1A2C3011091, funded by the Korean government (MSIT), and the Korea Astronomy and Space Science Institute under the R & D program (Project No. 2020-1-600-05) supervised by the Ministry of Science and ICT.

Facilities: KMTNet (Kim et al. 2016), WIT, DOAO, SAO, LOAO (Im et al. 2019).

Software: SExtractor (Bertin & Arnouts 1996), Scamp (Bertin 2006), SWarp (Bertin et al. 2002), Astronomy.net (Lang et al. 2010), SkyBoT (v3.0; Berthier et al. 2006), HOTPANTS (Becker 2015), LA-Cosmic (van Dokkum 2001).