Quasi-periodic pulsations in solar flares: new clues from theFermi Gamma-Ray Burst Monitor

Abstract

Aims. In the past four decades, it has been observed that solar flares display quasi-periodic pulsations (QPPs) from the lowest, i.e. radio, to the highest, i.e. gamma-ray, frequencies in the electromagnetic spectrum. It remains unclear which mechanism creates these QPPs. In this paper, we analyze four bright solar flares that display compelling signatures of quasi-periodic behavior and were observed with the Gamma-Ray Burst Monitor (GBM) onboard the Fermi satellite. Because GBM covers over three decades in energy (8 keV to 40 MeV), it is regarded as a key instrument in our attempt to understand the physical processes that drive solar flares. Methods. We tested for periodicity in the time series of the solar flares observed by GBM by applying a classical periodogram analysis. However, in contrast to previous authors, we did not detrend the raw light curve before creating the power spectral density (PSD) spectrum. To assess the significance of the frequencies, we used a method that is commonly applied to X-ray binaries and Seyfert galaxies. This technique takes into account the underlying continuum of the PSD, which for all of these sources has a P( f ) ∼ f −α dependence and is typically labeled red-noise. Results. We checked the reliability of this technique by applying it to observations of a solar flare that had been observed by the Reuven Ramaty High-Energy Solar Spectroscopic Imager (RHESSI ). These data contain, besides any potential periodicity from the Sun, a 4 s rotational period caused by the rotation of the spacecraft about its axis. We were unable to identify any intrinsic solar quasi-periodic pulsation but we did manage to reproduce the instrumental periodicity. Moreover, with the method adopted here, we do not detect significant QPPs in the four bright solar flares observed by GBM. We stress that for this kind of analyses it is of utmost importance to account appropriately for the red-noise component in the PSD of these astrophysical sources.

Above the Noise: The Search for Periodicities in the Inner Heliosphere

“…The fit of the noise model to the spectrum in the 193 Å case relies more heavily on the κ-component of Equation 1 than the 171 Å spectrum. Figure 3a is a closer fit to a simple power-law model (used by Gruber et al, 2011), while Figure 3b appears closer to a broken power law plus additional humps, as per Equation 1. The findings of the o1 box illustrate that the same underlying structures may yield relatively different background noise characteristics in different passbands. The same is true for box f2 (Figure 4), which contains our most likely positive detection of a periodic signature above confidence.…”

“…Vaughan, 2005;Gruber et al, 2011), but the technique has only recently started to gain traction in the analysis of light curves from solar environments/instruments. Several articles, particularly in the context of quasi-periodic pulsations (QPPs) associated with solar flares, now incorporate the fundamental power-law nature of power spectra in the analysis of coronal time series, but they use several different approaches (e.g.…”

“…Vaughan, 2005) could be achieved by a solution where B = 0 and C = 0, while a broken power law (e.g. Gruber et al, 2011) could be reproduced by a solution where B = 0 only. Global confidence levels to the resulting fit were then generated using the method described by Auchère et al (2016).…”

“…Several recent articles (e.g. Gruber et al, 2011;Inglis, Ireland, and Dominique, 2015;Ireland, McAteer, and Inglis, 2015;Auchère et al, 2016) have invalidated several reported QPP observations (e.g. from Inglis and Nakariakov, 2009;Nakariakov et al, 2010) that originally incorporated detrending and did not account for the power-law nature of the recovered spectra.…”

“…Several methods have been proposed, e.g. by Vaughan (2005), Gruber et al (2011), Auchère et al (2016), Pugh, Broomhall, and Nakariakov (2017), in order to identify periodic signals in red-noise-dominated spectra. These methods typically comprise two distinct components: first the mean power as a function of frequency has to be estimated (based on fitting a power law, plus additional components).…”