1:
2: % Draft 5
3: % 2/9/06
4:
5: \documentclass[preprint,12pt]{aastex}
6:
7: \begin{document}
8:
9: \shorttitle{{\it Swift}'s Burst Sensitivity}
10:
11: \title{Post-Launch Analysis of {\it Swift}'s Gamma-Ray Burst Detection Sensitivity}
12: %
13: \author{David~L.~Band\altaffilmark{1,2}}
14: %
15: \altaffiltext{1}{Code 661, NASA/Goddard Space Flight
16: Center, Greenbelt, MD 20771}
17: %
18: \altaffiltext{2}{Joint Center for Astrophysics, Physics Department,
19: University of Maryland, Baltimore County, 1000 Hilltop Circle,
20: Baltimore, MD 21250}
21: %
22: \email{dband@milkyway.gsfc.nasa.gov}
23:
24: \begin{abstract}
25: The dependence of {\it Swift}'s detection sensitivity on a burst's
26: temporal and spectral properties shapes the detected burst
27: population. Using simplified models of the detector hardware and the
28: burst trigger system I find that {\it Swift} is more sensitive to
29: long, soft bursts than {\it CGRO}'s BATSE, a reference detector
30: because of the large burst database it accumulated. Thus {\it Swift}
31: has increased sensitivity in the parameter space region into which
32: time dilation and spectral redshifting shift high redshift bursts.
33: \end{abstract}
34:
35: \keywords{gamma-rays: bursts --- instrumentation:
36: detectors}
37:
38: \section{Introduction}
39:
40: The gamma-ray bursts which {\it Swift} detects depend on the
41: physical properties of the Burst Alert Telescope (BAT), {\it
42: Swift}'s gamma-ray detector, and on the BAT's triggering system. The
43: dependence of the BAT's detection sensitivity on a burst's temporal
44: and spectral characteristics shapes the burst population that {\it
45: Swift} studies. While the {\it Swift} observations are revealing a
46: wealth of new phenomena through the study of individual bursts, we
47: also want to relate these bursts to the bursts studied by previous
48: missions. In particular, because of the large and statistically
49: well-defined sample of more than 2700 bursts it collected, the Burst
50: and Transient Source Experiment (BATSE) on the {\it Compton
51: Gamma-Ray Observatory (CGRO)} is the reference detector to which
52: subsequent detectors such as the BAT are compared. Therefore, in
53: this work I use the BAT's on-orbit calibration to gain a deeper
54: understanding of the detector's sensitivity to different types of
55: bursts. The insight from this study will help the design of future
56: missions, such as {\it EXIST} (Grindlay 2005).
57:
58: The BAT detects approximately 100 bursts per year. Compared
59: to BATSE's burst sample, a higher fraction of the bursts
60: the BAT detects are long duration ($T_{90}>2$~s) bursts
61: (see Figure~1), although the few short duration bursts that
62: have been detected have been particularly revelatory.
63: Understanding the observed duration distribution is a goal
64: of this work.
65:
66: Understanding the BAT's burst detection sensitivity requires the
67: sequence of events on board the spacecraft. Bursts are detected by
68: the BAT (Barthelmy et al. 2005), a large field-of-view
69: (FOV---1.4~sr), 15--150~keV coded mask detector with a 5200~cm$^{2}$
70: cadmium-zinc-teluride (CZT) detector plane. The BAT's detector
71: plane is sensitive to higher energy photons, but burst imaging and
72: spectroscopy have an effective high energy cutoff of $\sim 150$~keV.
73: Once the BAT detects a burst, the spacecraft slews autonomously
74: (within operational constraints) to place the burst location in the
75: center of the much smaller FOVs of the X-ray Ray Telescope
76: (XRT---Osborne et al. 2005) and the coaligned UV-Optical Telescope
77: (UVOT---Roming et al. 2005). Thus the BAT's trigger system
78: determines which bursts are detected, although the other two
79: detectors' performance and operational constraints affect whether
80: the afterglow is followed by {\it Swift} immediately after the
81: burst.
82:
83: The BAT's flight software detects bursts on board in two steps
84: (Fenimore et al. 2003, 2004; Palmer et al. 2004). A rate trigger
85: monitors the count rate from the CZT detectors for a statistically
86: significant increase; the BAT's rate trigger is complex, testing the
87: count rate from the detector plane (and subsections of the plane) on
88: timescales ranging from 0.004~s to 32~s using a variety of different
89: background estimates. Once a rate trigger occurs, an image is formed
90: through the coded mask system. To maximize the signal-to-noise ratio
91: of the counts used for imaging, the software varies: the energy
92: band; the `foreground' time period over which the counts are
93: accumulated; and the `background' time periods before, and perhaps
94: after, the foreground time period from which the background during
95: the foreground period is estimated. A number of images may be formed
96: using different foreground time periods before the significance
97: exceeds a detection threshold or the software concludes that a
98: detection is not possible. Only if a new statistically significant
99: point source is evident in an image is a burst considered to be
100: detected. Periodically (once every 64 and 320~s, and when the
101: spacecraft changes its orientation---Palmer et al. 2004; McLean et
102: al. 2004) an image is formed and checked for a new point source even
103: without a rate trigger. Because a burst detection requires the
104: imaging of a new point source, the threshold for the rate trigger is
105: set to permit many false positives that are subsequently rejected by
106: the imaging step. Consequently, the imaging step is usually the most
107: restrictive step and therefore determines the BAT's burst
108: sensitivity.
109:
110: The BAT's trigger system is complex with many triggers and
111: background estimates (Fenimore et al. 2003, 2004; Palmer et al.
112: 2004). The flight software turns triggers on and off based on the
113: computational load. While diagnostics are telemetered to the
114: ground, the telemetry stream cannot provide sufficient data to
115: reproduce on the ground the behavior of the trigger system precisely
116: at all times. The complexity of the trigger system maximizes the
117: BAT's sensitivity---{\it Swift}'s design goal---at the expense of
118: making an accurate determination of this sensitivity at a given time
119: very difficult if not impossible. In particular, the BAT achieves
120: high image sensitivity by accumulating counts over much longer
121: timescales than BATSE did, making the trigger sensitive to the
122: details of the burst lightcurve. In contrast to burst spectra whose
123: shapes are adequately described by two or three parameters,
124: lightcurves differ greatly from burst to burst when considered on
125: timescales greater than a second, and cannot be parameterized for
126: sensitivity calculations by only a few parameters.
127:
128: Despite all these caveats, I develop a semi-quantitative
129: understanding of the burst populations that the BAT detects using a
130: simplified model of the BAT's trigger system that captures the
131: essential features of the trigger. I assume that imaging is the
132: most restrictive step of the trigger, and therefore ignore the
133: complexity of the rate trigger. I use a signal-to-noise ratio
134: estimate of the sensitivity of the imaging step. This calculation
135: captures the fundamental dependence of the BAT's sensitivity on a
136: burst's hardness and duration. Calculations with greater
137: verisimilitude would result from applying the BAT trigger code to
138: simulated burst data; the BAT team maintains a working copy of the
139: trigger code incorporated in the flight software. Before launch such
140: simulations were run to verify the performance of the BAT and its
141: flight software (Fenimore et al. 2004) and to determine the trigger
142: system's initial settings (McLean et al. 2004). The simulated bursts
143: should be accurate representations of the bursts the BAT might
144: detect.
145:
146: I separate my evaluation of burst sensitivity into the dependencies
147: on the burst's spectrum and lightcurve. This is an approximation,
148: since a burst's spectrum changes during the burst---usually the
149: spectrum softens with time (Ford et al. 1995)---and the lightcurve
150: depends on the energy band---usually individual pulses and the
151: duration of the entire burst are shorter at high energy. After first
152: providing the formulae for the detection significance for rate and
153: image triggers (\S 2.1), I evaluate the BAT's energy-dependent (\S
154: 2.2) and duration-dependent (\S 2.3) burst sensitivity. I use these
155: results to understand the observed burst population (\S 3). While I
156: have discussed the factors affecting the BAT's sensitivity with the
157: members of the BAT instrument team, the conclusions are my own. I
158: use preliminary values for the performance of the instrument and the
159: mission, and thus my calculated sensitivities utilizing a simple
160: model of the BAT trigger should be regarded as illustrative, not
161: definitive.
162:
163: \section{The BAT's Burst Sensitivity}
164:
165: \subsection{Burst Triggers}
166:
167: While they involve very different operations, rate and imaging
168: triggers both analyze the counts accumulated by a burst detector
169: over an energy band $\Delta E$ and accumulation time $\Delta t$; the
170: BAT's flight software analyzes overlapping energy bands (see \S 2.2)
171: and accumulation times (see \S 2.3). For the BAT, the initial rate
172: trigger and the image with a statistically significant point source
173: detection need not use the same $\Delta E$ and $\Delta t$. If a
174: burst is present then the number of observed counts in $\Delta E$
175: and $\Delta t$ is the sum of the counts from the burst $C_s$ and the
176: background $B$.
177:
178: Rate and imaging triggers have similar dependencies on source and
179: background counts, and thus analogous methods can be used to
180: evaluate the resulting sensitivities. Before {\it HETE-II} and the
181: BAT most burst detectors, such as BATSE, used
182: %
183: \begin{equation}
184: S_r = {{C_s}\over \sqrt{B}}
185: \end{equation}
186: %
187: as the detection significance for a rate trigger: the increase in
188: the number of counts over the background is compared to the
189: background's fluctuation scale. For a trigger, $S_r$ must exceed a
190: threshold value. To mitigate difficulties that occur when $B$ is
191: very large or very small, the BAT's rate trigger replaces the
192: background $B$ in the denominator of eq.~1 with a sum $D$ of terms
193: (see eq.~3 of Fenimore et al. 2003). When $B$ is small $D$
194: asymptotes to a constant, while $D$ asymptotes to $B^2$ when $B$ is
195: large, converting the detection criterion from a signal-to-noise
196: ratio to a signal-to-background ratio. For intermediate values of
197: $B$, $D$ is approximately equal to $B$.
198:
199: The significance of a point source in a coded mask image is
200: (G.~Skinner 2005, personal communication)
201: %
202: \begin{equation}
203: S_i = {{f_m C_s}\over \sqrt{C_s+B}}
204: \end{equation}
205: %
206: where the factor $f_m$ compensates for the finite size of
207: the detector pixels relative to the mask elements. One
208: interpretation of this factor of $f_m$ is that the finite
209: size of the detector pixels smears images on the sky,
210: thereby lowering their significance. Skinner (2005,
211: personal communication) finds $f_m=0.73$ for the BAT, which
212: explains why the rate trigger significance is greater than
213: the image significance for the bursts the BAT detects
214: (D.~Palmer 2005, personal communication). The fluctuation
215: level for an image includes the source counts in addition
216: to the background counts (i.e., the denominator in eq.~2 is
217: $\sqrt{C_s+B}$ and not $\sqrt{B}$) because the burst counts
218: are merely background for positions on the sky other than
219: that of the burst (the origin of $C_s$).
220:
221: The rate trigger significance $S_r$ in eq.~1 is applicable
222: to BATSE but not to the BAT's rate trigger for which the
223: significance differs significantly from $S_r$ in eq.~1 for
224: both small and large numbers of background counts. However,
225: imaging almost always is the most restrictive step in the
226: BAT's burst detection algorithm, and therefore I only
227: consider the sensitivity resulting from imaging. In
228: comparing the BATSE and BAT sensitivities, I use eq.~1 for
229: BATSE and eq.~2 for the BAT. I also assume that the BAT
230: flight software successfully finds the `foreground' time
231: period (the time period used for imaging) that maximizes
232: the signal-to-noise ratio, thereby optimizing the image
233: step.
234:
235: In the analysis that follows I first consider the energy dependence
236: of $S_i$ holding the accumulation time $\Delta t$ fixed at 1.024~s
237: (\S 2.2), a background-dominated case that allows me to use the
238: methodology developed for rate triggers (Band 2003). Subsequently I
239: consider the dependence on burst duration (\S 2.3).
240:
241: \subsection{Energy Dependence}
242:
243: In this subsection I assume that $\Delta t$=1.024~s, for
244: which the background dominates the burst counts at
245: threshold (i.e., $B\gg C_s$). I calculate the number of
246: source counts $C_s$ by convolving the burst spectrum with
247: the effective area over the energy band $\Delta E$.
248: Figure~2 shows the current understanding of the BAT's
249: detector efficiency (D.~Hullinger, 2005, personal
250: communication), here defined as the effective area on-axis
251: divided by {\it half} of the area of the detector plane
252: (the detector plane area is divided by two to account for
253: the coded mask); thus this efficiency is equivalent to the
254: efficiency for a detector with half the area and no mask.
255:
256: Ideally the closed cells of the mask (consisting of lead
257: tiles) would be perfectly opaque, but at high energy (above
258: $\sim$100~keV) the optical depth through the lead tiles
259: decreases. Imaging with a coded mask relies on the shadow
260: cast by the closed mask cells. However, if flux leaks
261: through the closed mask cells, then the contrast between
262: the detector pixels that are illuminated by the source and
263: those that are shadowed is reduced; this leakage is
264: equivalent to no flux leaking through the closed mask cells
265: and the detection of less flux by the illuminated detector
266: pixels. The solid curve on Figure~2 is the net detector
267: efficiency, the efficiency for the difference between the
268: fluxes through the open and closed mask elements, which is
269: relevant for imaging. Because the imaging step is the most
270: restrictive part of the BAT's trigger, the net detector
271: efficiency is used to calculate the source counts for the
272: sensitivity.
273:
274: The dashed curve on Figure~2 is the gross detector efficiency, the
275: efficiency for the sum of the fluxes through the open and closed
276: mask elements. The product of the gross detector efficiency, the
277: incident flux, the total detector area, and the fraction of the
278: coded mask that is open (half for the BAT) results in the total
279: count rate. The gross detector efficiency is relevant for the BAT's
280: rate trigger.
281:
282: Thus the net detector efficiency is reduced at the energies where
283: the mask's lead tiles are partially transparent while the gross
284: detector efficiency is increased at the same energies because more
285: source photons reach the detector plane. Note that the rate trigger
286: and imaging use different detector efficiencies and will have
287: somewhat different energy dependencies.
288:
289: CZT has high quantum efficiency below 100~keV. However the optical
290: depth through the mask substrate that supports both the closed and
291: open mask cells results in the roll-off in the detector efficiency
292: at low energy ($<$40~keV). This low energy roll-off reduces the
293: aperture flux (i.e., the cosmic X-ray background), which dominates
294: the total background at low energy, but also decreases the BAT's
295: sensitivity to X-ray Flashes and X-ray rich bursts.
296:
297: The BAT's total on-orbit background rate is approximately
298: $\sim$10~kHz, consistent with pre-launch estimates. At low energies
299: the background is dominated by the aperture flux---the cosmic X-ray
300: background through the mask---while at high energy instrumental
301: background and aperture flux through the BAT's side shields (which
302: become transparent at $\sim 100$~keV) increase the background. The
303: background varies over an orbit, and I use the lowest observed
304: background rates in the energy bands used by the BAT's burst
305: trigger: $\sim$2300~Hz for $\Delta E=$15--25~keV; $\sim$4700~Hz for
306: $\Delta E=$15--50~keV; $\sim$4700~Hz for $\Delta E=$25--100~keV; and
307: $\sim$4700~Hz for $\Delta E=$50--500~keV. The background is near
308: this minimum about half the time; the reduction in sensitivity
309: resulting from higher backgrounds reduces the overall burst
310: detection rate by about 5\%. Note that for $B=4700$ counts in 1~s,
311: $C_s\sim S_{i,{\rm th}}\sqrt{B}/f_m\sim 700$, and thus $C_s \ll B$
312: at a threshold value of $S_{i,{\rm th}}=7$; the assumption that the
313: background dominates is valid.
314:
315: I parameterize the spectrum with the `Band' function (Band
316: et al. 1993) which is a smoothly broken power law with low
317: energy spectral index $\alpha$ ($N(E)\propto E^\alpha$) and
318: high energy spectral index $\beta$ ($N(E)\propto E^\beta$).
319: The characteristic energy is the peak energy $E_p$, the
320: photon energy of the peak of the $E^2N(E)\propto \nu f_\nu$
321: spectrum. The spectrum can be normalized by the flux
322: integrated over a specified energy band, which need not be
323: the same as $\Delta E$. I use $F_T$, the peak flux in the
324: 1--1000~keV band.
325:
326: Figure~3 compares the maximum sensitivity for BATSE (left) and the
327: BAT on-axis (right) for $\Delta t$=1.024~s. The sensitivity is the
328: threshold peak photon flux $F_T$ at which the detector triggers.
329: Because the threshold is expressed in the same units---the flux in
330: the 1--1000~keV band---regardless of $\Delta E$, the sensitivity of
331: different detectors and of different $\Delta E$ for the same
332: detector, can be compared (Band 2003). These thresholds depend on
333: the burst's spectral parameters that determine the shape of the
334: spectrum (not its normalization). The curves on Figure~3 show the
335: threshold flux as a function of $E_p$, holding the low and high
336: energy spectral indices $\alpha$ and $\beta$ fixed. The BAT runs its
337: trigger on four different $\Delta E$ simultaneously ($\Delta
338: E=$15--25~keV, 15--50~keV, 25--100~keV, and 50--500~keV), and the
339: detector sensitivity is the lowest threshold at any given $E_p$,
340: resulting in the scalloping of the BAT sensitivity curves.
341:
342: As can be seen, on-axis the BAT is less sensitive than BATSE's
343: maximum sensitivity for $E_p>100$~keV by a factor of $\sim$1.5, and
344: is more sensitive at lower $E_p$ values, again by a factor of
345: $\sim$1.5, for the same $\Delta t=1.024$~s. As will be discussed
346: below, the BAT's overall sensitivity depends on both its sensitivity
347: at fixed $\Delta t$ and the sensitivity resulting from triggering on
348: multiple values of $\Delta t$.
349:
350: \subsection{Duration Dependence}
351:
352: I now consider the BAT's sensitivity to bursts with
353: different durations. BATSE used a rate trigger with three
354: values of $\Delta t$---0.064, 0.256 and 1.024~s---while
355: after a rate trigger the BAT can form images on a variety
356: of timescales ranging from $0.004$~s to $26$~s. In
357: addition, the BAT forms images every 64 and 320~s without a
358: rate trigger.
359:
360: The relationship between burst duration and the detector
361: accumulation time $\Delta t$ is illustrated by considering a
362: constant flux burst of duration $T$ when the background dominates
363: the source. If $T > \Delta t$, that is, if the flux remains constant
364: over $\Delta t$, then the threshold flux is proportional to $\Delta
365: t^{-1/2}$ (i.e., fainter bursts will be detected as $\Delta t$
366: increases): the number of source counts increases as $\Delta t$, but
367: the square root of the background increases only as $\Delta
368: t^{1/2}$. However, when $T < \Delta t$---the lightcurve is a short
369: spike relative to the accumulation time---then the threshold flux is
370: proportional to $\Delta t^{1/2}$ (i.e., bursts must be brighter to
371: be detected for longer $\Delta t$): the number of source counts
372: remains constant but the square root of the background increases as
373: $\Delta t^{1/2}$.
374:
375: Imaging is the final, determining step of the BAT's
376: trigger, and for short $\Delta t$ the background may not
377: dominate the source counts $C_s$ in the denominator of
378: eq.~2; the addition of $C_s$ to $B$ increases the
379: denominator and therefore decreases $S_i$ relative to a
380: simple rate trigger (eq.~1). Assume that the lightcurve is
381: $P A h(t;T_{90})$ where $P$ is the instantaneous peak flux
382: accumulated over $\Delta E$, and $A$ is the detector area.
383: With a maximum value of one, the lightcurve function
384: $h(t;T_{90})$ parameterizes the lightcurve in terms of
385: duration $T_{90}$. The number of burst counts accumulated
386: in $\Delta t_i$ is therefore $C_s = \int_0^{\Delta t_i}P A
387: h(t;T_{90})dt$ where I assume that $h(t;T_{90})$ peaks
388: within $\Delta t$ and I ignore the issue of the
389: registration of the burst relative to the time bin
390: boundaries (i.e., I assume that $\Delta t$ begins at the
391: beginning of the burst, and ignore the possibility that the
392: fluent part of the burst lightcurve straddles two time
393: bins). Next, let $b$ be the background rate in $\Delta E$
394: for the entire detector; thus the number of background
395: counts is $B=b\Delta t$. For my calculations I use
396: $b=4700$~cts~s$^{-1}$.
397:
398: For a given $T_{90}$ and $\Delta t$ I calculate the threshold value
399: of $P_{th,r}(T_{90};\Delta t)$ for a rate trigger assuming a
400: threshold value of $S_r$ in eq.~1 and for an image trigger
401: $P_{th,i}(T_{90};\Delta t)$ assuming a threshold value of $S_i$ in
402: eq.~2. Note that $S_r$ is applicable to BATSE's rate trigger, but
403: not to the BAT's. When there are multiple accumulation times
404: $\{\Delta t_i\}$, then the resulting threshold peak flux $P_{th}$ is
405: the minimum $P_{th}$ for the different $\Delta t_i$ values at a
406: given $T_{90}$. Since BATSE established a very large statistically
407: homogeneous burst database for $\Delta t$=1.024~s, and many burst
408: distributions are normalized for this value of $\Delta t$, I
409: normalize $P_{th}$ for different $\Delta t$ values and trigger types
410: to the $P_{th,r}$ for a rate trigger with $\Delta t$=1.024~s.
411:
412: Figure~4a uses $h(t;T_{90})$=1 over the duration of the burst, while
413: Figure~4b uses $h(t)=\exp[-t/\tau]$ where $T_{90}=\tau \ln 10$. On
414: both figures the ratio $P_{th}(T_{90} ; \{\Delta
415: t_i\})/P_{th,r}(T_{90} ; \Delta t$=1.024~s) is plotted as a function
416: of $T_{90}$ for rate or image triggers and different sets of $\Delta
417: t$. The dashed curve is for a rate trigger with BATSE's three
418: values of $\Delta t$=0.064, 0.256 and 1.024~s; the plotted ratio is
419: $P_{th,r}(T_{90} ; \{\Delta t_i\}_{\rm BATSE})/P_{th,r}(T_{90} ;
420: \Delta t$=1.024~s). The decrease for durations $T_{90}$ less than
421: $\sim$1~s shows the increase in sensitivity to short duration bursts
422: that resulted from BATSE adding $\Delta t$=0.064 and 0.256~s to
423: $\Delta t$=1.024~s. The solid curve shows the ratio
424: $P_{th,i}(T_{90} ; \{\Delta t_i\}_{\rm BAT})/P_{th,r}(T_{90} ;
425: \Delta t$=1.024~s) for the BAT image trigger with $\Delta t$ ranging
426: from 0.004~s to 26~s. As can be seen, adding $\Delta t$ values both
427: greater than and less than BATSE's set increases the sensitivity to
428: both longer and shorter duration bursts. The increase in
429: sensitivity for short duration bursts is not very great because
430: $C_S\sim B$ (see the denominator of eq.~2); this is an unavoidable
431: feature of the imaging required to localize bursts.
432:
433: Because burst lightcurves have very different shapes, calculating a
434: general detector sensitivity as a function of $T_{90}$ is very
435: difficult. For example, Figures~4a and~b show that the increase in
436: sensitivity for long duration bursts occurs at longer durations for
437: the exponential lightcurves than for flat-top lightcurves. Applying
438: the BAT trigger code to an ensemble of typical observed burst
439: lightcurves (e.g., from BATSE) would provide a better estimate of
440: the sensitivity as a function of duration (see Fenimore et al.
441: 2004).
442:
443: In summary, the longer $\Delta t_i$ values increase significantly
444: BAT's sensitivity to long duration bursts. The BAT's increase in
445: sensitivity to short bursts relative to BATSE is not as great
446: because the number of source counts becomes comparable to the number
447: of background counts.
448:
449: \section{The Resulting Observed Burst Population}
450:
451: The BAT detects mostly long duration bursts, as shown by Figure~1
452: (although the few short bursts have been very revealing). On average
453: long duration bursts are softer than short duration bursts
454: (Kouveliotou et al. 1993). The BAT's detector efficiency shifts its
455: sensitivity to lower energies than BATSE's (Berger et al. [2005]
456: noted this factor), and its use of longer $\Delta t$ values
457: increases its sensitivity to long duration bursts. In addition, most
458: bursts show significant hard-to-soft spectral evolution (Ford et al.
459: 1995), and therefore their low energy emission lasts longer; this is
460: an effect not considered by studying the spectral and temporal
461: dependencies separately. Consequently, a longer accumulation time
462: increases the effectiveness of a lower energy trigger band for long
463: duration bursts.
464:
465: While the BAT's array of accumulation times increases its
466: sensitivity to both very short and long duration bursts relative to
467: BATSE's set of $\Delta t=0.064$, 0.256 and 1.024~s, the increase in
468: sensitivity is much greater for long bursts than for short bursts.
469: Whether BATSE's trigger truncated the duration distribution on the
470: short side was debated (e.g., Lee \& Petrosian 1996); unfortunately,
471: the BAT's relatively small increase in short duration sensitivity
472: (and its lower energy band) make it difficult to determine whether a
473: large population of short duration bursts exists. An analysis of
474: short duration rate triggers that do not result in successful image
475: triggers might address this issue.
476:
477: Combining the results of \S 2.2 and \S 2.3, Figures~5a, b and~c show
478: the ratio of the BAT to BATSE flux thresholds as a function of
479: duration $T_{90}$ and peak energy $E_p$ for three different sets of
480: spectral indices. These figures treat the energy and temporal
481: factors independently. Ratio values less than 1 indicate that the
482: BAT is more sensitive than BATSE. Also shown are a sample of BATSE
483: bursts for which values of $T_{90}$ and $E_p$ are available
484: (Mallozzi et al. 1998); the bursts in this sample provided enough
485: counts for spectral fits. As can be seen, the short, hard bursts are
486: in a region of parameter space where the BAT is less sensitive than
487: BATSE while the BAT is more sensitive to long, soft bursts. The
488: gradient of the contours shows that the BAT detects fewer short,
489: hard burst because its energy band is lower than BATSE's was, and
490: the BAT detects more long, soft bursts both because of its lower
491: energy band and its greater sensitivity to long bursts. This is
492: consistent with the shift in the duration distributions in Figure~1.
493: BAT's greater sensitivity to long duration bursts is consistent with
494: the average fluence of the {\it Swift} bursts being a factor of
495: $\sim2.5$ fainter than the average fluence of the BATSE bursts
496: (T.~Sakamoto, personal communication 2005).
497:
498: When a burst occurs at high redshift, its observed spectrum
499: is redshifted (i.e., becomes softer) and its observed
500: duration is dilated. Thus the burst is shifted towards the
501: parameter space region in which the BAT's sensitivity
502: increase is greatest. However, evolution of the average
503: burst's intrinsic spectrum and duration obviously
504: determines where the burst began, and therefore ends, in
505: parameter space.
506:
507: {\it BeppoSAX} and {\it HETE-II} also detect(ed) and
508: localize(d) bursts by forming images in low energy bands
509: with accumulation times longer than 1~s, and the bursts
510: detected by both detectors are almost exclusively long
511: duration bursts. {\it BeppoSAX} formed images in the
512: 1.8--28~keV band and {\it HETE-II} forms images in the
513: 2--25~keV band ({\it HETE-II} also forms images in a softer
514: band). Thus the same factors that favor the detection of
515: long bursts in the BAT's burst sample are relevant to these
516: two detectors.
517:
518: The burst detection rate depends on the spectral and temporal
519: sensitivities discussed in \S 2.2 and \S 2.3, respectively, and the
520: FOV. The BAT's sensitivity decreases off-axis, first as a result of
521: area foreshortening (i.e., the detector plane is not perpendicular
522: to the direction to an off-axis source) and then because the outer
523: regions of the FOV are only partially coded. In the partially coded
524: region the source flux falls on only part of the detector plane, but
525: the entire detector plane contributes background counts. To reduce
526: the dilution of source counts by background from sections of the
527: detector plane that are not illuminated by a source in the partially
528: coded region of the sky, the BAT detector plane is broken into
529: quadrants; each of the four quadrants, all four pairs of adjoining
530: quadrants and the entire detector plane are treated as independent
531: detectors simultaneously. Thus the burst detection sensitivity
532: varies across the FOV, and fainter bursts will be detected near the
533: center of the FOV, while only bright bursts will be detected near
534: the edges. Using BATSE's burst rate as a function of peak flux
535: (Band 2002) and the BAT's sensitivity across its FOV, compensating
536: for the BAT's different energy and temporal dependencies, and
537: accounting for various operational factors (e.g., the deadtime
538: resulting from slews and SAA passages) results in an estimated burst
539: detection rate that is consistent with the observed detection rate
540: of $\sim 100$~bursts per year.
541:
542: \section{Summary}
543:
544: {\it Swift}'s burst sensitivity depends on the imaging performance
545: of the BAT, {\it Swift}'s coded mask gamma-ray detector. Using a
546: simplified model of the BAT's trigger, my analysis focused on the
547: sensitivity to bursts' temporal and spectral properties separately.
548: Because of the large burst database it accumulated, BATSE is the
549: reference detector to which I compare the BAT.
550:
551: As expected from the detectors' detecting material (the BAT's CZT
552: vs. BATSE's NaI[{\it Tl}]), the BAT's energy band is shifted to
553: lower energies than BATSE's was. Thus, for same accumulation time
554: (e.g., $\Delta t$=1.024~s) the BAT is less sensitive than BATSE for
555: $E_p>$100~keV but is more sensitive for lower $E_p$. Note that the
556: relative sensitivity at fixed $\Delta t$ is only one component of
557: the comparison between detectors.
558:
559: The BAT forms images by accumulating counts on timescales
560: much longer (up to 26~s) than BATSE's rate trigger (up to
561: 1.024~s), increasing the BAT's sensitivity to long duration
562: bursts. Because the number of burst counts is comparable
563: to the number of background counts for short bursts, the
564: BAT's image trigger is not as sensitive to short bursts as
565: a simple rate trigger would be; however, a rate trigger
566: would not localize the bursts. A study of statistically
567: significant rate triggers of short bursts that did not
568: result in statistically significant point sources might
569: determine whether there is a large population of thus far
570: undetected short bursts.
571:
572: The longer accumulation times increase the BAT's
573: sensitivity to long duration bursts, particularly for
574: bursts with a high level of emission over an extended
575: period (as opposed to long duration bursts dominated by a
576: short spike). The BAT detects bursts in a lower energy band
577: than BATSE did, and long duration bursts are softer, on
578: average, than short duration bursts. Consequently the BAT
579: detects long duration bursts preferentially. Spectral
580: redshifting and time dilation of a burst's duration shift
581: high redshift bursts into the parameter region where BAT is
582: more sensitive. The same trigger characteristics explain
583: why {\it BeppoSAX} and {\it HETE-II} also detect(ed) long
584: duration bursts.
585:
586: I emphasize that my semi-analytic calculations use a
587: simplified model of the complex BAT trigger system and my
588: goal is to determine how the BAT's hardware and trigger
589: shape the burst population the BAT detects. My goal is
590: {\it not} to develop an accurate description of the
591: detection threshold, which is very difficult if not
592: impossible given the complex trigger and the time-varying
593: background.
594:
595: \acknowledgments I thank S.~Barthelmy, E.~Fenimore, N.~Gehrels,
596: D.~Hullinger, H.~Krimm, D.~Palmer, A.~Parsons, T.~Sakamoto,
597: G.~Skinner, and J.~Tueller for their assistance and advice, and for
598: the preliminary data about the BAT they made available.
599:
600: %\clearpage
601:
602: \begin{thebibliography}{}
603:
604: \bibitem[a7]{a7}Band, D. 2002, ApJ, 578, 806
605:
606: \bibitem[a6]{a6}Band, D. 2003, ApJ, 588, 945
607:
608: \bibitem[a1]{a1}Band, D., et al. 1993, ApJ, 413, 281
609:
610: \bibitem[a3]{a3}Barthelmy, S., et al. 2005, Space Sci Rev (in
611: press)
612:
613: \bibitem[a31]{a31}Berger, E., et al. 2005, ApJ, 634, 501
614: [astro-ph/0505107]
615:
616: \bibitem[a2]{a2}Burrows, D., et al. 2005, Space Sci Rev (in
617: press)
618:
619: \bibitem[a8]{a8}Fenimore, E. E., Palmer, D., Galassi, M., Tavenner, T.,
620: Barthelmy, S., Gehrels, N., Parsons, A., \& Tueller, J. 2003, ``The
621: Trigger Algorithm for the Burst Alert Telescope on Swift'' in
622: ``Gamma-Ray Burst and Afterglow Astronomy 2001'' AIPC 662, eds.
623: G.~Ricker and R.~Vanderspek (Melville, NY: American Institute of
624: Physics), 491
625:
626: \bibitem[a83]{a83}Fenimore, E. E., et al. 2004, ``Swift's Ability to
627: Detect Gamma-Ray Bursts'' in Proceedings of the 2003 JEMAM
628: Conference, Baltic Astronomy, in press
629:
630: \bibitem[a81]{a81}Ford, L. A., et al. 1995, ApJ, 439, 307
631:
632: \bibitem[a5]{a5}Gehrels, N., et al. 2004, ApJ, 611, 1005
633:
634: \bibitem[a515]{a515}Grindlay, J. 2005, NewAR, 49, 436
635:
636: \bibitem[a52]{a52}Kouveliotou, C., Meegan, C. A., Fishman, G. J.,
637: Bhat, N. P., Briggs,~M.~S., Koshut,~T.~M., Paciesas,~W.~S.,
638: \& Pendleton,~G.~N. 1993, ApJ, 413, 101
639:
640: \bibitem[a51]{a51}Lee, T., \& Petrosian, V. 1996, ApJ, 470, 479
641:
642: \bibitem[a661]{a661}Mallozzi, R., et al. 1998, in Gamma-Ray
643: Bursts, 4th Huntsville Symposium, AIPC 428, eds. C. Meegan, R.
644: Preece and T. Koshut (AIP: Woodbury, NY), 273
645:
646: \bibitem[a53]{a53}McLean, K., Fenimore, E. E., Palmer, D., Barthelmy,~S.,
647: Gehrels,~N., Krimm,~H., Markwardt,~C., \& Parsons~A. 2004, ``Setting
648: the Triggering Threshold on Swift'' in Gamma-Ray Bursts: 30 Years of
649: Discovery, AIP 727, eds. E. E. Fenimore and M. Galassi (Melville,
650: NY: American Institute of Physics), 667
651:
652: \bibitem[a44]{a44}Osborne, J., et al. 2005, proceedings of SPIE 5898,
653: UV, X-Ray and Gamma-Ray Space Instrumentation for Astronomy XIV (in
654: press) [astro-ph/0510489]
655:
656: \bibitem[a9]{a9}Palmer, D. M., et al. 2004 ``The BAT-Swift Science Software''
657: in Gamma-Ray Bursts: 30 Years of Discovery, AIPC 727, eds. E. E.
658: Fenimore and M. Galassi (Melville, NY: American Institute of
659: Physics), 663
660:
661: \bibitem[a4]{a4}Roming, P., et al. 2005, Space Sci Rev, 120, 95
662:
663: \end{thebibliography}
664:
665: \clearpage
666:
667: \begin{figure}
668: \plotone{f1.eps}
669: %\plotone{t90_swift_batse.eps}
670: \caption{$T_{90}$ distribution for {\it Swift} (solid) and BATSE
671: (dashed). An arbitrary normalization was used for the 2041 BATSE
672: bursts. {\it Swift} detects few of the short duration bursts that
673: BATSE detected. The {\it Swift} distribution includes bursts up to
674: mid-December 2005. The three short vertical dashed lines at the top
675: of the plot indicate the $\Delta t$ values used by the BATSE
676: trigger, while the solid line indicates the maximum $\Delta t$ value
677: used by the BAT's trigger system (the minimum value is less than the
678: smallest $T_{90}$ plotted).}
679: \end{figure}
680:
681: \begin{figure}
682: \plotone{f2.eps}
683: %\plotone{swift_eff.eps}
684: \caption{Detector efficiency of the BAT. Both curves are
685: calculated with the area of the detector plane divided by
686: two, accounting for the coded mask. Relevant for imaging,
687: the net detector efficiency (solid curve) is the efficiency
688: for the {\it difference} between the fluxes through the
689: open and closed mask cells. The transparency of the closed
690: mask cells at high energy reduces the source's coded
691: signal. The gross detector efficiency (dashed curve) is
692: the efficiency for the sum of the fluxes through the open
693: and closed mask cells. In this case the mask transparency
694: increases the number of source photons that reach the
695: detector plane, increasing the gross efficiency. }
696: \end{figure}
697:
698: \begin{figure}
699: \plottwo{f3a.eps}{f3b.eps}
700: %\plottwo{batse.eps}{swift_mask.eps}
701: \caption{Maximum
702: detection sensitivity for BATSE's LAD (left) and {\it
703: Swift}'s BAT (right) for $\Delta t=1.024$~s. Solid
704: line---$\alpha = -1$, $\beta = -2$; dashed line---$\alpha =
705: -0.5$, $\beta = -2$; dot-dashed line---$\alpha = -1$,
706: $\beta = -3$. }
707: \end{figure}
708:
709: \clearpage
710:
711: \begin{figure}
712: \figurenum{4} \caption{Ratio of the threshold peak flux for
713: a detector's set of accumulation times $\Delta t$ to the
714: peak flux for $\Delta t$=1.024~s as a function of the burst
715: duration $T_{90}$. The solid curve shows the ratio for the
716: BAT resulting from requiring the detection of a
717: statistically significant source in an image. The dashed
718: curve is the ratio for BATSE's set of $\Delta t$. Panel a:
719: a flat-top burst lightcurve; panel b: an exponential
720: lightcurve.}
721: \end{figure}
722: %\clearpage
723: %\begin{figure}
724: %\figurenum{4a}
725: \plotone{f4a.eps}
726: \centerline{Fig. 4a. ---}
727: %\plotone{flat_top.eps}
728: %\caption{}
729: %\end{figure}
730: \clearpage
731: %\begin{figure}
732: %\figurenum{4b}
733: \plotone{f4b.eps}
734: \centerline{Fig. 4b. ---}
735: %\plotone{exponential.eps}
736: %\caption{}
737: %\end{figure}
738:
739: \clearpage
740:
741: \begin{figure}
742: \figurenum{5} \caption{Contour plot of the ratio of the
743: sensitivities of the BAT and BATSE as a function of $E_p$
744: and $T_{90}$; a ratio less than one indicates that the BAT
745: is more sensitive than BATSE at that particular set of
746: $E_p$ and $T_{90}$. Also plotted are the $E_p$ and $T_{90}$
747: for a set of BATSE bursts with enough counts for spectral
748: fits. The energy and temporal effects were treated
749: separately; differences in the burst lightcurve in
750: different energy bands were not considered. Panel a:
751: $\alpha=$-1 and $\beta=$-2; panel b: $\alpha=$-1 and
752: $\beta=$-3; panel c: $\alpha=$-0.5 and $\beta=$-3. }
753: \end{figure}
754: %\clearpage
755: %\begin{figure}
756: %\figurenum{5a}
757: \plotone{f5a.eps}
758: \centerline{Fig. 5a. ---}
759: %\plotone{Swift_BATSE_sens_comp_1_2.eps}
760: %\caption{}
761: %\end{figure}
762: \clearpage
763: %\begin{figure}
764: %\figurenum{5b}
765: \plotone{f5b.eps}
766: \centerline{Fig. 5b. ---}
767: %\plotone{Swift_BATSE_sens_comp_1_3.eps}
768: %\caption{}
769: %\end{figure}
770: \clearpage
771: %\begin{figure}
772: %\figurenum{5c}
773: \plotone{f5c.eps}
774: \centerline{Fig. 5c. ---}
775: %\plotone{Swift_BATSE_sens_comp_p5_2.eps}
776: %\caption{}
777: %\end{figure}
778:
779: \end{document}
780: