1: \documentclass[12pt,preprint]{aastex}
2: \newcommand{\totalnum}{72}
3: \newcommand{\ldnum}{65}
4: \newcommand{\sdnum}{7}
5: \newcommand{\opticalagnum}{47} % 25 w/o OAG
6: \newcommand{\Xoverlapnum}{14}
7: \newcommand{\Xoverlapexp}{6.7}
8: \newcommand{\Xmontecarlo}{0.5\%}
9: \newcommand{\Xmedianprob}{6.9\%}
10:
11: \newcommand{\Emedianarea}{1.6\%}
12: \newcommand{\EmedianareaOAG}{1.6\%}
13: \newcommand{\Eexpected}{1.3}
14: \newcommand{\EexpectedOAG}{0.82}
15:
16: \newcommand{\Emedianareawcut}{1.4\%}
17: \newcommand{\EmedianareawcutOAG}{1.3\%}
18: \newcommand{\Eexpectedwcut}{1.2}
19: \newcommand{\EexpectedwcutOAG}{0.73}
20:
21: \newcommand{\Imedianarea}{1.0\%}
22: \newcommand{\ImedianOAG}{1.0\%}
23: \newcommand{\Iexpected}{0.86}
24: \newcommand{\IexpectedOAG}{0.54}
25:
26: \newcommand{\Imedianareawcut}{0.9\%}
27: \newcommand{\ImedianOAGwcut}{0.9\%}
28: \newcommand{\Iexpectedwcut}{0.79}
29: \newcommand{\IexpectedOAGwcut}{0.50}
30:
31: \shorttitle{GRBs and Host Galaxies}
32: \shortauthors{Cobb & Bailyn}
33:
34: \begin{document}
35:
36: \title{Connecting GRBs and galaxies: the probability of chance coincidence\altaffilmark{1}}
37:
38: \author{Bethany E. Cobb\altaffilmark{2} and Charles D. Bailyn\altaffilmark{2}}
39: \email{cobb@astro.yale.edu}
40:
41: \altaffiltext{1}{Using observations from the Small and Moderate Aperture Research Telescope System (SMARTS) consortium, which operates 4 telescopes at the
42: Cerro Tololo Inter-American Observatory}
43:
44: \altaffiltext{2}{Department of Astronomy, Yale University, P.O. Box 208101, New Haven, CT 06520}
45:
46: \begin{abstract}
47: Studies of GRB host galaxies are crucial to understanding GRBs.
48: However, since they are identified by the superposition
49: in the plane of the sky of a GRB afterglow and a galaxy there is always a
50: possibility that an association represents a chance
51: alignment, rather than a physical connection.
52: We examine a uniform sample of \totalnum\ GRB fields to explore
53: the probability of chance superpositions. There is
54: typically a $\sim1$\% chance that an optical afterglow
55: will coincide with a galaxy by chance.
56: While spurious host galaxy detections will, therefore, be rare, the possibility must be
57: considered when examining individual GRB/host galaxy examples.
58: It is also tempting to use the large and uniform collection of X-ray afterglow positions
59: to search for GRB-associated galaxies. However, we find
60: that approximately half of the 14 superpositions in our sample are likely to
61: occur by chance, so in the case of GRBs localized only by an X-ray afterglow, even
62: statistical studies are suspect.
63: \end{abstract}
64:
65: \keywords{gamma rays: bursts}
66:
67: \section{Introduction}
68: From the earliest associations between gamma-ray bursts (GRBs) and host galaxies, burst environments have revealed many
69: important clues to the nature of GRB progenitors \citep[e.g.][]{Sokolov+01,Bloom+02,Fynbo+03,LeFloch+03}. As \textit{Swift}
70: \citep{Gehrels+04} accurately localizes
71: hundreds of GRBs, statistical studies of GRB host galaxies are becoming increasingly critical to
72: understanding GRB formation mechanisms \citep[e.g.][]{Christensen+04,Wainwright+07}.
73:
74: A GRB host galaxy is identified by the superposition in the plane of the sky of the GRB afterglow (AG) and a galaxy.
75: With only visual associations between afterglows and galaxies used to identify GRB host galaxies, the possibility always
76: exists for an incorrect association.
77: Some afterglow and galaxy associations may be chance superpositions, with the galaxy either a
78: foreground or background galaxy that is not physically associated with the GRB.
79: These mistaken associations could cause confusion when analyzing data on GRB environments.
80: When dealing with a statistical sample of GRBs, rather than a single case-study, results will only
81: be impacted if a relatively significant number of the host galaxies have been mistakenly identified.
82: Understanding the likelihood of galaxy mis-identification in a given sample of GRBs is, therefore, imperative when
83: conducting such studies.
84: However, even a single mistaken association might generate confusion if the combination of
85: GRB and galaxy characteristics are anomalous.
86: For example, the potentially paradigm-shifting object GRB 060614 \citep[e.g.][]{Cobb+06,DellaValle+06,Gal-Yam+06,SX06},
87: a burst identified as low-redshift for which no SN was detected, could
88: be a typical object of no special interest if the burst's purported host galaxy is actually just a random galaxy
89: along the line-of-sight to the GRB.
90:
91: When an optical afterglow is detected, afterglow localization is precise to significantly less than an arcsecond.
92: The superposition between a galaxy and an optical afterglow is, therefore,
93: generally taken as proof that the galaxy in question is the host galaxy of the GRB.
94: Since the optical afterglow position can be determined to within sub-pixel accuracy,
95: the probability of the optical afterglow falling randomly on any galaxy-covered pixel
96: is determined by dividing the number of pixels covered by galaxies in each field by
97: the total number of pixels in that field.
98: If only an X-ray afterglow is detected, the GRB can only be localized to within an error region of radius $\sim2$ arcseconds.
99: If a galaxy is detected within this error
100: region it is generally assumed to be the host galaxy of the GRB, although confusion can arise when two or more sources
101: are detected within or around a single X-ray afterglow error region \citep[e.g.][]{Ferrero+07}.
102:
103: In this paper we investigate a uniform set of \totalnum\ GRB fields in order to examine the general
104: issue of associations between GRBs and host galaxies. In \S 2, we describe our data and analysis techniques.
105: In \S 3, the probability of chance coincidence between optical afterglows
106: and galaxies is determined by measuring the fractional area covered by galaxies.
107: In \S 4, localizations based on X-ray afterglows are considered by measuring the probability
108: of galaxies falling within randomly placed X-ray error regions. Afterglow-associated galaxies are compared
109: to field galaxies in \S 5. We discuss our results in \S 6 along with some
110: strategies for recognizing false hosts and conclude in \S 7.
111:
112: \section{Data}
113: Our I-band optical images were obtained using the ANDICAM instrument mounted on the 1.3m telescope at Cerro Tololo Inter-American
114: Observatory.\footnote{http://www.astronomy.ohio-state.edu/ANDICAM}
115: This telescope is operated as part of the Small and Moderate Aperture Research
116: Telescope System (SMARTS) consortium.\footnote{http://www.astro.yale.edu/smarts}
117:
118: \subsection{GRB Selection}
119: The \totalnum\ bursts included in these analyses were, in general, selected based on their observability from CTIO, limited
120: to those bursts with declination $\lesssim +35\degr$. Observations were also limited by each burst's right
121: ascension because all observations began within days post-burst.
122: Bursts were occasionally not observed due to telescope scheduling time limitations.
123: Both long-soft bursts (\ldnum) and short-hard bursts (\sdnum) were observed.
124: Due to large and uncertain values of galactic reddening near the galactic plane, no GRBs with galactic latitude
125: $\left|b\right|<10\degr$ are included in this sample.
126:
127: The redshifts, or redshift upper limits, of 37 bursts in our sample have been determined through either
128: optical afterglow observations or observations of possible host galaxies.
129: If only the 21 redshifts that are measured directly from GRB afterglow absorption are considered, then the median redshift is $z=2.3$.
130: For the entire sample of bursts, the median redshift is $z=1.3$ with a range from 0.089 to 6.6.
131: The median redshift is understandably reduced when including GRB redshifts obtained from associated galaxies,
132: because galaxies are generally easier to detect and observe spectrographically when at low redshift.
133:
134: The GRBs included in this sample are listed in Table 1.
135:
136: \subsection{Observations and Data Reduction}
137: The nightly data set for each burst consisted of 6 individual 360-second I-band observations, taken at slightly offset telescope positions.
138: Each individual image was reduced in the typical manner, with bias and dark subtraction, flat fielding,
139: and cosmic ray removal using the L.A. Cosmic program\footnote{http://www.astro.yale.edu/dokkum/lacosmic/} \citep{vanDokkum01}.
140: The individual images were median combined with maximum pixel rejection to produce a source-free initial fringe correction image.
141: This background image was scaled and subtracted from the individual images, which were then aligned and combined.
142: This combined image was used to produce a source mask and a new fringe correction image was
143: produced by recombining the original images -- only this time masking the sources in each image,
144: rather than using minmax rejection. This new background was scaled and subtracted from the individual images,
145: which were then aligned and combined to produce the final nightly science frame.
146:
147: For each GRB, between 1 and 11 usable nightly images were obtained (the mode being 4 images).
148: Particularly shallow images or nights with relatively poor seeing were excluded. Some images taken at early times post-burst
149: were also excluded as they contained optical afterglow light. All usable nightly images were aligned and combined
150: to produce a final deep frame for each GRB. The frame edges were cropped so that each final frame is of
151: uniform depth over the entire field. Field size varied slightly but is typically $5\farcm6\times5\farcm4$.
152: A few fields contained saturated stars with
153: diffraction spikes; the area immediately surrounding these stars is excluding from all of the following analyses.
154:
155: Secondary standard stars in each field were photometrically calibrated by comparison, on photometric nights,
156: with Landolt standard stars \citep{Landolt92}. Photometric calibration is typically
157: accurate to 0.05 magnitudes. No photometric observations
158: were obtained for 4 GRBs and those fields are calibrated using USNO-B1.0 I2
159: magnitudes \citep{Monet+03}, with typical calibration errors of 0.2 magnitudes.
160: The reddening corrected \citep{Schlegel+98} $3\sigma$ I-band limiting magnitudes for point sources in these images ranges from
161: I=21.2 mag to 23.3 mag, with a median of 22.4 mag. All magnitudes are given in the Vega system.
162:
163: The fields are astrometrically calibrated using USNO-B1.0 stars, with statistical error of $<0.2"$. The
164: CCD pixel scale is 0.37"/pixel, so that the astrometry is accurate to within a single pixel.
165: X-ray afterglow coordinates and 90\% confidence error radii for all bursts are taken from \cite{Butler07}.
166: Error region radii vary from $0\farcs5$ to $6\farcs9$, with a median of $1\farcs9$.
167:
168: The data are summarized in Table 1.
169:
170: \subsection{Galaxy Detection}
171: All objects in the field were cataloged using SExtractor \citep{BA96} with a 2$\sigma$ detection threshold.
172: The SExtractor parameter CLASS\_STAR was then used to distinguish the galaxies from the stars in the field. This parameter
173: is determined by a neural network and ranges from CLASS\_STAR=0.0 for extended objects to
174: CLASS\_STAR=1.0 for point-like objects. We consider objects with CLASS\_STAR$>0.8$
175: to be stars and these objects are not included in the following analyses.
176: This value is chosen to avoid contaminating the sample with stars that are mistakenly
177: classified as galaxies. Adopting a less strict value results in more galaxy identifications.
178: Defining galaxies to have CLASS\_STAR$<0.9$, for example, increases
179: chance coincidence probabilities by $\sim10$\%.
180: In crowded starfields, SExtractor's ability to differentiate between stars and galaxies is reduced.
181: This effect is minimized in our analyses, however, as none of the fields considered here are within $10\degr$ of the galactic plane.
182: Galaxy I-band magnitudes are given by MAG\_AUTO, with zeropoints appropriately
183: adjusted for each field. These magnitudes are generally accurate to within a few tenths of a
184: magnitude.
185:
186: Only galaxies detected by SExtractor are considered in these analyses. At
187: the position of a few afterglows there are low significance galaxies
188: that are not identified by SExtractor.
189: For consistency, these visual identifications are not included.
190:
191: Since our observations are limited to galaxies brighter than $\sim23$ mag, we would not expect
192: to be able to detect all the potential host galaxies in our images.
193: Galaxies go undetected primarily because of their large redshifts, though dwarf galaxies
194: might be missed even at low redshifts.
195: Examination of galaxies brighter than the median limiting magnitude of our sample ($I<22.4$)
196: in the VVDS-CDFS galaxy catalog \citep{LeFevre+04} suggests that nearly all of the galaxies detected in our
197: fields are likely to be at a redshift lower than $z=1.5$ and about 65\% have redshift $z\leq0.7$.
198:
199: To understand our ability to detect GRB host galaxies, it would be useful to know
200: the characteristics of a typical GRB host. In pre-\textit{Swift} GRB samples, however,
201: galaxies associated with GRBs have been noted to be a rather
202: heterogeneous group \citep[e.g.][]{Conselice+05,LeFloch+03} covering a wide range of galaxy types and absolute
203: magnitudes ($M_B\sim-16$ to $-21$) \citep{Sokolov+01,Conselice+05,Wainwright+07},
204: thought there is a tendency for GRBs to occur in faint galaxies \citep{Jaunsen+03,LeFloch+03,Fruchter+06}.
205: Assuming a typical galaxy color of $B-I=2$ \citep{Fukugita+95}, the brightest host galaxies could have
206: have $I=-23$ and might be detected in our images to $z\sim1$. Observed galaxy brightness would, of course,
207: depend on the necessary $k$-correction, which depends strongly on galaxy morphology.
208: The star-forming galaxies of long-duration bursts would be favored for detection at higher redshifts because
209: of their strong rest-frame UV emission. Unfortunately, redshift limits for detecting the dimmest host galaxies
210: are much more severe. These dwarf hosts may be common in the local universe, but many could not
211: be detected in our images to even moderate redshifts given our magnitude limits.
212:
213: \section{Optical Afterglow and Galaxy Coincidence}
214: There is no exact projected distance at which a GRB and a galaxy become associated,
215: and galaxies lack clear ``edges''. Therefore, we calculate the probability
216: of a chance optical afterglow/galaxy association below using two slightly different definitions of
217: galaxy area.
218:
219: \subsection{ISOAREAF\_IMAGE area}
220: One measure of the pixel area of each galaxy is given by the isophotal SExtractor parameter ISOAREAF\_IMAGE.
221: This value disregards the lower significance outskirts of each galaxy, so it is a conservative estimate of galaxy area.
222: If an optical afterglow were to fall within the ISOAREAF\_IMAGE area of a galaxy,
223: that galaxy would be regarded as a strong host-galaxy candidate.
224: This excludes cases in which GRBs fall at the outskirts of galaxies -- which are bound to occur.
225: Hence these figures serve strictly as a lower limit.
226:
227: The ISOAREAF\_IMAGE area covered by galaxies in each field ranges from 0.2\% to 3.5\% (see Figure 1 and Table 1), with a median of \Imedianarea\
228: (see Table 2).
229: Summing the overlap probability over all \totalnum\ fields yields an expectation of \Iexpected\ observed galaxy/optical afterglow
230: coincidences -- including even those fields for which no optical afterglow detection was made.
231: For the \opticalagnum\ fields in which optical afterglow was detected from either short- or long-duration bursts,
232: there is an expectation of \IexpectedOAG\ observed galaxy/optical afterglow coincidences.
233: In fact, this sample contains 7 such coincidences (GRBs 050416a, 050724, 050826, 060505, 060614, 061021 and 061121).
234: Based on a Monte Carlo simulations, the chance of having randomly observed that many coincidences is less than 0.004\%.
235: These results are summarized in Table 2.
236:
237: Since the total galaxy-covered area in each field depends on limiting magnitude, which varies from field to field, it
238: is useful to consider a more homogeneous selection of galaxies. As deeper observations
239: are obtained, more galaxies will be detected. Therefore, deeper imaging increases
240: the chance for a correct identification, but also increases the chance
241: for spurious associations because the fractional area covered by galaxies is increased.
242: By imposing a magnitude cutoff, therefore, we produce
243: a minimum value for the possibility of chance superposition.
244: For the homogeneous sample, we limit our analysis to galaxies with Galactic extinction corrected magnitudes brighter than $I=21.5$.
245: This value is determined by producing a magnitude histogram of all the galaxies being used in this analysis, with bin
246: size of half a magnitude, and selecting the midvalue of the bin containing the maximum number of galaxies (see Figure 2).
247: Beyond this magnitude, our galaxy sample is significantly affected by
248: incompleteness. In individual fields, this galaxy completeness ``turnover" value ranges from 20 to 22, with
249: a median and mode of 21.5. In the fields that are incomplete at 21.5 mag, there will be ``missing'' galaxies.
250: This means that the probability of chance superposition with a $I\leq21.5$ galaxy will be slightly underestimated.
251:
252: Not surprisingly, the exclusion of all $I>21.5$ galaxies reduces the field coverage of galaxies by only a small amount (see
253: Table 2). The most significant change is that only 4 of the 7 observed optical afterglow/galaxy associations in this sample
254: occur with galaxies brighter than 21.5 magnitudes (GRBs 050724, 050826, 060505, 061021). However, Monte Carlo
255: simulations indicate that the chance of having observed these 4 at random in the \opticalagnum\ fields in
256: which optical afterglow was detected is $<1$\%.
257: So even in this reduced sample, there remains a meaningful overdensity of galaxy/optical afterglow coincidences.
258: However, it is plausible that one or more of these associations could be a coincidence.
259:
260: \subsection{Ellipse Area}
261: Without fully understanding how all GRBs are produced, the exact placement of GRBs within their galaxies cannot
262: be accurately predicted. Long-duration GRBs, for example, may require the kind
263: of rapid star formation regions often found in galactic spiral arms \citep{Conselice+05,Fruchter+06}, while short-duration GRBs may
264: favor the fringes of galaxies if they are formed by the mergers of compact remnants \citep{Bloom+06a,Berger+07b}.
265: We have, therefore, been somewhat too restrictive in assuming that a galaxy will only
266: be identified as a host if the GRB occurs within the galaxy's ISOAREAF\_IMAGE area.
267:
268: To consider the more general situation of a optical afterglow at a given observed position
269: relative to a nearby galaxy, we define galaxy area as an ellipse having major and minor axes of
270: length $v\times$A\_IMAGE and $v\times$B\_IMAGE, where A\_IMAGE and B\_IMAGE
271: are SExtractor parameters that represent the maximum and minimum spatial \textit{rms} of each
272: galaxy profile and $v$ is a simple scaling factor. In general,
273: an ellipse with $v=3$ is visually coincident with the extent of the galaxy (see Figure 3).
274: When the shape of each galaxy is defined in this way, galaxy area is equal to $v^2\pi\times$A\_IMAGE$\times$B\_IMAGE.
275:
276: In Figure 4, we plot the probability of chance alignment for an afterglow contained
277: within a galaxy ellipse with scale factor $v$. The $v$ values of the 7 GRBs with optical afterglow/galaxy coincidences are marked with arrows.
278: The probability of chance alignment rises quadratically with $v$
279: until $v$ is so large that the galaxies significantly overlap with one another. We
280: plot both the entire sample of galaxies and only those galaxies with $I\leq21.5$.
281: Excluding the dimmer galaxies only changes the random overlap probability by a few tenths
282: of a percent at large $v$. On the right hand side of the graph, the number of expected chance coincidences in our sample
283: of \totalnum\ GRBs is shown.
284:
285: Setting the scale factor to $v=3$, we use the ellipse galaxy area to
286: repeat the analyzes that were done in \S3.1 with the ISOAREAF\_IMAGE area. These results are
287: shown in Table 2. For individual galaxies the ellipse area is generally larger than the ISOAREAF\_IMAGE area
288: so using the ellipse area results in a slightly increase probability of chance coincidence.
289:
290: \subsection{Comparing Short- and Long-Duration GRB Fields}
291: A comparison of short- and long-duration fields is of interest, although
292: the sample of short duration bursts in this sample is limited to only \sdnum\ fields (versus
293: \ldnum\ long-duration bursts). The galaxy area coverage spans a similar range of values between the two samples.
294: The median area (ISOAREAF\_IMAGE area) covered by all galaxies in short-GRB fields ($\sim0.009$) is only slightly lower than the median
295: area covered in long-GRB fields ($\sim0.01$). Since short bursts may occur further outside
296: their galaxies than long-duration bursts, the galaxy area over which a true association
297: would be physically plausible may be larger for short bursts than for long bursts.
298:
299: In this sample, three short-GRBs had detected optical afterglows, but there is only
300: one short-burst galaxy/optical afterglow association.
301: This association frequency (1 of 3) is somewhat larger than that of the long-GRBs (6 of 44), but this may
302: be an artifact of the limited short-burst sample.
303: While the exact nature of short bursts is not yet clear, if short bursts are associated with early-type galaxies, they
304: may be preferentially located in local galaxy clusters \citep[e.g.][]{Pedersen+05,Bloom+06a}.
305: In that case, a higher frequency of associations might be expected for short bursts than long bursts.
306: However, the correlation between short bursts and clusters is not yet confirmed.
307: The observations presented here are consistent with recent evidence that suggests that not all short bursts
308: are limited to local clusters \citep{Berger+07a,Berger+07b}.
309:
310: \section{X-ray Afterglow and Galaxy Coincidence}
311: For a non-negligible population of \textit{Swift} GRBs (nearly 50\%), no optical afterglow is detected.
312: Underluminous optical afterglows and heavy line-of-sight extinction may account for many ``dark'' bursts.
313: An optical afterglow might also go undetected due to observing constraints, such
314: as the relative position of the sun or the moon to the burst's coordinates or poor weather conditions
315: at optimum observing sites. Regardless of the reason, such a non-detection generally
316: means that the burst will only by localized to within a few arcseconds by XRT observations of the burst's X-ray afterglow.
317: XRT observations comprise a dataset that is significantly more uniform than deep, ground-based optical afterglow observations that,
318: by necessity, are obtained from a large number of different instruments at varying times post-burst.
319: This homogeneity, combined with the fact that the X-ray localized dataset contains nearly double
320: the number of optically localized GRBs, makes it tempting to analyze GRB host galaxies based exclusively
321: on X-ray AG positions. We, therefore, examine our data to determine how significantly such
322: an analysis might be impacted by the presence of falsely identified hosts.
323:
324: Each X-ray afterglow error region of the GRBs in this sample was examined for coincident galaxies,
325: and \Xoverlapnum\ are found to overlap with one or more SExtracted galaxies.\footnote{The X-ray afterglow
326: error region of GRB 060505 overlaps with 2 galaxies.} Any overlap between
327: the circle defined by the X-ray afterglow error region and a galaxy ellipse with $v=3$ is considered to be a coincidence
328: (see Figure 3).
329: This definition of galaxy area is used instead of the ISOAREAF\_IMAGE area because an
330: ellipse defines a clear ``edge'' to the galaxy while SExtractor does not output the exact
331: boundaries of a galaxy.
332:
333: We then investigated the probability that one or more galaxies will fall within any random
334: region that is the same size as a burst's X-ray afterglow error region. For each GRB field, 500 random positions were selected
335: and each position was assigned a region with a radius equivalent to the X-ray afterglow error radius of the corresponding GRB.
336: All region/galaxy overlaps were then counted. Only a handful of overlaps occur in
337: some fields, while in others more than a quarter of all random regions contain a galaxy.
338: The median overlap probability is \Xmedianprob. The probability
339: of one or more galaxies overlapping any region the size of the burst's X-ray afterglow error region
340: is given in column 8 of Table 1.
341: The overlap probability in each field strongly depends on both the radius of the random regions
342: and the galaxy population density.
343: The burst with the largest X-ray error region is GRB 050412,
344: with a radius of $6\farcs9$. In this field, any randomly placed region of that size has a 32\% chance of
345: overlapping with a galaxy.
346: Note that the density of the galaxy population is dependent on
347: image depth, as shallower images will contain fewer dim galaxies. Chance probabilities will, therefore,
348: increase with image depth.
349: Figure 5 shows a histogram of the probability of region/galaxy overlap in the \totalnum\ GRB fields examined.
350: Arrows on the graph indicate the fields in which the X-ray afterglow error region did coincide
351: with one or more galaxies.
352:
353: The expected number of observed overlaps created by chance in this GRB sample is \Xoverlapexp, which is obtained by summing the overlap probability over
354: all fields. This is significantly less than the \Xoverlapnum\ actually observed.
355: Using a Monte Carlo simulation (see Figure 6), the probability of having observed \Xoverlapnum\
356: overlaps in this sample is found to be only \Xmontecarlo.
357: Thus, there is a clear over-density of galaxies in GRB X-ray afterglow error regions.
358: But the galaxies associated with the X-ray afterglow error regions do not have to be the true host galaxies of the GRBs.
359: In fact, of the X-ray afterglow error regions in this sample that contain a galaxy, it is likely that
360: approximately half contain a galaxy that is not associated with the GRB.
361: This does not appear to be an artifact of images in our sample with large X-ray afterglow error region radii.
362: When we consider only fields with small error radii, $\leq2"$, we observed 5 overlaps when only 2 are expected.
363:
364: Approximately 250 \textit{Swift} bursts have been detected and well-localized by their X-ray afterglows using the XRT. Based simply
365: on the median probability of overlap in these fields, there is only a $10^{-8}$ chance that \textit{no} ``false hosts'' have been detected.
366: In fact, if our sample represents a typical distribution then there could be over 20 such detections
367: in the entire \textit{Swift} sample. Improving the limiting magnitudes in each field would only serve to significantly
368: increase the number of galaxy detections. Indeed, as image depth increases
369: the detection of one or several galaxies within a given X-ray afterglow error region is inevitable.
370: Clearly, caution is required when identifying a galaxy in a X-ray afterglow error region
371: as the host galaxy of the GRB. Note, in particular, that relatively few optical afterglow have been
372: associated with short-duration bursts, yet many claims as to the nature of short-duration
373: bursts have recently been made on the basis of their X-ray afterglow and galaxy associations \citep[e.g.][]{Pedersen+05,Bloom+06a}.
374: The likelihood of misidentified hosts is, therefore, of particular concern when considering short bursts.
375:
376:
377: \section{Field Galaxies versus Afterglow-Associated Galaxies}
378: Having collected the observable properties of all the galaxies in these \totalnum\ fields, we
379: can determine if the galaxies associated with either X-ray or optical afterglows represent
380: an anomalous sample of the field galaxies. The observable properties considered are magnitude, pixel area (either ISOAREAF\_IMAGE
381: area or ellipse area) and ellipticity ($e$), where $e\equiv1$-(B\_IMAGE/A\_IMAGE). These observable properties
382: are, of course, a function of intrinsic size, shape, brightness and redshift.
383:
384: Figure 7a shows galaxy ellipse area, with $v=3$, versus magnitude. Monte Carlo simulations show that
385: the galaxies associated with X-ray afterglow error regions are a typical sample of all
386: the galaxies in the field over this dimension. This result is not surprising, since nearly
387: half these galaxies could actually be just field galaxies, rather than GRB-associated galaxies.
388: The galaxies associated with optical afterglow, however, do appear slightly different than the field
389: galaxies at the $\sim2\sigma$ level. This seems to be due to the fact that these galaxies
390: are somewhat brighter than typical field galaxies: $\sim30$\% of optical-afterglow-associated galaxies
391: have $I<18$ mag, while this is true of only $\sim3$\% of field galaxies.
392: To test this, we split the entire sample of galaxies into three magnitude bins:
393: bright: $I<18$, intermediate: $18\leq I<22$ and faint: $I\geq22$. Of the 7 galaxies associated with optical afterglow,
394: 2 are bright, 3 are intermediate and 2 are faint. We then repeat the Monte Carlo
395: simulation but require that for each run the 7 galaxies randomly picked from the entire
396: sample include 2 bright, 3 intermediate and 2 faint galaxies.
397: With the addition of this magnitude selection, the optical afterglow-associated galaxy sample
398: becomes indistinguishable from the field galaxies.
399:
400: Similar analyses are performed for galaxy ISOAREAF\_IMAGE area versus magnitude,
401: ellipticity versus magnitude, galaxy ellipse area versus ellipticity and
402: galaxy ISOAREAF\_IMAGE area versus ellipticity (see Figure 7b-e). The
403: X-ray afterglow error regions are a typical sample of all the galaxies in the field over these dimensions.
404: The optical afterglow-associated galaxies are atypical on the $\sim2\sigma$ level when comparing
405: ISOAREAF\_IMAGE area versus magnitude and ellipticity versus magnitude. Again, this seems to
406: be due to the fact that these galaxies are somewhat brighter than the average field galaxies.
407: %Since 6 out of 7 of these galaxies are associated with long-duration GRBs, the
408: %fact that long-duration GRB host generally have strong ongoing star-formation may account for
409: %their relative brightness when compared with the field galaxies in general.
410:
411: \section{Discussion}
412: Unlike the fairly uniform images considered here, most observations of GRB fields are obtained
413: with a wide range of depth and resolution. Improving depth and resolution
414: will result in the detection of more galaxies, in which case the values calculated
415: here serve only as a lower limit to the probability of chance association.
416: While increased resolution may seem to make the ``edge'' of a galaxy more defined,
417: we would caution that knowing the exact optical extent of the galaxy
418: is not helpful for identifying a GRB's host without first fully understanding
419: how GRBs trace light.
420:
421: An additional complication is lensing, which will always occur
422: when a background GRB is observed through a foreground galaxy.
423: Lensing will increase the brightness of the GRB and its afterglow
424: and shift the observed position of the GRB's afterglow relative to the lensing
425: galaxy \citep{Paczynski86,Mao92,Grossman+Nowak94}.
426: The magnification produced by this lensing, of course, depends strongly on galaxy mass,
427: observer to lens distance, lens to source distance, and the extent of the alignment.
428: This lensing effect actually increases the likelihood of detecting a chance association
429: because it increases the observed brightness of a GRB's optical afterglow.
430: Host galaxy/optical afterglow associations, by definition, require the detection of an optical afterglow,
431: and brighter optical afterglows are more likely to be detected than dim afterglows.
432:
433: If pseudo-redshifts derived from GRB luminosity relationships eventually become accurate enough to use
434: in evaluating afterglow/galaxy coincidences, lensing would present a complication because lensed
435: GRBs might not follow expected luminosity relationships.
436: The Amati relationship \citep{Amati+02,Amati06}, for example, correlates gamma-ray spectral peak energy with a burst's isotropic equivalent
437: radiated energy. From measurements of peak spectral energy and gamma-ray flux, therefore, a
438: GRB's redshift can be estimated. Lensing should not alter the gamma-ray burst spectral shape so that
439: peak energy remains unchanged. The increase in gamma-ray flux produced by lensing, however, would result
440: in an underestimated burst distance ($z\propto$ luminosity/flux). This is problematic for
441: detecting inconsistent GRB and galaxy redshifts because lensing will move the
442: pseudo-redshift of the GRB toward the redshift of the foreground galaxy.
443:
444: Treating the lensing galaxy as a single isothermal sphere, the strongest lensing
445: effect would be produced by a low-redshift galaxy lying directly along the line-of-sight
446: to a high-redshift GRB (angular offset between source and lens $\lesssim0\farcs3$).
447: Random chance, however, favors less exact alignments. At greater angular separations ($>0\farcs3$, where magnification only depends weakly
448: on angular separation),
449: only very large galaxies ($>$L$_*$ or $\sigma>100$km/s) can produce more than a factor of 2 in magnification of the
450: source. Consider a scenario in which a 0.1L$_*$ galaxy at z=0.3 lenses a GRB at z=3, with an angular
451: separation of $0\farcs5$. This produces only a 15\% increase in brightness.
452: Lensing, therefore, only becomes a concern when a burst aligns very closely with a $>$L$_*$ galaxy.
453: However, there is only a $~10^{-4}$ chance that the line-of-sight to a GRB at
454: redshift $z=5$ would pass within $0\farcs3$ of the gravitational center of
455: a $>$L$_*$ galaxy. For bursts that occur at lower redshift, this probability
456: only decreases. Hence, while lensing will always occur with chance alignments, the effects
457: will generally be of little consequence.
458:
459: Currently, there are only a few situations in which any kind of mis-match between a GRB and proposed host is likely
460: to be noted. One example of this would be when no SN is detected in a low-redshift galaxy associated with a long-duration GRB.
461: A second example would be the association of a long-duration burst with an elliptical galaxy with little to no ongoing star-formation.
462: In either event, it is difficult to make a firm conclusion because of the possibility
463: that not all GRBs are produced in the canonical fashion. One can question the cause of a GRB as easily as question its redshift.
464:
465: To avoid contaminating host galaxy samples with false-hosts, it would be useful to have a
466: way to separate true and false hosts. Unfortunately, no exact method of doing so exists.
467: Some strategies are:
468: \begin{itemize}
469: \item \textbf{Gamma-ray luminosity indicators:}
470: While luminosity indicators \citep[e.g.][]{Amati+02,Ghirlanda+04,Liang+Zhang05} cannot currently
471: produce accurate redshifts for individual bursts, they may eventually be able to
472: be used in such a manner.
473: In that case, however, a discrepancy could either indicate a chance superposition
474: or a very unusual GRB, so that interpretation of such a mismatch
475: is unclear. Lensing effects could also invalidate this method.
476:
477: \item \textbf{Detection (or non-detection) of SNe:}
478: This is only meaningful if long-duration bursts are all formed in core-collapse SNe
479: and are similar in brightness to GRB 980425/SN 1998bw, which is assumed to be the archetypal GRB-related SN.
480: This is further limited to GRBs that occur at relatively low redshifts ($z\lesssim0.7$).
481: Host-galaxy extinction is also a complicating factor that may
482: obscure an underlying SN.
483:
484: \item \textbf{Optical afterglow spectral absorption features:}
485: A GRB must occur behind the absorbing material, so line
486: detection actually only gives a lower limit on redshift (though
487: this is generally taken to be the GRB's redshift).
488: Redshifts derived in this manner can be compared to
489: redshifts measurements of the proposed host galaxy after
490: the afterglow has faded. Only a very bright afterglow
491: could possibly show absorption features independent
492: of the superimposed galaxy, however. The exception is high redshift bursts whose
493: redshifts can be accurately determined by detection of the Lyman-$\alpha$ break.
494: Alternatively,
495: an afterglow with no absorption features might
496: be assumed to be at low redshift because
497: of the complete lack of absorption features
498: along the line-of-sight.
499:
500: \item \textbf{Visual detection of true host galaxy:}
501: In deep images, the true host galaxy may be detectable
502: through the foreground galaxy. An example of this situation is
503: GRB 060912a which occurred in a $z=0.937$ galaxy that
504: overlaps with the outskirts of a large $z=0.0936$ galaxy \citep{Levan+07}.
505:
506: \item \textbf{X-ray afterglow spectral emission lines:}
507: Such emission lines have been suggested to be produced by material associated
508: with the GRB (outflows, disks, etc.) so they could serve as accurate
509: redshift indicators. This is very unproven method, however,
510: because line identification is inexact, reported lines
511: have been at only low significance, and few GRB X-ray
512: spectra contain these lines \citep[e.g.][]{Sako+05}.
513:
514: \item \textbf{Strong lensing of GRB afterglows:}
515: This would be extremely unlikely for a single foreground galaxy,
516: (as shown in Appendix A) but could potentially be caused by a foreground galaxy cluster.
517: \end{itemize}
518:
519: \section{Conclusions}
520: We have studied the galaxy population surrounding a sample of \totalnum\ GRBs.
521: Typically 1\% of the sky near the positions of GRBs is covered
522: by galaxies with $I\leq21.5$.
523: With $\sim125$ \textit{Swift} GRBs with detected optical afterglows, the probability
524: that no chance alignments have been detected is $(0.99)^{125} = 28$\%. Indeed,
525: approximately 1 superposition between the GRB and a foreground galaxy
526: is expected. While it is possible that no such chance alignments have yet been observed, as \textit{Swift} detects
527: more and more GRBs with optical afterglows the likelihood of such an event only increases.
528: While most GRB/galaxy associations noted in the literature are almost certainly
529: correct, caution is required when making sweeping conclusions from only one or two
530: GRB/host galaxy associations.
531:
532: We have also considered galaxies associated with X-ray afterglows.
533: Over 250 X-ray afterglows, with typical error radii of $\sim2''$, have been detected by \textit{Swift}.
534: These numbers guarantee that some galaxies associated with X-ray afterglow will be falsely identified as hosts.
535: In fact, approximately half of the 14 X-ray afterglow error region/galaxy coincidences in our sample may exist only by chance.
536: Even with large samples, therefore, using X-ray afterglow-identified host galaxies to draw conclusions about GRBs causes confusion.
537:
538: \acknowledgments
539: We thank SMARTS observers D. Gonzalez, J. Espinoza and A. Pasten for their dedication
540: and S. Tourtellotte for assistance with optical data reduction.
541: We are also grateful to P. G. van Dokkum and P. Natarajan for
542: useful discussions about galaxies and lensing.
543: This work is supported by NSF Graduate Fellowship DGE0202738
544: to BEC and NSF/AST grants 0407063 and 0707627 and \textit{Swift}
545: grant NNG05GM63G to CDB.
546:
547: \begin{thebibliography}{}
548: \bibitem[Amati(2006)]{Amati06} Amati, L.\ 2006, \mnras, 372, 233
549: \bibitem[Amati et al.(2002)]{Amati+02} Amati, L., et al.\ 2002, \aap, 390, 81
550: \bibitem[Berger(2006)]{Berger06} Berger, E.\ 2006, GRB Coordinates Network, 5965
551: \bibitem[Berger et al.(2005)]{Berger+05} Berger, E., Gladders, M., \& Oemler, G.\ 2005, GRB Coordinates Network, 3201
552: \bibitem[Berger et al.(2007a)]{Berger+07a} Berger, E., Shin, M.-S., Mulchaey, J.~S., \& Jeltema, T.~E.\ 2007, \apj, 660, 496
553: \bibitem[Berger et al.(2007b)]{Berger+07b} Berger, E., et al.\ 2007, \apj, 664, 1000
554: \bibitem[Berger \& Gladders(2006)]{Berger+Gladders06} Berger, E., \& Gladders, M.\ 2006, GRB Coordinates Network, 5170
555: \bibitem[Berger \& Mulchaey(2005)]{Berger+Mulchaey05} Berger, E., \& Mulchaey, J.\ 2005, GRB Coordinates Network, 3122
556: \bibitem[Bertin \& Arnouts(1996)]{BA96} Bertin, E., \& Arnouts, S.\ 1996, \aaps, 117, 393
557: \bibitem[Bloom et al.(2002)]{Bloom+02} Bloom, J.~S., Kulkarni, S.~R., \& Djorgovski, S.~G.\ 2002, \aj, 123, 1111
558: \bibitem[Bloom et al.(2006a)]{Bloom+06a} Bloom, J.~S., et al.\ 2006, \apj, 638, 354
559: \bibitem[Bloom et al.(2006b)]{Bloom+06b} Bloom, J.~S., Perley, D.~A., \& Chen, H.~W.\ 2006, GRB Coordinates Network, 5826
560: \bibitem[Butler(2007)]{Butler07} Butler, N.~R.\ 2007, \aj, 133, 1027
561: \bibitem[Castro-Tirado et al.(2006)]{Castro-Tirado+06} Castro-Tirado, A.~J., et al. \ 2006, GRB Coordinates Network, 5218
562: \bibitem[Cenko et al.(2005)]{Cenko+05} Cenko, S.~B., Kulkarni, S.~R., Gal-Yam, A., \& Berger, E.\ 2005, GRB Coordinates Network, 3542
563: \bibitem[Chen et al.(2005)]{Chen+05} Chen, H.-W., et al. \ 2005, GRB Coordinates Network, 3709
564: \bibitem[Chen et al.(2007)]{Chen+07} Chen, H.-W., et al. \ 2007, GRB Coordinates Network, 6217
565: \bibitem[Christensen et al.(2004)]{Christensen+04} Christensen, L., Hjorth, J., \& Gorosabel, J.\ 2004, \aap, 425, 913
566: \bibitem[Cobb et al.(2006)]{Cobb+06} Cobb, B.~E., Bailyn, C.~D., van Dokkum, P.~G., \& Natarajan, P.\ 2006, \apjl, 651, L85
567: \bibitem[Conselice et al.(2005)]{Conselice+05} Conselice, C.~J., et al.\ 2005, \apj, 633, 29
568: \bibitem[D'Elia et al.(2006)]{DElia+06} D'Elia et al. \ 2006, GRB Coordinates Network, 5637
569: \bibitem[Della Valle et al.(2006)]{DellaValle+06} Della Valle, M., et al.\ 2006, \nat, 444, 1050
570: \bibitem[de Pasquale et al.(2007)]{dePasquale+07} de Pasquale, M., et al.\ 2007, \mnras, 377, 1638
571: \bibitem[Ferrero et al.(2007)]{Ferrero+07} Ferrero, P., et al.\ 2007, \aj, 134, 2118
572: \bibitem[Foley et al.(2005)]{Foley+05} Foley, R.~J., et al. \ 2005, GRB Coordinates Network, 3483
573: \bibitem[Fruchter et al.(2006)]{Fruchter+06} Fruchter, A.~S., et al.\ 2006, \nat, 441, 463
574: \bibitem[Fugazza et al.(2006)]{Fugazza+06} Fugazza, D., et al.\ 2006, GRB Coordinates Network, 5513
575: \bibitem[Fukugita et al.(1995)]{Fukugita+95} Fukugita, M., Shimasaku, K., \& Ichikawa, T.\ 1995, \pasp, 107, 945
576: \bibitem[Fynbo et al.(2003)]{Fynbo+03} Fynbo, J.~P.~U., et al.\ 2003, \aap, 406, L63
577: \bibitem[Fynbo et al.(2005)]{Fynbo+05} Fynbo, J.~P.~U., et al.\ 2005, GRB Coordinates Network, 3176
578: \bibitem[Fynbo et al.(2006)]{Fynbo+06} Fynbo, J.~P.~U., et al.\ 2006, GRB Coordinates Network, 5651
579: \bibitem[Fynbo et al.(2007)]{Fynbo+07} Fynbo, J.~P.~U., et al. \ 2007, GRB Coordinates Network, 6759
580: \bibitem[Gal-Yam et al.(2006)]{Gal-Yam+06} Gal-Yam, A., et al.\ 2006, \nat, 444, 1053
581: \bibitem[Gehrels et al.(2004)]{Gehrels+04} Gehrels, N., et al.\ 2004, \apj, 611, 1005
582: \bibitem[Ghirlanda et al.(2004)]{Ghirlanda+04} Ghirlanda, G., Ghisellini, G., \& Lazzati, D.\ 2004, \apj, 616, 331
583: \bibitem[Grossman \& Nowak(1994)]{Grossman+Nowak94} Grossman, S.~A., \& Nowak, M.~A.\ 1994, \apj, 435, 548
584: \bibitem[Halpern \& Mirabal(2006)]{Halpern+Mirabal06} Halpern, J.~P., \& Mirabal, N.\ 2006, GRB Coordinates Network, 5982
585: \bibitem[Jakobsson et al.(2006a)]{Jakobsson+06a} Jakobsson, P., et al.\ 2006, GRB Coordinates Network, 5319
586: \bibitem[Jakobsson et al.(2006b)]{Jakobsson+06b} Jakobsson, P., et al. \ 2006, GRB Coordinates Network, 5320
587: \bibitem[Jakobsson et al.(2006c)]{Jakobsson+06c} Jakobsson, P. et al. \ 2006, GRB Coordinates Network, 5617
588: \bibitem[Jakobsson et al.(2007a)]{Jakobsson+07a} Jakobsson, P, \ 2007, GRB Coordinates Network, 6997
589: \bibitem[Jakobsson et al.(2007b)]{Jakobsson+07b} Jakobsson, P., et al.\ 2007, GRB Coordinates Network, 6398
590: \bibitem[Jaunsen et al.(2003)]{Jaunsen+03} Jaunsen, A.~O., et al.\ 2003, \aap, 402, 125
591: \bibitem[Jaunsen et al.(2007)]{Jaunsen+07} Jaunsen, A.~O., et al. \ 2007, GRB Coordinates Network, 6202
592: \bibitem[Kelson \& Berger(2005)]{Kelson+Berger05} Kelson, D., \& Berger, E.\ 2005, GRB Coordinates Network, 3101
593: \bibitem[Landolt(1992)]{Landolt92} Landolt, A.~U.\ 1992, \aj, 104, 340
594: \bibitem[Le F{\`e}vre et al.(2004)]{LeFevre+04} Le F{\`e}vre, O., et al.\ 2004, \aap, 428, 1043
595: \bibitem[Le Floc'h et al.(2003)]{LeFloch+03} Le Floc'h, E., et al.\ 2003, \aap, 400, 499
596: \bibitem[Levan et al.(2007)]{Levan+07} Levan, A.~J., et al.\ 2007, \mnras, 378, 1439
597: \bibitem[Liang \& Zhang(2005)]{Liang+Zhang05} Liang, E., \& Zhang, B.\ 2005, \apj, 633, 611
598: \bibitem[Mao(1992)]{Mao92} Mao, S.\ 1992, \apjl, 389, L41
599: \bibitem[Monet et al.(2003)]{Monet+03} Monet, D.~G., et al.\ 2003, \aj, 125, 984
600: \bibitem[Oates et al.(2006)]{Oates+06} Oates, S.~R., et al.\ 2006, \mnras, 372, 327
601: \bibitem[Ofek et al.(2006)]{Ofek+06} Ofek, E.~O., et al. \ 2006, GRB Coordinates Network, 5123
602: \bibitem[Osip et al.(2006)]{Osip+06} Osip, D., Chen, H.-W., \& Prochaska, J.~X.\ 2006, GRB Coordinates Network, 5715
603: \bibitem[Paczynski(1986)]{Paczynski86} Paczynski, B.\ 1986, \apjl, 308, L43
604: \bibitem[Pedersen et al.(2005)]{Pedersen+05} Pedersen, K., et al.\ 2005, \apjl, 634, L17
605: \bibitem[Pellizza et al.(2006)]{Pellizza+06} Pellizza, L.~J., et al.\ 2006, \aap, 459, L5
606: \bibitem[Piranomonte et al.(2006a)]{Piranomonte+06a} Piranomonte, S., et al.\ 2006, GRB Coordinates Network, 4520
607: \bibitem[Piranomonte et al.(2006b)]{Piranomonte+06b} Piranomonte, S., et al.\ 2006, GRB Coordinates Network, 4583
608: \bibitem[Price et al.(2006)]{Price+06} Price, P.~A., Berger, E., \& Fox, D.~B.\ 2006, GRB Coordinates Network, 5275
609: \bibitem[Prochaska et al.(2005a)]{Prochaska+05a} Prochaska, J.~X., et al. \ 2005, GRB Coordinates Network, 3390
610: \bibitem[Prochaska et al.(2005b)]{Prochaska+05b} Prochaska, J.~X., et al. \ 2005, GRB Coordinates Network, 3700
611: \bibitem[Roming et al.(2006)]{Roming+06} Roming, P.~W.~A., et al.\ 2006, \apj, 651, 985
612: \bibitem[Sako et al.(2005)]{Sako+05} Sako, M., Harrison, F.~A., \& Rutledge, R.~E.\ 2005, \apj, 623, 973
613: \bibitem[Schlegel et al.(1998)]{Schlegel+98} Schlegel, D.~J., Finkbeiner, D.~P., \& Davis, M.\ 1998, \apj, 500, 525
614: \bibitem[Schaefer \& Xiao(2006)]{SX06} Schaefer, B.~E., \& Xiao, L.\ 2006, (astro-ph/0608441)
615: \bibitem[Sokolov et al.(2001)]{Sokolov+01} Sokolov, V.~V., et al.\ 2001, \aap, 372, 438
616: \bibitem[Thoene et al.(2006a)]{Thoene+06a} Thoene, C.~C., et al.\ 2006, GRB Coordinates Network, 5373
617: \bibitem[Thoene et al.(2006b)]{Thoene+06b} Thoene, C.~C., Fynbo, J.~P.~U., \& Jakobsson, P.\ 2006, GRB Coordinates Network, 5747
618: \bibitem[Thoene et al.(2007)]{Thoene+07} Thoene, C.~C., Perley, D.~A., \& Bloom, J.~S.\ 2007, GRB Coordinates Network, 6663
619: \bibitem[van Dokkum(2001)]{vanDokkum01} van Dokkum, P.~G.\ 2001, \pasp, 113, 1420
620: \bibitem[Wainwright et al.(2007)]{Wainwright+07} Wainwright, C., Berger, E., \& Penprase, B.~E.\ 2007, \apj, 657, 367
621: \end{thebibliography}
622:
623: \clearpage
624: \input{tab1}
625: \clearpage
626:
627: \begin{deluxetable}{lllllll}
628: \tabletypesize{\small}
629: \tablecolumns{7}
630: \tablewidth{0pc}
631: \tablecaption{Optical afterglow chance coincidence probabilities}
632: \tablehead{
633: \colhead{} &
634: \colhead{Mag.} &
635: \colhead{} &
636: \colhead{Median} &
637: \colhead{\# expected} &
638: \colhead{\# observed} &
639: \colhead{Prob. of observed} \\
640: \colhead{Galaxy area} &
641: \colhead{cutoff} &
642: \colhead{Sample\tablenotemark{a}} &
643: \colhead{prob.\tablenotemark{b}} &
644: \colhead{coincidences} &
645: \colhead{coincidences} &
646: \colhead{coincidences\tablenotemark{c}}
647: }
648: \startdata
649: ISOAREAF\_IMAGE & none & total& \Imedianarea & \Iexpected & 7& 0.004\% \\
650: & none & OAG & \ImedianOAG & \IexpectedOAG & 7& 0.0002\% \\
651: & $I\leq21.5$ & total& \Imedianareawcut& \Iexpectedwcut & 4& 0.98\% \\
652: & $I\leq21.5$ & OAG & \ImedianOAGwcut & \IexpectedOAGwcut& 4& 0.18\% \\
653: Ellipse ($v=3$) & none & total & \Emedianarea & \Eexpected & 7& 0.04\% \\
654: & none & OAG & \EmedianareaOAG & \EexpectedOAG & 7& 0.002\% \\
655: & $I\leq21.5$ & total & \Emedianareawcut& \Eexpectedwcut & 4& 3.2\% \\
656: & $I\leq21.5$ & OAG & \EmedianareawcutOAG& \EexpectedwcutOAG& 4& 0.68\% \\
657: \enddata
658: \tablenotetext{a}{The sample of GRB fields used. ``Total'' refers to all 72 GRBs. ``OAG'' refers to the 47 GRBs
659: for which optical afterglows were detected.}
660: \tablenotetext{b}{The median probability of a chance coincidence between an optical afterglow and a galaxy.}
661: \tablenotetext{c}{Probability of having observed the given number of optical afterglow/galaxy coincidences
662: at random given the expected number of random coincidences, calculated using Monte Carlo simulations.}
663: \end{deluxetable}
664:
665: \clearpage
666:
667: \begin{figure}
668: \includegraphics[width=1\textwidth]{f1.eps}
669: \caption{Histogram of the probability of any pixel in a GRB field falling
670: inside of a galaxy (\textit{solid line} - median of \Imedianarea) or inside a
671: galaxy with magnitude $I\leq21.5$ (\textit{dotted line} - median of \Imedianareawcut).
672: The area of each galaxy is given by the SExtractor output ISOAREAF\_IMAGE.
673: }
674: \end{figure}
675:
676: \begin{figure}
677: \includegraphics[width=1\textwidth]{f2.eps}
678: \caption{Magnitude histogram of all galaxies in this sample. The magnitude cut
679: of I=21.5 is chosen as the midvalue of the bin containing the maximum number
680: of galaxies. The sample of dimmer galaxies is significantly affected by
681: incompleteness.
682: }
683: \end{figure}
684:
685: \begin{figure}
686: \includegraphics[width=1\textwidth]{f3.eps}
687: \caption{Sections of each of the 14 fields in which the X-ray afterglow error region
688: overlaps a galaxy. The X-ray afterglow error regions are shown as green circles.
689: SExtracted galaxies are indicated as yellow ellipses. Each ellipse has major and minor axes
690: of 3$\times$A\_IMAGE and 3$\times$B\_IMAGE. For bursts with detected optical afterglow,
691: the optical afterglow position is marked by a red x.
692: }
693: \end{figure}
694:
695: \begin{figure}
696: \includegraphics[width=1\textwidth]{f4.eps}
697: \caption{
698: The probability of chance overlaps for \textit{(solid squares)} all galaxies or \textit{(open circles)} galaxies
699: with $I\leq21.5$ for increasing values of $v$, where $v$ is the scale factor used in defining galaxy size.
700: On the right is shown the corresponding expectation value for the 72 GRBs in this sample.
701: The $v$ values of the 7 GRBs with optical afterglow/galaxy coincidences are marked with arrows.
702: }
703: \end{figure}
704:
705:
706: \begin{figure}
707: \includegraphics[width=1\textwidth]{f5.eps}
708: \caption{Histogram of the probability of a random X-ray error region/galaxy overlap
709: in the \totalnum\ GRB fields examined. The median probability is \Xmedianprob. The
710: arrows indicate the actual probability of each field in which the X-ray
711: afterglow error region did coincide with one or more galaxies.
712: The arrows over $>0.3$ indicate fields having actual probabilities of 32\% and 57\%.
713: A Kolmogorov-Smirnov (K-S) test on these data gives $P\sim0.4$, which indicates that
714: the distribution of probabilities in the 14 fields containing X-ray afterglow-associated galaxies
715: is not significantly different from the distribution of the entire sample.
716: }
717: \end{figure}
718:
719:
720: \begin{figure}
721: \includegraphics[width=1\textwidth]{f6.eps}
722: \caption{Monte Carlo simulation of X-ray afterglow error region/galaxy coincidences within our \totalnum\ GRB fields.
723: Approximately 7 overlaps are expected, while \Xoverlapnum\ are observed (indicated
724: by the arrow). The probability of observing \Xoverlapnum\ or more overlaps
725: is only \Xmontecarlo.
726: }
727: \end{figure}
728:
729: \begin{figure}
730: \includegraphics[width=1\textwidth]{f7.eps}
731: \caption{Observed galaxy properties:
732: \textit{(a)} galaxy ellipse area versus magnitude, \textit{(b)} ISOAREAF\_IMAGE area versus magnitude,
733: \textit{(c)} ellipticity versus magnitude, \textit{(d)} galaxy ellipse area versus ellipticity
734: and \textit{(e)} galaxy ISOAREAF\_IMAGE area versus ellipticity.
735: The plot axes limits are chosen to emphasize the afterglow-associated galaxies and exclude some of the galaxies
736: in this sample. While not plotted, the galaxies are included in the Monte Carlo
737: simulations. The galaxies associated with
738: X-ray afterglows (\textit{filled and open squares}) are a typical
739: sample of all the field galaxies (\textit{gray points}).
740: The galaxies associated with both X-ray and optical afterglows (\textit{open squares})
741: differ from the field galaxy population at the $\sim2\sigma$ level
742: for all relationships involving magnitude (\textit{a,b,c}).}
743: \end{figure}
744:
745: \end{document}
746: ~
747: