All-sky search in early O3 LIGO data for continuous gravitational-wave signals from unknown neutron stars in binary systems

Rapidly spinning neutron stars are promising sources of continuous gravitational waves. Detecting such a signal would allow probing of the physical properties of matter under extreme conditions. A significant fraction of the known pulsar population belongs to binary systems. Searching for unknown neutron stars in binary systems requires specialized algorithms to address unknown orbital frequency modulations. We present a search for continuous gravitational waves emitted by neutron stars in binary systems in early data from the third observing run of the Advanced LIGO and Advanced Virgo detectors using the semicoherent, GPU-accelerated, binaryskyhough pipeline. The search analyzes the most sensitive frequency band of the LIGO detectors, 50–300 Hz. Binary orbital parameters are split into four regions, comprising orbital periods of three to 45 days and projected semimajor axes of two to 40 light seconds. No detections are reported. We estimate the sensitivity of the search using simulated continuous wave signals, achieving the most sensitive results to date across the analyzed parameter space.

Figure 1

Binary orbital parameters considered by the present search. Solid color regions denote parameter space regions in which a search was performed; blue dots mark binary orbital parameters corresponding to the known binary pulsar population. Regions A, B, C, and D were covered by the low-frequency analysis, while region B was covered by the high-frequency analysis as well. Time of ascending node passage is taken into account according to the orbital frequency, as explained in Sec. 2. Pulsar population data was taken from [14] using [24].

Figure 2

Explicit description of the binaryskyhough toplist construction process.

Figure 3

Example of 95% sensitivity depth interpolation. Five sensitivity depths were selected at 124.625 Hz in region H1B, and 200 injections were injected at each of these depths, applying the criteria exposed in the text in order to label injections as detected or not detected. Blue dots represent the fraction of detected injections; the sigmoid fit is represented by a blue line; fit uncertainties at one, two, and three sigmas are represented by pale yellow shades. The interpolated 95% sensitivity depth ${\mathcal{D}}^{95\%}=21\pm 0.4$ is marked using a star.

Figure 4

Average 95% sensitivity depths obtained in the low-frequency (top panel) and high-frequency (bottom panel) bands. Data points correspond to the interpolated results obtained through the sigmoid fit of the efficiencies at the selected frequency bands (five bands randomly selected in each 25 Hz). Error bars correspond to 95% efficiency uncertainties toward low depth values. Shaded regions show the averaged results with their uncertainties, as summarized in Table 6. In the top panel, shading is only shown for the results obtained for the binary parameter space region B.

Figure 5

Implied 95% efficiency amplitude from the obtained sensitivity depth values. The $h_0^{95\%}$ amplitude estimates are obtained from the 95% efficiency depth values shown in Table 6 and the inverse squared averaged PSD using Eq. (15). Low-frequency results are shown for binary parameter space region B.

Figure 6

Recovered $2\widehat{\mathcal{F}}$ values using a set of injections labeled as detected by the sensitivity estimation criteria. The amplitudes of these injections were distributed using the five sensitivity depth values explained in Sec. 6. The top panel shows the results for the four regions of the low-frequency search, involving 33757 injections; the bottom panel shows the same result for the high-frequency search, using 37549 injections. The horizontal orange line marks the threshold $2{\widehat{\mathcal{F}}}_{\mathrm{th}}=2500$.

Figure 7

MCMC follow-up results. Dots represent cluster centers not vetoed by the identified line veto, and a horizontal line represents the imposed threshold $2{\widehat{\mathcal{F}}}_{\mathrm{th}}=2500$, calibrated by software injections (see Fig. 6). The horizontal axis represents the frequency value associated to each cluster center, and the vertical axis represents the maximum $2\widehat{\mathcal{F}}$ value sampled by the MCMC run.

Figure 8

Example of outliers produced by instrumental noise in one of the Advanced LIGO detectors. Each pair of panels corresponds to a different outlier; the segment-wise $2{\tilde{\mathcal{F}}}_{\alpha }$ statistic is shown at the top of each pair; the cumulative semicoherent $2\widehat{\mathcal{F}}$ as described in Eq. (22) is shown at the bottom of each pair. Dashed lines denote an H1-only analysis, dotted blue lines an L1-only analysis, and solid black lines a multidetector analysis. The $2{\widehat{\mathcal{F}}}_{\mathrm{th}}=2500$ threshold is shown as a horizontal line.

Figure 9

Example of an outlier vetoed by the multidetector consistency veto. Dashed lines denote an H1-only analysis, dotted blue lines an L1-only analysis, and solid black lines a multi-detector analysis. The $2{\widehat{\mathcal{F}}}_{\mathrm{th}}=2500$ threshold is shown as a horizontal line.

Figure 10

Example of an outlier surpassing the $2{\widehat{\mathcal{F}}}_{\mathrm{th}}=2500$ threshold later vetoed by inspection of the detector spectra. The combined ASD of both detectors around the frequency of this group of outliers is shown in Fig. 11. Candidates related to this signal saturated the top list in the four regions LA, LB, LC, LD. Dashed lines denote an H1-only analysis, dotted blue lines an L1-only analysis, and solid black lines a multidetector analysis. The $2{\widehat{\mathcal{F}}}_{\mathrm{th}}=2500$ threshold is shown as a horizontal line.

Figure 11

Location of the manually inspected outliers (vertical dashed line) with respect to the amplitude spectral density of the detector, here represented as a multidetector inverse squared average (orange line). A diamond and a square mark the twin peaks’ frequency, 59.53 Hz and 60.47 Hz, respectively.

Figure 12

Astrophysical reach (left) and equatorial ellipticity (right) implied by the search sensitivity $h_0^{95\%}$. These results were obtained assuming a canonical neutron star moment of inertia $I_{\mathrm{z}}={10}^{38}\mathrm{kg}·{\mathrm{m}}^2$. Shaded areas denote regions excluded due to the spin-down limit implied by the maximum spin-down value considered in this search.