astro-ph0611554/ms.tex
1: \documentclass[useAMS,usenatbib]{mn2e}
2: 
3: \usepackage{graphicx}
4: \usepackage{amssymb}
5: %\usepackage{natbib}
6: 
7: 
8: %%%%% AUTHORS - PLACE YOUR OWN MACROS HERE %%%%%
9: \hyphenation{Taglia-ferri}
10: 
11: \newcommand{\mynote}[1]{[\textsf{\textit{#1}}]}
12: \newcommand{\thisgrb}{GRB\,050716}
13: \newcommand{\swift}{\textit{Swift}}
14: \newcommand{\Tzero}{\ensuremath{\mathrm{T}_0}}
15: \newcommand{\Tnine}{\ensuremath{\mathrm{T}_{90}}}
16: \newcommand{\etal}{{et.~al.}}
17: \newcommand{\plusminus}[2]{\ensuremath{^{+#1}_{-#2}}}
18: \newcommand{\nh}{\ensuremath{N_\mathrm{H}}}
19: \newcommand{\ebv}{\ensuremath{E_{B-V}}}
20: \newcommand{\av}{\ensuremath{A_V}}
21: \newcommand{\nuc}{\ensuremath{\nu_{\mathrm{c}}}}
22: \newcommand{\num}{\ensuremath{\nu_{\mathrm{m}}}}
23: \newcommand{\alphao}{\ensuremath{\alpha_{\mathrm{o}}}}
24: \newcommand{\betao}{\ensuremath{\beta_{\mathrm{o}}}}
25: \newcommand{\chisqred}{\ensuremath{\chi^2_{\mathrm{red}}}}
26: \newcommand{\Ec}{\ensuremath{E_{\mathrm{c}}}}
27: \newcommand{\Ep}{\ensuremath{E_{\mathrm{p}}}}
28: \newcommand{\nodata}{---}
29: %\newcommand{\arcsec}{\mbox{\ensuremath{^{\prime\prime}}}}
30: 
31: 
32: \newcommand{\aj}{{AJ}}         % Astronomical Journal
33: \newcommand{\actaa}{{Acta Astron.}}    % Acta Astronomica
34: \newcommand{\araa}{{ARA\&A}}          % Annual Review of Astron and Astrophys
35: \newcommand{\apj}{{ApJ}}         % Astrophysical Journal
36: \newcommand{\apjl}{{ApJ}}          % Astrophysical Journal, Letters
37: \newcommand{\apjs}{{ApJS}}          % Astrophysical Journal, Supplement
38: \newcommand{\ao}{{Appl.~Opt.}}          % Applied Optics
39: \newcommand{\apss}{{Ap\&SS}}          % Astrophysics and Space Science
40: \newcommand{\aap}{{A\&A}}          % Astronomy and Astrophysics
41: \newcommand{\aapr}{{A\&A~Rev.}}          % Astronomy and Astrophysics Reviews
42: \newcommand{\aaps}{{A\&AS}}          % Astronomy and Astrophysics, Supplement
43: \newcommand{\mnras}{{MNRAS}}          % Monthly Notices of the RAS
44: \newcommand{\na}{{New A}}  % New Astronomy
45: \newcommand{\pasp}{{PASP}}          % Publications of the ASP
46: \newcommand{\nat}{Nature}       % Nature
47: \newcommand{\science}{Science}   % Science
48: \newcommand{\jcp}{JCP}         % The Journal of Chemical Physics}
49: 
50: 
51: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
52: 
53: \title[\thisgrb]{The early and late-time spectral and temporal evolution of \thisgrb}
54: 
55: \author[E. Rol et al.]
56: {E.~Rol,$^1$ 
57: J.~P.~Osborne,$^1$
58: K.~L.~Page,$^1$ 
59: K.~E.~McGowan,$^2$ 
60: A.~P.~Beardmore,$^1$
61: \newauthor
62: P.~T~.O'Brien,$^1$ 
63: A.~J.~Levan,$^3$ 
64: D.~Bersier,$^{4,5}$
65: C.~Guidorzi,$^{5,6}$
66: F.~Marshall,$^7$
67: \newauthor
68: A.~S.~Fruchter,$^4$
69: N.~R.~Tanvir,$^{1,3}$ 
70: A.~Monfardini,$^5$
71: A.~Gomboc,$^{5,8}$
72: \newauthor
73: S.~Barthelmy,$^7$ and
74: N.~P.~Bannister$^1$
75: \\
76: $^1$University of Leicester, University Road, Leicester LE1 7RH, UK \\
77: $^2$School of Physics \& astronomy, University of Southamption, Southampton, SO17 3BJ, UK \\
78: $^3$Department of Physics, Astronomy and Mathematics, University of Hertsfordshire, College Lane, Hatfield, Herts, AL9 10AB, UK  \\
79: $^4$Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA \\
80: $^5$Astrophysics Research Institute, Liverpool John Moores University, Twelve Quays House, Birkenhead, CH41 1LD, UK \\
81: $^6$Osservatorio Astronomico di Brera, INAF, Via Brera 28, 20121, Milano, IT. \\
82: $^7$GSFC, USA \\
83: $^8$Faculty of Mathematics and Physics, University of Ljubljana, Jadranska 19, 1000 Ljubljana, SI.
84: }
85: 
86: \begin{document}
87: 
88: \date{}
89: 
90: \pagerange{\pageref{firstpage}--\pageref{lastpage}} \pubyear{}
91: 
92: \maketitle
93: 
94: \label{firstpage}
95: 
96: \begin{abstract}
97: 
98:   We report on a comprehensive  set of observations of Gamma Ray Burst
99:   050716,   detected  by  the   \swift\  satellite   and  subsequently
100:   followed-up rapidly in X-ray,  optical and near infra-red wavebands.
101:   The  prompt emission is  typical of  long-duration bursts,  with two
102:   peaks in a  time interval of $\Tnine = 68$ seconds  (15 -- 350 keV).
103:   The  prompt emission  continues at  lower flux  levels in  the X-ray
104:   band, where several smaller flares can be seen, on top of a decaying
105:   light~curve that exhibits an  apparent break around 220 seconds post
106:   trigger. This  temporal break is roughly coincident  with a spectral
107:   break.  The latter  can be related to the  extrapolated evolution of
108:   the  break  energy  in  the  prompt $\gamma$-ray  emission,  and  is
109:   possibly the manifestation  of the peak flux break  frequency of the
110:   internal  shock  passing  through  the  observing  band. A
111:     possible 3~$\sigma$ change in  the X-ray absorption column is also
112:     seen  during this  time.   The late-time  afterglow behaviour  is
113:   relatively standard,  with an electron  distribution power-law index
114:   of $p = 2$; there is no  noticable temporal break out to at least 10
115:   days.   The broad-band  optical/nIR  to X-ray  spectrum indicates  a
116:   redshift  of  $z \gtrsim  2$  for  this  burst, with  a  host-galaxy
117:   extinction  value of  $\ebv \approx  0.7$ that  prefers  an SMC-like
118:   extinction curve.
119: 
120: \end{abstract}
121: 
122: \begin{keywords}
123: gamma-rays: bursts -- radiation mechanisms: non-thermal
124: \end{keywords}
125: 
126: \section{Introduction}
127: 
128: 
129: One    of     the    main    goals    of     the    \swift\    mission
130: \citep{gehrels2004:apj611:1005}  is to  obtain  early-time information
131: for  gamma-ray   burst  (GRB)  afterglows  in   X-ray  and  optical/UV
132: wavelengths,  using its  rapid and  autonomous slewing  capability. In
133: addition,  longer term  monitoring, especially  in X-rays,  has become
134: more feasible than previously, with a dedicated facility like \swift.
135: 
136: 
137: Results of the early observations include the rapid decline of (X-ray)
138: afterglows      in      the       first      tens      of      minutes
139: \citep{tagliaferri2005:nature436,  goad2006:aa449:89}.  An explanation
140: for this  behaviour is prompt  emission seen from angles  further away
141: from our  line of sight  \citep{kumar2000:apj541}.  It is  likely that
142: emission from  the afterglow itself has  not risen enough  so early to
143: contribute significantly to the measured flux, although there are some
144: exceptions where  the (X-ray) flux  of the late-time  external forward
145: shock     is     already    visible     from     early    times     on
146: \citep[eg][]{obrien2006:apj647:1213}.   A  thorough understanding  of
147: this phenomenon requires detailed descriptions of the light~curve (and
148: broad-band spectral) behaviour early on, preferentially with as little
149: as possible  interference from other  phenomena such as  X-ray flares.
150: These   X-ray    flares   are   seen    in   a   number    of   bursts
151: \citep[eg][]{burrows2005:science309},    and    are   now    generally
152: interpreted   as    continuing   activity   of    the   inner   engine
153: (\citealt{king2005:apj630:113};  but  see also  \citealt{piro2005:apj623:314},
154: who  interpreted  the  X-ray  rebrightenings  as  the  result  of  the
155: beginning  of the  afterglow).  Late time  monitoring  allows the
156: construction of multiple  epochs of combined near infra-red/optical/UV
157: and  X-ray  broad-band spectra  (occasionally  including radio  data),
158: building a more complete picture of GRB afterglows than before.
159: 
160: 
161: We present here a full  analysis of \thisgrb. This includes the prompt
162: (burst)  emission in  gamma-rays and  early X-ray  data, for  which we
163: performed time-resolved  spectroscopy. We present the  late time X-ray
164: behaviour of the afterglow, as well as late time optical observations,
165: which we then combine to form a broad-band spectrum. The observations
166: and their analysis are described in Section \ref{section:obsana}, and in
167: Section   \ref{section:discussion}  we  discuss   the  results   of  the
168: observations. In Section \ref{section:summary}, we summarise our findings
169: and draw our conclusions.
170: 
171: In the following, $1\,\sigma$ errors are used except where
172: noted. The temporal and spectral power~law indices $\alpha$ and $\beta$
173: are defined by $F \propto t^{-\alpha} \nu^{-\beta}$, and the photon
174: index $\Gamma = 1+\beta$. All \swift\ data have been reduced using the
175: \swift\ software version 2.4 within the HEAsoft software (version
176: 6.0.5), and the corresponding CALDB files.
177: 
178: 
179: 
180: \section{Observations and analysis} \label{section:obsana}
181: 
182: \thisgrb\  was detected  by the  \swift\ Burst  Alert  Telescope (BAT,
183: \citealt{barthelmy2005:ssrv120:143}) on  July 16, 2005,  at 12:36:04 UT.
184: All times mentioned  in this text are relative  to this \Tzero, except
185: where noted.  The spacecraft slewed and started observing in the X-ray
186: and optical/UV bands  96 and 99 seconds later,  respectively.  The XRT
187: \citep{burrows2005:ssrv120:165}  was  able  to  immediately  locate  a
188: position  for  the  X-ray  afterglow on-board,  allowing  ground-based
189: telescopes to perform rapid, deep follow-up observations.
190: 
191: 
192: \subsection{BAT analysis} \label{section:bat}
193: 
194: BAT data were reduced using  the BAT software tools within the \swift\
195: software package.  A  PHA to PI energy conversion  was performed first
196: with the  \texttt{bateconvert} tool,  which ensures the  conversion is
197: quadratic, rather than linear.  We subsequently corrected the data for
198: hot  pixels. An  additional  systematic error  correction was  applied
199: before  spectral fitting.   The 15--350  keV light~curve  is  shown in
200: Figure \ref{figure:batlc}.
201: 
202: 
203: The  light~curve shows an  increase similar  to the  typical Fast-Rise
204: Exponential-Decay (FRED) behaviour seen  for many other bursts, with a
205: peak around 9  seconds after trigger and the start of  the rise of the
206: light~curve some 30 seconds earlier.  A notable difference is that the
207: rise  is not  as  steep as  usually seen  for  a FRED.   A similar  FRED
208: light~curve is  superposed on the tail  of the first  one, producing a
209: second peak 39  seconds after the trigger.  After 80  seconds, most of the
210: prompt  emission  has disappeared  in  the  15--350  keV energy  band.
211: \Tnine\ in this energy band for  this burst is $66 \pm 1$ seconds.
212: A 4-second binned light~curve  shows some emission still apparent past
213: 80 seconds,  up to  200 seconds,  which is mostly  found in  the lower
214: energy bands.
215: 
216: \begin{figure}
217: \includegraphics[width=\columnwidth]{fig1}
218: 
219: \caption{\textbf{Top  panel}  The   BAT  15--350  keV  mask-weighted
220: light~curve, with  the two peaks around  9 and 39  seconds. Points are
221: 0.5-second  binned  light~curve, while  the  overplotted  line is  the
222: 4-second  binned  light~curve.  The  latter shows  some  evidence  for
223: possible emission after 80 seconds. \textbf{Bottom panel} The spectral
224: evolution of  the prompt emission in  the 15--150 keV  range; shown is
225: the photon index for a power~law model with an exponential cutoff.
226: \label{figure:batlc}
227: }
228: 
229: \end{figure}
230: 
231: We have  subdivided the  gamma-ray data into  4 temporal  sections and
232: obtained  spectral fits  for each  of  these. The  spectral shape  was
233: fitted  with a  power~law,  which  results in  an  acceptable fit.   A
234: power~law with an exponential cut-off, however, provides a better fit,
235: while  a fit with  a Band  model \citep{band1993:apj413:281}  does not
236: improve  the fit  compared  to either  one  of those  models. The  BAT
237: spectral energy  range and low number  of energy bins do  not allow an
238: accurate  determination of  the  cut-off energy  \Ec\  in the  cut-off
239: power~law model (for the first section,  it is in fact a lower limit),
240: but there  is an indication  that the spectrum  softens, with
241: \Ec\ gradually becoming lower,  and the photon index $\Gamma$ changing
242: from about  0.4 to  0.9 during the  prompt gamma-ray emission,  over a
243: period of $\approx 50$ seconds.
244: 
245: To  verify this  spectral evolution,  we  have fitted  the later  time
246: sections  (C  \&  D)  with   the  model  parameters  (other  than  the
247: normalisation) fixed at those obtained from the earlier sections (A \&
248: B).  By  then freeing the parameters  and applying an  F-test, we find
249: that  there is  indeed a  change in  the spectrum  at a  99.9\% level.
250: Performing a  similar test to  the first and  last sections (A  \& D),
251: with the results obtained from the fit to the total section (-20 -- 80
252: s), also indicates a changing spectrum at a 99.9\% confidence level.
253: 
254: Such  softening behaviour  of the  prompt spectrum  has  been commonly
255: found before in other bursts \citep{ford1995:apj439:307}.  The results
256: are  detailed in Table  \ref{table:gammaray-spectra}, and  include the
257: result  of   a  spectral   fit  to  the   whole  time   span.   Figure
258: \ref{figure:batlc} shows the spectral evolution of the prompt emission
259: between 15 and 150 keV as well.
260: 
261: 
262: 
263: 
264: \begin{table}
265: \caption{Results of spectral fits to four sections of the prompt
266: gamma-ray emission, as well as that of the combined fit.
267: \label{table:gammaray-spectra}
268: }
269: \begin{tabular}{lrlll}
270: Section &
271: Time &
272: $\Gamma$ &
273: cutoff energy &
274: \chisqred 
275: \\
276:  &
277: (sec) &
278:  &
279: (keV) & 
280: \\
281: \hline
282: 
283: A &
284: -20 -- 0 &
285: 0.42 \plusminus{0.21}{0.60} &
286: 221  ($> 6.7$) &
287: 1.05
288: \\
289: 
290: B &
291: 0 -- 20 &
292: 0.74 \plusminus{0.27}{0.29} &
293: 125 \plusminus{178}{49} &
294: 0.86 
295: \\
296: 
297: C &
298: 20 -- 40 &
299: 0.64 \plusminus{0.40}{0.45} &
300: 105 \plusminus{305}{48} &
301: 0.86 
302: \\
303: 
304: D &
305: 40 -- 80 &
306: 0.87 \plusminus{0.33}{0.36} &
307: 65.6 \plusminus{47.5}{20.9} &
308: 0.95 
309: \\
310: 
311: A -- D &
312: -20 -- 80 &
313: 0.85 \plusminus{0.21}{0.23} &
314: 150 \plusminus{169}{54} &
315: 0.76
316: \\
317: \hline
318: 
319: \end{tabular}
320: \end{table}
321: 
322: 
323: \subsection{XRT analysis} \label{section:xrt}
324: 
325: XRT data were reduced from level 1 to level 2 with the \swift\
326: software task \texttt{xrtpipeline}. Data obtained in Windowed Timing
327: (WT) mode and Photon Counting (PC) mode have been used in all our
328: following analysis. Counts were grouped by 20 per  bin for the spectra.
329: WT mode  data for the  light curve have  been grouped by 40  counts per
330: bin, and  PC mode light curve data  have been grouped by  20 counts per
331: bin.
332: 
333: We  have  subdivided  our  analysis  into an  early-time  part  and  a
334: late-time part. We designate the early-time part as the section of the
335: X-ray afterglow  that shows flares  (probably related to  inner engine
336: activity, see  also the discussion  below), while the  late-time part
337: shows a  relatively smooth decay and  is presumed to  be the afterglow
338: emission from a forward shock. The  data for the early part extends up
339: to almost 1ks, with most of  the data having been obtained in WT mode.
340: The late-time part  starts some 4ks after the  trigger and extends out
341: to almost three weeks.
342: 
343: 
344: \subsection{Early XRT data} \label{section:earlyxrt}
345: 
346: WT mode  observations started at 103  seconds, and extended  up to 516
347: seconds. After this, the count-rate  was low enough for the instrument
348: to automatically switch to PC  mode. WT mode data were extracted using
349: a rectangular region with a length of 93\arcsec\ centred on the source
350: and  along  the  readout  direction,  with  an  equally  sized  region
351: off-source serving as the  background determination. PC mode data have
352: been  extracted with  a 47\arcsec\  circular aperture  instead  of the
353: default 71\arcsec\,  because of  a contaminating close-by  source. The
354: first few  100 seconds  of PC mode  data are  piled-up and we  used an
355: annular  extraction  region, with  a  12\arcsec\  inner  radius and  a
356: 47\arcsec\ outer  radius.  In addition,  some bad columns  are located
357: near the source centre.  We used  default grades for WT mode (grades 0
358: to 2),  while we used only  grade 0 for  the first few 100  seconds of
359: PC-mode observations.
360: 
361: To determine  the combined correction factor for  the annular aperture
362: and bad  column loss, we  modelled the PSF using  \textsf{ximage}, and
363: calculated the ratio of  the integrated response between an unmodified
364: PSF and  that masked  out by  an inner 12\arcsec\  circle and  the bad
365: columns. In the same manner, we correct for the expected loss of using
366: only a 47\arcsec\ circular aperture. The combined correction factor is
367: 4.47.
368: 
369: 
370: \subsubsection{Spectral evolution}
371: 
372: Similar to the BAT observations, we  find evidence for a change in the
373: spectral slope over time.  Modelling  the spectrum with a simple power
374: law, we  find that the  X-ray spectrum starts  with $\Gamma =  0.9 \pm
375: 0.1$ (0.3 -- 10 keV), and the  slope then evolves to $\Gamma = 2.0 \pm
376: 0.1$, with a change between 250 and 400 seconds post trigger.
377: 
378: If a break is expected in the X-ray band, a broken power~law model for
379: the spectra around this time is preferred. Fits with broken power~law
380: models, however, do not show a significant improvement in the fit, and
381: we have therefore used the results from the single power~law fits above.
382: 
383: Because there is still some flux in the BAT light~curve coincident
384: with the first 100 seconds or so of WT mode data, we have performed a
385: joined fit to the BAT and XRT data between 100 and 200 seconds post
386: trigger. A single power law fit has a \chisqred of $1.98$ (degrees of
387: freedom (DOF) = 61), but a Band model does a much better job
388: ($\chisqred = 0.90$, DOF = 59), and results in $E_0 =
389: 12.2\plusminus{5.4}{2.4}$~keV, or a peak energy of $\Ep =
390: 13.1\plusminus{5.8}{2.6}$~keV.  The spectral indices are $\alpha =
391: -0.93\plusminus{0.08}{0.13}$ and $\beta =
392: -2.56\plusminus{0.37}{1.51}$.
393: 
394: Because  of the  low number  of counts  in the  BAT data,  the  fit is
395: dominated by  the XRT  data and the  outcome should be  taken somewhat
396: cautiously.
397: 
398: \subsubsection{Light curve} \label{sect:light-curve}
399: 
400: Since  there is a  change in  the spectral  slope halfway  through the
401: light~curve, we  cannot simply fit the count-rate  light~curve: a flux
402: calibrated light~curve  has to be  used.  To derive the  count-rate to
403: flux conversion factors, we have  used the spectra between 103 and 223
404: seconds   and  between   263   and  513   seconds   (see  also   Table
405: \ref{table:nh}), which provide the conversion values for the first and
406: last part of the light~curve,  respectively. For the middle section we
407: have used  the average of these  values: the rapid  change in spectrum
408: does not allow for a good  spectral fit. In all our following fits, we
409: have  modelled  the  two   most  pronounced  flares  with  two  simple
410: Gaussians. We also left out the last data point of the light~curve: it
411: is likely that  this point is the onset  of the later-time light~curve
412: (see  Section \ref{section:latexrt}),  or possibly  even the  start of
413: another flare.
414: 
415: A fit to the flux light curve with a single power~law results in
416: $\chisqred = 2.06$ (DOF = 123), while a broken power~law fit gives
417: $\chisqred = 1.83$ (DOF = 121).  Neither fit is good, but the broken
418: power~law does provide a significant improvement. A fit with a
419: smoothly broken power~law (either with the sharpness of the break
420: fixed or free) hardly improves this ($\chisqred = 1.78$ for
421:   both, DOF = 121 and 120, respectively), and an exponential fit is
422: worse ($\chisqred = 2.07$, DOF = 123).
423: 
424: The broken power~law fit results  in decay parameters $\alpha_1 = 0.91
425: \plusminus{0.30}{0.09}$ and  $\alpha_2 = 3.78 \plusminus{0.19}{0.15}$.
426: The temporal  break is found at $220  \plusminus{27}{4}$ seconds, just
427: before the  time of the spectral break.   This coincidence strengthens
428: the suggestion that  a broken power~law is a  reasonable model for the
429: underlying  light~curve. See Section  \ref{section:early} for  a more
430: detailed exploration of this break.
431: 
432: The  light~curve  and  the  spectral  evolution are  shown  in  Figure
433: \ref{figure:earlylc}.   As commonly seen  in other  early X-ray
434: light~curves
435: \citep[eg][]{burrows2005:science309,nousek2006:apj642:389,obrien2006:apj647:1213},
436: this one exhibits  a few flares, which most  likely indicates activity
437: of  the burst  engine  itself.   This would  agree  with the  possible
438: emission  seen  in the  BAT  15--350 keV  range  between  100 and  200
439: seconds. Also visible is the  deviation of the underlying decay from a
440: power~law.
441: 
442: \begin{figure}
443: \includegraphics[width=\columnwidth]{fig2}
444: 
445: \caption{ \textbf{Top  panel} 0.3--10 keV  early-time light~curve. The
446: last three points  are PC mode data, the  previous data points consist
447: of WT mode data.  A broken
448: power~law fit with two Gaussians is shown as the continuous line, with
449: its break time around 200  seconds post trigger (the Gaussians account
450: for the  two main flares, to  obtain a better overall  fit; the broken
451: power~law without  Gaussians is shown  as the dashed line).   The last
452: data  point  has  not been  fit,  since  this  likely belongs  to  the
453: later-time   light~curve. \textbf{Middle  panel}  The   fit  residuals,
454: expressed  in sigma  deviation. Even  with  the Gaussian  fits to  two
455: flares, the overall light  curve shows significant deviations from the
456: fit. \textbf{Bottom panel} The  evolution of the photon index $\Gamma$
457: of the  X-ray spectrum.
458: \label{figure:earlylc}
459: }
460: 
461: \end{figure}
462: 
463: \subsubsection{X-ray absorption change?}
464: 
465: For the combined WT-mode data, we find no evidence for excess \nh\
466: above that of the estimated Galactic column of $1.1 \times 10^{21}
467: \mathrm{cm}^{-2}$ \citep{dickey1990:araa28:215}. Subdividing the WT-mode
468: into a pre- and a post- spectral break section indicates, however, a
469: modest amount of excess absorption at early times, while no excess
470: absorption is measured past the break: $\nh =
471: 2.0\,(\pm\,0.2)\,\cdot\,10^{21}\,\mathbf{\mathrm{cm}^{-2}}$ and $\nh =
472: 1.1 (\pm 0.2) \cdot 10^{21} \mathbf{\mathrm{cm}^{-2}}$ for the total
473: \nh\ at the two respective epochs.  While the observed change in
474: photon index could correlate with this change in \nh, a contour plot
475: (Figure \ref{figure:contour}) indicates it is likely that the column
476: density did change.  Something similar has been seen, for example, in
477: the early (prompt) X-ray emission for GRB\,000528
478: \citep{frontera2004:apj614:301}, or claimed for GRB\,050730
479: \citep{starling2005:aa442:21}.  An F-test for a model with \nh\ fixed
480: at the later time value (see also Section \ref{section:latexrt}),
481: compared to a model with \nh\ free to fit, gives a probability of
482: 6.7e-5 for the two models to be consistent.  See the second part of
483: Table \ref{table:nh} for more details.
484: 
485: 
486: 
487: \begin{figure}
488: \includegraphics[angle=-90,width=\columnwidth]{fig3}
489: 
490: \caption{Contour  plots for  the  1-$\sigma$, 90\%  and 99\%  combined
491: confidence  intervals for  the  photon index  $\Gamma$  and \nh.   The
492: contours  upper-left  are for  the  WT-mode  data  after the  observed
493: spectral  break (303-513  seconds),  the contours  in the  lower-right
494: corner  are pre-break  (103-253  seconds).  The  middle part  (253-303
495: seconds) is left out, since here  the change in spectral break is most
496: pronounced,  indicating  $\Gamma$   is  highly  variable.   Note  that
497: $\Gamma$  for the  first  data  section is  slightly  higher than  the
498: previously quoted $0.9$,  which is for the very  first part of WT-mode
499: data only.
500: \label{figure:contour}
501: }
502: 
503: \end{figure}
504: 
505: \begin{table*}
506: \caption{Results of spectral fits for the XRT WT-mode data. The second half
507: of the table details the results on variation of the column
508: density. See also Figure \ref{figure:contour}. A simple power~law was
509: used, with the Teubingen-Boulder model
510: \citep[][\texttt{tbabs}]{wilms2000:apj542:914} for the absorption.
511: \label{table:nh}
512: }
513: \begin{tabular}{llll}
514: \multicolumn{1}{c}{time}  & 
515: \multicolumn{1}{c}{$\Gamma$} & 
516: \multicolumn{1}{c}{\nh} & 
517: \multicolumn{1}{c}{$\chi^2_{\mathrm{red}}$} \\
518: \multicolumn{1}{c}{(sec)} &
519:  &
520: \multicolumn{1}{c}{($10^{21}$ cm$\mathbf{^{-2}}$)} &
521:  \\
522: 
523: \hline
524: 103 -- 123  &  
525: $0.95 \pm 0.12$  &
526: $1.37 \plusminus{0.73}{0.66}$  &
527: 0.509 (12.2/24)  \\
528: 
529: 123 -- 143  &
530: $1.31 \plusminus{0.15}{0.14}$  &
531: $1.87 \plusminus{0.74}{0.67}$  &
532: 0.899 (15.3/17)  \\
533: 
534: 143 -- 163  &  
535: $1.26 \pm 0.14$  &
536: $1.18 \plusminus{0.66}{0.61}$  &
537: 0.518 (9.84/19)  \\
538: 
539: 163 -- 183  &  
540: $1.18 \plusminus{0.13}{0.12}$  &
541: $1.11 \plusminus{0.61}{0.57}$  &
542: 0.698 (16.8/24)  \\
543: 
544: 183 -- 243  &  
545: $1.23 \pm{0.11}$  &
546: $1.65 \plusminus{0.42}{0.39}$  &
547: 0.801 (25.6/32)  \\
548: 
549: 243 -- 363  &  
550: $1.66 \pm{0.07}$  &
551: $1.51 \plusminus{0.25}{0.23}$  &
552: 1.09 (63.3/58)  \\
553: 
554: 363 -- 403  &  
555: $2.16 \plusminus{0.23}{0.21}$  &
556: $1.88 \plusminus{0.56}{0.53}$  &
557: 1.12 (13.5/12)  \\
558: 
559: 403 -- 513  &  
560: $2.05 \plusminus{0.17}{0.16}$  &
561: $1.09 \plusminus{0.39}{0.36}$  &
562: 1.09 (22.9/21)  \\
563: 
564: \hline
565: \multicolumn{4}{l}{\hspace{-0.5cm}\nh\ variation} \\
566: 
567: 103 -- 223  &  
568: $1.26 \pm 0.05$  &
569: $2.04 \plusminus{0.21}{0.20}$  &
570: 0.890 (111.2/125) \\
571: 
572: 263 -- 513  &
573: $1.86 \plusminus{0.08}{0.07}$  &
574: $1.13 \plusminus{0.19}{0.18}$  &
575: 1.053 (73.69/70)  \\
576: 
577: \hline
578: 
579: \end{tabular}
580: \end{table*}
581: 
582: 
583: 
584: 
585: \subsection{Late time XRT observations}
586: \label{section:latexrt}
587: 
588: The XRT  data after 4ks are  not piled-up, and were  extracted using a
589: circular aperture  of 47\arcsec  radius and the  default grades  (0 to
590: 12).   The  resulting 0.3\,--\,10  keV  light~curve  follows a  smooth
591: power-law decline,  with a  power-law decay index  of 0.99.   The full
592: light~curve  is shown  in Figure  \ref{figure:xrtlc} and  includes the
593: early time light~curve  as well.  There is no  indication of a shallow
594: part  ($\alpha \lesssim  0.5$) in  the light~curve  between  the steep
595: early decay and  the later time decay, as often  seen around this time
596: in the light~curve  \citep[eg][]{nousek2006:apj642:389}; this may have
597: been  missed if  it was  relatively short  and between  the  early and
598: late-time part.   There has been, however,  no need to  include such a
599: section  in  our  fits to  obtain  good  results.   There is  also  no
600: indication of a late-time break  before the light~curve is lost in the
601: background noise,  up to  $10^6$ seconds. A  spectral fit to  the data
602: between 10ks and 100ks, with \nh\ fixed at its Galactic value, results
603: in a good fit ($\chi^2_{\mathrm{red}} = 0.54$), with a photon index of
604: $\Gamma = 2.01 \plusminus{0.11}{0.10}$.
605: 
606: \subsection{Optical observations}
607: \label{section:optical}
608: 
609: Optical  observations were  automatically performed  with  the \swift\
610: Ultra       Violet       and       Optical      Telescope       (UVOT,
611: \citealt{roming2005:ssrv120:95})  in the  $UBVW1\,M2\,W2$  filters and
612: without  a filter.   The  prompt dissemination  of  the position  also
613: triggered  the Faulkes Telescope  North (FTN),  which began  the first
614: observations     of      its     automatic     follow-up     procedure
615: \citep{guidorzi2006:pasp188:288}   243  seconds   after   the  \swift\
616: trigger.  Finally,  near infra-red observations at  the United Kingdom
617: Infrared  Telescope (UKIRT) started  56 minutes  post trigger,  in the
618: $JHK$ bands,  and Gemini NIRI  $K$-band observations were  obtained 22
619: hours after \Tzero.  The log of our observations can be found in Table
620: \ref{table:optical-log}.
621: 
622: A faint,  fading source was detected  in the UKIRT  data, just outside
623: the  90\%  XRT  error  circle,  and identified  as  the  IR  afterglow
624: \citep{tanvir2005:gcn3632}.  The IR data  also gives the best position
625: for   the  GRB   and   its  afterglow.    This  is   RA~=~22:34:20.73,
626: Dec~=~+38:41:03.6 (J2000), with an estimated error of 0.4\arcsec. This
627: position  is derived  from the  NIRI $K$-band  image,  calibrated with
628: respect to the 2MASS survey \citep{skrutskie2006:aj131:1163}.
629: 
630: A single power~law fit to the three $K$-band points results in a
631: power~law decay index of $\alphao = 0.80\plusminus{0.13}{0.07}$, with
632: $\chi^2_{\mathrm{red}} = 0.689$. While different from the X-ray decay
633: slope at these times, the paucity of data points does not allow a
634: strong quantification of this difference. A fit with $\alphao$ fixed
635: at $\alpha_X = 0.99$, however, results in $\chi^2_{\mathrm{red}} =
636: 2.58$.
637: 
638: 
639: We calculated the $JHK$ spectral index by shifting the data to a
640: common epoch using the previously calculated decay index, and
641: correcting for the estimated Galactic extinction
642: \citep{schlegel1998:apj500}. Conversion to flux was performed using
643: the Vega fluxes from \citet{fukugita1995:pasp107} for the optical and
644: from \citet{tokunaga2005:pasp117} for the nIR filters.  The fit shows
645: the spectrum to be inconsistent with a power-law spectrum ($\betao =
646: 2.9$, but $\chisqred = 4.2$), indicative of intrinsic reddening. We
647: address this in Section \ref{section:broadband}.
648: 
649: 
650: \begin{table*}
651: \caption{\label{table:optical-log}   The   results   of  our   optical
652: follow-up campaign, covering the  wavelength range from near infra-red
653: to the near  ultraviolet. Other data collected through  the GCN network
654: provide only  upper limits similar to  those of the FTN  and UVOT, and
655: have therefore not been repeated here.}
656: \begin{tabular}{cccccc}
657:    Start date         &   days since burst  &  exposure time  &  filter  &
658:   magnitude    &  telescope/instrument  \\
659: (days UT, July 2005)  &      (mid-time)     &    (seconds)    &          & 
660:               &             \\
661: \hline
662: 16.529991 &   0.01954   &  400    &    $i'$   &  $ > 21.5 $       &  FTN         \\
663: 16.527721 &   0.00306   &  30     &    $R$    &  $ > 20.3 $       &  FTN         \\
664: 16.527721 &   0.02239   &  540    &    $R$    &  $ > 22.3 $       &  FTN         \\
665: 16.529911 &   0.02236   &  520    &    $B$    &  $ > 23.0 $       &  FTN         \\
666: 16.526169 &   0.4315    &  5799   &    $V$    &  $ > 21.2 $       &  \swift/UVOT  \\
667: 16.527859 &   0.4042    &  5507   &    $B$    &  $ > 22.1 $       &  \swift/UVOT  \\
668: 16.527697 &   0.3988    &  5477   &    $U$    &  $ > 21.6 $       &  \swift/UVOT  \\
669: 16.527534 &   0.4416    &  5053   &   $W1$    &  $ > 21.5 $       &  \swift/UVOT  \\
670: 16.527373 &   0.4374    &  5486   &   $M2$    &  $ > 21.8 $       &  \swift/UVOT  \\
671: 16.528044 &   0.4084    &  4383   &   $W2$    &  $ > 21.8 $       &  \swift/UVOT  \\
672: 19.055751 &   2.7699    &  9480   &  'white'  &  $ > 22.5 $       &  \swift/UVOT  \\
673: 16.579109 &   0.05594   &  300    &    $J$    &  $20.76 \pm 0.08$ &  UKIRT/UFTI  \\
674: 16.581678 &   0.05937   &  450    &    $H$    &  $19.31 \pm 0.05$ &  UKIRT/UFTI  \\
675: 16.564178 &   0.04272   &  600    &    $K$    &  $17.74 \pm 0.15$ &  UKIRT/UFTI  \\
676: 16.583727 &   0.06055   &  300    &    $K$    &  $18.22 \pm 0.15$ &  UKIRT/UFTI  \\
677: 17.421337 &   0.9275    & 3600    &    $K$    &  $20.5 \pm 0.2$   &  Gemini/NIRI \\
678: \hline
679: \end{tabular}
680: \end{table*}
681: 
682: 
683: \begin{figure}
684: \includegraphics[width=\columnwidth]{fig4}
685: 
686: \caption{\label{figure:xrtlc}
687: The complete X-ray light~curve between
688: 0.3 and 10 keV. The continuous line is modelled using a broken
689: power~law and 2 Gaussians for the early time light~curve (see Section
690: \ref{section:earlyxrt}), and a single power~law added to that for the
691: late-time light~curve. The latter has a decay index of $0.99 \pm
692: 0.02$.
693: \textbf{Bottom panel} The residuals of the fit,
694: expressed in sigma deviation.
695: }
696: \end{figure}
697: 
698: 
699: 
700: 
701: 
702: 
703: \section{Discussion} \label{section:discussion}
704: 
705: 
706: \begin{table*}
707: \caption{\label{table:summary}
708:   Summary of the main results obtained in Section \ref{section:obsana}. 
709:   We have converted the spectral results from $\Gamma$ to $\beta$.
710: }
711: \begin{tabular}{llll}
712: \multicolumn{1}{c}{data section}  & 
713: \multicolumn{1}{c}{$\beta$} & 
714: \multicolumn{1}{c}{$\alpha$} & 
715:  \\
716: \hline
717: 
718: X-ray, 103 -- 223 seconds & 
719: 0.26 $\pm$ 0.05 &
720: 0.91 \plusminus{0.30}{0.09} 
721: \\
722: 
723: X-ray, 263 -- 513 seconds &
724: 0.86 \plusminus{0.08}{0.07} &
725: 3.78 \plusminus{0.19}{0.15} 
726: \\
727: 
728: X-ray, post 1 hour &
729: 1.01 \plusminus{0.11}{0.10} &
730: 0.99 $\pm$ 0.02 
731: \\
732: 
733: optical, post 1 hour &
734: 2.94$^1$ &
735: %\nodata &
736: 0.80 \plusminus{0.13}{0.07} 
737: \\
738: \hline
739: $^1$ $\chisqred = 4.2$, therefore no error has been calculated.
740: \end{tabular}
741: \end{table*}
742: 
743: Here we discuss  the results obtained above. A  short summary of the
744: results is given in Table \ref{table:summary}.
745: 
746: \subsection{Early emission}
747: \label{section:early}
748: 
749: \thisgrb\ shows  characteristics as seen in  previous (\swift) bursts,
750: both   in   its   prompt    emission   and   the   (early)   afterglow
751: \citep{nousek2006:apj642:389,obrien2006:apj647:1213}.    The  flares,
752: combined  with the  multiple peaks  in the  gamma-ray  emission, imply
753: several  shocks,  likely   indicative  of  prolonged  engine  activity
754: \citep{king2005:apj630:113}.  The  early X-ray data  show a
755: power-law decline, that appears to be unrelated to the later afterglow
756: emission and  which is often explained as  prompt emission originating
757: from higher latitudes  at the emission source,  after the emission at
758: smaller angles has  faded away.
759: 
760: 
761: In the context of this model, we can test the curvature effect $\alpha
762: = 2  + \beta$  \citep{kumar2000:apj541}. We used  the values  pre- and
763: post-break  in  the XRT  data,  showing  that  by this  relation,  the
764: observed  $\alpha_1$   is  too  shallow  and   $\alpha_2$  too  steep.
765: \citet{zhang2006:apj642:354}  list several  possible  causes why  this
766: relation would  not hold. Most important  are the effect  of using the
767: trigger time  instead of the time  of the last  emission peak (thereby
768: assigning an  incorrect zeropoint  for a power  law fit), and  that of
769: superposition of tails of several emission peaks. Both causes could be
770: valid here, since the last  significant emission peaks some 40 seconds
771: after  the  trigger. However,  applying  such  a  correction for  both
772: effects would effectively produce  an ever shallower decay, making the
773: discrepancy larger. Possible causes for $\alpha$ being too shallow are
774: the    fact   that   one    is   looking    at   a    structured   jet
775: \citep{rossi2002:mnras332:945} or when the observed (X-ray) wavelength
776: regime is still below the cooling break. In the latter case, the decay
777: is given by $\alpha  \sim 1 + 3 \beta/2$ \citep{zhang2006:apj642:354}.
778: Only the very  early section of the X-rays  agrees with this relation,
779: with    $\alpha    =   0.91$    and    $\beta    =   -0.1$    (Section
780: \ref{sect:light-curve},  Table \ref{table:nh}).   Post-break, however,
781: the discrepancy  is even larger  in this scenario.   Potentially, both
782: the peak frequency and the cooling break have passed the X-ray band at
783: this  time. While  this  could explain  the  relatively long  spectral
784: transition, there is no direct  evidence for this.  The discrepancy at
785: later times, for either relation, might more likely be attributed to
786: an increasing influence of  the external shock (which likely dominates
787: the last  data), and the  last flare, which  hamper a good fit  to the
788: steeper tail of the early light curve.
789: 
790: Evolution of the early emission in the X-ray to gamma-ray energy range
791: has  been studied before,  eg by  \citet{frontera2000:apjs127:59}, who
792: looked at  the spectral  evolution for eight  \textit{Beppo}SAX bursts
793: which had been detected by both the Gamma-Ray Burst Monitor and one of
794: the Wide-Field Cameras. While their findings are similar, an important
795: advantage here is that we have  an early X-ray light curve which, with
796: provisions for some  flaring, can be fit with a  broken power law; the
797: break time  of this  power law is  roughly coincident with  the moment
798: \Ep\ passes through the X-ray  band. The WFC data generally cover only
799: the very  early part of the  burst, and any  underlying (broken) power
800: law is  not visibile.  As a  result, the \Ep\ evolution  could only be
801: deduced from the spectral changes for the \textit{Beppo}SAX data.
802: 
803: Variable X-ray absorption at early times in the X-ray spectra of GRBs
804: has been noted before in other GRBs
805: \citep{frontera2000:apjs127:59,amati2000:sci290:953,intzand2001:apj559:710,starling2005:aa442:21}.
806: \citet{lazzati2002:mnras330:383} have suggested that ionisation
807: strongly modifies the absorption properties of the surrounding
808: material. For this effect to be noticable, the material should be in a
809: compact region surrounding the GRB ($< 5$ parsec), and have an initial
810: column density $\nh > 10^{21} \mathrm{cm}^{-2}$. The change detected
811: here is $0.8 \cdot 10^{21} \mathrm{cm}^{-2}$ at redshift 0.  For a
812: redshift of $z = 2$ (see below), the column density would be $\nh
813: \approx 1.5 \cdot 10^{22} \mathrm{cm}^{-2}$ (at Solar meticallicity),
814: large enough to fulfill the $\nh > 10^{21} \mathrm{cm}^{-2}$
815: criterion.
816: 
817: 
818: 
819: \subsection{The late-time light~curve and broad-band spectrum}
820: \label{section:broadband}
821: 
822: 
823: From the combination of the X-ray temporal and spectral index around
824: 5ks, we can infer a electron power~law index of $p = 2.0$ within the
825: standard fireball model
826: \citep[eg][]{meszaros1997:apj476:232,sari1999:apj519:17,chevalier1999:apj520:29}.
827: This assumes that 1) the afterglow is in the slow-cooling regime
828: ($\num < \nuc$), 2) \nuc\ is below the X-rays and 3) the jet-break has
829: not manifested itself at this time. Other scenarios are easily ruled
830: out by invalidating relations between the two indices (see eg Table 1
831: in \citealt{zhang2004:ijmp19:2385}).  There is still a degeneracy
832: between the type of circumburst environment (a constant density medium
833: or a medium with a $R^{-2}$ dependency ($R$ the distance from the
834: centre) such as a stellar wind), which can potentially be resolved by
835: looking at the temporal and spectral indices below \nuc; in our case
836: the only candidate for that is the optical band, which is generally
837: found to be below \nuc\ at these times. To verify this is indeed the
838: case, we compared the X-ray and optical decay indices.
839: 
840: 
841: There is a 1.5$\sigma$ difference between the afterglow decay slope in
842: optical and X-rays, which is only marginally significant and does
843: allow \nuc\ to be below optical wavelengths. However, a cooling break
844: between the optical and X-ray waveband makes the picture
845: self-consistent: for an ISM-like, the decay index in optical should be
846: 0.25 lower than that in X-ray, 0.74, which is certainly compatible
847: with $\alphao$ obtained from our fit to the $K$-band data. On the
848: other hand, for a wind-like medium, the optical decay index would have
849: to be 0.25 higher than that in X-ray, 1.24, and is therefore
850: incompatible with \alphao.
851: 
852: The non power-law behaviour of the near infra-red (nIR) spectrum
853: indicates reddening, which we estimate by examining the broad-band
854: spectral behaviour, by combining the X-ray, optical and nIR data.  For
855: this, we have extrapolated the optical and X-ray data to a common
856: epoch: we choose 0.04 days (3.5ks) after trigger, which is central to
857: the optical FTN and nIR UKIRT data, and lies just at the beginning of
858: the late-time X-ray data. We obtained an unabsorbed X-ray flux from
859: the data between 4ks and and 12ks (orbits 2 and 3 combined), which
860: were then extrapolated back to 0.04 days using the power~law decay
861: index of $0.99$.  The optical and nIR data were corrected for Galactic
862: extinction and converted to fluxes as before.  We then extrapolated to
863: 0.04 days post trigger using our estimate for the optical power~law
864: decay index.
865: 
866: \begin{figure}
867: \includegraphics[width=\columnwidth]{fig5}
868: 
869: \caption{The X-ray to optical/nIR broad-band spectrum of GRB\,050716, 0.04 days post burst. The best fit is shown for a broken power~law connecting the two
870:   wavelength ranges, with host-galaxy absorption for an
871:   SMC-like extinction curve at $z = 2$. The X-ray spectrum has been
872:   represented relatively simply by a single data point. This
873:     has the purpose of enhancing the \chisqred\ sensitivity on the
874:     nIR data instead of on the X-ray data, since we try to calculate
875:     the estimated optical/UV host-galaxy extinction.
876: \label{fig:bbspec}
877: }
878: 
879: \end{figure}
880: 
881: With  the scaled  fluxes, we  constructed the  broad-band  spectrum of
882: \thisgrb\ at 0.04 days (Figure \ref{fig:bbspec}). We then tried to fit
883: the  data with models  available within  the standard  fireball model,
884: allowing for intrinsic (UV/optical)  extinction within the host galaxy
885: and  a Lyman  cutoff  resulting from  its  redshift. No  spectroscopic
886: redshift  is known,  however.  While  in principal  one  can obtain  a
887: photometric redshift, there are too  many free parameters with the few
888: available data  to do this  accurately. We have therefore  obtained an
889: estimate of  the intrinsic extinction  for an afterglow  positioned at
890: various redshifts  (1, 2, 3,  \ldots, 7). We have  further constrained
891: the power~law slope connecting the  X-ray and optical/nIR regime to be
892: equal to that obtained by the  X-ray spectral fit alone, $\beta = 1.0$.
893: 
894: A single power~law, absorbed by host-galaxy extinction or
895: intergalactic Lyman extinction, is not able to reproduce the observed
896: broad-band spectrum. A spectral break between optical and X-ray is
897: required, as already found from the fits to the optical and X-ray
898: light~curves.  With the assumption that this break is indeed the
899: cooling break (with $\Delta \beta = 0.5$), we have kept the lower end
900: of this break fixed at a power~law index of $\beta_{\mathrm{opt}} =
901: 0.5$, and the higher end at $\beta_{\mathrm{X}} = 1.0$).  Reasonable
902: fits are then obtained for redshifts above $z = 2$, with host-galaxy
903: extinction that follows an SMC-like extinction curve (\ebv\ = 0.5 --
904: 0.7, \av\ = 1.5 -- 2.1, with lower values for higher redshifts. See
905: Figure \ref{fig:bbspec}). This estimate for the host-galaxy is higher
906: than on average found for GRB host galaxies, but not unknown
907: \citep[eg][]{levan2006:apj647:471}.  However, localised extinction or
908: a line-of-sight through an edge-on galaxy could easily explain the
909: somewhat larger than average host extinction.
910: 
911: The reduced $\chi^2$ is in all cases $\lesssim 1$ (for redshifts above
912: 4, it becomes much less than 1), resulting in a lower limit on the
913: redshift of $z \gtrsim 2$.  A Galactic-type extinction curve
914: results in a steeper nIR slope than measured here, and
915:   provides a worse fit. The same gradual nIR slope also implies that
916: host-galaxy extinction should be present, and cannot be explained by a
917: Lyman break alone.  We further looked at the resultant extinction
918: value for small changes ($\pm 0.1$) in our assumed spectral index,
919: which results in changes of about 0.1 magnitude in \ebv (the value for
920: \chisqred, however, also increases considerably doing so). Finally,
921: from our best fit we find that the \nuc\ cannot be located directly
922: below the X-ray frequencies, but is restricted to $\sim 10^{16}$ Hz.
923: 
924: An \ebv\ of $0.7$ with an SMC-like extinction curve would imply an
925: \nh\ of $3.2 \cdot 10^{22}$ \citep{martin1989:aa215:219} in the host
926: galaxy.  Correcting for an assumed SMC-metallicity of $1/4$ Solar and
927: adjusting for the redshift of $z = 2$, this turns out to be not
928: measurable above the estimated Galactic extinction in our data, and
929: indeed, no excess extinction is measured past 300 seconds.
930: 
931: 
932: We can obtain a lower limit on the fluence and burst energy from the
933: BAT spectrum, which results in $5.9 \cdot 10^6$ erg cm$^{-2}$ and $5.6
934: \cdot 10^{52}$ erg, respectively. Both values are in the 15 -- 150 keV
935: energy range in the observers frame.  With a lower limit of $z \gtrsim
936: 2$, the relation found by Amati
937: \citep{amati2002:aa390:81,amati2006:mnras372:233} gives $\Ep
938: \gtrsim 76$ keV in the observers frame, while we find $\Ep \gtrsim
939: 170$ keV from the spectral fit (Table \ref{table:gammaray-spectra},
940: after conversion of the cut-off energy to \Ep).  Note that the latter
941: uses a power law with an exponential cut-off, not the Band function.
942: Using the previous fit values with a Band function, fixing $\beta =
943: -2$, and extrapolating outside the BAT spectral range (0.001 --
944: $10^{4}$ keV), we obtain an estimate of the isotropic energy of
945: $E_{\mathrm{iso}} \approx 2.3 \cdot 10^{53} \mathrm{erg}$, which is
946: almost four times larger than the 15 -- 150 keV estimate.  Using the
947: Amati relation, this would put the peak energy at $\sim 150$ keV in
948: the observer frame, assuming $z~\approx~2$.  This redshift estimate
949: therefore agrees reasonably well with the results expected from the
950: Amati relation, and indicates $z~\approx~2$ is a likely good estimate
951: for the redshift of this burst.
952: 
953: 
954: \section{Summary} \label{section:summary}
955: 
956: We have presented here a comprehensive analysis of the data available
957: on GRB\,050716.  The most remarkable feature of the early emission is
958: the spectral change, clearly visible in the X-rays, which can possibly
959: be traced back to the gamma-ray emission.  The temporal break
960: coincident with the spectral break points to the peak-flux frequency
961: passing through the X-rays at this time.  In addition, the X-ray
962: absorption before the break appears to be higher than after.  While
963: this could be an artificial result related to the break, care has been
964: taken to eliminate instrumental effects and the result appears to be
965: significant, possibly indicating ionisiation of surrounding
966:   medium.
967: 
968: At later times, our multi-wavelength data of \thisgrb\ shows a fairly
969: standard afterglow: the X-ray light curve and spectrum indicate $p =
970: 2$, with no evidence for a break related to the broadening of the jet
971: outflow.  The optical data are rather limited, partly because the
972: optical afterglow is relatively faint: host-galaxy extinction with
973: \av\ = 1.5 -- 2.1 can account for the obtained
974: upper limits.
975: Though no  spectroscopic redshift has  been obtained for  \thisgrb, we
976: obtain a  lower limit of  $z \gtrsim 2$, which  appears to  be in
977: agreement with the Amati relation.
978: 
979: 
980: 
981: \section*{Acknowledgements}
982: 
983: We thank the referee for a careful reading of the manuscript, which
984: improved it overall.
985: ER, JPO, KLP, APB, AJL and NRT acknowledge financial support from
986: PPARC. The  authors like  to thank Simon  Vaughan for  additional help
987: with  parts of  the  data analysis.  This  work is  also supported  at
988: Pennsylvania State  University (PSU)  by NASA contract  NAS5-00136 and
989: NASA grant  NNG05GF43G, and at  the Osservatorio Astronomico  di Brera
990: (OAB) by  funding from  ASI on grant  number I/R/093/04.   "The United
991: Kingdom Infrared  Telescope is operated by the  Joint Astronomy Centre
992: on  behalf  of  the  U.K.   Particle Physics  and  Astronomy  Research
993: Council."  This  publication makes use  of data products from  the Two
994: Micron All Sky  Survey, which is a joint project  of the University of
995: Massachusetts    and   the    Infrared    Processing   and    Analysis
996: Center/California  Institute  of Technology,  funded  by the  National
997: Aeronautics  and   Space  Administration  and   the  National  Science
998: Foundation. We gratefully appreciate  the contributions of all members
999: of the \swift\ team.
1000: 
1001: 
1002: \bibliographystyle{mn2e}
1003: \bibliography{references}
1004: 
1005: 
1006: \label{lastpage}
1007: 
1008: 
1009: \end{document}
1010: