A resampling algorithm to detect continuous-wave signals from neutrons stars in binary systems

Rapidly rotating neutron stars (NSs) with non-axisymmetric deformations are an interesting type of continuous-wave (CW) source for the advanced LIGO-Virgo detectors. Within the sensitivity band of these detectors, more than half of the known pulsars (i.e., ≳ 1200) in our galaxy are in binary systems. CWs are received at Earth-based detectors with a Doppler modulation due to the relative motion between the source and the detector. Consequently, the observed phase evolution depends on the intrinsic signal frequency, frequency time derivatives and source sky position. Besides the aforementioned modulations, CW signals coming from binary sources are also Doppler modulated due to the source orbital motion, which spreads the signal in several frequency bins resulting in a significant loss of signal-tonoise ratio, making thus the signal much harder to be detected. Searching for these weak signals requires coherent integration of long data stretches and precise knowledge of these modulations. Electromagnetic observations have provided us with the best possible parameter estimates for Scorpius X-1, which is the brightest and closest low-mass x-ray binary. Due to the absence of knowledge of its rotating frequency, and relatively large uncertainties in binary orbital Keplerian parameters, however, these are not enough to model precisely its phase evolution [1, 2]. The Doppler correction for unknown parameters implies an extensive computational burden, as the computation cost follow a geometrical progression with the number of unknown parameters. Time-domain correction using the generalized 5-vector methodology [3], for low eccentric binary NSs [1] in the context of directed/narrowband searches, is more affordable, and we investigate its robustness over uncertainty in all orbital parameters. We also present the 90% confidence-level sensitivity estimation for a real Scorpius X-1 search and 8 targeted binary millisecond pulsars across the most sensitive region of interferometer (i.e., [100-199] Hz). Using the same methodology, we empirically estimate a formulation, which is curve fitted to the detection statistics [4] as a function of binary parameter uncertainties.