Source-linked AI summary
The First Survey of X-ray Flares from Gamma Ray Bursts Observed by Swift: Temporal Properties and Morphology
G. Chincarini, A. Moretti, P. Romano, A. D. Falcone, D. Morris, J. Racusin, S. Campana, C. Guidorzi, G. Tagliaferri, D. N. Burrows, C. Pagani, M. Stroh, D. Grupe, M. Capalbi, G. Cusumano, N. Gehrels, P. Giommi, V. La Parola, V. Mangano, T. Mineo, J. A. Nousek, P. T. O'Brien, K. L. Page, M. Perri, E. Troja, R. Willingale, B. Zhang
TL;DR
The paper addresses whether Swift/XRT X-ray flares reflect prolonged central-engine activity and how they relate to prompt gamma-ray pulses. It analyzes a comprehensive flaring-GRB sample and finds temporal properties and pulse-ratio behavior consistent with a common, predominantly internal origin, while recognizing selection limits and alternative refreshed-shock explanations for some events.
Problem
The study asks whether X-ray flares are linked to prolonged central-engine activity and whether their properties connect them to BAT prompt-emission pulses.
Method
The authors analyze XRT flares in a sample of 33 GRBs and cross-check their timing and intensity properties against BAT prompt pulses using simulations to assess selection biases.
Results
The mean flare ratio is ⟨Δt/t⟩ = 0.13 ± 0.10; pulse and flare counts show no correlation, but successive-event intensity-ratio distributions match, while 29/69 flares may fit refreshed shocks and 10/69 require internal shocks.
Takeaways & Limitations
The evidence supports a common origin for gamma-ray pulses and X-ray flares, with flares related to central-engine activity and arising from internal rather than external shocks within the standard fireball scenario.
Takeaways & Limitations
Flares can be blended and XRT time resolution degrades during follow-up, so the measured low Δt/t is an upper limit on intrinsic flare sharpness.
Abstract
from arXiv · showhide
We present the first systematic investigation of the morphological and timing properties of flares in GRBs observed by Swift/XRT. We consider a large sample drawn from all GRBs detected by Swift, INTEGRAL and HETE-2 prior to 2006 Jan 31, which had an XRT follow-up and which showed significant flaring. Our sample of 33 GRBs includes long and short, at low and high redshift, and a total of 69 flares. The strongest flares occur in the early phases, with a clear anti-correlation between the flare peak intensity and the flare time of occurrence. Fitting each X-ray flare with a Gaussian model, we find that the mean ratio of the width and peak time is <Delta t / t > = 0.13+/-0.10, albeit with a large scatter. Late flares at times > 2000 seconds have long durations, Delta t>300 s, and can be very energetic compared to the underlying continuum. We further investigated if there is a clear link between the number of pulses detected in the prompt phase by BAT and the number of X-ray flares detected by XRT, finding no correlation. However, we find that the distribution of intensity ratios between successive BAT prompt pulses and that between successive XRT flares is the same, an indication of a common origin for gamma-ray pulses and X-ray flares. All evidence indicates that flares are indeed related to the workings of the central engine and, within the standard fireball scenario, originate from internal shocks rather than external shocks. While all flares can be explained by long-lasting engine activity, 29/69 flares may also be explained by refreshed shocks. However, 10 can only be explained by prolonged activity of the central engine.
1. Introduction
Swift revealed that X-ray flares are frequent, energetic, rapidly varying features with properties favoring a mechanism distinct from the afterglow. This work systematically tests their temporal behavior and connection to prompt gamma-ray pulses.
- Swift/XRT observations revealed flares in a large percentage of GRB X-ray light curves.
- Flare fluences can reach 100% of the prompt fluence, with large flux variations occurring on short timescales.Reported examples include ΔF/F values of approximately 6, 500, and 25 for three events.
- The underlying continuum can retain the same slope before and after a flare, leaving no trace of energy injection.
- Very short Δt/tpeak, large ΔF/F, and harder-to-softer flare spectra generally disfavor external-shock explanations.
- The study performs a comprehensive temporal analysis of flaring GRBs and cross-checks XRT flares against BAT prompt-emission pulses.
2. Sample definition
The sample was built from GRBs detected by Swift, INTEGRAL, and HETE-2 before 2006 January 31, then reduced to events suitable for full timing analysis. Completeness for faint flares was established mainly at late times.
- 119 detected events yielded 99 GRBs with XRT positions, from which light curves were searched for flares above an underlying power law.
- The full sample contains 33 GRBs selected for timing analysis, with redshifts, T90 values, and BAT fluences tabulated when available.
- The flare sample is complete for faint flares only at late times, typically around 10^3 seconds after the trigger.
- Three events were excluded from full analysis because of low statistics, while another lacked sufficient XRT coverage to characterize its flare.
- The restricted sample contains 30 GRBs, and its membership differs from a spectroscopic-analysis sample because the analysis requirements differ.
3. Data Reduction
XRT event data were calibrated, screened, and converted into corrected light curves using standard processing choices for observing mode, extraction, and detector effects.
- Level 1 XRT event lists were reprocessed with XRTDAS and xrtpipeline using calibration and standard filtering criteria.
- An approximately 0.2 keV central-pixel threshold was applied to reduce bright-Earth-limb and CCD-dark-current background.
- Observation modes were selected according to source count rates to minimize pile-up, with special handling for calibration-phase and bright observations.
- Source extraction used circular or annular regions adjusted for source brightness and pile-up, with PSF fitting determining the inner annular radius when needed.
- Light curves covered 0.2–10 keV with at least 20 source counts per bin and corrections for background, pile-up, vignetting, exposure, and PSF losses.
4. Data Analysis
The analysis models X-ray flares as Gaussian components superimposed on broken-power-law continua, then quantifies their timing, shape, energetics, and observational biases. Across 69 Gaussian-modeled flares, the width-to-peak-time ratio has mean ⟨∆t/t⟩=0.13±0.10, while sampling and blending can bias measured sharpness upward.
- Measured properties: The analysis measured ∆t/t, αfall, ∆tfall/∆trise, flare energetics, and flare-to-burst flux ratios using approaches adapted to each flare’s statistics.Rise and fall times were defined relative to an assumed underlying power-law continuum.
- Equivalent widths: 48 flares had measurable equivalent widths, but discrete sampling and flare faintness limited this measurement.Equivalent width compares flare fluence with the time-integrated underlying continuum and is reported in seconds.
- Continuum and flare modeling: The underlying continuum was modeled with simple, broken, or doubly-broken power laws, while superimposed flares were represented by iteratively added Gaussians.A joint fit provided continuum and flare parameters, including peak flux relative to the underlying continuum.
- Gaussian flare timing: 69 flares were fit with Gaussians, using σ as ∆t and the Gaussian peak time as t, yielding ⟨∆t/t⟩=0.13±0.10.The distribution peaks at 0.13, with fitted peak times from 95 s to approximately 75 ks.
- Selection effects: Blended flares overestimate EW, ∆t, and ∆t/t, so the measured low ∆t/t is an upper limit on intrinsic flare sharpness.Time resolution also causes early short flares to resemble steep power laws and late short flares to be smeared below detection.
- Selection effects: Simulations found detection probability above 90% for ∆t/t > 2×10^-3 and t > 10^4 s, while observations were generally not biased against ∆t/t ≲0.1.The simulations were repeated 14,000 times and compared with the observed sampling and detection procedure.
5. XRT flares vs. BAT pulses
The study finds no relationship between how many BAT prompt pulses and XRT flares a burst has, but successive-event intensity ratios match across the two classes.
- 46 pulses were identified in 28 gamma-ray profiles, with negligible contamination from statistical fluctuations.
- No clear correlation exists between the number of gamma-ray pulses and the number of X-ray flares.A KS test found no basis to infer flare counts from pulse counts, or vice versa.
- The pulse-count and flare-count distributions show no significantly different origin, with a 30% probability of being drawn from the same distribution.
- The distributions of intensity ratios between successive gamma-ray pulses and successive X-ray flares are consistent.For bursts with at least two events, the merged ratios follow a log-normal distribution with mean −0.258 and σlog = 0.68.
- 10^-0.157 ≃ 0.7: on average, each successive event peaks at about 0.7 times the preceding event, with scatter from 0.3 to 1.8.This shared relation is evidence for a common origin of gamma-ray pulses and X-ray flares.
6. Results
The analysis characterizes flare morphology, timing, energetics, and correlations across a 33-burst sample, finding diverse flares with narrow relative widths and weaker but longer late events.
- Late flares are less intense than early flares but last much longer, allowing their fluence to become very large.The peak-intensity versus occurrence-time correlation has rs = −0.539 for 63 points, while EW and tpeak have rs = 0.729 for 48 points.
- The decay-time ratio shows a tentative correlation with τ90, with rs = 0.543 for 24 flares.
- 33 light curves were gathered, yielding 69 Gaussian-modeled flares; equivalent widths were calculated for 48, decay slopes for 35, and τ90-related measures for 24.The analysis modeled the underlying continuum with multiply-broken power laws and flares with analytical functions.
- Flares can be modeled with Gaussians over a multiply-broken power-law continuum, although asymmetric exponential or power-law profiles are often required for accurate fits.
- Flares occur in long and short GRBs, high-energy-peaked bursts and XRFs, and both early and late XRT light curves.The summary lists 32 long and 2 short GRBs, and 32 high-energy-peaked bursts versus 2 XRFs.
- 8 s–7×10^5 s: equivalent widths span this range, measuring flare fluence relative to the underlying continuum.
- ⟨∆t/t⟩ = 0.13 ± 0.10: flare width-to-peak-time ratios show substantial scatter, with simulations supporting sensitivity to ∆t/t < 0.1.Simulations also show no sharp flares at large times.
- 1.3–6.8: decay slopes generally agree with the curvature effect, while decay-to-rise time ratios range from 0.5 to 8.
7. Discussion
The discussion links X-ray flares to prompt-like central-engine activity rather than predominantly external-shock mechanisms. Statistical comparisons and kinematic tests support internal shocks, while refreshed shocks remain viable for a subset.
- Prompt–flare connection: No correlation was found between prompt-emission characteristics and X-ray-flare frequency, although the sample may include random flare events after initial prompt flickering.The authors note that contamination from early XRT flares being late prompt-emission tails does not change this result.
- Flare energetics: Late flares are not as intense as early flares but can be highly energetic because of their longer durations.The analysis reports no detection bias favoring high-intensity late flares.
- Central-engine activity: The flare behavior resembles prompt emission superimposed on a standard light curve in both long and short bursts.This similarity is presented as evidence for activity associated with the central engine.
- Shock interpretation: 29/69 flares fall in the refreshed-shock region, whereas 10/69 can only be explained by internal shocks; only one is consistent with patchy shells and three with density fluctuations.These classifications use the kinematically allowed regions defined for afterglow variability.
- Engine models: A fragmented accretion disk can produce the observed peak-luminosity–flare-epoch anticorrelation because more distant blobs accrete later and at lower rates.A magnetic barrier modulating continuous accretion is described as another way to retain this behavior.
Number of cases
The supplied figure captions identify distributions and relationships among flare timing, morphology, energetics, and peak intensity, but do not provide numerical case counts.
- Peak ratios: The successive-event peak-ratio distribution compares X-ray flares, gamma-ray pulses, and their combined sample.The caption specifies red, green, and blue for the three classes, respectively.
- Peak intensity over time: The Gaussian flare peaks are plotted as a function of time with a best fit and 95% confidence limits.The solid line is the fit and dashed lines mark the confidence limits.
- Flare duration and energetics: The EW/tpeak versus ∆t/tpeak plot includes the plane bisector as a reference line.The caption identifies the plotted variables but gives no outcome or case count.
- Afterglow-variability tests: The ∆F/F–∆t/t scatter plot places flare measurements within kinematically allowed regions for afterglow variability.It uses Gaussian FWHM for ∆t, Gaussian peak time for t, and peak-to-continuum flux ratios for ∆F/F.