A semicoherent glitch-robust continuous-gravitational-wave search method

Isolated nonaxisymmetric rotating neutron stars producing continuous-gravitational-wave signals may undergo occasional spin-up events known as glitches. If unmodeled by a search, these glitches can result in continuous wave signals being missed or misidentified as detector artifacts. We outline a semicoherent glitch-robust search method that allows identification of continuous wave signal candidates that contain glitches and inferences about the model parameters. We demonstrate how this can be applied to the follow-up of candidates found by wide-parameter space searches. We find that a Markov chain Monte Carlo method outperforms a grid-based method in speed and accuracy.

Figure 1

Grid-corner plot showing various marginalizations of the glitch-robust semicoherent $\mathcal{F}$-statistic computed using a grid-based search over the prior ranges in Table 2. Solid lines indicate the simulated signal parameters.

Figure 2

Corner plot showing various marginalizations of the exponential of the glitch-robust semicoherent $\mathcal{F}$-statistic computed using a MCMC-based search. Solid lines indicate the simulated signal parameters while dashed lines (on the one-dimensional histograms) indicate $1\text{−}\sigma$ quantiles. The figure was generated using the corner [27] package.

Figure 3

Comparison of the relative maximum $\widehat{2\mathcal{F}}$ found by each method compared to ${\widehat{2\mathcal{F}}}_{\text{inject}}$, the value calculated at the simulated signal parameters ${\mathbf{\lambda }}_s$. All timings were performed on an Intel Core i7-7820HQ CPU @ 2.90 GHz processor.

Figure 4

Monte Carlo study of the $B_{\mathrm{gS}/\mathrm{S}}(\mathit{x},N_{\mathrm{g}}=1)$ Bayes factor as a function of the simulated glitch magnitude. A dashed vertical line indicates the value of Eq. (13) for the 50-day duration used in this study.

Figure 5

The fully coherent $\widetilde{2\mathcal{F}}$ computed over a grid in $f$ and $\dot{f}$ for the simulated glitching signal.

Figure 6

Illustration of the frequency sliding window, a useful diagnostic tool for identifying glitching signals. In this example, a 10-day window is slid in increments of 1 day.