PaperlayerView exact revision on arXiv

Source-linked AI summary

Methods and results of an automatic analysis of a complete sample of Swift-XRT observations of GRBs

P. A. Evans, A. P. Beardmore, K. L. Page, J. P. Osborne, P. T. O'Brien, R. Willingale, R. L. C. Starling, D. N. Burrows, O. Godet, L. Vetere, J. Racusin, M. R. Goad, K. Wiersema, L. Angelini, M. Capalbi, G. Chincarini, N. Gehrels, J. A. Kennea, R. Margutti, D. C. Morris, C. J. Mountford, C. Pagani, M. Perri, P. Romano, N. Tanvir

arXiv:0812.3662v3astro-ph

TL;DR

Rapid, uniform analysis is needed because GRBs fade quickly and follow-up observers must decide promptly whether to invest observing time. The paper develops automated Swift-XRT products and applies them homogeneously to GRBs, finding diverse light-curve morphologies consistent with a single underlying behaviour but posing difficulties for the forward-shock interpretation.

  • Problem

    GRBs fade rapidly, making rapid, reliable, uniform data analysis desirable for follow-up decisions about observing time.

  • Method

    The Swift-XRT team developed software that automatically produces positions, light curves, hardness ratios, and spectra, then used it for a homogeneous analysis of observed GRBs.

  • Results

    The analysis found diverse light-curve morphologies, with the canonical curve accounting for less than half of Swift light curves, and showed that a two-component model can explain this range.

  • Takeaways & Limitations

    A single underlying behaviour involving prompt and afterglow components, with varying emission ratios and energy injection rates, can explain the observed morphology range.

  • Takeaways & Limitations

    Reconciling the data with the forward-shock model requires explaining spectrally invariant temporal breaks and significant energy injection lasting days to weeks after the explosion.

Abstract

from arXiv · show

We present a homogeneous X-ray analysis of all 318 Gamma Ray Bursts detected by the X-ray Telescope on the Swift satellite up to 2008 July 23; this represents the largest sample of X-ray GRB data published to date. In Sections 2--3 we detail the methods which the Swift-XRT team has developed to produce the enhanced positions, light curves, hardness ratios and spectra presented in this paper. Software using these methods continues to create such products for all new GRBs observed by the Swift-XRT. We also detail web-based tools allowing users to create these products for any object observed by the XRT, not just GRBs. In Sections 4--6 we present the results of our analysis of GRBs, including probability distribution functions of the temporal and spectral properties of the sample. We demonstrate evidence for a consistent underlying behaviour which can produce a range of light curve morphologies, and attempt to interpret this behaviour in the framework of external forward shock emission. We find several difficulties, in particular that reconciliation of our data with the forward shock model requires energy injection to continue for days to weeks.

1 INTRODUCTION

The paper introduces a homogeneous, automated analysis of Swift-XRT GRB data designed to deliver rapid, uniform products for follow-up and model testing. It applies these tools to a catalogue of 318 XRT-detected GRBs and examines the results in the context of GRB afterglow models.

  • Swift-XRT observations: Swift automatically slews to new bursts, allowing XRT and UVOT observations typically within ∼60–90 seconds of the trigger.The BAT detects the burst before Swift repoints the narrower-field instruments.
  • Motivation: Rapid, reliable, and uniform analysis is important because GRBs fade quickly and follow-up observers must decide promptly whether to invest observing time.Swift data are processed and made publicly available minutes to hours after downlink.
  • Automated analysis: The software processes TDRSS and Malindi data to improve positions, light curves, and spectra, with TDRSS products available earlier but intended mainly for rapid use.TDRSS processing provides better positions and more reliable light curves and spectra than onboard products, while SPER spectra remain quick-look products because they are not fully calibrated.
  • Scope: The study presents tools and a catalogue covering all 318 GRBs detected by Swift-XRT up to GRB 080723B.The paper also discusses implications for afterglow science using the fireball model.
  • GRB afterglow models: The paper interprets the X-ray data within an external-shock afterglow framework, while noting that the early X-ray light curve can include a shallow-decay plateau with competing interpretations.Proposed interpretations include energy injection, reverse-shock emission, and dust scattering.

2 AUTOMATED DATA PRODUCTS

The authors developed three automated XRT data products—enhanced positions, light curves, and spectra—and provide automatic processing for GRBs alongside user-customisable tools.

  • Data products: The software produces enhanced positions, light curves, and spectra from XRT observations.These products form the paper’s core automated data-analysis outputs.
  • Method updates: The position-enhancement algorithm improves precision by approximately a factor of two, while the light-curve code receives minor modifications and new user functionality.The position method builds on the previously documented G07 algorithm.
  • Processing workflow: BAT-triggered GRBs are processed automatically, whereas Swift Target-of-Opportunity bursts must be manually registered for automatic analysis.Registration is usually performed when the Target of Opportunity is uploaded.
  • Processing workflow: Before product generation, XRT data are reprocessed at the UK Swift Science Data Centre using the latest xrtpipeline release.This processing version can differ from the one used for quick-look and archive event lists.

2.1 Spectra

The spectral pipeline constructs and fits XRT spectra from segmented observations, combines interval-specific responses using count weighting, and validates the automated products against manual processing. It also provides web-accessible spectra for GRBs and arbitrary XRT targets, with time-sliced extraction available on demand.

  • Spectrum construction: Spectra are built from source intervals divided by observing snapshot, instrument mode, and pile-up state, using source regions selected according to the data.Pile-up is identified from count-rate thresholds and, in PC mode, comparison with the calibrated PSF.
  • Response combination: The pipeline combines interval spectra and ARFs, weighting each ARF by the fraction of total source counts contributed by its interval.Count weighting accounts for changing source brightness and detector effective area across snapshots.
  • Validation: Automatic spectra were compared with manually produced spectra for more than thirty cases, and parameter differences were much smaller than the corresponding uncertainties.The authors conclude that automatic spectrum generation is reliable.
  • Spectral fitting: The software automatically fits spectra with an absorbed power-law model using X-statistic fitting in xspec, including Galactic and variable absorption components.Reported redshifts can replace the second absorption component with a redshifted absorption model.
  • User access: Time-averaged spectra are generated automatically, while users can request spectra over arbitrary time intervals through the web interface.The tools can also construct spectra for any Swift-XRT target, not only GRBs.
  • SPER spectra: SPER spectra are available rapidly but should be treated as quick-look products because SPER data are not fully calibrated.Additional processing is required before scientific spectral extraction.

2.2 ‘Enhanced’ positions

The enhanced-position pipeline combines improved PSF fitting, UVOT-based aspect correction, and broader early-time data use to produce more precise XRT localisations. These positions are available rapidly, support non-GRB targets, and achieve agreement with UVOT positions at calibrated error levels.

  • The revised enhancement process reduces XRT position error radii by 30–50%, with SPER positions available 10–20 minutes after a trigger.Enhanced positions are typically available within hours, while the SPER implementation provides prompt localisations.
  • Multiple observations beginning within 12 hours can be combined, while limiting the software to v, b, and white UVOT filters gives the best results.The multi-observation option is especially useful when automatic observations are interrupted.
  • The pipeline uses UVOT-corrected aspect information because UVOT photons are shifted for spacecraft motion, whereas XRT detector images are not.This makes the UVOT-derived aspect solution effective at the start of the UVOT exposure and improves sky-coordinate PSF fitting.
  • Each XRT image is fitted with six PSF profiles, and the profile with the lowest C-statistic determines the GRB position.Additional piled-up-source profiles are modelled with a King−Gaussian function.
  • The enhanced positions have error radii ≤1.5′′ for 50% of bursts and ≤2.0′′ for 90%, requiring a 1.36′′ systematic uncertainty for 90% UVOT agreement.The authors describe these as the best available XRT positions for most GRBs.
  • Enhanced-position software is also available through a web tool for user-selected non-GRB XRT objects.The online catalogue contains enhanced positions for Swift-observed GRBs.

2.3 Light curves and hardness ratios

The revised software automatically produces XRT light curves and hardness ratios with improved position handling, binning, flux conversion, event-list access, and early SPER processing. User-facing rebinning tools expose alternative binning choices while retaining automation as the default.

  • Using PSF-fit positions instead of xrtcentroid avoids inaccurate source positions when CCD bad columns intersect the PSF.This change makes pile-up and bad-column corrections more reliable and caused small changes to a few light curves.
  • The final low-count light-curve bin is tested with the Kraft, Burrows & Nousek Bayesian method rather than automatically plotted as an upper limit.The Bayesian test determines whether the source is detected at the 3-σ level.
  • The software uses a logarithmic-mean event time and reallocates incomplete snapshot events to maximize fractional exposure, reducing low-fractional-exposure bins.The logarithmic-mean time better reflects the distribution of counts within a bin.
  • Automatic products include flux-unit light curves, downloadable WT and PC source/background event lists, and improved SPER light curves available within minutes.SPER light curves are background-subtracted and use the same binning method as Malindi data, but lack GTI information.
  • Hardness-ratio bins now require C counts per band instead of 2C, giving significantly better time resolution.Here C is the number of counts needed to complete a main-light-curve bin.
  • The web interface lets users change counts-per-bin criteria, choose fixed or dynamic binning, select grades and energy bands, and adjust hardness-ratio bands.The default automated criteria provide a useful, valid representation but not necessarily the best one.

2.4 Automatic light curve fitting

The fitting pipeline identifies and removes deviations such as flares before modelling the remaining light curve with broken power-law segments. It automates deviation detection and model selection while retaining manual review and accounts for finite bins and fractional exposure.

  • The fitting procedure identifies deviations, excludes flare intervals, fits power-law series, selects fits with an F-test, and allows manual repetition.Only the first three steps are automated; a human checks the results.
  • Identifying ‘deviations’: A possible deviation begins after at least two consecutive rising bins produce a count-rate increase of at least 2σ.The algorithm then locates the peak and evaluates subsequent decay indices.
  • Identifying ‘deviations’: A deviation ends after a qualifying observation gap or when two of three bins satisfy the specified decay-index conditions.Later flares may span snapshots because their duration-to-midpoint ratio is typically about 0.1.
  • Identifying ‘deviations’: A candidate deviation is confirmed only with fewer than 10 mode changes, at least 2% of component bins before the peak, and significance 1.8.For deviations within 2 ks of the trigger, the required significance increases to 3.
  • Fitting power laws: After deviations are removed, the remaining light curve is fitted with unsmoothed power-law segments separated by zero to five breaks.Upper limits are excluded, and the fitting must account for finite bin durations and fractional exposures.
  • Fitting power laws: The compromise fitting method groups GTIs into long GTIs, integrates the model over each, applies fractional-exposure corrections, and sums the results.An LGTI is a cluster of GTIs separated by less than 30 s of dead time.

LGTIs

The analysis groups closely spaced Good Time Intervals into LGTIs to make light-curve fitting practical, while automatically fitting and validating broken power-law models with targeted human intervention.

  • LGTI method: Mode switching can create many short GTIs, making direct integration over every interval impractical for large samples.Applying the fully detailed integration method to all light curves could take several months.
  • LGTI method: LGTIs group consecutive GTIs separated by less than 30 s of dead-time within each light-curve bin.The model is integrated across each LGTI, corrected by fractional exposure, and summed across the bin.
  • Automatic fitting: The automatic script starts with an unbroken power law and adds up to five breaks by locating intervals with systematic deviations from the previous fit.Fewer breaks are used when there is insufficient freedom, and candidate break locations are chosen to reduce local-minimum problems.
  • Automatic fitting: A break is selected as significant when the F-test assigns the χ2 improvement a probability below 0.3%, but the test is used operationally rather than as a 3σ confirmation.The best fit is the model with the greatest number of breaks deemed significant by this procedure.
  • Human intervention: 23% of cases required manual flare correction, while about 5% required manual adjustment because automated fitting did not identify the true best fit.Most flare-identification failures were false positives; roughly 1% of light curves contained breaks supported only at the 90–99% level.

3 NON-GRBS

The paper adapts Swift-XRT analysis tools for non-GRB sources through a public web interface, adding source-appropriate choices for spectra, positions, and light-curve binning while documenting important caveats.

  • Non-GRB tools: The three GRB analysis tools were adapted for non-GRB sources and made available through an on-demand public web interface.The tools produce enhanced positions, light curves, and spectra for sources beyond GRBs.
  • Spectra: Non-GRB spectral analysis lets users select observations and up to four time intervals, then automatically fits absorbed power-law models.Spectral files are also downloadable for fitting with other models.
  • Positions: Non-GRB position enhancement uses a two-pass strategy because filter-of-the-day observations often use UV filters associated with larger position errors.For GRBs, the tool instead uses only the v, b, and white filters.
  • Light curves: For non-GRB light curves, users can specify fixed bin durations, use one bin per snapshot or observation, or retain GRB-style count binning.The default GRB observation-selection limit of 12 hours is retained for speed but can be changed for non-GRB sources.
  • Caveats: Bins with fewer than 15 counts may have inaccurate Gaussian uncertainties after background subtraction, so users must choose adequate bin sizes.The software warns when any bin falls below this threshold.
  • Caveats: WT-mode bins with fewer than 15 counts are excluded by default because they may be spurious, while low-exposure final snapshot bins can exaggerate statistical fluctuations.Users can change the WT threshold and are advised to inspect or reject low-fractional-exposure points.
  • Light curves: Fixed-width binning additionally produces an OGIP-compliant FITS light-curve file.

4 RESULTS

The paper applies automated XRT processing to a complete GRB sample, validates rapid SPER products against Malindi data, and characterizes temporal and spectral population properties. The results reveal structured light-curve behavior, broadly useful online products, and important limitations in interpreting breaks and spectra.

  • Validation of SPER results: 90% of enhanced SPER positions agree with UVOT positions, confirming that the enhanced SPER 90% confidence error radius is correctly calibrated.SPER and Malindi light curves also showed good agreement when compared directly.
  • Validation of SPER results: SPER spectral column densities and spectral indices agree well with Malindi fits, but fluxes agree within their 90% errors only 70% of the time.The discrepancy may reflect missing SPER GTI information or underestimated flux uncertainties from covariance matrices.
  • Malindi data: 665 temporal-index values form a fairly tight α distribution with steep probability drops near α = 0.5 and 1.5.The authors associate these features with distinct phases in canonical and other GRB light-curve morphologies.
  • Malindi data: Break-time PDFs peak near ∼1–300 s and 10^4 s, while substantial break probability extends across approximately 10^2–10^5 s.The peaks correspond to common plateau start and end times, but late breaks and breaks near the 2–4 ks observing gap are harder to constrain.
  • Time-resolved analysis: The (α, β) distributions show that many normal-decay segments fall outside standard afterglow closure-relation predictions, whereas post-jet-break points agree better.Steep-decay and plateau segments are not expected to follow the standard relations because they are associated with prompt-tail emission and energy injection, respectively.

5 A CANONICAL LIGHT CURVE?

The canonical four-phase X-ray light curve is the most common morphology but appears in fewer than half of well-covered bursts. Comparisons of temporal, spectral, and break-time distributions show that several alternative morphologies are difficult to reconcile with a single standard afterglow interpretation without prolonged energy injection.

  • The canonical light curve comprises steep decay, plateau, normal decay, and sometimes post-jet-break phases.
  • 162 bursts had sufficient coverage to test for the canonical shape; 68 (42%) were canonical, while 38 (24%) were oddballs.The same subsample included 49 (30%) one-break and 7 (4%) no-break light curves.
  • The canonical morphology, although most common, occurs in less than half the bursts in which it could be identified.
  • 5.1 Light curves with no breaks: No-break bursts have temporal-index distributions broadly consistent only with the canonical normal-decay phase, although the small sample makes the K-S result uncertain.The paper interprets these bursts as possibly showing only the afterglow power-law phase, without dominant energy injection.
  • 5.2 Light curves with one break: type b: Type b light curves match canonical steep-decay behavior more plausibly than canonical plateaus, but their shallow-decay indices and break times differ significantly from canonical distributions.The K-S tests give <0.1% for the shallow-decay comparison and 0.3% for plateau-start versus type-b break times.
  • 5.3 Light curves with one break: type c: Type c light curves can match canonical phase pairs in temporal-spectral indices, but proposed identifications create timing or fluence difficulties.Identifying them with normal and post-jet-break phases would require 7/25 (28%) jet breaks within 1000 s, while the plateau-normal interpretation lacks the expected fluence trend.

6 UNDERSTANDING THE X-RAY AFTERGLOW

The paper tests a four-phase interpretation of GRB X-ray afterglows against closure relationships, finding that the phases impose specific constraints on energy injection, breaks, and jet-break timing. The forward-shock model remains consistent after reorganization, but requires unresolved mechanisms and prolonged energy injection.

  • Physical interpretation of the phases: The proposed sequence assigns steep decay to high-latitude prompt emission, plateau to an externally shocked jet with energy injection, and normal decay to the same shock without injection.The post-jet-break phase is likewise modeled as an externally shocked jet without energy injection, with the jet edge visible to the observer.
  • Constraints from closure relationships: Energy must be injected into the shock during the plateau phase.The plateau closure relationships are consistent with an energy-injection scenario.
  • Constraints from closure relationships: The so-called post-jet-break phase is better explained as pre-jet-break emission after energy injection ceases.This interpretation fits the closure-relationship comparison better than the standard post-jet-break interpretation.
  • Constraints from closure relationships: A spectrally invariant mechanism must steepen the light curve independently of energy injection.The plateau-to-normal break cannot always be attributed to the cessation of energy injection because many normal-phase points remain above the permitted closure band.
  • Constraints from closure relationships: The ordering of the unknown-origin break and energy-injection cessation determines whether normal-phase bursts lie within or above the closure band.Some bursts require no energy injection during normal decay, whereas many others require injection during that phase.
  • Remaining difficulties: Significant energy injection may need to persist for days to weeks after the trigger, while the forward-shock interpretation remains unproven.The authors identify both prolonged energy injection and a spectrally invariant temporal break as unresolved difficulties.

7 CONCLUSIONS

The paper delivers automated, homogeneous Swift-XRT products and analyzes 318 GRBs, finding diverse light-curve morphologies that can arise from a shared two-component behavior but challenge the forward-shock interpretation.

  • Data products: Automated software produces enhanced positions, light curves, hardness ratios, and spectra for Swift-XRT GRBs, with general-purpose web tools for other XRT targets.Preliminary products can appear within minutes, while full versions are available within a few hours.
  • Light-curve morphologies: The 162-burst coverage-controlled subsample contains 8 (5 %) no-break, 49 (30 %) one-break, 67 (41 %) canonical, and 38 (24 %) oddball light curves.
  • Interpretation: The range of observed morphologies can be explained by a single underlying two-component behavior combining prompt and afterglow emission with varying ratios and energy-injection rates.
  • Interpretation: Under the external forward-shock model, many cases require energy injection to continue for days to weeks after the plateau and a break mechanism that leaves the spectrum unchanged.
  • Usage policy: Users are advised to sanity-check tool results, particularly light-curve binning, despite the authors’ efforts to verify the software.
Loading 0812.3662v3…