Detecting long-duration narrow-band gravitational wave transients associated with soft gamma repeater quasiperiodic oscillations

D. Murphy, Maggie Tse, Péter Raffai, I. Bartos, R. Khan, Zsuzsa Márka, Luca Matone, K. Redwine, Szabolcs MárkaPhysical 2013 · DOI · arXiv

Abstract

We have performed an in-depth concept study of a gravitational wave data analysis method which targets repeated long quasimonochromatic transients (triggers) from cosmic sources. The algorithm concept can be applied to multitrigger data sets in which the detector-source orientation and the statistical properties of the data stream change with time, and does not require the assumption that the data is Gaussian. Reconstructing or limiting the energetics of potential gravitational wave emissions associated with quasiperiodic oscillations observed in the x-ray lightcurve tails of soft gamma repeater flares might be an interesting endeavor of the future. Therefore we chose this in a simplified form to illustrate the flow, capabilities, and performance of the method. We investigate performance aspects of a multitrigger based data analysis approach by using $\mathcal{O}(100\text{ }\text{ }\mathrm{s})$ long stretches of mock data in coincidence with the times of observed quasiperiodic oscillations, and by using the known sky location of the source. We analytically derive the probability density function of the background distribution and compare to the results obtained by applying the concept to simulated Gaussian noise, as well as off-source playground data collected by the 4-km Hanford detector during LIGO's fifth science run (S5). We show that the transient glitch rejection and adaptive differential energy comparison methods we apply succeed in rejecting outliers in the fifth science run background data. Finally, we discuss how to extend the method to a network containing multiple detectors, and as an example, tune the method to maximize sensitivity to soft gamma repeater $1806\ensuremath{-}20$ flare times.

Introduction

Soft Gamma Repeaters (SGRs) are magnetars—neutron stars with incredibly intense magnetic fields—that occasionally emit massive flares of X-rays and gamma rays. Following these flares, astronomers have observed Quasi-Periodic Oscillations (QPOs) in the X-ray data, which are thought to be the star's crust and core vibrating like a bell. If these vibrations occur, they should emit gravitational waves (GWs). However, these signals are expected to be weak, narrow-band, and relatively long-lived (lasting hundreds of seconds).

Standard GW searches usually target short "bursts" or continuous waves. This paper, published in 2013, addresses this gap by presenting a robust "multi-trigger" data analysis framework designed to detect these elusive intermediate-duration transients. The method achieves high sensitivity even in the presence of non-Gaussian detector noise and changing instrument orientations by using an adaptive energy-comparison method and combining data from multiple events.

Methodology: The Multi-Trigger Pipeline

The core of the proposed analysis is a pipeline based on an excess energy statistic. The process is designed to handle the "glitchy" and non-stationary noise characteristic of real-world detectors like LIGO.

Overview of the multi-trigger data analysis pipeline for both background and flare regions.

The pipeline follows several key steps:

  1. Channel Splitting: Data is filtered into three frequency bands: a "QPO channel" for the expected signal and two adjacent "up" and "down" channels to measure the local background noise.
  2. Adaptive Weighting ($w$): To account for the curvature of the LIGO noise floor, the algorithm calculates a local weighting factor ($w$).
  3. Differential Energy ($\Delta E_{excess}$): The pipeline compares the excess energy during a flare window to the energy in a "background" window immediately preceding it. This handles noise that changes over time.
  4. Glitch Rejection: A "veto" system divides data into small tiles and rejects those with outlier energy levels, cleaning the data of instrumental artifacts without assuming Gaussian noise.
  5. Statistical Combining: The two-sample Kolmogorov-Smirnov (K-S) test compares flare vs. background distributions, and Fisher’s method combines results from multiple flares or detectors into a single significance value.

Performance and Validation

The authors validated the pipeline using both simulated Gaussian noise and real "playground" data from LIGO’s fifth science run (S5). Numerical simulations matched the analytically derived probability density function (PDF) with a mean percentage error of only ~1.25%.

A critical challenge in GW astronomy is the non-Gaussian nature of detector noise, which often contains "glitches." The transient glitch rejection method implemented here proved highly effective.

Impact of data quality (DQ) measures and glitch vetoes on the distribution of excess energy in real LIGO S5 data.

As shown in the results, applying Data Quality (DQ) flags and the specific glitch veto successfully removed outliers from real S5 data. This cleaning process made the real-world data behave almost exactly like theoretical Gaussian noise, which is essential for reliable statistical inference.

Key Insights and Results

  • The Power of Stacking: By "stacking" the statistical significance of many triggers, the pipeline can detect signals much fainter than what a single event would allow.
  • Optimal Subsets: The study found that including every available trigger isn't always optimal. Because the Earth rotates, detectors are sometimes "blind" to specific sky locations. Excluding triggers with poor detector-source alignment actually maximizes overall network sensitivity.
  • Sensitivity Limits: For the H1-L1 detector network, the authors estimated a sensitivity ($h_{rss}$) of $5.42 \times 10^{-22} \text{ Hz}^{-1/2}$. Assuming a distance of 14 kpc for SGR 1806−20, this could set an upper limit on isotropic GW emission of $1.87 \times 10^{46}$ ergs—comparable to the total electromagnetic energy released in giant magnetar flares.

Conclusion and Future Directions

While the current framework assumes GW signal frequency and duration match X-ray QPOs, real-world signals might have frequency offsets or time lags. Future work will involve refining the "excess energy" approach for more complex waveforms and studying how varying signal durations across different flares affects the "stacking" logic. Nonetheless, this study provides a robust foundation for targeted searches of long-duration GW transients in the advanced detector era.

Summary updated

Abstract

We have performed an in-depth concept study of a gravitational wave data analysis method which targets repeated long quasimonochromatic transients (triggers) from cosmic sources. The algorithm concept can be applied to multitrigger data sets in which the detector-source orientation and the statistical properties of the data stream change with time, and does not require the assumption that the data is Gaussian. Reconstructing or limiting the energetics of potential gravitational wave emissions associated with quasiperiodic oscillations observed in the x-ray lightcurve tails of soft gamma repeater flares might be an interesting endeavor of the future. Therefore we chose this in a simplified form to illustrate the flow, capabilities, and performance of the method. We investigate performance aspects of a multitrigger based data analysis approach by using $\mathcal{O}(100\text{ }\text{ }\mathrm{s})$ long stretches of mock data in coincidence with the times of observed quasiperiodic oscillations, and by using the known sky location of the source. We analytically derive the probability density function of the background distribution and compare to the results obtained by applying the concept to simulated Gaussian noise, as well as off-source playground data collected by the 4-km Hanford detector during LIGO's fifth science run (S5). We show that the transient glitch rejection and adaptive differential energy comparison methods we apply succeed in rejecting outliers in the fifth science run background data. Finally, we discuss how to extend the method to a network containing multiple detectors, and as an example, tune the method to maximize sensitivity to soft gamma repeater $1806\ensuremath{-}20$ flare times.
BibTeX
@inproceedings{murphy2013detecting,
  author = {D. Murphy and Maggie Tse and Péter Raffai and I. Bartos and R. Khan and Zsuzsa Márka and Luca Matone and K. Redwine and Szabolcs Márka},
  title = {{Detecting long-duration narrow-band gravitational wave transients associated with soft gamma repeater quasiperiodic oscillations}},
  booktitle = {Physical},
  year = {2013},
  doi = {10.1103/physrevd.87.103008},
}
Figure