Source-linked AI summary
The First Survey of X-ray Flares from Gamma Ray Bursts Observed by Swift: Spectral Properties and Energetics
A. D. Falcone, D. Morris, J. Racusin, G. Chincarini, A. Moretti, P. Romano, D. N. Burrows, C. Pagani, M. Stroh, D. Grupe, S. Campana, S. Covino, G. Tagliaferri, R. Willingale, N. Gehrels
TL;DR
Swift observations raised the question of whether late X-ray flares arise from the afterglow or renewed activity of the GRB engine. The paper analyzes a statistical sample using temporal and spectral comparisons, finding evidence that most flares reflect late-time internal-engine activity, although some relationships remain poorly constrained.
Problem
The paper asks whether late X-ray flares are afterglow-related or instead reflect renewed activity of the GRB internal engine.
Method
The authors analyze all significant flares in the first 110 Swift GRBs, fit temporal models, and compare four flare spectral models after accounting for the underlying afterglow.
Results
The sample contains 33 GRBs and 77 flares; some favor Band-function spectra, and differing flare and afterglow photon-index distributions support an internal-engine origin.
Takeaways & Limitations
Most X-ray flares are supported as late-time activity of the internal engine rather than an afterglow-related effect.
Takeaways & Limitations
The Epeak–Eiso relationship cannot be firmly determined because only 18 flares have suitable redshifts and constrained spectra, with large error bars.
Abstract
from arXiv · showhide
Observations of gamma ray bursts (GRBs) with Swift produced the initially surprising result that many bursts have large X-ray flares superimposed on the underlying afterglow. The flares were sometimes intense, had rapid rise and decay phases, and occurred late relative to the ``prompt'' phase. Some remarkable flares are observed with fluence comparable to the prompt GRB fluence. Many GRBs have several flares, which are sometimes overlapping. Short, intense, repetitive, and late flaring can be most easily understood within the context of the standard fireball model with the internal engine that powers the prompt GRB emission in an active state at late times. However, other models for flares have been proposed. Flare origin can be investigated by comparing the flare spectra to that of the afterglow and the initial prompt emission. In this work, we have analyzed all significant X-ray flares from the first 110 GRBs observed by Swift. From this sample 33 GRBs were found to have significant X-ray flares, with 77 flares that were detected above the 3$σ$ level. In addition to temporal analysis presented in a companion paper, a variety of spectral models have been fit to each flare. In some cases, we find that the spectral fits favor a Band function model, which is more akin to the prompt emission than to that of an afterglow. We find that the average fluence of the flares is 2.4e-7 erg/cm^2/s in the 0.2-10 keV energy band, which is approximately a factor of ten below the average prompt GRB fluence. These results, when combined with those presented in the companion paper on temporal properties of flares, supports the hypothesis that most X-ray flares are late-time activity of the internal engine that spawned the initial GRB; not an afterglow related effect.
1. Introduction
Swift revealed that many GRBs contain late, sometimes intense and rapidly varying X-ray flares superimposed on the afterglow. This paper addresses their origin by statistically comparing flare properties with prompt and afterglow emission.
- The standard fireball model attributes prompt emission to internal shocks and the afterglow to shocks with the ambient material.
- Continued internal-engine activity was proposed to explain energy release after the classical prompt-emission timeframe, potentially lasting much longer.
- Before Swift, observations provided indications of late X-ray emission, but individual detections and ensemble analyses left multiple explanations possible.
- Swift observations found X-ray flares that were intense, rapidly rising and decaying, late relative to the prompt phase, and sometimes overlapping.Some individual flares had fluences comparable to the prompt GRB emission.
- Comparing general flare properties with prompt and afterglow emission can test their emission mechanism and whether flares form multiple classes.
- This work presents the first statistical temporal and spectral study of Swift X-ray flares, using all bursts through 2006 January 24 with at least one significant flare.
2. The Sample
The sample was selected from Swift-XRT light curves by identifying deviations from the underlying decay and retaining flares with signal-to-noise above 3. The resulting dataset contains 33 GRBs and 77 flare intervals.
- The initial sample comprised Swift-XRT light curves through 2006 January 24 that showed hints of deviations from typical power-law decay.
- A broken power-law underlying decay was fit together with power-law rise and decay components for candidate flares.
- S/N > 3 was required for retention, yielding 33 GRBs with at least one significant flare and 77 flare time intervals.
- Some of the 77 intervals overlap, so the beginning and ending of individual flares are not always unambiguous.
- The flare sample is listed in Table 1, while temporal and spectral analyses can involve different subsets of flares.
3. Analysis
The analysis separates flare emission from the underlying afterglow, defines flare intervals from light-curve fits, and compares four spectral models after estimating the afterglow contribution.
- 3.1. Light Curve Analysis: Flare intervals were defined where power-law rise and decay fits intersected the underlying decay curve, allowing spectral extraction despite incomplete flare coverage.
- 3.1. Light Curve Analysis: The underlying decay was modeled with multiply broken power laws representing the phases of the GRB and afterglow light curve.
- 3.1. Light Curve Analysis: Spectral extraction used flare and underlying-light-curve time regions, sometimes multiple regions to improve statistics or account for incomplete and overlapping coverage.
- 3.2. Spectral Analysis: Spectra were fit from 0.3 to 10.0 keV using Xspec, with 3% systematic error and bins containing at least 20 photons.
- 3.2. Spectral Analysis: The underlying afterglow was fit with an absorbed power law, with parameters constrained from data before and/or after the flare and normalized using temporal extrapolation.
- 3.2. Spectral Analysis: The frozen afterglow model was added to four flare models: a simple power law, exponentially cutoff power law, power law plus blackbody, and Band function.
4. Spectral Results
The spectral analysis finds that simple power laws fit many Gold flares, while Band functions provide superior fits for a significant subset. The mean unabsorbed flare fluence is 2.4 × 10^-7 erg cm^-2 in 0.2–10 keV.
- Gold flares were selected by requiring more than 15 degrees of freedom in the power-law spectral fit.
- The underlying afterglow has a mean photon index of 1.9 with a standard deviation of 0.3, consistent with typical GRB afterglows.
- Simple power laws provide reasonable fits for many flares, while Band functions provide superior fits for some cases.The analysis also applied exponentially cutoff power-law and blackbody-plus-power-law models.
- 9 of 47 Gold flares have Δχ2 > 9.0 between power-law and Band-function fits, compared with an expected 0.23 by chance from simulated power-law spectra.The observed distribution is skewed toward positive Δχ2 values, although power laws remain reasonable for many flares.
- Extending the typical flare spectrum from 0.2–10 keV to 0.2–150 keV increases the fluence by only 1.4%.The authors therefore use the observed XRT energy band for fluence calculations.
- 2.4 × 10^-7 erg cm^-2 is the mean unabsorbed 0.2–10.0 keV fluence from Band-function fits, with no evidence for a bimodal distribution.Reported fluences include only the flare component and exclude the modeled underlying afterglow contribution.
5. Flare Fluence versus Prompt Fluence
Flare fluence is typically about ten times lower than prompt GRB fluence, although the distributions overlap and at least one flare matches its burst’s prompt fluence.
- 10× lower: the mean flare fluence in 0.2–10 keV is approximately a factor of ten below the mean prompt fluence in 15–150 keV.The mean prompt fluence is 2.4 × 10−6 erg cm−2.
- At least one flare, from GRB 050502b, matches the fluence of its prompt GRB emission.The flare and prompt fluences are measured in their respective Swift energy bands.
- The flare and prompt fluence distributions overlap despite the lower average flare fluence.
6. Flare Properties versus Underlying Afterglow Properties
Flare spectra differ from the underlying afterglows, with flare power-law indices spanning a wider distribution than afterglow indices.
- Flare power-law photon indices have a wider distribution than those of the underlying afterglows.
- The comparison uses power-law fits to both flare spectra and underlying afterglow data.
7. Temporal Evolution of Flare Properties
The study tests whether flare spectral peak energy and total energy release vary with rest-frame flare time across Gold flares. No clear ensemble relationship is found for total energy release, while burst-to-burst scaling may obscure individual-burst trends.
- Spectral evolution: All Gold flares with known redshift are plotted against rest-frame time relative to prompt T0 without scaling by prompt Epeak.
- Energy evolution: No clear relationship appears between total flare energy release and rest-frame flare time across the combined sample.
- Energy evolution: Individual-GRB temporal relationships may require a scaling factor dependent on prompt GRB parameters.
8. Epeak versus Eiso
The paper estimates flare Eiso from Band-function fits and examines its relationship with redshift-corrected Epeak. The flares’ peak energies are lower than typical prompt values, but any Epeak–Eiso relationship remains uncertain.
- Eiso is calculated in the 0.2 keV–10 MeV band from Band-function spectra fitted in the observed 0.2–10 keV band.The calculation uses unabsorbed observed fluence, redshift, a k-correction, and luminosity distance.
- The analysis plots Eiso against redshift-corrected Epeak for Gold flares with measured redshifts.
- Flare peak energies are significantly lower than the hundreds of keV typical of initial GRB prompt emission.
- A strong Epeak–Eiso relationship is uncertain because of large error bars and the limited sample.
- The possible relationship will be investigated with more flares and broader spectral coverage.
9. Redshift Distribution
Among 14 flaring GRBs with measured redshifts, the mean redshift is z=2.6, consistent with the overall Swift GRB sample. The redshift distribution is therefore not significantly different from that of all Swift GRBs.
- z=2.6 is the mean redshift for the 14 flaring GRBs with measured redshifts.
- The flaring-GRB mean is consistent with the overall Swift mean redshift of 2.5–2.8.
- Figure 14 shows the redshift distribution of GRBs with flares.
10. Discussion & Conclusions
The paper systematically analyzes flare spectra while correcting for underlying afterglow contamination. Its results show distinct flare and afterglow spectral behavior, substantial late-time energetics, and support for late internal-engine activity, though the Epeak–Eiso relation remains uncertain.
- Sample and analysis: 77 X-ray flares from 33 of the first 110 Swift GRBs were analyzed systematically, including correction for the underlying afterglow.
- Spectral analysis: When underlying-lightcurve fluence is a significant fraction of flare fluence, underlying photons significantly affect the flare spectrum.
- Spectral analysis: Some flares favor curved Band-function spectra, although curvature may also result from spectral evolution averaged over the flare.
- Interpretation: Different flare and underlying-afterglow photon-index distributions indicate distinct flare and afterglow photon populations or mechanisms.
- Interpretation: The 14-GRB redshift sample has an average matching all Swift GRBs, arguing against late flares being merely redshifted multi-peaked prompt emission.
- Energetics: Flare fluences are sometimes comparable to prompt emission and typically about 10× lower, requiring late X-ray emission within a substantial energy budget.
- Limitations: Only 18 flares constrain the Epeak–Eiso relation, so limited sample size and large error bars prevent a firm conclusion.