Gamma-Ray Bursts (GRBs) are the most luminous transient phenomena in the observable Universe: brief flashes of gamma-ray radiation, released at cosmological distances (up to z ~ 9.4) with isotropic-equivalent energies from ~10^48 to a few 10^54 erg. Their bimodal duration distribution motivates the classification into short GRBs (< 2 s), associated with the coalescence of compact objects, and long GRBs (> 2 s), produced by the core collapse of massive stars. GRBs display a highly variable prompt emission (keV-MeV), followed by a long-lasting broadband afterglow. In the standard fireball model both phases are powered by an ultra-relativistic collimated outflow, with bulk Lorentz factor Γ ~ 100-1000: the prompt emission arises from internal dissipation within the jet, the afterglow from the shock that the outflow drives into the circumburst medium.Despite five decades of observations, the physical origin of the prompt emission remains an open question: how the jet energy is dissipated, and which radiative process converts it into the observed photons. A powerful observable in this context is the spectral lag, the time delay τ between the light curves of the same burst in two different energy bands, whose measurement requires far fewer photons than a full spectral characterization and does not rely on any assumed spectral model. The lag is defined as positive when the hard photons arrive before the soft ones, and negative in the opposite case. Long GRBs systematically show positive lags in the keV-MeV range, while at GeV energies negative lags have been reported in a few bright bursts. Short GRBs, on the other hand, are consistently found to have lags compatible with zero. The interpretation of these delays is not yet settled: they may reflect the way the spectrum of a single component evolves in time, or the coexistence of distinct emission processes contributing to the observed radiation, such as the prompt emission and the afterglow. The spectral lag can therefore tell us what happens at the spectral level during the emission of the burst, and which components are at play.In order to understand the link between time lags and spectral evolution in GRBs, during my PhD I computed spectral lags over the widest energy baseline considered so far, from the soft X-rays to 100 MeV, and combined them with time-resolved spectral analysis.I performed a systematic study of GRB spectral lags, extending for the first time the analysis into the 30-100 MeV domain by exploiting Fermi LAT Low-Energy (LLE) data jointly with Fermi/GBM data for a sample of 70 GRBs. Cross-correlating the light curves in four energy bands, I found that the lags measured within the GBM range (10 keV - 1 MeV) are predominantly positive, with no significantly negative case. The lag between the lowest GBM band and the LLE band, instead, splits the sample almost evenly between significantly positive (40%) and negative (37%) values. A positive lag would be consistent with a single component, softening in time, that dominates across the whole energy range, whereas a negative lag might suggest that a distinct and harder component emerges in the LLE band while the prompt emission is fading.To clarify the spectral origin of this behavior, I performed a time-integrated spectral analysis of the LLE data, comparing the spectral slope measured in the LLE band with the extrapolation of the keV-MeV spectrum. Bursts with a positive high-energy lag tend to show a softer LLE spectrum, consistent with a single component softening in time, whereas bursts with a negative lag tend to show a harder one. These results led me to propose a phenomenological scenario in which the sign of the high-energy lag reflects the origin of the observed emission: two components can dominate the 30-100 MeV band, the high-energy tail of the prompt emission and the onset of the afterglow, and a negative lag marks the delayed emergence of the latter.To test this scenario, I performed a time-resolved spectral analysis and, in parallel, a time-resolved measurement of the lags, following the evolution of the spectral parameters and of the time delays across the same time intervals. I applied this analysis to GRB 160625B and GRB 190114C, two of the brightest bursts of the sample, having a positive and a negative lag respectively, and whose photon statistics allow the spectrum to be constrained interval by interval.In GRB 160625B a single spectral component softens continuously in time and all lags remain positive; interpreting its high-energy cutoff as due to pair-production opacity, I derived Γ ~ 120-260 and an emission radius R ~ 10^14 cm, well above the photosphere. In GRB 190114C an additional power-law component emerges at ~2.5 s after trigger and dominates in the LLE band, thus producing a negative lag throughout the burst. The flux of this power-law component is broadly consistent with the early afterglow, although its steep rise (faster than t^3) leaves an internal origin equally viable. The measurement of the lag thus provides an indication of which component dominates the LLE band, at a much lower cost than the time-resolved spectral analysis that would otherwise be required.I explored with numerical simulations a third scenario for the origin of negative lags. When the high-energy emission is initially suppressed by gamma-gamma pair-production opacity and is progressively released as the outflow expands and becomes transparent, the high-energy photons reach the observer later than the soft ones, even when the underlying spectrum does not evolve at all. I found that this process can operate in sufficiently compact sources, having a low bulk Lorentz factor Γ combined with a high luminosity L. However, since Γ is observed to correlate with L, the two conditions lie at opposite ends of the population, and the mechanism is confined to the outliers of this correlation, whose low Γ makes them intrinsically faint at high energies: their absence in LLE-selected samples is thus a selection effect rather than evidence against the mechanism.The proposed scenario also makes a prediction at the opposite end of the energy range: where no additional hard component is expected to contribute, the lag should always be positive, since it would be driven solely by the spectral evolution of the prompt emission. To test this prediction, I computed for the first time the spectral lags of the prompt emission between the XRT (0.3-10 keV) and BAT (15-150 keV) detectors onboard Swift, for 33 GRBs with simultaneous detections of the prompt emission. For the 19 bursts with a measured redshift I adopted fixed rest-frame energy bands. I found that 26 out of 33 lags are significantly positive and that none is significantly negative, in agreement with the expectation for an energy range where a single prompt component dominates. A time-resolved XRT+BAT spectral analysis of four bursts spanning the extremes of the lag distribution (~0.6 to ~15.5 s) revealed what the lag magnitude reflects. The lag is null or very small when the peak energy of the nuFnu spectrum, Epeak, stays constant, or varies too little to cross the observing bands; it becomes sizeable when Epeak migrates from one band to the other, as in GRB 100906A. The several-second lags measured here imply that the spectral evolution proceeds considerably more slowly than in the GBM range.This thesis thus builds a coherent and predictive framework, tested from 0.3 keV to 100 MeV, in which the lag sign is a proxy of the nature of the emission components at play and the lag magnitude measures the amplitude of their spectral evolution.Identifying the number of components also matters for the applications of time lags for fundamental physics. Tests of Lorentz Invariance Violation exploit the fact that, if the speed of light depended on photon energy, photons of different energies emitted simultaneously at cosmological distances would accumulate a measurable delay along their path. Any observed lag is therefore the sum of two terms: an extrinsic one, due to propagation, and an intrinsic one, imprinted at the source by the physics of the emission. Constraining the former requires the latter to be understood, and these tests implicitly assume that the photons of the two bands are produced by the same component and are thus co-spatial. The sign of the high-energy lag provides a way to test this assumption, since a negative lag signals that two distinct components, produced in different regions, dominate the two bands.Spectral lags have long been regarded as a by-product of the spectral evolution. This thesis shows that they are a diagnostic in their own right, one that reads the physics of the emission from the arrival times of the photons alone. With the larger and better characterized samples that the coming missions will provide, what is here a framework tested on a handful of bursts can become a standard tool for probing the physics of GRBs.
I Gamma-Ray Burst (GRB) sono i fenomeni transienti più luminosi dell'Universo osservabile: brevi lampi di radiazione gamma, emessi a distanze cosmologiche (fino a z ~ 9.4) con energie isotrope equivalenti da ~10^48 a qualche 10^54 erg. La distribuzione bimodale delle loro durate motiva la classificazione in GRB brevi (< 2 s), associati alla coalescenza di oggetti compatti, e GRB lunghi (> 2 s), prodotti dal collasso del nucleo di stelle massicce. I GRB presentano un'emissione prompt altamente variabile (keV-MeV), seguita da un afterglow a banda larga di lunga durata. Nel modello standard a fireball entrambe le fasi sono alimentate da un flusso collimato ultrarelativistico, con fattore di Lorentz Γ ~ 100-1000: l'emissione prompt nasce dalla dissipazione interna al getto, l'afterglow dallo shock che il flusso guida nel mezzo circumburst.Nonostante cinque decenni di osservazioni, l'origine fisica dell'emissione prompt resta una questione aperta: come l'energia del getto venga dissipata, e quale processo radiativo la converta nei fotoni osservati. Un'osservabile potente in questo contesto è il lag spettrale, il ritardo temporale τ fra le curve di luce dello stesso burst in due diverse bande energetiche, la cui misura richiede un numero di fotoni molto inferiore a quello necessario per una caratterizzazione spettrale completa e non dipende da alcun modello spettrale assunto. Il lag è definito positivo quando i fotoni duri arrivano prima di quelli soffici, e negativo nel caso opposto. I GRB lunghi mostrano sistematicamente lag positivi nell'intervallo keV-MeV, mentre alle energie del GeV sono stati riportati lag negativi in alcuni burst brillanti. I GRB brevi, invece, presentano in modo consistente lag compatibili con zero. L'interpretazione di questi ritardi non è ancora stabilita: essi possono riflettere il modo in cui lo spettro di una singola componente evolve nel tempo, oppure la coesistenza di processi di emissione distinti che contribuiscono alla radiazione osservata, come l'emissione prompt e l'afterglow. Il lag spettrale può quindi dirci cosa avviene a livello spettrale durante l'emissione del burst, e quali componenti sono in gioco.Per comprendere il legame fra lag temporali ed evoluzione spettrale nei GRB, durante il mio dottorato ho calcolato i lag spettrali sul più ampio intervallo energetico considerato finora, dai raggi X soffici fino a 100 MeV, combinandoli con l'analisi spettrale tempo-risolta.Ho condotto uno studio sistematico dei lag spettrali dei GRB, estendendo per la prima volta l'analisi al dominio 30-100 MeV tramite l'uso congiunto dei dati Fermi LAT Low-Energy (LLE) e Fermi/GBM, su un campione di 70 GRB. Cross-correlando le curve di luce in quattro bande energetiche, ho trovato che i lag misurati all'interno dell'intervallo del GBM (10 keV - 1 MeV) sono prevalentemente positivi, senza alcun caso significativamente negativo. Il lag fra la banda GBM più bassa e la banda LLE, invece, divide il campione quasi equamente fra valori significativamente positivi (40%) e negativi (37%). Un lag positivo sarebbe coerente con una singola componente, che si ammorbidisce nel tempo, dominante su tutto l'intervallo energetico, mentre un lag negativo potrebbe indicare che una componente distinta e più dura emerge nella banda LLE mentre l'emissione prompt si affievolisce.Per chiarire l'origine spettrale di questo comportamento, ho eseguito un'analisi spettrale tempo-integrata dei dati LLE, confrontando la pendenza spettrale misurata nella banda LLE con l'estrapolazione dello spettro keV-MeV. I burst con lag di alta energia positivo tendono a mostrare uno spettro LLE più soffice, coerentemente con una singola componente che si ammorbidisce nel tempo, mentre quelli con lag negativo tendono a mostrarne uno più duro. Questi risultati mi hanno portato a proporre uno scenario fenomenologico in cui il segno del lag di alta energia riflette l'origine dell'emissione osservata: due componenti possono dominare la banda 30-100 MeV, la coda di alta energia dell'emissione prompt e l'inizio dell'afterglow, e un lag negativo segnala l'emergere ritardato della seconda.Per verificare questo scenario, ho eseguito un'analisi spettrale tempo-risolta e, in parallelo, una misura tempo-risolta dei lag, seguendo l'evoluzione dei parametri spettrali e dei ritardi temporali sugli stessi intervalli di tempo. Ho applicato questa analisi a GRB 160625B e GRB 190114C, due dei burst più brillanti del campione, caratterizzati rispettivamente da un lag positivo e da uno negativo, e con una statistica di fotoni sufficiente a vincolare lo spettro intervallo per intervallo.In GRB 160625B una singola componente spettrale si ammorbidisce in modo continuo nel tempo e tutti i lag restano positivi; interpretando il suo cutoff di alta energia come dovuto all'opacità per produzione di coppie, ho ricavato Γ ~ 120-260 e un raggio di emissione R ~ 10^14 cm, ben al di sopra della fotosfera. In GRB 190114C una componente aggiuntiva a legge di potenza emerge a ~2.5 s dal trigger e domina nella banda LLE, producendo così un lag negativo per tutta la durata del burst. Il flusso di questa componente è ampiamente compatibile con l'afterglow iniziale, benché la sua crescita ripida (più rapida di t^3) renda altrettanto plausibile un'origine interna. La misura del lag fornisce dunque un'indicazione su quale componente domini la banda LLE, a un costo molto inferiore a quello dell'analisi spettrale tempo-risolta che sarebbe altrimenti necessaria.Ho esplorato con simulazioni numeriche un terzo scenario per l'origine dei lag negativi. Quando l'emissione di alta energia è inizialmente soppressa dall'opacità gamma-gamma per produzione di coppie e viene progressivamente liberata mentre il flusso si espande e diventa trasparente, i fotoni di alta energia raggiungono l'osservatore più tardi di quelli soffici, anche quando lo spettro sottostante non evolve affatto. Ho trovato che questo processo può operare in sorgenti sufficientemente compatte, con un fattore di Lorentz Γ basso combinato a una luminosità L elevata. Tuttavia, poiché Γ risulta correlato a L, le due condizioni si collocano agli estremi opposti della popolazione, e il meccanismo è confinato agli outlier di questa correlazione, il cui basso Γ li rende intrinsecamente deboli alle alte energie: la loro assenza nei campioni selezionati sulla rivelazione LLE è dunque un effetto di selezione, e non un'evidenza contro il meccanismo.Lo scenario proposto fornisce anche una previsione all'estremo opposto dell'intervallo energetico: dove non si attende il contributo di alcuna componente dura aggiuntiva, il lag dovrebbe essere sempre positivo, essendo determinato unicamente dall'evoluzione spettrale dell'emissione prompt. Per verificare questa previsione, ho calcolato per la prima volta i lag spettrali dell'emissione prompt fra i rivelatori XRT (0.3-10 keV) e BAT (15-150 keV) a bordo di Swift, per 33 GRB con rivelazione simultanea dell'emissione prompt. Per i 19 burst con redshift misurato ho adottato bande energetiche fissate nel sistema di riposo. Ho trovato che 26 lag su 33 sono significativamente positivi e che nessuno è significativamente negativo, in accordo con quanto atteso per un intervallo energetico in cui domina una sola componente prompt. Un'analisi spettrale tempo-risolta XRT+BAT di quattro burst che coprono gli estremi della distribuzione dei lag (da ~0.6 a ~15.5 s) ha rivelato di cosa sia indice il valore del lag. Il lag è nullo o molto piccolo quando l'energia di picco dello spettro nuFnu, Epeak, resta costante, oppure varia troppo poco per attraversare le bande di osservazione; diventa apprezzabile quando Epeak migra da una banda all'altra, come in GRB 100906A. I lag di diversi secondi qui misurati implicano che l'evoluzione spettrale procede in modo considerevolmente più lento che nell'intervallo del GBM.Questa tesi costruisce quindi un quadro coerente e predittivo, verificato da 0.3 keV a 100 MeV, in cui il segno del lag è un indicatore della natura delle componenti di emissione in gioco e il valore del lag misura l'ampiezza della loro evoluzione spettrale.Identificare il numero di componenti è rilevante anche per le applicazioni dei lag temporali alla fisica fondamentale. I test di violazione dell'invarianza di Lorentz sfruttano il fatto che, se la velocità della luce dipendesse dall'energia dei fotoni, fotoni di energie diverse emessi simultaneamente a distanze cosmologiche accumulerebbero un ritardo misurabile lungo il loro percorso. Ogni lag osservato è quindi la somma di due termini: uno estrinseco, dovuto alla propagazione, e uno intrinseco, impresso alla sorgente dalla fisica dell'emissione. Vincolare il primo richiede che il secondo sia compreso, e questi test assumono implicitamente che i fotoni delle due bande siano prodotti dalla stessa componente e siano quindi co-spaziali. Il segno del lag di alta energia offre un modo per verificare questa assunzione, poiché un lag negativo indica che due componenti distinte, prodotte in regioni diverse, dominano le due bande.I lag spettrali sono stati lungamente considerati un prodotto secondario dell'evoluzione spettrale. Questa tesi mostra che sono invece una diagnostica a sé stante, capace di leggere la fisica dell'emissione dai soli tempi di arrivo dei fotoni. Con i campioni più ampi e meglio caratterizzati che le prossime missioni forniranno, quello che qui è un quadro verificato su una manciata di burst può diventare uno strumento standard per sondare la fisica dei GRB.
Maraventano, C. (2026). SPECTRAL LAGS AS A PROBE OF THE PROMPT EMISSION PHYSICS IN GAMMA-RAY BURSTS. (Tesi di dottorato, Università degli Studi di Palermo, 2026).
SPECTRAL LAGS AS A PROBE OF THE PROMPT EMISSION PHYSICS IN GAMMA-RAY BURSTS
MARAVENTANO, CLAUDIA
2026-12-01
Abstract
Gamma-Ray Bursts (GRBs) are the most luminous transient phenomena in the observable Universe: brief flashes of gamma-ray radiation, released at cosmological distances (up to z ~ 9.4) with isotropic-equivalent energies from ~10^48 to a few 10^54 erg. Their bimodal duration distribution motivates the classification into short GRBs (< 2 s), associated with the coalescence of compact objects, and long GRBs (> 2 s), produced by the core collapse of massive stars. GRBs display a highly variable prompt emission (keV-MeV), followed by a long-lasting broadband afterglow. In the standard fireball model both phases are powered by an ultra-relativistic collimated outflow, with bulk Lorentz factor Γ ~ 100-1000: the prompt emission arises from internal dissipation within the jet, the afterglow from the shock that the outflow drives into the circumburst medium.Despite five decades of observations, the physical origin of the prompt emission remains an open question: how the jet energy is dissipated, and which radiative process converts it into the observed photons. A powerful observable in this context is the spectral lag, the time delay τ between the light curves of the same burst in two different energy bands, whose measurement requires far fewer photons than a full spectral characterization and does not rely on any assumed spectral model. The lag is defined as positive when the hard photons arrive before the soft ones, and negative in the opposite case. Long GRBs systematically show positive lags in the keV-MeV range, while at GeV energies negative lags have been reported in a few bright bursts. Short GRBs, on the other hand, are consistently found to have lags compatible with zero. The interpretation of these delays is not yet settled: they may reflect the way the spectrum of a single component evolves in time, or the coexistence of distinct emission processes contributing to the observed radiation, such as the prompt emission and the afterglow. The spectral lag can therefore tell us what happens at the spectral level during the emission of the burst, and which components are at play.In order to understand the link between time lags and spectral evolution in GRBs, during my PhD I computed spectral lags over the widest energy baseline considered so far, from the soft X-rays to 100 MeV, and combined them with time-resolved spectral analysis.I performed a systematic study of GRB spectral lags, extending for the first time the analysis into the 30-100 MeV domain by exploiting Fermi LAT Low-Energy (LLE) data jointly with Fermi/GBM data for a sample of 70 GRBs. Cross-correlating the light curves in four energy bands, I found that the lags measured within the GBM range (10 keV - 1 MeV) are predominantly positive, with no significantly negative case. The lag between the lowest GBM band and the LLE band, instead, splits the sample almost evenly between significantly positive (40%) and negative (37%) values. A positive lag would be consistent with a single component, softening in time, that dominates across the whole energy range, whereas a negative lag might suggest that a distinct and harder component emerges in the LLE band while the prompt emission is fading.To clarify the spectral origin of this behavior, I performed a time-integrated spectral analysis of the LLE data, comparing the spectral slope measured in the LLE band with the extrapolation of the keV-MeV spectrum. Bursts with a positive high-energy lag tend to show a softer LLE spectrum, consistent with a single component softening in time, whereas bursts with a negative lag tend to show a harder one. These results led me to propose a phenomenological scenario in which the sign of the high-energy lag reflects the origin of the observed emission: two components can dominate the 30-100 MeV band, the high-energy tail of the prompt emission and the onset of the afterglow, and a negative lag marks the delayed emergence of the latter.To test this scenario, I performed a time-resolved spectral analysis and, in parallel, a time-resolved measurement of the lags, following the evolution of the spectral parameters and of the time delays across the same time intervals. I applied this analysis to GRB 160625B and GRB 190114C, two of the brightest bursts of the sample, having a positive and a negative lag respectively, and whose photon statistics allow the spectrum to be constrained interval by interval.In GRB 160625B a single spectral component softens continuously in time and all lags remain positive; interpreting its high-energy cutoff as due to pair-production opacity, I derived Γ ~ 120-260 and an emission radius R ~ 10^14 cm, well above the photosphere. In GRB 190114C an additional power-law component emerges at ~2.5 s after trigger and dominates in the LLE band, thus producing a negative lag throughout the burst. The flux of this power-law component is broadly consistent with the early afterglow, although its steep rise (faster than t^3) leaves an internal origin equally viable. The measurement of the lag thus provides an indication of which component dominates the LLE band, at a much lower cost than the time-resolved spectral analysis that would otherwise be required.I explored with numerical simulations a third scenario for the origin of negative lags. When the high-energy emission is initially suppressed by gamma-gamma pair-production opacity and is progressively released as the outflow expands and becomes transparent, the high-energy photons reach the observer later than the soft ones, even when the underlying spectrum does not evolve at all. I found that this process can operate in sufficiently compact sources, having a low bulk Lorentz factor Γ combined with a high luminosity L. However, since Γ is observed to correlate with L, the two conditions lie at opposite ends of the population, and the mechanism is confined to the outliers of this correlation, whose low Γ makes them intrinsically faint at high energies: their absence in LLE-selected samples is thus a selection effect rather than evidence against the mechanism.The proposed scenario also makes a prediction at the opposite end of the energy range: where no additional hard component is expected to contribute, the lag should always be positive, since it would be driven solely by the spectral evolution of the prompt emission. To test this prediction, I computed for the first time the spectral lags of the prompt emission between the XRT (0.3-10 keV) and BAT (15-150 keV) detectors onboard Swift, for 33 GRBs with simultaneous detections of the prompt emission. For the 19 bursts with a measured redshift I adopted fixed rest-frame energy bands. I found that 26 out of 33 lags are significantly positive and that none is significantly negative, in agreement with the expectation for an energy range where a single prompt component dominates. A time-resolved XRT+BAT spectral analysis of four bursts spanning the extremes of the lag distribution (~0.6 to ~15.5 s) revealed what the lag magnitude reflects. The lag is null or very small when the peak energy of the nuFnu spectrum, Epeak, stays constant, or varies too little to cross the observing bands; it becomes sizeable when Epeak migrates from one band to the other, as in GRB 100906A. The several-second lags measured here imply that the spectral evolution proceeds considerably more slowly than in the GBM range.This thesis thus builds a coherent and predictive framework, tested from 0.3 keV to 100 MeV, in which the lag sign is a proxy of the nature of the emission components at play and the lag magnitude measures the amplitude of their spectral evolution.Identifying the number of components also matters for the applications of time lags for fundamental physics. Tests of Lorentz Invariance Violation exploit the fact that, if the speed of light depended on photon energy, photons of different energies emitted simultaneously at cosmological distances would accumulate a measurable delay along their path. Any observed lag is therefore the sum of two terms: an extrinsic one, due to propagation, and an intrinsic one, imprinted at the source by the physics of the emission. Constraining the former requires the latter to be understood, and these tests implicitly assume that the photons of the two bands are produced by the same component and are thus co-spatial. The sign of the high-energy lag provides a way to test this assumption, since a negative lag signals that two distinct components, produced in different regions, dominate the two bands.Spectral lags have long been regarded as a by-product of the spectral evolution. This thesis shows that they are a diagnostic in their own right, one that reads the physics of the emission from the arrival times of the photons alone. With the larger and better characterized samples that the coming missions will provide, what is here a framework tested on a handful of bursts can become a standard tool for probing the physics of GRBs.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi_phd_Claudia.pdf
accesso aperto
Descrizione: Tesi di dottorato Claudia Maraventano
Tipologia:
Tesi di dottorato
Dimensione
15.24 MB
Formato
Adobe PDF
|
15.24 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


