GWTC-2: Compact Binary Coalescences Observed by LIGO and Virgo during the First Half of the Third Observing Run

We report on gravitational-wave discoveries from compact binary coalescences detected by Advanced LIGO and Advanced Virgo in the first half of the third observing run (O3a) between 1 April 2019 $15:00$ UTC and 1 October 2019 $15:00$ UTC. By imposing a false-alarm-rate threshold of two per year in each of the four search pipelines that constitute our search, we present 39 candidate gravitational-wave events. At this threshold, we expect a contamination fraction of less than 10%. Of these, 26 candidate events were reported previously in near-real time through gamma-ray coordinates network notices and circulars; 13 are reported here for the first time. The catalog contains events whose sources are black hole binary mergers up to a redshift of approximately 0.8, as well as events whose components cannot be unambiguously identified as black holes or neutron stars. For the latter group, we are unable to determine the nature based on estimates of the component masses and spins from gravitational-wave data alone. The range of candidate event masses which are unambiguously identified as binary black holes (both objects $\ge 3M_{\odot }$) is increased compared to GWTC-1, with total masses from approximately $14M_{\odot }$ for GW190924_021846 to approximately $150M_{\odot }$ for GW190521. For the first time, this catalog includes binary systems with significantly asymmetric mass ratios, which had not been observed in data taken before April 2019. We also find that 11 of the 39 events detected since April 2019 have positive effective inspiral spins under our default prior (at 90% credibility), while none exhibit negative effective inspiral spin. Given the increased sensitivity of Advanced LIGO and Advanced Virgo, the detection of 39 candidate events in approximately 26 weeks of data (approximately 1.5 per week) is consistent with GWTC-1.

Popular Summary

In 2015, the U.S.-based Laser Interferometer Gravitational-Wave Observatory (LIGO) detected the first gravitational waves—ripples in spacetime from the merging of two massive black holes in a distant galaxy. Since then, LIGO and its European counterpart Virgo have racked up 50 detections that span a wide range of astrophysical sources consistent with binary black holes, binary neutron stars, and black hole–neutron star mergers. Here, we present an updated catalog—GWTC-2, or the gravitational-wave transient catalog 2—that includes all gravitational-wave detections up to October 1, 2019, which marks the end of the first half of the third observing period (O3a) for LIGO and Virgo.

O3a ran in 2019 from April 1 to October 1 and added 39 gravitational-wave events to the 11 confirmed events listed in GWTC-1. Remarkably, O3a produced about 3 times as many confidently detected gravitational-wave events as the two previous observing periods combined. Furthermore, the Virgo detector was able to join the two LIGO detectors for the full duration of O3a, with at least one detector observing for about 97% of the time, and at least two observing for about 82% of the time. Some especially interesting O3a events include the second-ever gravitational-wave observation consistent with a binary neutron star merger, the first events with unequivocally unequal masses, and a very massive black hole binary with a total mass of about 150 times the mass of the Sun.

In this report, we describe the improvements within LIGO and Virgo that make this possible as well as the significance of these events to the field of astrophysics.

Figure 1

The number of compact binary coalescence detections versus the effective VT to which the gravitational-wave network is sensitive to BNS coalescences. The effective VT is defined as the Euclidean sensitive volume [30] of the second-most-sensitive detector in the network at a given time, multiplied by the live time of that network configuration. The Euclidean sensitive volume of the network is the volume of a sphere with a radius given by the BNS inspiral range [29, 30] (shown in Fig. 3) of the second-most-sensitive detector in the network. To account for the addition of single-detector candidates in O3a, a single-detector Euclidean sensitive volume is also included, defined using the inspiral range of the most sensitive detector divided by 1.5. The effective BNS VT does not account for differences in sensitivity across the entire population of signals detected, necessary cosmological corrections, or changes to analysis pipeline efficiency between observing runs but, as shown in this figure, is consistent with the currently observed rate of detections. The colored bands indicate the three runs: O1, O2, and O3a. The black line is the cumulative number of confident detections of all compact binary coalescences (including black holes and neutron stars) for GWTC-1 [8] and this catalog. The blue line, dark blue band, and light blue band are the median, 50% confidence interval, and 90% confidence interval, respectively, of draws from a Poisson fit to the number of detections at the end of O3a. The increase in detection rate is dominated by improvements to the sensitivity of the LIGO and Virgo detectors with changes in analysis methods between observing runs being subdominant.

Figure 2

Representative amplitude spectral density of the three detectors’ strain sensitivity (LIGO Livingston 5 September 2019 $20:53$ UTC, LIGO Hanford 29 April 2019 $11:47$ UTC, and Virgo 10 April 2019 $00:34$ UTC). From these spectra, we compute BNS inspiral ranges of 109, 136, and 50 Mpc for LIGO Hanford, LIGO Livingston, and Virgo, respectively.

Figure 3

The BNS range of the LIGO and Virgo detectors. Left: the evolution in time of the range over the entire duration of O3a. Each data point corresponds to the median value of the range over one-hour-long time segments. Right: distribution of the range and the median values for the entire duration of O3a.

Figure 4

The rate of single interferometer glitches with SNR $>6.5$ and frequency between 10 and 2048 Hz identified by Omicron [84, 85] in each detector during O3a. Each point represents the average rate over a 2048 s interval. Dotted black lines show the median glitch rate for each detector in O2 and O3a.

Figure 5

Top: time-frequency representation of the data surrounding event GW190701_203306 at LIGO Livingston, showing the presence of scattered light glitches (modulated arches). Bottom: the same data after glitch identification and subtraction as described in Sec. 3d. In both plots, the time-frequency track of the matched-filter template used to identify GW190701_203306 is overlaid in orange. Investigations identify seven candidate events in coincidence with similar scattering glitches and require mitigation before further analysis. Despite the clear overlap of the signal with the glitch, the excess power from the glitch is successfully modeled and subtracted.

Figure 6

Credible region contours for all candidate events in the plane of total mass $M$ and mass ratio $q$. Each contour represents the 90% credible region for a different event. We highlight the previously published candidate events: GW190412, GW190425, GW190521, and GW190814, the potential NSBH GW190426_152155, and, finally, GW190924_021846, which is most probably the least massive system with both masses $>3M_{\odot }$. The dashed lines delineate regions where the primary or secondary can have a mass below $3M_{\odot }$. For the region above the $m_2=3M_{\odot }$ line, both objects in the binary have masses above $3M_{\odot }$.

Figure 7

Credible region contours for all candidate events in the plane of chirp mass $\mathcal{M}$ and effective inspiral spin ${\chi }_{\mathrm{eff}}$. Each contour represents the 90% credible region for a different event. We highlight the previously published candidate events (cf. Fig. 6), as well as GW190517_055101 and GW190514_065416, which have the highest probabilities of having the largest and smallest ${\chi }_{\mathrm{eff}}$, respectively.

Figure 8

Marginal posterior distributions on primary mass $m_1$, secondary mass $m_2$, mass ratio $q$, effective inspiral spin ${\chi }_{\mathrm{eff}}$, and the luminosity distance $d_L$ for all candidate events in O3a. The vertical extent of each colored region is proportional to one-dimensional marginal posterior distribution at a given parameter value for the corresponding event.

Figure 9

Marginalized distributions for the combined dimensionless tidal deformability parameter $\tilde{\mathrm{\Lambda }}$ (top) and for mass ratio (bottom) for BNS event GW190425. The solid lines correspond to the high-spin prior, while the dashed to the low-spin prior. In the top panel, the vertical lines mark the corresponding 90% upper bound on $\tilde{\mathrm{\Lambda }}$. seobnrv4t_surrogate constrains the mass ratio better toward equal masses for high larger spins. The plot and quoted 90% upper limits for $\tilde{\mathrm{\Lambda }}$ are obtained by reweighing posterior distribution as described in Appendix F of Ref. [32].

Figure 10

Estimates of the dimensionless spin parameters ${\overrightarrow{\chi }}_i=c{\overrightarrow{S}}_i/(G{m}_i^2)$ for merger components of selected sources. Each pixel’s radial distance from the circle’s center on the left (right) side of each disk corresponds to the spin magnitude $|\overrightarrow{\chi }|$ of the more (less) massive component, and the pixel’s angle from the vertical axis indicates the tilt angle ${\theta }_{\mathrm{LS}}$ between each spin and the Newtonian orbital angular momentum. Pixels have equal prior probability, and shading denotes the posterior probability of each pixel, after marginalizing over azimuthal angles.

Figure 11

Posterior and prior distributions on the effective precession spin parameter ${\chi }_p$ for select events. The ${\chi }_p$ prior shown on the right half of each leaf is conditioned on the posterior of ${\chi }_{\mathrm{eff}}$, since the default prior in sampling contains correlations between ${\chi }_p$ and ${\chi }_{\mathrm{eff}}$.

Figure 12

$p$-value plot for the candidate events reconstructed by the minimally modeled pipelines in O3a. The upper panel is from the BW analysis, and the lower panel is from the cWB analysis. The cWB analysis includes the candidate events that are both detected and reconstructed (listed in Table 4) and those that are detected but reconstructed offline. (GW190513_205428, GW190707_093326, GW190728_064510, GW190814, and GW190828_065509). The BW analysis uses the same selection of BBH candidate events as are used for testing general relativity [36]. The $p$ values are sorted in increasing order and plotted vs the order number (which is also the cumulative number of candidate events). Each $p$ value is obtained from the observed on-source match value and the corresponding off-source distribution of the match values from off-source injections. The green band indicates the theoretical 90% symmetric distribution about the null hypothesis (dark green line). Only deviations below the green 90% confidence band indicate disagreement, and we see that there are none. A few of the cWB $p$ values are above the band, indicating either a statistical fluctuation or an overfitting, which we attribute to a small asymmetry between the way the on-source and off-source matches are computed.

Figure 13

Jensen-Shannon divergence $D_{\mathrm{JS}}$ of one-dimensional marginal distributions using different waveform models. The vertical axes show $D_{\mathrm{JS}}$ between one-dimensional posteriors using seobnrv4p and either seobnrv4phm or nrsur7dq4. The horizontal axes show $D_{\mathrm{JS}}$ between one-dimensional posteriors using seobnrv4p and imrphenompv2. The gray regions show the selection criteria from Eq. (a4a): Any events in these regions are presented using higher-order multipole moment results in Sec. 7.

Figure 14

Marginal posterior distributions on the mass ratio $q$, effective inspiral spin ${\chi }_{\mathrm{eff}}$, and source-frame chirp mass $\mathcal{M}$ for GW190924_021846 and GW190521_074359, with the imrphenompv2 and seobnrv4p waveform families.

Figure 15

Marginal posterior distributions on mass ratio $q$, effective inspiral spin ${\chi }_{\mathrm{eff}}$, and source-frame chirp mass $\mathcal{M}$, for GW190519_153544, GW190602_175927, GW190706_222641, and GW190929_012149 with the seobnrv4p, seobnrv4phm, and nrsur7dq4 waveform families.

Figure 16

The upper panels show waveform reconstructions for GW190519_153544 in the LIGO Livingston detector. The waveform posterior from the template-based analysis (shown in orange) is compared to the BW analysis in the upper left and to the cWB analysis in the upper right. The panels also show the 90% credible bands for the Bayesian LALInference and BW algorithms and the 90% confidence band for cWB derived from off-source injections, i.e., by injecting samples from the template-based analysis into data surrounding the event and repeating the analysis multiple times (these bands are computed on an individual time sample basis). The lower two panels show the distribution of overlap values when running BW and cWB on waveforms drawn from the template-based analysis that are injected into data surrounding the event. The fraction of runs with matches below that of the on-source analysis give the $p$ value for the event.

Figure 17

Off-source vs on-source match values for the candidate events in O3a. The upper panel displays the results of the BW analysis, and the lower panel shows the results of the cWB analysis. The on-source match on the horizontal axis is the value obtained comparing the maximum likelihood waveform from parameter estimation with point estimates from the minimally modeled waveform reconstructions. The off-source match on the vertical axis is the median value of the match distribution obtained from off-source injection of sample waveforms from the template-based posterior distribution. In both panels, the errors bars denote the 90% equal-tailed confidence interval.