astro-ph0506442/ms.tex
1: \documentclass{emulateapj}
2: %\documentclass[12pt,preprint]{aastex}
3: \shorttitle{Untriggered GRB Afterglows}
4: \shortauthors{Rykoff et al.}
5: 
6: \begin{document}
7: 
8: \title{A Search for Untriggered GRB Afterglows with ROTSE-III}
9: 
10: \author{Rykoff,~E.~S.\altaffilmark{1}, 
11: Aharonian,~F.\altaffilmark{2},
12: Akerlof,~C.~W.\altaffilmark{1},
13: Alatalo,~K.\altaffilmark{1},
14: Ashley,~M.~C.~B.\altaffilmark{3}, 
15: G\"{u}ver, T.\altaffilmark{4},
16: Horns,~D.\altaffilmark{2},
17: Kehoe,~R.~L.\altaffilmark{5},
18: K{\i}z{\i}lo\v{g}lu,~\"{U}.\altaffilmark{6},
19: McKay,~T.~A.\altaffilmark{1},
20: \"{O}zel,~M.\altaffilmark{7},
21: Phillips,~A.\altaffilmark{3}, 
22: Quimby,~R.~M.\altaffilmark{8},
23: Schaefer,~B.~E.\altaffilmark{9}, 
24: Smith,~D.~A.\altaffilmark{1},
25: Swan,~H.~F.\altaffilmark{1},
26: Vestrand,~W.~T.\altaffilmark{10}, 
27: Wheeler,~J.~C.\altaffilmark{8},
28: Wren,~J.\altaffilmark{10}, 
29: Yost,~S.~A.\altaffilmark{1} }
30: 
31: \altaffiltext{1}{University of Michigan, 2477 Randall Laboratory, 450 Church
32:         St., Ann Arbor, MI, 48104, erykoff@umich.edu, akerlof@umich.edu,
33:         kalatalo@umich.edu, tamckay@umich.edu, donaldas@umich.edu,
34:         hswan@umich.edu, sayost@umich.edu}
35: \altaffiltext{2}{Max-Planck-Institut f\"{u}r Kernphysik, Saupfercheckweg 1,
36:         69117 Heidelberg, Germany, Felix.Aharonian@mpi-hd.mpg.de,
37:         horns@mpi-hd.mpg.de}
38: \altaffiltext{3}{School of Physics, Department of Astrophysics and Optics,
39:         University of New South Wales, Sydney, NSW 2052, Australia,
40:         mcba@phys.unsw.edu.au, a.phillips@unsw.edu.au}
41: \altaffiltext{4}{Istanbul University Science Faculty, Department of Astronomy
42:         and Space Sciences, 34119, University-Istanbul, Turkey, 
43:         tolga@istanbul.edu.tr}
44: \altaffiltext{5}{Dept. of Physics, Southern Methodist University, Dallas, TX
45:         75275, kehoe@physics.smu.edu}
46: \altaffiltext{6}{Middle East Technical University, 06531 Ankara, Turkey,
47:         umk@astroa.physics.metu.edu.tr}
48: \altaffiltext{7}{\c{C}anakkale Onsekiz Mart \"{U}niversitesi, Terzio\v{g}lu
49:         17020, \c{C}anakkale, Turkey, me\_ozel@ibu.edu.tr}
50: \altaffiltext{8}{Department of Astronomy, University of Texas, Austin, TX
51:         78712, quimby@astro.as.utexas.edu, wheel@astro.as.utexas.edu}
52: \altaffiltext{9}{Department of Physics and Astronomy, Louisiana State
53:         University, Baton Rouge, LA 70803, schaefer@lsu.edu}
54: \altaffiltext{10}{Los Alamos National Laboratory, NIS-2 MS D436, Los Alamos, NM
55:         87545, vestrand@lanl.gov, jwren@nis.lanl.gov}
56: 
57: \begin{abstract}
58: We present the results of a search for untriggered gamma-ray burst (GRB)
59: afterglows with the Robotic Optical Transient Search Experiment-III (ROTSE-III)
60: telescope array.  This search covers observations from September 2003 to March
61: 2005.  We have an effective coverage of $1.74\,\mathrm{deg}^2\,\mathrm{yr}$ for
62: rapidly fading transients that remain brighter than $\sim17.5$ magnitude for
63: more than 30 minutes.  This search is the first large area survey to be able to
64: detect typical untriggered GRB afterglows.  Our background rate is very low and
65: purely astrophysical.  We have found 4 previously unknown cataclysmic variables
66: (CVs) and 1 new flare star.  We have not detected any candidate
67: afterglow events or other unidentified transients.  We can place an upper limit
68: on the rate of fading optical transients with quiescent counterparts dimmer
69: than $\sim\,20^{th}$ magnitude at a rate of less than
70: $1.9\,\mathrm{deg}^{-2}\,\mathrm{yr}^{-1}$ with 95\% confidence.  This places
71: limits on the optical characteristics of off-axis (orphan) GRB afterglows.  As
72: a byproduct of this search, we have an effective
73: $\sim52\,\mathrm{deg}^2\,\mathrm{yr}$ of coverage for very slowly decaying
74: transients, such as CVs.  This implies an overall rate of outbursts from high
75: galactic latitude CVs of $0.1\,\mathrm{deg}^{-2}\,\mathrm{yr}^{-1}$.
76: \end{abstract}
77: \keywords{gamma rays:bursts, stars: cataclysmic variables}
78: 
79: \section{Introduction}
80: \label{sec:intro}
81: 
82: There is much circumstantial evidence that gamma-ray burst (GRB) outflows are
83: highly relativistic and collimated.  An achromatic break has been seen in light
84: curves for several GRB afterglows, with the canonical example being
85: GRB~990510.~\citep{hbfsk99, sgkpt99} These breaks are naturally explained by a
86: geometric constraint on the outflow.  The jet opening angle has been inferred
87: for GRB~990510 and several other GRBs, and appears to range from $2^{\circ}$ to
88: $30^{\circ}$.~\citep{fksdb01, pk01} Therefore, the true rate of GRBs must be
89: $\gtrsim100$ times that detected by satellite experiments such as BATSE,
90: HETE-2, INTEGRAL, and Swift.  Although the Earth will not receive $\gamma$-ray
91: emission from these bursts, they might be detectable at longer wavelengths.  It
92: remains an open question what these off-axis ``orphan'' afterglows should look
93: like. In the simplest model, an orphan afterglow looks like a standard
94: afterglow, except that it becomes visible after a delay of
95: $\sim0.5\,\mathrm{day}$.  As the ejecta cools, the relativistic beaming angle
96: increases until the afterglow can be seen off-axis, long after the
97: $\gamma$-rays have ceased.  \citep{rhoad97} However, this assumes that there is
98: no significant optical emission outside the $\gamma$-ray beaming angle.
99: \citet{np03} have suggested that the beaming angle of the optical emission
100: might be different than that of the $\gamma$-ray emission.  They refer to these
101: afterglows as ``on-axis orphan afterglows.''  A detectable rate of orphan
102: afterglows gives an orthogonal approach to measuring typical GRB collimation.
103: 
104: Whether or not orphan (off-axis) afterglows are detectable, there must be
105: untriggered afterglows from normal GRBs that have simply not been seen by a
106: $\gamma$-ray satellite.  The Swift satellite can detect approximately two GRBs
107: per week in its field of view, while an extrapolation of the BATSE event
108: trigger rate for the entire sky suggests that there are around 2 GRBs per day
109: visible to the Earth, corresponding to
110: $0.018\,\mathrm{deg}^{-2}\,\mathrm{yr}^{-1}$.~\cite{fehkl93} How much solid
111: angle would a survey need to cover to observe an untriggered GRB afterglow
112: serendipitously?  We know from ROTSE-I and LOTIS prompt follow-ups to BATSE
113: triggers that the preponderance of early afterglows do not get as bright as
114: $14^{th}$ magnitude. \citep{abbbb00,kabbb01,ppwab99} Currently, we do not have
115: enough data on early afterglows with deeper imaging to provide any firm
116: predictions of the rate of detectable bursts.  We expect this to change now
117: that Swift is operational.  Rapid follow-up to a small number of HETE-2
118: triggers has shown that approximately 50\% of bursts might be brighter than
119: $\sim18.5$ for 30 minutes or more.~\citep{lrabb04} Thus, around
120: $9\times10^{-3}\,\mathrm{bursts}\,\mathrm{deg}^{-2}\,\mathrm{yr}^{-1}$ should
121: be visible to an instrument capable of reaching this magnitude, such as
122: ROTSE-III.  Thus, an exposure of $\sim110\,\mathrm{deg}^2\,\mathrm{yr}$ is
123: required for a high probability of finding an afterglow independent of any
124: $\gamma$-ray trigger.
125: 
126: To date, there have only been a few published searches for untriggered and
127: orphan GRBs and other short duration transients.  These searches have all
128: probed different magnitude ranges and timescales.  In general, the wide-field
129: instruments cover more solid angle but cannot go very deep.  The ROTSE-I
130: transient search covered $3.5\,\mathrm{deg}^2\,\mathrm{yr}$, to a limiting
131: magnitude of 15.7. \citep{kabbc02} The RAPTOR array covers the entire visible
132: sky several times each night to a limiting magnitude of 12, and is sensitive to
133: very fast transients, on the order of minutes. \citep{vbcfg04} The Deep Lens
134: Survey (DLS) transient search covered $0.01\,\mathrm{deg}^2\,\mathrm{yr}$ with
135: sensitivity to $24^{th}$ magnitude, and found a couple of tantalizing
136: unidentified transients. \citep{bwbcd04} In spite of the relatively rapid
137: detection of these transients, they were too faint for spectroscopic follow-up
138: and could not be positively identified as extragalactic or associated with
139: GRBs.  \citet{vlwbl02} performed a color-selected transient search with the
140: Sloan Digital Sky Survey to a limiting magnitude of 19 and detected one unusual
141: AGN.
142: 
143: The search reported in this paper was specifically designed to detect
144: untriggered and orphan GRB afterglows.  This search is based upon the
145: assumption that an orphan afterglow might have an optical behavior similar to
146: that of observed afterglows.  As a result, we search for transients which meet
147: two criteria: first, the quiescent counterpart or host galaxy would have $m_R >
148: 20$, and would not be detectable by ROTSE-III; second, the transient must be
149: brighter than our limiting magnitude for at least 30 minutes.  Other known
150: astrophysical sources fall into this category, including: cataclysmic variables
151: (CVs) and novae in the galactic halo that burst by $>2$ magnitudes; faint flare
152: stars that brighten on short timescales by several magnitudes; and active
153: galactic nuclei (AGN), blazars, and quasars that display optically violent
154: variability (OVV), occasionally flaring by several magnitudes on very short
155: timescales.
156: 
157: The rapid identification of new transients is essential for a search of this
158: nature.  Only spectroscopic follow-up can positively identify an orphan
159: afterglow or a new type of astrophysical phenomenon. As ROTSE-III can identify
160: transients while they are still relatively bright, this enables follow-up with
161: telescopes with modest apertures.
162: 
163: \section{Observations \& Data Reduction}
164: 
165: The ROTSE-III systems are described in detail in \citet{akmrs03}.  The
166: ROTSE-III telescopes are installed at four sites around the globe:
167: Coonabarabran, Australia; Ft. Davis, Texas; Mt. Gamsberg, Namibia; and
168: Bakirlitepe, Turkey.  They have a wide ($1\fdg85 \times 1\fdg85$) field of view
169: imaged onto an E2V $2048\times2048$ back-illuminated thinned CCD, and operate
170: without filters.  The camera has a fast readout cycle of 6~s.  The limiting
171: magnitude for a typical 60~s exposure is around $m_{R}\sim18.5$, which is well
172: suited for study of GRB afterglows during the first hour or more.  The typical
173: FWHM of the stellar images is $<2.5$ pixels ($8\farcs1$).
174: 
175: In September 2003 we initiated analysis of our nightly sky patrol images for
176: rapid identification of fast transients.  The region patrolled includes $370\times
177: 3.4\,\mathrm{deg}^2$ fields in the equatorial stripe with declination of
178: $|\delta|<2\fdg64$.  To avoid field crowding and galactic opacity, all the
179: fields are at high galactic latitude with $|b|>30^{\circ}$.  Because asteroids
180: were a significant background early in our search, after May 2004 we only
181: imaged fields with ecliptic latitudes of $|\beta|>10^{\circ}$. These fields
182: were chosen for two primary reasons.  First of all, they are visible to all
183: four ROTSE-III locations, two of which are in the northern hemisphere, and two
184: of which are in the southern hemisphere.  Furthermore, this region has public
185: Sloan Digital Sky Survey (SDSS) data with 5 color imaging to well below our
186: limiting magnitude. \citep{aaaaa05} This allows calibration of our fields to a
187: set of well-measured stars, as well as providing easy identification of flares
188: from objects such as CVs and quasars.
189: 
190: Our standard observing sequence includes a pair of closely spaced 60 s
191: exposures, followed after 30 minutes by a second pair of 60 s exposures.  We
192: define a \emph{set} of images as four consecutive images taken with this
193: interval.  The second image of each pair is offset by $\sim 10$ pixels to
194: reduce the impact of bad pixels.  During bright moon conditions (lunar
195: illumination $>70\%$) we reduce the exposure length to 20~s to prevent the
196: background sky from saturating the images.  The 30 minute interval was chosen
197: to enhance sensitivity to rapidly fading transients while still allowing a
198: large solid angle coverage.  
199: 
200: After each image is recorded, it is dark subtracted and flat fielded by an
201: automated pipeline.  Twilight flats are generated every night, and are updated
202: for the pipeline on a monthly basis.  Our back illuminated thinned CCD imager,
203: combined with broadband filterless optics, consequently imposes an interference
204: fringe pattern on all images.  We therefore must correct for these fringing
205: effects.  The pattern is stable, although the amplitude varies with the
206: brightness of night sky lines.  We have created a fringe map for each CCD by
207: comparing a twilight flat image, which does not display a fringe pattern, with
208: a sky flat image, which does display a fringe pattern.  After flat fielding,
209: the sky pixels in the central subregion of the image are fit via linear
210: regression with the corresponding pixels in the fringe pattern.  The fringe
211: pattern is then scaled accordingly and subtracted from the image.  If the fit
212: is poor (reduced $\chi^2>3$), due to scattered moonlight or clouds in the
213: image, the fringe pattern is not subtracted.  In the worst case, the fringe
214: pattern can introduce photometric errors as large as $5\%$, and can produce
215: occasional false detections of faint objects.
216: 
217: The pipeline then runs SExtractor~\citep{ba96} to perform initial object
218: detection, measure centroid positions, and determine aperture magnitudes.  A
219: separate pipeline program written in IDL ({\tt idlpacman}) correlates the
220: object list with stars brighter than $15^{th}$ magnitude in the USNO~A2.0
221: catalog to determine an astrometric solution as well as an approximate
222: magnitude zero-point for the field.
223: 
224: The limiting magnitude of each image is estimated from the background noise,
225: FWHM of the point spread function (PSF), and the zero-point offset, which is
226: essentially a measure of the transparency of the sky.  With these three
227: values, we can estimate the magnitude at which we can detect a star 90\% of the
228: time with our SExtractor cuts in an uncrowded region of sky. As will
229: be shown below in \S~\ref{sec:analysis}, our detection efficiency declines
230: very rapidly in crowded regions.
231: 
232: After each pair of images is calibrated, we pair-match the object lists.
233: Objects that are detected in both of the pair of images are considered real.
234: All objects that are detected only in a single image are rejected.  This
235: strategy removes most spurious detections caused by cosmic rays, pixel defects,
236: satellite glints, and noise spikes.  However, some backgrounds still remain:
237: noise spikes usually due to imperfect fringe subtraction when the sky is not
238: fully transparent; some cosmic ray coincidences; and asteroids.
239: 
240: One of the great difficulties in calibrating a large area survey of this type
241: is the lack of standard stars uniformly distributed across the sky.  The
242: USNO~A2.0 catalog provides excellent astrometric solutions for any field as
243: there are typically $>1000$ stars with $R<15$.  However, the
244: photometric zero-points as determined from USNO A2.0 $R$-band magnitudes have
245: typical systematic errors of up to $0.30$ magnitude \citep{mcdgh98} making
246: field-to-field comparisons difficult.  
247: 
248: For around 90\% of our fields, we have significant overlap with SDSS 5-color
249: data.  We have thus decided to recalibrate all of our fields relative to the
250: SDSS $r'$-band, as the field-to-field variations are around
251: 2\%. \citep{aaaaa05} For each of our overlapping fields we compare all
252: ROTSE-III template stars that have counterparts in SDSS between $15<r'<17$ with
253: $g'-r' < 1.0$.  We find that the typical offset is $0.22\pm0.16$ when
254: converting from $m_{\mathrm{ROTSE(USNO)}}$ to $m_{\mathrm{ROTSE(SDSS)}}$.  The
255: scatter in offsets is primarily due to systematic errors in the USNO~A2.0
256: $R$-band zero points.  For the remaining $\sim$10\% of sky patrol fields
257: without SDSS calibration data, we have offset the zero-points obtained from
258: USNO~A2.0 calibration by $0.22$ magnitudes.  For the remainder of this paper,
259: the magnitudes quoted are calibrated relative to the SDSS $r'$-band.
260: 
261: Through March 2005 we have searched over 23000 \emph{sets} of images for new
262: transients as described in \S~\ref{sec:intro}.  Figure~\ref{fig:wmlim} is a
263: histogram of the number of \emph{sets} searched as a function of limiting
264: magnitude at the second (return) epoch.  Since we demand that a new transient
265: be present in an entire \emph{set}, the limiting magnitude at the second epoch
266: is the primary constraint for detecting rapidly fading transients.  The solid
267: histogram describes all \emph{sets} imaged, and the dashed histogram describes
268: the contribution from 20~s exposures taken during bright lunar phases.
269: 
270: 
271: \begin{figure}
272: \rotatebox{90}{\scalebox{0.85}{\plotone{f1.eps}}}
273: \caption{\label{fig:wmlim}Histogram of number of \emph{sets} searched as a
274:   function of limiting magnitude ($m_{\mathrm{ROTSE(SDSS)}}$) at the second
275:   (return) epoch.  The limiting magnitude at the second epoch is the primary
276:   constraint for detecting quickly fading transients.  The solid histogram
277:   describes all \emph{sets} imaged, and the dashed histogram describes the
278:   contribution from 20~s exposures taken during bright lunar phases.}
279: \end{figure}
280: 
281: 
282: 
283: \section{Analysis}
284: \label{sec:analysis}
285: 
286: We have chosen to identify transients by the appearance of a new object
287: compared to a template list of ROTSE-III objects.  This strategy has distinct
288: advantages for our survey relative to image subtraction. As most of our fields
289: are relatively uncrowded, we only lose $\sim4\%$ of our solid angle to direct
290: starlight.  Due to bandwidth limitations at our remote observatory sites, we
291: need a strategy that is fast and has relatively few false positives.  Image
292: subtraction routines are very sensitive to variations in point spread function
293: (PSF) which are difficult to keep absolutely steady in an automated telescope
294: over a wide range of conditions.  Finally, as will be shown below, our solid
295: angle efficiency is over 80\% for detection of transients in most of our high
296: galactic latitude fields, comparable to the search strategy employed by the
297: DLS.~\citep{bwbcd04}
298: 
299: Our basic detection algorithm is quite simple.  We demand that a new object is
300: detected in at least four consecutive images, or an entire \emph{set}.  The
301: position resolution of each of the images must be less than 0.3 pixels
302: ($1''$).  This removes images that are out of focus, and images with
303: anomalously large point spread functions common on windy nights. The new object
304: must be more than 5 pixels ($16\farcs2$) from the nearest ROTSE-III template
305: object.  We have no cuts on the shape of the light curve so as not to limit the
306: types of transient we can detect, although we do demand that the brightest
307: detection is brighter than $18^{th}$ magnitude.  For each candidate we examine
308: its PSF to ensure that it is comparable to the PSF of its neighbors.  We next
309: place an aperture at the same location in two of our best template images to
310: check if there is an object that had not been properly deblended by SExtractor.
311: 
312: After these cuts have been performed, typically $\ll$1 transient remains in each
313: \emph{set}.  Occasionally there are more detections, usually due to
314: instrumental effects such as artifacts near saturated stars, or due to
315: deblending problems caused by bad focus or wind-degraded images.  Therefore, if
316: more than 5 transient candidates remain in a \emph{set}, then it is rejected as
317: a bad \emph{set}.
318: 
319: Thumbnail images of the remaining transient candidates are then copied to a web
320: page at the University of Michigan for hand scanning.  These candidates are
321: usually faint stars that are just at our detection threshold, and are clearly
322: visible in the MAST Digitized Sky Survey.~\citep{lasker98} Occasionally there
323: is a minor planet near opposition that has a proper motion of $<1''$ during the
324: 30 minute interval between exposure pairs. These minor planets are usually in
325: the Minor Planet Center MPChecker database, allowing easy exclusion.
326: 
327: In order to measure our overall efficiency, we must be able to parameterize our
328: coverage for each field.  We have used typical test fields to estimate the
329: detection efficiency as a function of distance to the nearest template star and
330: thus derive the effective solid angle covered for each of our sky patrol
331: fields.  We can also calculate the probability of detecting a transient
332: relative to the limiting magnitude of the field.
333: 
334: Figure~\ref{fig:effvsdist} shows the detection efficiency as a function of
335: distance to the nearest template star.  To obtain this plot we ran a Monte
336: Carlo simulation.  50000 objects were uniformly distributed in magnitude from
337: 10.0 to 19.0 at random positions in a pair of sample images.  The objects were
338: generated with the IDL astronomy library function {\tt psf\_gaussian}, with the
339: median FWHM of a star in the field.  We then added the pixel counts at the
340: appropriate positions in the image.  These new images were reprocessed in the
341: standard analysis pipeline.  Only simulated objects that were detected in each
342: of the pair of images were considered.  For this plot, any object that is
343: within 2 pixels of a template star is vetoed; in our actual search we place the
344: cut at 5 pixels which incurs a minimal penalty in solid angle coverage while
345: greatly decreasing the number of false detections from deblending issues.  The
346: solid histogram describes simulated objects more than 0.5 magnitudes brighter
347: than the limiting magnitude.  The dashed histogram describes simulated objects
348: within 0.5 magnitudes of the limiting magnitude, where the detection efficiency
349: is reduced by $\sim$20\%.
350: 
351: 
352: \begin{figure}
353: \rotatebox{90}{\scalebox{0.85}{\plotone{f2.eps}}}
354: \caption{\label{fig:effvsdist}Detection efficiency in a pair of images
355:   vs. distance from nearest template star.  The solid line describes simulated
356:   objects more than 0.5 magnitude brighter than the limiting magnitude; the
357:   dashed line describes simulated objects within 0.5 magnitude of the limiting
358:   magnitude, where our detection efficiency is reduced.}
359: \end{figure}
360: 
361: 
362: 
363: Figure~\ref{fig:areahist} shows a histogram of the available solid angle
364: coverage for each of our sky patrol fields.  To determine these values, we
365: calculated the distance from each pixel to the nearest template star in each
366: field.  The solid angle lost in each field is primarily due to the 5 pixel
367: exclusion radius around each template star.  Additional pixels are masked out
368: in the wings of very bright stars and saturation bleed trails from bright
369: stars.  These typically cover $\sim$1\% of each field.  In most of our fields
370: the efficiency is greater than 80\%.  The low-coverage tail comprises dense
371: fields including those containing globular clusters.
372: 
373: 
374: \begin{figure}
375: \rotatebox{90}{\scalebox{0.85}{\plotone{f3.eps}}}
376: \caption{\label{fig:areahist}Histogram of solid angle coverage for our sky
377:   patrol fields.  The maximum available solid angle in our field is
378:   $3.42\,\mathrm{deg}^2$, and the efficiency in most of the fields is $>$80\%.}
379: \end{figure}
380: 
381: \section{Results}
382: 
383: \subsection{Transient Detections}
384: 
385: Through March 2005, we have found three new cataclysmic variables, one flare
386: star, and one blazar.  This is a comprehensive list of all of our transient
387: detections that were not identified as asteroids. Each of these objects was
388: well below our detection limits in quiescence.  A brief description of these
389: objects follows.
390: 
391: \begin{itemize}
392: \item{CV ROTSE3 J151453.6+020934.2:} This CV was detected by ROTSE-IIIb on 28
393:   March, 2004, at 16$^{th}$ magnitude, over 3 magnitudes brighter than
394:   quiescence.  It remained around 17$^{th}$ magnitude for two weeks before
395:   fading below our detection threshold.  During quiescence we obtained $UBVI$
396:   measurements from the MDM Hiltner 2.4m telescope on Kitt Peak, Arizona.  Its
397:   colors were consistent with a dwarf nova at minimum light \citep{raagg04}.
398: 
399: \item{CV ROTSE3 J221519.8-003257.2:} This CV was detected by ROTSE-IIId on 8
400:   July, 2004, at 17.5 magnitude, around 3 magnitudes brighter than quiescence.
401:   The outburst lasted over two days before fading below our detection
402:   threshold.  As with the previous CV, its colors during quiescence were
403:   consistent with a dwarf nova at minimum light. \citep{raagg04}  This CV
404:   burst again on 4 October, 2004.  It was detected by ROTSE-IIId at 17.2
405:   magnitude, and was again discovered by our transient detection pipeline.
406: 
407: \item{Flare Star ROTSE3 J220806.9+023100.3:} This flare star was detected by
408:   ROTSE-IIIb on 12 November, 2004, at $16^{th}$ magnitude.  It faded by 0.8
409:   magnitudes in 30 minutes, mimicking our expected signature of an untriggered
410:   GRB afterglow.  However, the counterpart was clearly visible in 2MASS at
411:   $J\sim15$.  In addition, the $J-H$ and $H-K$ colors of 0.65 and 0.32
412:   respectively, suggest a very red object.  Finally, the USNO B-1.0 catalog
413:   measured proper motion of the quiescent counterpart of
414:   $32\,\mathrm{mas}/\mathrm{yr}$.  These observations are consistent with a
415:   flare from a nearby M-Dwarf type star.
416: 
417: \item{CV 2QZ J142701.6-012310:} This CV was detected by ROTSE-IIIc on 23
418:   January, 2005, at $15^{th}$ magnitude, around 5 magnitudes brighter than
419:   quiescence.  It remained bright around $16^{th}$ magnitude for about 7 days
420:   before fading below our detection threshold.  A spectrum previously had been
421:   obtained during quiescence by the 2dF redshift survey.  In addition, a
422:   spectrum of the object during outburst was obtained by the Hobby-Eberly
423:   Telescope (HET) at McDonald Observatory on 25 January, which showed a blue
424:   contiuum with no obvious emission or absorption features.~\citep{r05}
425:   Follow-up observations of the object during quiescence at the University of
426:   Cape Town revealed this to be a rare Am CVn type doubly-degenerate helium
427:   transferring binary.~\citep{wwr05}
428: 
429: \item{CV ROTSE3 J100932.2-020155:} This CV was detected by ROTSE-IIIc on 20
430:   February 2005, at $14^{th}$ magnitude, around 6 magnitudes brighter than
431:   quiescence. It faded slowly over the next 27 days before dropping below our
432:   detection threshold.  Two spectra had previously been obtained during
433:   quiescence by the 2dF redshift survey.  A spectrum from the HET on 21
434:   February displays prominent H-alpha emission expected from a CV, and the
435:   object is consistent with a SU UMa Dwarf Nova during a super
436:   outburst. \citep{rq05}
437: 
438: \end{itemize}
439: 
440: \subsection{Transient Detection Efficiency}
441: 
442: We calculate our detection efficiency using different methods for two different
443: types of transients.  The first method is suitable for slowly decaying
444: transients such as cataclysmic variables and other novae that rise rapidly and
445: are roughly constant over our observation interval.  These objects typically
446: stay brighter than our limiting magnitude for over 1 day.  Our effective time
447: coverage for these transients is therefore quite high.
448: 
449: In the second method, suitable for short duration (rapidly fading) transients,
450: we parameterize the transients as GRB afterglows with a simple decaying power
451: law.  This determines our sensitivity to rapidly varying transients but
452: describes a relatively small effective time coverage.  Specifically, the time
453: that the transient is brighter than our limiting magnitude must be $>30$
454: minutes.
455: 
456: To calculate our sensitivity for slowly decaying transients, we examine the
457: limiting magnitudes of each \emph{set} of images.  First, we account for the
458: solid angle covered, as plotted in Figure~\ref{fig:areahist}.  We then generate
459: 5000 simulated transients uniformly distributed from 9.5 to $20^{th}$
460: magnitude.  If an object is $>$ 0.5 magnitudes brighter than the limiting
461: magnitude in an individual image it is considered a detection.  If it is $<$
462: 0.5 magnitudes brighter than the limiting magnitude there is a 90\% chance that
463: it will be detected.  An object must be detected in all four of a \emph{set} of
464: images, and at least one detection must be brighter than $m_{ROTSE} = 18.0$.
465: This decreases our efficiency for faint CVs, but greatly decreases the false
466: detections as well as simplifying the interpretation of the data.
467: 
468: Figure~\ref{fig:staticcov} shows our total solid angle coverage for new slowly
469: decaying sources.  The detection efficiency before cuts is shown with the
470: dashed histogram.  After we apply saturation cuts and our magnitude cut, the
471: result is the solid histogram.  This plot was made using all of our \emph{sets}
472: with limiting magnitudes deeper than $\sim$17, with over
473: $47000\,\mathrm{deg}^2$ of coverage.  For transients that remain roughly
474: constant for over 0.5 days, this results in $\sim
475: 52\,\mathrm{deg}^2\,\mathrm{yr}$ of coverage for CVs that peak between
476: $13^{th}$ and $16^{th}$ magnitude.
477: 
478: 
479: \begin{figure}
480: \rotatebox{90}{\scalebox{0.85}{\plotone{f4.eps}}}
481: \caption{\label{fig:staticcov}Histogram of total solid angle coverage for new
482:   slowly decaying sources.  The solid histogram describes objects that pass our
483:   cuts.  The dashed histogram describes objects that are detected but do not
484:   pass our cuts, either due to saturation at the bright end or due to falling
485:   below our magnitude threshold.}
486: \end{figure}
487: 
488: 
489: To calculate our sensitivity to short duration transients, we parameterize each
490: transient as a fading powerlaw, $f=f_{0}t^{-\alpha}$.  Each transient is
491: assigned a peak magnitude, $m_{60}$, at $t=t_0 + 60$s, and a decay constant
492: $\alpha$.  The effective coverage time for each \emph{set} is the time period
493: when a transient outburst would be detected in our search.  We have decided to
494: assume an effective coverage time of 30 m for each \emph{set} of four
495: observations, although our efficiency depends on peak magnitude and decay
496: constant for each transient.  The effective coverage time refers to the
497: \emph{preceding} 30 minute interval, as our search is insensitive to to
498: transients that appear between our two observation epochs. In order to achieve
499: $\sim100\%$ efficiency for a transient type in our observation window, it must
500: remain above our limiting magnitude for at least 1 hour.
501: 
502: As before with slowly decaying transients, we first account for the solid angle
503: covered in each field.  We ran a Monte Carlo simulation with 5000 objects per
504: \emph{set}.  The peak magnitude $m_{60}$ was uniformly distributed from 7.5 to
505: 18.5, the decay constant $\alpha$ was uniformly distributed from 0.3 to 2.5,
506: and the burst time $t_0$ was set at a random time at most 30 minutes
507: \emph{prior} to the first image in the \emph{set}.  Essentially, we are
508: calculating the detection efficiency when assuming that our coverage time is 30
509: minutes for each \emph{set}.  As mentioned above, objects $<$ 0.5 magnitudes
510: brighter than the limiting magnitude have a 90\% chance of detection.
511: 
512: Figure~\ref{fig:transcov} shows our detection efficiency for all the
513: \emph{sets} in the search with limiting magnitudes at the second epoch deeper
514: than 17.5, assuming 30 minutes of coverage. The integrated coverage is
515: $1.74\,\mathrm{deg}^2\,\mathrm{yr}$ for these \emph{sets}.  The 20\%, 50\%, and
516: 90\% contours are shown.  The contour lines roll over at the bright end where
517: our saturation cuts take effect -- these extraordinarily bright transients
518: would be vetoed in our pipeline.  However, the ROTSE-I transient search covered
519: much of this parameter space (shaded in gray) without finding any candidate
520: afterglows with approximately twice as much coverage. \citep{kabbc02}
521: Overplotted are approximate peak magnitudes and decay slopes, averaged over the
522: first hour after the burst time, for ten of the twelve GRB afterglows that have
523: been detected in the first hour after the burst.  The remaining two bursts are
524: too faint to be placed on this plot.  We are sensitive to $\sim40\%$ of these
525: afterglows, which have been detected for $\sim50\%$ of all promptly localized
526: GRBs.  Also plotted is an inferred peak magnitude and decay slope for the flare
527: star ROTSE3 J220806.9+023100.3, which is contained in the locus of afterglow
528: points.
529: 
530: 
531: \begin{figure}
532: \rotatebox{90}{\scalebox{0.9}{\plotone{f5.eps}}}
533: \caption{\label{fig:transcov}Detection efficiency for fading transients,
534:   parameterized by their peak magnitude at $t=t_0+60$s, $m_{60}$, and their
535:   decay constant $\alpha$.  The efficiencies calculated assume 30 minutes of
536:   effective coverage for each \emph{set} for an integrated coverage of
537:   $1.74\,\mathrm{deg}^2\,\mathrm{yr}$.  All \emph{sets} with limiting
538:   magnitudes deeper than 17.5 were used for this plot.  The contours describe
539:   our 20\%, 50\%, and 90\% efficiency levels.  The contours roll over at the
540:   bright end where many transients would be saturated and rejected.  The gray
541:   area describes the coverage in the ROTSE-I transient search.  Overplotted are
542:   approximate peak magnitudes and decay constants (averaged over the first
543:   hour) for six GRB afterglows with early
544:   detections.~\citep{abbbb99,fyktk03,fox04,lfcj03,rspaa04,rsq05,rys05c,rys05,yssa05,fbcsc05}}
545: \end{figure}
546: 
547: 
548: 
549: \section{Discussion}
550: 
551: To date, this is the deepest wide-field search for untriggered and orphan GRBs.
552: With a total coverage of $1.74\,\mathrm{deg}^2\,\mathrm{yr}$, we have not found
553: any candidate afterglows or other unknown transients.  The primary reason we
554: have been able to positively identify each of our transient candidates is the
555: 5-color SDSS data of the fields in our search.  Other transient searches such
556: as the DLS that can image much deeper do not have this advantage.
557: 
558: We have also determined that the backgrounds for a search of this type are
559: purely astrophysical.  We have not had any difficulty from satellite glints or
560: any mysterious terrestrial flashes.  Our primary background consists of
561: asteroids with small proper motion, but the MPChecker database combined with
562: follow-up observations can eliminate these.  Outside the solar system, but
563: within the galaxy, we see a small but measurable rate of CVs and flare stars,
564: as well as one extragalactic blazar.  The flare star ROTSE3 J220806.9+023100.3
565: is the only object that we have found that had a light curve that mimicked our
566: expected signature of an untriggered GRB afterglow.
567: 
568: We can place an upper limit on the rate of fading optical transients with
569: quiescent counterparts dimmer than $\sim20^{th}$ magnitude, as described by our
570: 90\% coverage region shown in Figure~\ref{fig:transcov}.  To calculate an upper
571: limit on the rate of transients in this region, we follow the method of
572: \citet{bwbcd04}.  The observed rate is
573: \begin{equation}
574: \eta =
575: \frac{N}{{\langle}{\cal E}{\rangle}E}\,\mathrm{events}\,\mathrm{deg}^{-2}\,\mathrm{yr}^{-1}\label{eqn:rate},
576: \end{equation}
577: where $N$ is the number of events, $E$ is the exposure, and ${\langle}{\cal
578: E}{\rangle}$ is the efficiency.  With the observed number of transients
579: $N_{\mathrm{obs}} = 0$, Poisson statistics place an upper limit of
580: $N_{\mathrm{max}} < 3.0$ with 95\% confidence.  Therefore, we can place a 95\%
581: confidence upper limit of $\eta_{\mathrm{max}} <
582: 1.9\,\mathrm{deg}^{-2}\,\mathrm{yr}^{-1}$ in our coverage region.
583: 
584: The launch of Swift will allow us to probe the early afterglow phase of GRBs
585: far more systematically than has been achieved to date.  We have learned some
586: details from the variety of early afterglows detected so far.  It is clear that
587: very bright prompt flashes like that from GRB~990123~\citep{abbbb99} are not
588: the norm.  It also appears that extrapolating late time afterglows to the early
589: time generally over-predicts the brightness, as with GRB~030418 and
590: GRB~030723.~\citep{rspaa04}  These data certainly make our search more
591: difficult, as only $\sim20\%$ of GRBs have afterglows that fall within our
592: sensitivity region.
593: 
594: If we assume that the $\gamma$-ray emission of GRBs is confined within a double
595: jet with a cone half-angle of $\theta_{\mathrm{max}}$, while the optical
596: emission is isotropic, then a given GRB is visible in a small fraction of the
597: sky approximated by $\theta^2/2$.  Therefore, the true rate of GRB events
598: within the observable universe must be
599: $\sim1500/\theta_{\mathrm{max}}^2\,\mathrm{events}\,\mathrm{yr}^{-1}$.  The
600: 95\% confidence limit of $<78000\,\mathrm{events}\,\mathrm{yr}^{-1}$ is for a
601: region that is sensitive to $\sim20\%$ of GRBs, and therefore our assumption of
602: isotropic optical emission is tenable as long as
603: $\theta_{\mathrm{max}}>3.6^{\circ}$.  As this is an approximate estimate of the
604: GRB jet angle, the present limits cannot set stringent bounds on the properties
605: of these objects.  However, if programs such as ours continue to reduce the
606: upper bounds for the orphan afterglow rate, the isotropic emission hypothesis
607: will become incompatible with what we know about the structure of GRB jets.
608: Although this is not an enormous surprise, it does represent a sanity check of
609: the accepted model of GRBs by completely independent reasoning.
610: 
611: Our search has also detected several previously unknown high galactic latitude
612: cataclysmic variables with dim quiescent counterparts.  We have detected 3 new
613: CVs with $\sim52\,\mathrm{deg}^2\,\mathrm{yr}$ of coverage for slow decay
614: transients that peak between $13^{th}$ and $16^{th}$ magnitude.  This implies a
615: rate of $0.06\,\mathrm{deg}^{-2}\,\mathrm{yr}^{-1}$.  Assuming Poisson
616: statistics, the 95\% confidence upper limit on the rate is
617: $0.17\,\mathrm{deg}^{-2}\,\mathrm{yr}^{-1}$.  We have found 1 new CV with
618: $\sim25\,\mathrm{deg}^2\,\mathrm{yr}$ of coverage for CVs that peak between
619: $16^{th}$ to $18^{th}$ magnitude.  This implies a rate of $0.04\,
620: \mathrm{deg}^{-2}\,\mathrm{yr}^{-1}$.  Assuming Poisson statistics, the 95\%
621: confiedence upper limit on the rate is
622: $0.21\,\mathrm{deg}^{-2}\,\mathrm{yr}^{-1}$.  If we extrapolate the rate over
623: the whole sky, this would produce $\sim2000$ high galactic latitude CV
624: outbursts every year.  In the ``Living Edition'' catalog of CVs there are only
625: $\sim 130$ CVs in our sensitivity range as of March 2005.~\citep{dwsrk01} None
626: of the CVs detected by ROTSE-III had been known previously.  Our search results
627: imply that many more CVs remain to be discovered.
628: 
629: The ROTSE-III transient search is an ongoing project.  We expect to continue to
630: patrol the sky over the lifetime of the Swift instrument (at least 2 years), as
631: we wait for GRB triggers.  We will thus gain another factor of 2-3 in coverage,
632: and will achieve a modest improvement towards the goal of
633: $\sim110\,\mathrm{deg}^2\,\mathrm{yr}$, the threshold for having a high
634: probability of finding an untriggered afterglow.  In this paper, we have
635: demonstrated that such a search is feasible, as the background rate of unknown
636: transients is very low.
637: 
638: \acknowledgements
639: 
640: This work has been supported by NASA grants NNG-04WC41G and F006794, NSF grants
641: AST-0119685 and 0105221, the Australian Research Council, the University of New
642: South Wales, and the University of Michigan.  Work performed at LANL is
643: supported through internal LDRD funding.  Special thanks to the
644: observatory staff at the ROTSE sites, including David Doss, Toni Hanke, and
645: Tuncay \"{O}z{\i}\c{s}{\i}k.
646: 
647: %\include{biblio}
648: \begin{thebibliography}{}
649: 
650: \bibitem[\protect\astroncite{{Abazajian} {\rm et~al.\/}}{2005}]{aaaaa05}
651:  {Abazajian}, K., et~al.
652: \newblock Mar. 2005, \aj, 129, 1755
653: 
654: \bibitem[\protect\astroncite{{Akerlof} {\rm et~al.\/}}{1999}]{abbbb99}
655:  {Akerlof}, C., et~al.
656: \newblock 1999, \nat, 398, 400
657: 
658: \bibitem[\protect\astroncite{{Akerlof} {\rm et~al.\/}}{2000}]{abbbb00}
659:  {Akerlof}, C., et~al.
660: \newblock Mar. 2000, \apjl, 532, L25
661: 
662: \bibitem[\protect\astroncite{{Akerlof} {\rm et~al.\/}}{2003}]{akmrs03}
663:  {Akerlof}, C.~W., et~al.
664: \newblock Jan. 2003, \pasp, 115, 132
665: 
666: \bibitem[\protect\astroncite{{Becker} {\rm et~al.\/}}{2004}]{bwbcd04}
667:  {Becker}, A.~C., et~al.
668: \newblock Aug. 2004, \apj, 611, 418
669: 
670: \bibitem[\protect\astroncite{{Bertin} \& {Arnouts}}{1996}]{ba96}
671: {Bertin}, E. \& {Arnouts}, S.
672: \newblock June 1996, \aaps, 117, 393
673: 
674: \bibitem[\protect\astroncite{{Downes} {\rm et~al.\/}}{2001}]{dwsrk01}
675: {Downes}, R.~A., {Webbink}, R.~F., {Shara}, M.~M., {Ritter}, H., {Kolb}, U., \&
676:   {Duerbeck}, H.~W.
677: \newblock June 2001, \pasp, 113, 764
678: 
679: \bibitem[\protect\astroncite{{Falcone} {\rm et~al.\/}}{2005}]{fbcsc05}
680:  {Falcone}, A., et~al.
681: \newblock 2005, GCN Circ. No. 3330
682: 
683: \bibitem[\protect\astroncite{{Fenimore} {\rm et~al.\/}}{1993}]{fehkl93}
684:  {Fenimore}, E.~E., et~al.
685: \newblock Nov. 1993, \nat, 366, 40
686: 
687: \bibitem[\protect\astroncite{{Fox}}{2004}]{fox04}
688: {Fox}, D.~B.
689: \newblock 2004, GCN Circ. No. 2741
690: 
691: \bibitem[\protect\astroncite{{Fox} {\rm et~al.\/}}{2003}]{fyktk03}
692:  {Fox}, D.~W., et~al.
693: \newblock Mar. 2003, \nat, 422, 284
694: 
695: \bibitem[\protect\astroncite{{Frail} {\rm et~al.\/}}{2001}]{fksdb01}
696:  {Frail}, D.~A., et~al.
697: \newblock Nov. 2001, \apjl, 562, L55
698: 
699: \bibitem[\protect\astroncite{{Harrison} {\rm et~al.\/}}{1999}]{hbfsk99}
700:  {Harrison}, F.~A., et~al.
701: \newblock Oct. 1999, \apjl, 523, L121
702: 
703: \bibitem[\protect\astroncite{{Kehoe} {\rm et~al.\/}}{2001}]{kabbb01}
704:  {Kehoe}, R., et~al.
705: \newblock 2001, \apjl, 554, L159
706: 
707: \bibitem[\protect\astroncite{{Kehoe} {\rm et~al.\/}}{2002}]{kabbc02}
708:  {Kehoe}, R., et~al.
709: \newblock Oct. 2002, \apj, 577, 845
710: 
711: \bibitem[\protect\astroncite{{Lamb} {\rm et~al.\/}}{2004}]{lrabb04}
712:  {Lamb}, D.~Q., et~al.
713: \newblock Apr, 2004, New Astronomy Review, 48, 423
714: 
715: \bibitem[\protect\astroncite{{Lasker}}{1998}]{lasker98}
716: {Lasker}, B.~M.
717: \newblock May 1998, Bulletin of the American Astronomical Society, 30, 912
718: 
719: \bibitem[\protect\astroncite{{Li} {\rm et~al.\/}}{2003}]{lfcj03}
720: {Li}, W., {Filippenko}, A.~V., {Chornock}, R., \& {Jha}, S.
721: \newblock Mar. 2003, \apjl, 586, L9
722: 
723: \bibitem[\protect\astroncite{{Monet} {\rm et~al.\/}}{1998}]{mcdgh98}
724:  {Monet}, D.~B.~A., et~al.
725: \newblock Oct. 1998, VizieR Online Data Catalog, 1252, 0
726: 
727: \bibitem[\protect\astroncite{{Nakar} \& {Piran}}{2003}]{np03}
728: {Nakar}, E. \& {Piran}, T.
729: \newblock Feb. 2003, New Astronomy, 8, 141
730: 
731: \bibitem[\protect\astroncite{{Panaitescu} \& {Kumar}}{2001}]{pk01}
732: {Panaitescu}, A. \& {Kumar}, P.
733: \newblock Oct. 2001, \apjl, 560, L49
734: 
735: \bibitem[\protect\astroncite{{Park} {\rm et~al.\/}}{1999}]{ppwab99}
736:  {Park}, H.~S., et~al.
737: \newblock 1999, \aaps, 138, 577
738: 
739: \bibitem[\protect\astroncite{{Rhoads}}{1997}]{rhoad97}
740: {Rhoads}, J.~E.
741: \newblock 1997, \apjl, 487, L1
742: 
743: \bibitem[\protect\astroncite{{Rykoff}}{2005}]{r05}
744: {Rykoff}, E.
745: \newblock Jan. 2005, The Astronomer's Telegram, 403, 1
746: 
747: \bibitem[\protect\astroncite{{Rykoff} \& {Quimby}}{2005}]{rq05}
748: {Rykoff}, E. \& {Quimby}, R.
749: \newblock Feb. 2005, The Astronomer's Telegram, 423, 1
750: 
751: \bibitem[\protect\astroncite{{Rykoff} {\rm et~al.\/}}{2005a}]{rsq05}
752: {Rykoff}, E., {Schaefer}, B., \& {Quimby}, R.
753: \newblock 2005, GCN Circ. No. 3116
754: 
755: \bibitem[\protect\astroncite{{Rykoff} {\rm et~al.\/}}{2004a}]{raagg04}
756:  {Rykoff}, E.~S., et~al.
757: \newblock Aug. 2004, Informational Bulletin on Variable Stars, 5559, 1
758: 
759: \bibitem[\protect\astroncite{{Rykoff} {\rm et~al.\/}}{2004b}]{rspaa04}
760:  {Rykoff}, E.~S., et~al.
761: \newblock Feb. 2004, \apj, 601, 1013
762: 
763: \bibitem[\protect\astroncite{{Rykoff} {\rm et~al.\/}}{2005b}]{rys05c}
764: {Rykoff}, E.~S., {Yost}, S.~A., \& {Smith}, D.~A.
765: \newblock 2005, GCN Circ. No. 3165
766: 
767: \bibitem[\protect\astroncite{{Rykoff} {\rm et~al.\/}}{2005c}]{rys05}
768: {Rykoff}, E.~S., {Yost}, S.~A., \& {Swan}, H.
769: \newblock 2005, GCN Circ. No. 3465
770: 
771: \bibitem[\protect\astroncite{{Stanek} {\rm et~al.\/}}{1999}]{sgkpt99}
772: {Stanek}, K.~Z., {Garnavich}, P.~M., {Kaluzny}, J., {Pych}, W., \& {Thompson},
773:   I.
774: \newblock Sept. 1999, \apjl, 522, L39
775: 
776: \bibitem[\protect\astroncite{{Vanden Berk} {\rm et~al.\/}}{2002}]{vlwbl02}
777:  {Vanden Berk}, D.~E., et~al.
778: \newblock Sept. 2002, \apj, 576, 673
779: 
780: \bibitem[\protect\astroncite{{Vestrand} {\rm et~al.\/}}{2004}]{vbcfg04}
781:  {Vestrand}, W.~T., et~al.
782: \newblock 2004, Astronomische Nachrichten, 325, 549
783: 
784: \bibitem[\protect\astroncite{{Woudt} {\rm et~al.\/}}{2005}]{wwr05}
785: {Woudt}, P.~A., {Warner}, B., \& {Rykoff}, E.
786: \newblock May 2005, \iaucirc, 8531, 3
787: 
788: \bibitem[\protect\astroncite{{Yost} {\rm et~al.\/}}{2005}]{yssa05}
789: {Yost}, S.~A., {Swan}, H., {Schaefer}, B.~A., \& {Alatalo}, K.
790: \newblock 2005, GCN Circ. No. 3322
791: 
792: \end{thebibliography}
793: 
794: \end{document}
795: