Detecting and reconstructing gravitational waves from the next galactic core-collapse supernova in the advanced detector era

We performed a detailed analysis of the detectability of a wide range of gravitational waves derived from core-collapse supernova simulations using gravitational-wave detector noise scaled to the sensitivity of the upcoming fourth and fifth observing runs of the Advanced LIGO, Advanced Virgo, and KAGRA. We use the coherent WaveBurst algorithm, which was used in the previous observing runs to search for gravitational waves from core-collapse supernovae. As coherent WaveBurst makes minimal assumptions on the morphology of a gravitational-wave signal, it can play an important role in the first detection of gravitational waves from an event in the Milky Way. We predict that signals from neutrino-driven explosions could be detected up to an average distance of 10 kpc, and distances of over 100 kpc can be reached for explosions of rapidly-rotating progenitor stars. An estimated minimum signal-to-noise ratio of 10–25 is needed for the signals to be detected. We quantify the accuracy of the waveforms reconstructed with coherent WaveBurst and we determine that the most challenging signals to reconstruct are those produced in long-duration neutrino-driven explosions, and models that form black holes a few seconds after the core bounce.

Figure 1

GW energy as a function of peak frequency (maximum of $d{E}_{\mathrm{GW}}/df$) for 82 analyzed waveforms. The signals from 2D models are shown with hollow symbols. Spectra of some waveforms are wide band and the peak frequencies could not be accurately determined. For the majority of the signals, the peak frequencies lay between 300 Hz and 1000 Hz that usually corresponds to the proto-neutron star oscillations. The typical energy is in the range from ${10}^{-10}$ to ${10}^{-7}{M}_{\odot }{c}^2$ that is smaller than 0.01% of a typical CCSN explosion energy. The current GW energy constraints are $4.27\times {10}^{-4}{M}_{\odot }{c}^2$ at 235 Hz and $1.28\times {10}^{-1}{M}_{\odot }{c}^2$ at 1304 Hz [16].

Figure 2

Examples of the GW energy evolution. The $\mathrm{Abd}+14$ waveforms are short and energetic core-bounce GW signals. For the neutrino-driven explosions most of the energy is emitted after around 100 ms.

Figure 3

The GW signals from CCSNe are typically broadband with the majority of the energy at higher frequencies. The peak frequency can be difficult to determine for some waveforms, like for $\mathrm{Abd}+14$ A3O01.0.

Figure 4

The projected noise amplitudes of the LIGO, Virgo, and KAGRA detectors for O4 and O5. The characteristic strains are broadband with the majority of the energy in higher frequencies that usually is emitted by PNS oscillations. $\mathrm{Abd}+14$ represents the core-bounce signal that is strong but broadband. The GW150914 signal is shown for comparison.

Figure 5

Example O2 LIGO Hanford detector sensitivity rescaled to O4 and O5 designs. The rescaling procedure preserves all features of the real noise.

Figure 6

The comparison between an example blip glitch and core-bounce waveforms for $\mathrm{Dim}+08$ and $\mathrm{Rich}+17$.

Figure 7

Detection efficiency curves for example waveforms are presented in panels (a) and (b). The numbers in the brackets are distances at 50% detection efficiencies. Panel (c) shows the distances at 10%, 50%, and 90% detection efficiencies for all waveforms analyzed in this paper. The predicted detection ranges for O5 are typically between 1 kpc and 100 kpc. This range contains the distances to the Galactic center and the Large Magellanic Cloud (LMC) that hosted SN 1987A. The detectability of GW signals coming from 3D simulations can reach almost 100% detection efficiency at close distances, while linearly polarized waveforms (hollow symbols) reach only 70% of the detection efficiency.

Figure 8

Detection efficiency curves as a function of injected SNR (${\mathrm{SNR}}_{\mathrm{inj}}$) for example waveforms are presented in panels (a) and (b). The numbers in the brackets are SNRs at 50% detection efficiencies. Panel (c) shows the SNRs at 10%, 50%, and 90% detection efficiencies for all waveforms analyzed in this paper. The waveforms are typically detectable at SNR of 10–25. The $\mathrm{Cer}+13$ fiducial waveform is almost 2 s long which makes it challenging to detect. The dominant GW emission periods for $\mathrm{Rad}+19$ s25 are 0.6 s apart, making it challenging to group them by the algorithm. The linearly-polarized waveforms (hollow symbols) reach only 70% detection efficiency.

Figure 9

Panel (a) shows an example $\mathrm{Oco}+18$ mesa20 waveform with an overlay of the cWB reconstruction. The waveform is relatively long and broadband and for small SNR values the algorithm does not reconstruct a large part of the signal. Panel (b) quantifies the SNR difference as a function of injected SNR. The cWB search reconstructs well the signals that are short in duration and its capabilities decrease with the increasing signal complexity in their time-frequency evolution.

Figure 10

Waveform overlap as a function of the injected SNR. The waveforms that are short, like $\mathrm{Rad}+17$ are accurately reconstructed for small SNRs, while waveforms that are long and broadband, like $\mathrm{Oco}+18$, need to be relatively strong to be reconstructed accurately.

Figure 11

Waveform overlaps (reconstruction accuracy) for all waveforms analyzed in this paper at injected SNR of 20, 40, and 60. The accuracy of reconstruction is close to unity for short waveforms such as $\mathrm{Abd}+14$, $\mathrm{Dim}+08$, and $\mathrm{Ric}+17$. It decreases with waveform length and its complexity in the time-frequency evolution. The signals from 2D models are shown with hollow symbols.

Figure 12

Detection efficiency curves as a function of (a) distance and (b) injected SNR. The numbers in the brackets show detection range and minimum detectable SNR, respectively. The three- and four-detector networks are significantly less sensitive than for two-detector networks.