Implementing a search for aligned-spin neutron star-black hole systems with advanced ground based gravitational wave detectors

We study the effect of spins on searches for gravitational waves from compact binary coalescences in realistic simulated early advanced LIGO data. We construct a detection pipeline including matched filtering, signal-based vetoes, a coincidence test between different detectors, and an estimate of the rate of background events. We restrict attention to neutron star-black hole (NS-BH) binary systems, and we compare a search using nonspinning templates to one using templates that include spins aligned with the orbital angular momentum. To run the searches we implement the binary inspiral matched-filter computation in PyCBC, a new software toolkit for gravitational-wave data analysis. We find that the inclusion of aligned-spin effects significantly increases the astrophysical reach of the search. Considering astrophysical NS-BH systems with nonprecessing black hole spins, for dimensionless spin components along the orbital angular momentum uniformly distributed in $(-1,1)$, the sensitive volume of the search with aligned-spin templates is increased by $\sim 50\%$ compared to the nonspinning search; for signals with aligned spins uniformly distributed in the range (0.7,1), the increase in sensitive volume is a factor of $\sim 10$.

Figure 1

Sensitivity curves used in this work. The black solid curve corresponds to the recolored data used for testing the search.

Figure 2

Flowchart of the search pipeline. Data from the Hanford (H) and Livingston (L) detectors are processed by the main search engine pycbc_inspiral which computes the SNR and ${\chi }_r^2$ time series for a common template bank which is also fixed in time. This results in a list of unclustered single-detector triggers which then pass through a coincidence step followed by clustering over a suitable time window. The single-detector triggers can additionally be clustered independently in each detector.

Figure 3

Mass boundaries used in constructing our nonspinning and spinning template banks. The nonspinning bank includes templates with NS masses above the usual NS mass range. As explained in the text, these templates are able to detect spinning NS-BH signals with a NS mass in the usual range.

Figure 4

Template density of the spinning stochastic bank in $({\tau }_0,{\tau }_3)$ coordinates. The black contour delimits the nonspinning region of NS-BH parameter space and the dashed lines show the additional NS mass range allowed by the nonspinning bank. Templates above and below the black contour correspond to ${\chi }_{\mathrm{BH}}<0$ and ${\chi }_{\mathrm{BH}}>0$, respectively.

Figure 5

Fitting factor of SpinTaylorT2 NS-BH signals with fixed masses ($m_{\mathrm{BH}}=7.8M_{\odot }$, $m_{\mathrm{NS}}=1.35M_{\odot }$) and different BH spins and nonspinning template banks constructed for different choices of sensitivity and lower cutoff frequency $f_L$ (see Fig. 1). The nonspinning bank considered in the rest of this paper corresponds to the early advanced-LIGO sensitivity (top right plot).

Figure 6

Mass parameters recovered in the nonspinning bank for SpinTaylorT2 signals with fixed masses ($m_{\mathrm{BH}}=7.8M_{\odot }$, $m_{\mathrm{NS}}=1.35M_{\odot }$, dashed lines) and variable dimensionless BH spin (x axes).

Figure 7

Mass and spin parameters recovered in the spinning bank for SpinTaylorT2 signals with fixed masses ($m_{\mathrm{BH}}=7.8M_{\odot }$, $m_{\mathrm{NS}}=1.35M_{\odot }$) and variable dimensionless BH spin (x axes). Dashed lines show the true parameters.

Figure 8

Performance of the two template banks across the whole nonprecessing NS-BH parameter space. The color scale shows the lowest fitting factor found in each hexagonal bin (note the different color scales). The region of poor match in the spinning bank is likely due to the different termination conditions of templates and test waveforms.

Figure 9

Rate of single-detector false alarms for the spinning and nonspinning searches as a function of the threshold on SNR and reweighted SNR.

Figure 10

Rate of coincident false alarms for the spinning and nonspinning searches as a function of the threshold on combined reweighted SNR. The shaded dotted curve shows the nonspinning curve multiplied by the relative number of templates of the two searches ($\sim 5\times$).

Figure 11

Top row: fraction of optimal combined SNR and reweighted SNR recovered by the nonspinning search for each simulated (and found) source, as a function of the BH spin (compare with Fig. 5). Middle row: combined SNR and reweighted SNR recovered by the nonspinning search for each found source vs the optimal combined SNR. The color distinguishes between high ($>0.4$) and low ($<0.4$) BH spin magnitude. Bottom row: same as middle row, for the spinning search.

Figure 12

ROC curves for the spinning and nonspinning searches, comparing the relative sensitive volume (or number of detections) at fixed false-alarm rate. The four panels assume NS-BH systems with different limits on a uniform distribution of BH spins.