1: %\documentclass[12pt,preprint]{aastex}
2: \documentclass{emulateapj}
3: %\documentclass[12pt,preprint2]{aastex}
4: %\documentclass[]{aastex}
5: %\usepackage{emulateapj5}
6: %\documentstyle[emulateapj5,12pt,onecolfloat]{aastex}
7:
8: \begin{document}
9: \newcommand{\Lya}{{$Ly\alpha$}}
10: \newcommand{\acsg}{{$g_{475}$}}
11: \newcommand{\acsi}{{$i_{775}$}}
12: \newcommand{\acsz}{{$z_{850}$}}
13: \newcommand{\nicJ}{{$J_{110}$}}
14: \newcommand{\nicH}{{$H_{160}$}}
15: \newcommand{\hst}{{\it HST}}
16:
17:
18: \title{An Overdensity of \boldmath $i$-dropouts Among A Population of Excess Field Objects in the Virgo Cluster\altaffilmark{1}}
19:
20: \author{Haojing Yan\altaffilmark{2}, Nimish P. Hathi\altaffilmark{3} \& Rogier A. Windhorst\altaffilmark{4}}
21:
22: \altaffiltext{1}{Based on observations made with the NASA/ESA Hubble Space
23: Telescope, obtained at the Space Telescope Science Institute, which is
24: operated by the Association of Universities for Research in Astronomy, Inc.,
25: under NASA contract NAS 5-26555. These observations are associated with
26: programs \#9780, 9488, 9575, 9584, and 9984}
27:
28: \altaffiltext{2} {The Observatories of the Carnegie Institution of Washington,
29: 813 Santa Barbara Street, Pasadena, CA 91101; yhj@ociw.edu}
30: \altaffiltext{3} {Department of Physics, Arizona State University, Tempe, AZ 85287}
31: \altaffiltext{4} {School of Earth \& Space Exploration, Arizona State University, Tempe, AZ 85287}
32:
33: \begin{abstract}
34:
35: Using a set of deep imaging data obtained by the Advanced Camera for Surveys
36: (ACS) on the {\it Hubble Space Telescope} (\hst) shortly after its deployment,
37: Yan, Windhorst \& Cohen (2003) found a large number of F775W-band dropouts
38: ($i$-dropouts), which are consistent with being galaxies at $z\approx 6$. The
39: surface density of $i$-dropouts thus derived, however, is an order of magnitude
40: higher than those subsequent studies found in other deep ACS fields, including
41: the {\it Hubble} Ultra-Deep Field (HUDF). Here we revisit this problem, using
42: both the existing data and the new data obtained for this purpose. We confirm
43: that the large overdensity of $i$-dropouts does exist in this field, and that
44: their optical-to-IR colors are similar to those in the HUDF. However, we have
45: discovered that the $i$-dropout overdensity is accompanied with an even larger
46: excess of faint field objects in this region and its vicinity. This large
47: excess of field objects is most likely caused by the fact that we have resolved
48: the faint diffuse light extending from an interacting galaxy pair in the Virgo
49: Cluster, M60/NGC4647, which lies several arcminutes away from the region where
50: the excess is found. The integrated light from the excess is a few percent of
51: the luminosity of the galaxy pair. We argue that this excess population is most
52: likely tidal ``debris'' and/or halo stars related to the galaxy pair rather
53: than to the Virgo Cluster in general. The $i$-dropouts in this field are within
54: the magnitude range where this excess of field objects occurs. The spatial
55: distribution of the $i$-dropouts seems to follow the same gradient as the
56: excess field population. This excess population is also red in color, and the
57: red wing of its color distribution continuously extends to the regime where the
58: $i$-dropouts reside. While we still cannot completely rule out the possibility
59: that the overdensity of $i$-dropouts might be a genuine large-scale structure
60: of galaxies at $z\approx 6$, we prefer the interpretation that most of them are
61: part of the excess stellar population related to M60/NGC4647. Future
62: spectroscopic work will be needed to unambiguously identify the nature of this
63: $i$-dropout overdensity.
64:
65: \end{abstract}
66: \keywords{galaxies: high-redshift --- galaxies: halos --- galaxies: interactions
67: --- stars: AGB and post-AGB}
68:
69: \section{Introduction}
70:
71: The so-called ``Lyman-break'' technique (Steidel \& Hamilton 1992) is now
72: widely used to select galaxies at high redshifts. This method relies on
73: multi-band imaging to identify the characteristic discontinuity --- the
74: {\it Lyman-break} --- in the spectral energy distributions (SED) of
75: high-redshift galaxies, which is largely caused by the Lyman-limit
76: and \Lya\, absorptions of intervening neutral hydrogen along the sightlines to
77: such galaxies (e.g., Madau 1995). The technique requires imaging in at least
78: two passbands, one to the blue side of the break and the other to the red side.
79: Lyman-break makes high-redshift galaxies much fainter in the blue band than in
80: the red one, or in other words, it makes them seem to ``drop-out'' from the
81: blue band. For this reason, this method is also known as the ``drop-out''
82: selection, and the candidates found in this way are usually referred to as
83: ``dropouts''.
84:
85: As its first application in the $z\approx 6$ regime, Yan, Windhorst \& Cohen
86: (2003; hereafter YWC03) used this technique in a deep field observed by the
87: Advanced Camera for Surveys (ACS; Ford et al. 2003) in the default pure parallel
88: mode (Sparks et al. 2001) soon after its deployment on-board the Hubble Space
89: Telescope (\hst). In this redshift range, Lyman-break occurs at around 8512\AA\,
90: in observer's frame (restframe 1216\AA), which is well targeted by the F775W
91: and F850LP filters through which this deep parallel field was imaged. YWC03
92: found a large number of F775W-band dropouts, and argued that the vast majority
93: of them were very likely galaxies at $z\approx 6$.
94:
95: The number density of $z\approx 6$ galaxies inferred from YWC03, however,
96: seem to be much higher than those in other deep ACS fields (e.g.,
97: Bouwens et al. 2003; Stanway, Bunker \& McMahon 2003; Dickinson et al. 2004;
98: Bouwens et al. 2004; Bunker et al. 2004; Yan \& Windhorst 2004, hereafter YW04).
99: While the cosmic variance could result in significant fluctuation in the number
100: density (e.g., Somerville et al. 2004; Bouwens et al. 2004), it cannot account
101: for the large difference between the result of YWC03 and those of others. As
102: the true number density of galaxies at $z\approx 6$ is closely related to a
103: series of important cosmological questions (e.g., the formation of early
104: galaxies; the reionization history of the universe; etc.), it is prudent to
105: closely examine the cause of the observed difference.
106:
107: In this paper, we present new results that shed light to the nature
108: of this overdensity. We first describe the existing and new \hst\, observations
109: in \S 2 and the data reduction in \S 3. The overdensity of F775W-band dropouts
110: in YWC03 is scrutinized in \S 4, and a new interpretation is given in \S 5. We
111: conclude with a summary in \S 6. For simplicity, we refer to the \hst\, F475W,
112: F775W, F850LP, F110W, and F160W passbands as \acsg, \acsi, \acsz, \nicJ, and
113: \nicH, respectively. The F775W-band dropouts are then referred to as
114: $i$-dropouts. The magnitudes are in AB system. All coordinates quoted are in
115: J2000. The following cosmological parameters from the first-year Wilkinson
116: Microwave Anisotropy Probe (WMAP) results in Spergal et al. (2003) are
117: adopted: $(\Omega_M, \Omega_\Lambda, H_0)=(0.27,0.73,71)$. Using the three-year
118: WMAP results (Spergal et al. 2007) would not change any of our results.
119:
120: \section{Observations}
121:
122: In addition to the ACS parallel field discussed in YWC03 (hereafter referred
123: to as ``Par1''), we have acquired two sets of new observations to help tackle
124: the question at hand. To follow-up a subset of the $i$-dropouts found in YWC03,
125: two regions in Par1 were imaged by the Near Infrared Camera and Multi-Object
126: Spectrometer (NICMOS) Camera 3 (NIC3). During the course of these NIC3
127: observations, the ACS instrument was working in parallel, which resulted in a
128: new ACS parallel field (hereafter referred to as ``Par2'') that is several
129: arcminutes away from the original one. The layout of these two ACS fields are
130: show in Fig. 1. Note that they are in the general direction of the Virgo
131: Cluster. We describe all these observations below.
132:
133: \subsection{ACS parallel observations of Par1}
134:
135: The center of Par1 is at $RA=12^h43^m32^s$, $Dec=11^o40'32''$. The ACS data
136: were taken during \hst\, Cycle 11 soon after the instrument was installed in
137: March, 2002. In fact, the observations spanned from April 28 to June 19, 2002,
138: and were acquired during the execution of the Guest Observer (GO) program ID
139: 9043 (PI. Tonry), when it was using the Wide-Field Planetary Camera 2 (WFPC2) as
140: the primary instrument to image NGC4647, which is a member of the Virgo
141: Cluster. Only \acsi\, and \acsz\, filters were used for these ACS parallels.
142: In total, 15 images were taken in \acsi\, and 27 images were taken in \acsz. The
143: total exposure time in these two bands is 2.65 hours and 4.28 hours,
144: respectively.
145:
146: \subsection{NIC3 observations in Par1}
147:
148: Our NIC3 observations of Par1 were carried out as a GO program (PID 9780) in
149: \hst\, Cycle-12. Two regions in Par1, designated as fields ``Sub1'' and
150: ``Sub2'', were observed in $J_{110}$ and $H_{160}$. Due to an unfortunate
151: system glitch that caused NICMOS enter a short safing-mode period in early
152: August 2003, our program missed its originally planned visit windows, and its
153: execution had to stretch from January to December 2004 through rescheduled
154: visits.
155:
156: For each field, the observations were done at 15 dithering positions in
157: $J_{110}$ and 25 positions in $H_{160}$. The observation mode was MULTIACCUM
158: with SPARS64 sequence. The exposure time at each dithering position is 512
159: seconds, resulting in a total integration time of 2.13 hours in $J_{110}$ and
160: 3.56 hours in $H_{160}$, respectively.
161:
162: \subsection{ACS parallel observations of Par2}
163:
164: While our NIC3 program was not executed in August 2003 as originally
165: scheduled, part of its ACS parallel observations were still automatically
166: carried out on August 5 and 6, 2003 as planned in default.
167: This resulted in Par2, which is about 8$'$ to the
168: north of Par1. The field center is $RA=12^h43^m30^s$, $Dec=11^o49'21^s$.
169: In total, 5 images in $z_{850}$, 12 images in $i_{775}$ and
170: 6 images in $g_{475}$ were taken in this field, giving a net exposure time of
171: 1.00, 1.80 and 0.89 hours in these three bands, respectively.
172:
173: \section{Data Reduction and Photometry}
174:
175: In this section, we discuss the data reduction and photometry of all
176: relevant \hst\, images.
177:
178: \subsection{ACS data reduction}
179:
180: When the YWC03 paper was written (September 2002), the HST ``On-the-Fly
181: Reprocessing'' (OTFR) calibration pipeline could only process ACS data up to
182: the step of geometric distortion correction. The {\it Multidrizzle} package
183: (Koekemoer et al. 2002), now a standard tool for mosaicing ACS images using
184: the drizzle algorithm (Fruchter \& Hook 2002), was then not yet well known
185: and tested. Therefore, the ACS mosaics of Par1 used in YWC03 were created by
186: stacking the flat-fielded, distortion-corrected images utilizing the
187: well-tested {\it IMCOMBINE} task of IRAF that had been used by the community
188: for decades.
189:
190: For this current study, the ACS data of Par1 were retrieved again from the
191: \hst\, data archive. In so doing, the requested data were automatically
192: reprocessed by the OTFR pipeline using the best reference files (bias, dark
193: current, flat field, etc.) currently available. New mosaics were then made from
194: these reprocessed images by using the now-fully-matured {\it Multidrizzle}
195: routine as available in STSDAS under PyRAF. The reduction of the Par2 ACS data
196: followed the same procedure.
197:
198: Some ACS/WFC OTFR-processed images are known to show a quadrant-to-quadrant
199: ``jump'' in background because of the varing residual bias levels in the four
200: ACS/WFC readout amplifiers (Pavlovsky et al. 2005).
201: While this bias offset is
202: minimal in the data of Par2 ($\sim 0.1 e^{-}$ level), it is obvious in most
203: of the Par1 images ($\sim 10 e^{-}$ level). Although such ``jumps'' do not
204: have significant impact on the photometry of compact sources in general, they
205: do affect the sources that are close to quadrant boundaries. Therefore, before
206: stacking the Par1 data, such jumps were removed by subtracting the background
207: from each quadrant using our home-grown routines. The quadrant offsets were
208: reduced to $\sim 1 e^{-}$ level after this process.
209:
210: The final mosaics were created following the standard drizzling procedures
211: implemented in {\it Multidrizzle}, and were normalized to unit exposure time
212: (count per second). A square kernel was used in drizzling, and
213: the linear size of drop was set to 0.9 (the ``final\_pixfrac'' parameter of
214: {\it Multidrizzle}). To get a finer spatial resolution, the drizzle scale (the
215: ``final\_scale'' parameter) was set to 0.6, which resulted in an output pixel
216: scale of 0.03$''$/pixel. For each field, the output mosaics in different
217: passbands were all registered to the same reference position in the process.
218: The absolute astrometry of the final mosaics was calibrated by using the
219: compact objects visible in the Digital Sky Survey images.
220:
221: In order to provide a reference frame for Par1 NIC3 data reduction, another
222: version of Par1 ACS mosaics was also created by setting drizzle scale to 1.0,
223: i.e., preserving the native pixel scale (0.05$''$/pixel) of the ACS/WFC. We
224: will refer to these lower-resolution mosaics and the higher-resolution ones
225: mentioned above as ``LOWRS'' and ``HIGHRS'' ACS mosaics, respectively. They
226: each serve different purposes in the analysis process.
227:
228: As the original pixels were resampled during drizzling (and during
229: geometric distortion correction as well), the final mosaics all have correlated
230: pixel noise. The signal-to-noise ratio (S/N) of a given object derived directly
231: from the final mosaic would therefore be overestimated, and hence its
232: photometric error would be underestimated. We used a routine, kindly provided
233: to one of us (HY) by Dr. Mark Dickinson, to calculate the correlation amplitude
234: and hence to calculate proper statistical error associated with each pixel. The
235: error map (so-called ``RMS'' map) derived in this way are widely used,
236: e.g., in the data analysis of GOODS (e.g., Dickinson et al. 2004).
237:
238: \subsection{NIC3 data reduction}
239:
240: All data obtained by NICMOS show a number of persistent anomalies that
241: cannot currently be handled by the OTFR pipeline. Therefore, the OTFR-processed
242: NIC3 data of Sub1 and Sub2 fetched from the archive were further reduced before
243: stacking.
244:
245: An additional count-rate and wavelength dependent non-linearity of
246: NICMOS was recently identified by the STScI NICMOS team (NICMOS Instrument
247: Science Reports 2006-001, 002 and 003; see also
248: http://www.stsci.edu/hst/nicmos/performance
249: /anomalies/nonlinearity.html).
250: This non-linearity was removed from the data by using the routine provided
251: by the team.
252:
253: The quadrant bias of NICMOS also has a stochastic behavior similar to
254: what described above for ACS. As a result, a bias-corrected NICMOS image using
255: the standard bias reference file usually show ``pedestals'' from quadrant to
256: quadrant. Such pedestals were removed by using the {\it PESKY} task included
257: in the STSDAS package.
258:
259: All OTFR-processed NIC3 data have a persistent, additive pattern across
260: entire field. This pattern is different in different passbands, but is rather
261: stable with respect to time. To remove this pattern, a ``pattern image'' in
262: each band was first created by stacking all the pedestal-corrected images in
263: this band. The pattern image was then subtracted from each individual image.
264:
265: After all the above steps, the individual images were combined using the
266: series of drizzle tasks included in the {\it DITHER} package of STSDAS in
267: IRAF. Each input science image was weighted by a weight image calculated from
268: its associated error array stored in the [ERR,2] extension of the
269: OTFR-processed image. For the purpose of matched-aperture photometry that will
270: be discussed below, the LOWRS Par1 ACS $z_{850}$-band mosaic was binned in
271: $2\times 2$ and then used as the reference. The NIC3 images were all mapped
272: (i.e., rotated and registered)
273: to this binned ACS image during the drizzling process. The drizzle
274: scale was set to 0.5, which resulted in a final resolution of 0.10$''$/pixel,
275: i.e., the same as the binned LOWRS ACS image. The final mosaics were
276: normalized to unit exposure time as well.
277:
278: The associated RMS maps of all these NIC3 mosaics were derived in the way
279: similar to what described above for the ACS mosaics.
280:
281: \subsection{ACS source detection}
282:
283: As compared to the ACS mosaics used in YWC03, the new mosaics created
284: through the imporved reduction process (see \S 3.1) are of much higher quality,
285: and hence we can push to a significantly lower thresold for source detection.
286:
287: SExtractor (Bertin \& Arnouts 1996) was run on the HIGHRS mosaics in
288: dual-image mode to extract source catalogs for both Par1 and Par2.
289: ``MAG\_AUTO'' magnitudes were used to best approach the total magnitudes.
290: A number of source catalogs were
291: obtained for different purposes. The \acsi-band and \acsz-band mosaics
292: were used alternatively as the detection images to perform ``matched-aperture
293: photometry'', i.e., measuring the magnitudes of a given source in different
294: bands through the same aperture defined by its appearance on the detection
295: image. We will refer to these catalogs as ``$i_{775}$-based'' and
296: ``$z_{850}$-based'' ACS catalogs, respectively. Source detection was performed
297: at the threshold of 0.8~$\sigma$, using a $5\times 5$ Gaussian filter with a
298: full-width-at-half-maximum (FWHM) of 2.5 pixels. A minimum number of 4
299: connected pixels were required for a source to be included.
300:
301: As Par1 was observed in parallel mode during the very early stage of ACS,
302: the data were all taken at a high gain setting of 4.0~$e^-$/ADU, probably for
303: safety reason (i.e., avoiding saturation). In order to make the zeropoints
304: derived at a gain setting of 1.0~$e^-$/ADU universally applicable, the ACS OTFR
305: pipeline multiplies the flat-field reference file by the gain of the object
306: image before flat-field correction. In the end, an OTFR-processed ACS image
307: always has an effective gain of 1.0~$e^-$/ADU. When YWC03 was written,
308: unfortunately, the authors were not aware of this --- at that time
309: poorly-documented --- step in the OTFR processing,
310: and made a redundant correction for the gain by adding
311: $2.5\times log10(4.0)=1.505$ to the zeropoints. As a result, while the
312: candidate selection in YWC03 remains valid, all magnitudes reported in that
313: paper were too faint by 1.505 mag. This error has been reported in
314: a footnote in YW04.
315:
316: In this current study, the zeropoints published in the latest ACS
317: handbook were used; namely, 26.068, 25.654 and 24.862 mag for $g_{475}$,
318: $i_{775}$ and $z_{850}$, respectively. The 3~$\sigma$ limits within a circular
319: aperture of 0.1$^{''}$ in radius are 29.1 and 28.3 mag in \acsi\, and \acsz\,
320: in Par1, and 28.9, 28.7 and 27.9 mag in \acsg, \acsi\, and \acsz\, in Par2,
321: respectively.
322:
323: \subsection{NIC3 source detection}
324:
325: Photometry of Sub1 and Sub2 NIC3 mosaics in Par1 was also done in
326: dual-image mode, using $J_{110}$ and $H_{160}$ mosaics alternatively as the
327: detection images.
328: Similarly, we will refer to these catalogs as ``\nicJ-based'' and
329: ``\nicH-based'' catalogs, respectively. The detection filter was a
330: $5\times 5$ Gaussian filter with a FWHM of 3 pixels, the detection threshold
331: was set to 0.4~$\sigma$, and a minimum of 2 connecting pixels were required.
332: Again, MAG\_AUTO magnitudes were adopted.
333: As the magnitude zeropoints
334: provided by the OTFR pipeline were no longer valid after the non-linearity
335: correction described in \S 3.2, we followed the recipe given by the NICMOS
336: team to derive the zeropoints as 23.242 and 23.139 mag in $J_{110}$ and
337: $H_{160}$, respectively.
338:
339: During this process, ACS magnitudes of the NIC3 sources were also extracted
340: in order to properly measure their ACS-to-NICMOS colors for later analysis.
341: As the NIC3 mosaics were created by registering to the reference frame defined
342: by the $2\times 2$ binned LOWRS ACS mosaics, matched-aperture photometry between
343: NIC3 and ACS passbands was fairly straightforward. The detection images were
344: still the NIC3 mosaics, and the object images were $2\times 2$ binned ACS
345: mosaics that were convolved by the PSF of the \nicJ\, images (derived using the
346: TinyTim tool). The binning operation was sum, which preserves the original ACS
347: magnitude zeropoints.
348:
349: \section{Overdensity of \boldmath $i$-dropouts}
350:
351: Using the newly produced ACS mosaics of Par1, new NIC3 data obtained at two
352: pointings within Par1, and a new ACS parallel field Par2 for comparison, here
353: we will re-examine the overdensity of $i$-dropouts in Par1 reported by YWC03.
354:
355: \subsection{Overdensity of \boldmath $i$-dropouts in Par1}
356:
357: YWC03 reported 30 $i$-dropouts in Par1, whose \acsz\, magnitudes range from
358: 26.8 to 28.3 mag. Using the correct zeropoints, their true $z_{850}$-band
359: magnitude range should be 25.3 to 26.8 mag (see \S 3.3). The
360: corresponding cumulative surface density thus is $\sim$ 2.7 per arcmin$^2$
361: to $z_{850}=26.8$ mag. Taken at its face value, this number is 3--4 times
362: higher than those in other deep, high galactic-latitude ACS fields, including
363: the Hubble Ultra Deep Field (HUDF; Beckwith et al. 2006).
364: The $i$-dropouts in YWC03 were selected based on their invisibility in the
365: \acsi\, image, i.e., by selecting sources that have $S/N\geq 5$ in
366: $z_{850}$ but have $S/N<2$ in $i_{775}$. As a result, about 2/3 of those
367: objects have $i_{775}-z_{850}>2.0$ mag. Other studies utilizing ACS data
368: (including YW04) all have adopted less stringent selection criterion of
369: $i_{775}-z_{850}\geq 1.3$ mag or $i_{775}-z_{850}\geq 1.5$ mag.
370: Therefore, the high surface density of YWC03 is even more difficult to be
371: reconciled with all the other later work.
372:
373: To study this problem in detail, here we reselect $i$-dropouts in Par1
374: from the new $z_{850}$-based ACS catalog as described in \S 3.3. We only
375: consider sources with $S/N\geq 5$ in \acsz, and adopt the color criterion of
376: $i_{775}-z_{850}\geq 1.3$ mag for ease of comparison with other results. The
377: candidates thus selected are visually examined to eliminate false detections
378: such as residual cosmic rays and image defects. The final $i$-dropout sample
379: consists of 356 objects to $z_{850}=28.0$ mag. This is much deeper than what
380: we achived in YWC03, and is the result of the much improved data quality. In
381: particular, 198 of them have $z_{850}\leq 26.8$ mag (among which 21 have
382: $i_{775}-z_{850}\geq 2.0$ mag), which is a factor of 20 more than the
383: $i$-dropouts selected in the HUDF
384: (YW04; Bunker et al. 2004) to the same depth. This confirms -- and reinforces
385: -- the unusually large overdensity of $i$-dropouts seen by YWC03. The
386: distribution of these $i$-dropouts in Par1 are shown in Fig. 2.
387:
388: A necessary condition for an $i$-dropout being a legitimate candidate of
389: galaxy at $z\approx 6$ is that it must be a dropout in bluer bands.
390: However, Par1 does not have any data to the blue of \acsi.
391: Fortunately, as a subset of
392: these $i$-dropouts have NIC3 observations, we can use their ACS-to-NIC3 colors
393: to judge whether they are consistent with being galaxies at $z\approx 6$.
394:
395: \subsection{Optical-to-IR colors of \boldmath $i$-dropouts in Par1}
396:
397: In total, 21 $i$-dropouts are detected in our NIC3 data. Fig. 3 displays
398: the image cut-outs of one such objects. The optical-to-IR color-color diagrams
399: of all these sources are shown in the left panel of Fig. 4.
400: For comparison, the right panel of Fig. 4 shows the similar color-color
401: diagrams from the HUDF.
402:
403: It is clear that the $i$-dropouts in Par1 reside in the similar regions in
404: the optical-to-IR color space as the $i$-dropouts in the HUDF. In particular,
405: they are well separated from brown dwarf stars and E/S0 galaxies at
406: lower redshifts, which are the most common contaminators to $z\approx 6$
407: galaxy samples selected by the drop-out technique.
408:
409: \subsection{A similar overdensity of \boldmath $i$-dropouts in Par2}
410:
411: In order to compare with the results obtained in Par1, we perform
412: $i$-dropout selection in Par2 in a similar way. As this field has $g_{475}$
413: data, an additional constraint of non-detection in $g_{475}$ at 2~$\sigma$
414: level is imposed for the selection. The final $i$-dropout sample consists of
415: 20 objects to $z_{850}=26.5$ mag, which, while much less than the ones in Par1,
416: is still a factor of $\sim 7$ more than those found in the HUDF to the same
417: depth; in other words, the overdensity of $i$-dropouts also seems to present in
418: Par2.
419:
420: \section{Alternative Interpretation of the Overdensity}
421:
422: The analysis in \S 4 indicates that the $i$-dropouts in Par1 are consistent
423: with being galaxies at $z\approx 6$, and that the overdensity is also present
424: in Par2. If this interpretation is true, we will have to draw the conclusion
425: that the overdensity of $i$-dropouts in this region is due to an extremely
426: unusual concentration of $z\approx 6$ galaxies over $\sim 10$~Mpc scale. In this
427: section, however, we present another possibility.
428:
429: \subsection{Excess of field population in Par1 \& Par2}
430:
431: The alternative interpretation is triggered by an unusual field population
432: seen in both Par1 and Par2. This is illustrated in Fig. 5, which shows the
433: \acsz\, vs. \acsi$-$\acsz\, color-magnitude diagrams in these fields using the
434: \acsz-based catalogs, together with the corresponding diagrams constructed in
435: the HUDF based on the photometry from YW04. The left panel only plots the
436: sources at $S/N>5$, while the right panel includes all the sources at $S/N>3$.
437: A striking feature in this figure is the large concentration of objects at the
438: faint end in the two parallel fields, which is not seen in the HUDF (and other
439: deep ACS fields). Due to the shallower depth in Par2, this population is at
440: the verge of being detectable at $S/N>5$; however, it gets very prominent if
441: the threshold is lowered to $S/N>3$.
442: YWC03 noticed only a hint of this excess in Par1; the
443: source detection back then did not reach a sufficient depth that could
444: unambiguously confirm its existence.
445:
446: This excess of faint field objects in Par1 and Par2 is present in all
447: passbands. To quantify this excess, Fig. 6 compares the surface densities of
448: the detected sources in these two fields and the ones in the HUDF, derived
449: at two different threshold levels of $S/N>3$ (left panel) and $S/N>5$ (right
450: panel). Again, while the excess in Par2 is only detectable when the sources
451: of low $S/N$ are included, the excess in Par1 is obvious even at a high
452: source detection threshold. This excess starts at about 26.0 mag in both Par1
453: and Par2, and extends to a much fainter level. In particular, the
454: deep mosaics in Par1 allow this excess to be traced to at least $\sim 28.2$ and
455: 27.2 mag in \acsi\, and \acsz, respectively. To further ensure the existence of
456: the excess, we split the available images of Par1 into two groups and
457: created shallower stacks in both bands. The same excess was detected as well.
458:
459: The $i$-dropout surface densities are also superposed in Fig. 6. We point
460: out that the $i$-dropout excess in the two parallel fields resides in the
461: region where the field population excess happens. This is further discussed
462: later in the paper.
463:
464: \subsection {Resolving diffuse light from interacting galaxy pair}
465:
466: What is the nature of this field population excess? The answer probably
467: lies in the proximity of Par1 and Par2 to the galaxy pair M60/NGC4647, which
468: is several arcminutes away to the South (see Fig. 1). Extended diffuse
469: emissions of different natures have been detected around some nearby galaxies,
470: for example in NGC 5907 (Shang et al. 1998; Zheng et al. 1999) and in M31
471: (e.g., Ferguson et al. 2002). While there is no previous report in the
472: literature, we speculate that such extended diffuse emission also exists around
473: M60/NGC4647, and propose that the excess population is due to the fact that we
474: have resolved the diffuse light extending from this interacting pair.
475: If we adopt an average distance of $\sim 15.7$~Mpc
476: to M60/NGC4647 (from NASA/IPAC Extragalactic Database), Par1 and Par2 are
477: $\sim$ 32 and 69~kpc away from the center of pair, respectively. Therefore,
478: what we have detected are most likely the halo stars of the two galaxies and/or
479: the tidal ``debris'' repelled from this pair during the course of the
480: interaction.
481:
482: Here we exam the property of this excess more closely. We only use Par1 for
483: this purpose, as the data in this field are much deeper than in Par2. Fig. 7
484: plots the surface densities of $S/N>5$ sources in Par1 as functions of
485: magnitudes in \acsi\, (left) and \acsz\,(right), and compares them with the
486: results in the HUDF. The insets show the net excess after subtracting the
487: surface density calculated in the HUDF. Tab. 1 lists the surface densities
488: of the net excess as a function of magnitudes.
489:
490: Taken at face value, the net excess in Par1 as shown in Fig. 7 has a total
491: magnitude of 15.0 and 15.7 mag in \acsi\,
492: and \acsz, respectively. The average surface brightness from the net excess
493: (i.e., spreading the total integrated light of the net excess over the entire
494: field of Par1), on the other hand, is 26.6 and 27.2 mag/arcsec$^2$ in \acsi,
495: and \acsz, respectively. If this excess is
496: isotropic with respect to M60/NGC4647, the total magnitude integrated over
497: an annulus 3.4$^{'}$ in width (i.e., about the width of one ACS field) at the
498: distance of Par1 (centering on the galaxy
499: pair) is 12.2 mag in \acsi\, and 12.9 mag in \acsz. For comparison, the total
500: flux of M60 in Johnson I-band is 8.46 mag (or 0.925~Jy, i.e., 8.98 mag in AB)
501: within 84$^{''}$ aperture (Boroson, Strom \& Strom 1983). Therefore, the total
502: diffuse light within this annulus is about a few percent of the central galaxies.
503:
504: The net excess peaks at about 27.1 and 26.6 mag in \acsi\, and \acsz.
505: If these objects are indeed at the distance of Virgo Cluster, these values
506: correspond to absolute magnitudes of $-3.9$ and $-4.4$ mag, respectively,
507: which are in the regime of giants/supergiants or low luminosity globular
508: clusters.
509:
510: We shall point out that the excess that we detected relates more closely to
511: M60/NGC4546 rather than to the Virgo Cluster in general. A supporting evidence
512: comes from the spatial gradient of this excess, which decreases in amplitude
513: when moving away from the pair. This is demonstrated in Fig. 8, which plots the
514: spatial distribution of all $S/N>5$ sources at \acsz$>26.0$~mag detected in the
515: \acsz-based catalog of Par1. The gradient is obvious: there are more sources to
516: the south, which is the direction where the galaxy pair lies.
517:
518: Another piece of evidence
519: is that the amplitude of this excess (see Fig. 7 and Tab. 1) cannot be
520: attributed to the intracluster stellar populations of the Virgo Cluster alone.
521: We only consider Virgo's intracluster globular clusters (IGC) and red giants
522: (IRG), as they are the only two type of stellar objects that are within the
523: luminosity range of the detected excess. The {\it total} contribution from
524: Virgo's intracluster globular clusters (IGC) at $> 26.0$~mag, calculated using
525: the Virgo's IGC surface density at the bright-end (e.g., Williams et al. 2007a;
526: Cohen et al. 2003) and the symmetry of globular cluster luminosity function, is
527: only $\sim 0.3$--3.0~arcmin$^{-2}$ (for any reasonable color transformation
528: from F814W to F775W), depending on the proximity to a member galaxy. On the
529: other hand, the deep \hst\, surveys of Virgo's IRG (e.g., Ferguson, Tanvir \&
530: von Hippel 1998; Durrell et al. 2002; Williams et al. 2007b) show that their
531: cumulative surface density to $\sim 28.0$ mag can be as high as
532: $\sim 110$~arcmin$^{-2}$ (again for any reasonable F814W to F775W
533: transformation). However, this is still a factor of 5--6 lower than
534: what shown in Fig. 7 and Tab. 1.
535:
536:
537: \subsection{Relation between $i$-dropout overdensity and field object overdensity}
538:
539: We suggest that the $i$-dropout overdensity observed in Par1 \& Par2
540: might be related to the field object overdensity in the same regions.
541: The field object excess becomes significant at 26.0 mag in both bands, and
542: extends to at least 28.2 mag in \acsi\, and 27.2 mag in \acsz\, in the deeper
543: Par1 images (see the insets in Fig. 8) where the excess can be traced to a
544: faint level at high confidence. The $i$-dropouts selected in these fields
545: resides in the \acsz\, magnitude range that the field object excess spans
546: (Fig. 6). The spatial distribution of the $i$-dropouts also seems to follow
547: the gradient that seen in the distribution of the excess field population
548: (see Fig. 2 \& 7).
549:
550: This excess field population is also red in color. Fig. 9 shows the surface
551: densities of the sources in the above mentioned magnitude ranges as a function
552: of \acsi$-$\acsz\, color. For comparison, the surface densities of the HUDF
553: sources in the same magnitude ranges are also shown. The insets plot the net
554: excess using the HUDF result as the reference. The excess population peaks at
555: \acsi$-$\acsz$\approx 0.35$ mag in the \acsi\, image, and peaks at
556: \acsi$-$\acsz$\approx 0.60$ mag in the \acsz\, image. Note that the distribution
557: extends continuously beyond \acsi$-$\acsz$\geq 1.3$ mag, which is the regime
558: where the $i$-dropouts are selected.
559:
560: If the $i$-dropouts are indeed some stellar objects related to
561: M60/NGC4647, what type of stars could they be? Thus far we have
562: not yet been able to identify the satisfactory candidates. As Fig. 4 shows,
563: brown dwarfs are not likely to contribute much to our $i$-dropout sample
564: because their optical-to-IR colors are different. In addition, their
565: luminosities are too low. The only possibility then left is dusty stellar
566: population, for example, Asymptotic Giant Branch (AGB) stars. However, the
567: optical-to-IR colors of AGB stars are quite different from the $i$-dropouts as
568: well. This is illustrated in Fig. 10, where the synthesized colors of
569: AGB stars generated from the model spectra of Lancon \& Mouhcine (2002)
570: are superposed on the color-color diagrams of the $i$-dropouts in Par1.
571: We should also point out that no external dust extinction to any known types
572: of stars could produce such colors; no know extinction law could create such
573: a red color in $i_{775}-z_{850}$ and yet keep $z_{850}-J_{110}$ and
574: $J_{110}-H_{160}$ colors around zero.
575:
576: The red, excess field objects that we have discovered here seem to have some
577: connection to the so-called ``Red Halo Phenomenon'' (Zackrisson et al. 2006),
578: which referrs to the very red colors of the halos found around some galaxies
579: (e.g., Zibetti, White \& Brinkmann 2004; Zibetti \& Ferguson 2004; Taylor et al.
580: 2005). Such red halos cannot be easily explained by standard stellar
581: populations. The $i_{775}-z_{850}$ color shown in Fig. 9 quantitively agrees
582: with that of the halo around an edge-on disk galaxy in the HUDF reported by
583: Zibetti \& Ferguson (2004). Therefore, the excess field objects that we see
584: might represent the first example where such red halos have been resolved, and
585: the $i$-dropouts might be the most extreme sources among this population that
586: is most difficult to explain.
587:
588: \section{Summary}
589:
590: In this paper, we use new and existing \hst\, data to closely examine the
591: large overdensity of $i$-dropouts that YWC03 found in a deep ACS parallel
592: field. The new data include NIC3 imaging of a subset of the $i$-dropouts, and
593: a set of new ACS parallel data in a nearby region obtained during the NIC3
594: observations. We confirm that a factor of $\sim 20$ overdensity of $i$-dropouts
595: does exist in the parallel field of YWC03, and that the optical-to-IR colors of
596: these objects are similar to those selected in other deep ACS fields such as
597: the HUDF.
598:
599: However, these $i$-dropouts might {\it not} be galaxies at $z\approx 6$ as
600: one would expect. During the course of our investigation, we have found a large
601: excess of discrete sources in both the field of YWC03 and the new ACS parallel
602: field. This excess is most likely due to a stellar population that is related
603: to M60/NGC4647, an interacting galaxy pair in the Virgo Cluster, which lies
604: several arcminutes to the south of the ACS parallel fields. At the distance
605: of the Virgo Cluster, the ACS parallel fields are several tens of kpc away
606: from the galaxy pair, and hence the stellar objects are likely tidal ``debris''
607: expelled from the pair and/or distant halo stars belong to these two galaxies.
608: Both the spatial distribution and the amplitude of the excess indicate that
609: it is a population related more closely to the galaxy pair rather than to the
610: Virgo's intracluster populations.
611:
612: The $i$-dropouts in the field of YWC03 are within the magnitude range where
613: the field object excess occurs. Their distribution also shows similar spatial
614: gradient of the field object excess. Most importantly, the field object excess
615: is red in color, and the red wing of its color distribution extends well into
616: the regime where the $i$-dropouts were selected. For these reasons, we believe
617: these $i$-dropouts are part of the stellar objects that constitute the field
618: object excess rather than galaxies at $z\approx 6$.
619:
620: Nevertheless, we should caution that we still cannot completely rule out
621: the possibility that these $i$-dropouts might indeed be $z\approx 6$ galaxies.
622: On one hand, we do not have any definite evidence against this interpretation.
623: On the other hand, we have not yet found a stellar population that can
624: reasonably explain the colors of these $i$-dropouts. To unambiguously identify
625: their nature, spectroscopic observations of these objects are in demand. As the
626: brightest sources in our sample are brighter than 26.0 mag in \acsz, this
627: might be feasible with the existing spectrographs at our current 8--10m class
628: telescopes.
629:
630: \acknowledgments
631:
632: The authors thank the referee for the useful comments. We also thank Seth
633: Cohen, Inese Ivans, Daisuke Kawata, Huub R\"{o}ttgering and Fran\c{c}ois
634: Schweizer for helpful discussions.
635: Support for program HST-GO-09780.* was provided by NASA through a grant from
636: the Space Telescope Science Institute, which is operated by the Association of
637: Universities for Research in Astronomy, Inc., under NASA contract NAS 5-26555.
638: This research has made use of the NASA/IPAC Extragalactic Database (NED) which
639: is operated by the Jet Propulsion Laboratory, California Institute of
640: Technology, under contract with the National Aeronautics and Space
641: Administration.
642:
643: \begin{thebibliography}{}
644:
645: \bibitem[]{659} Beckwith, S. V. W., et al. 2006, AJ, 131, 1729
646: \bibitem[]{660} Bertin, E. \& Arnouts, S. 1996, A\&AS, 117, 393
647: \bibitem[]{661} Bouwens, R. J., et al. 2003, ApJ, 595, 589
648: \bibitem[]{662} Bouwens, R. J., et al. 2004, ApJ, 606, L25
649: \bibitem[]{663} Bruzual, A. G. \& Charlot, S. 1993, ApJ, 405, 538
650: \bibitem[]{664} Bunker, A. J., Stanway, E. R., Ellis, R. S., \& McMahon, R. G. et al. 2004, MNRAS, 355, 374
651: \bibitem[]{665} Cohen, S. H., et al. 2003, AJ, 125, 1762
652: \bibitem[]{666} Dickinson, M., et al. 2004, ApJ, 600, L99
653: \bibitem[]{667} Durrell, P. R., et al. 2002, ApJ, 570, 119
654: \bibitem[]{684} Ferguson, A. M. N., et al. 2002, AJ, 124, 1452
655: \bibitem[]{668} Ferguson, H. C., Tanvir, N. R. \& von Hippel, T. 1998, Nature, 391, 29
656: \bibitem[]{669} Ford, H., et al. 2003, in ``Future EUV and UV Visible Space Astrophysics Missions and Instrumentation'', eds. J. C. Blades \& O.H. Siegmund, Proc. SPIE, Vol. 4854, 81
657: \bibitem[]{670} Fruchter, A. S. \& Hook, R. N. 2002, PASP, 114, 144
658: \bibitem[]{671} Koekemoer, A. M., Fruchter, A. S., Hook, R. N., \& Hack, W. 2002, in ``The 2002 HST Calibration Workshop : Hubble after the Installation of the ACS and the NICMOS Cooling System'', eds. Santiago Arribas, Anton Koekemoer, and Brad Whitmore, Baltimore, USA, p.337
659: \bibitem[]{672} Lancon, A. \& Mouhcine, M. 2002, A\&A, 393, 167
660: \bibitem[]{673} Madau, P. 1995, ApJ, 441, 18
661: \bibitem[]{674} Pavlovsky, C., et al. 2005, "ACS Data Handbook", Version 4.0, Baltimore: STScI.
662: \bibitem[]{692} Shang, Z., et al. 1998, ApJ, 504, L23
663: \bibitem[]{675} Somerville, R. S., et al. 2004, ApJ, 600, L171
664: \bibitem[]{676} Stanway, E. R., Bunker, A. J., \& McMahon, R. G. 2003, MNRAS 342,439
665: \bibitem[]{677} Sparks, W. B., et al. 2001, ACS Default (Archival) Pure Parallel Program (ISR01-06)
666: \bibitem[]{678} Spergel, D. N., et al. 2003, ApJS, 148, 175
667: \bibitem[]{679} Spergel, D. N., et al. 2007, ApJS, 170, 377
668: \bibitem[]{680} Steidel, C. C. \& Hamilton, D. 1992, AJ, 104, 941
669: \bibitem[]{699} Taylor, V. A., Jansen, R. A., Windhorst, R. A., Odewahn, S. C., \&
670: Hibbard, J. 2005, ApJ 630, 78
671: \bibitem[]{681} Williams, B. F., et al. 2007a, ApJ, 654, 835
672: \bibitem[]{682} Williams, B. F., et al. 2007b, ApJ, 656, 756
673: \bibitem[]{683} Yan, H., Windhorst, R. \& Cohen, S. 2003, ApJ, 585, L93 (YWC03)
674: \bibitem[]{684} Yan, H. \& Windhorst, R., 2004, ApJ, 600, L1 (YW04)
675: \bibitem[]{705} Zackrisson, E., et al. 2006, ApJ, 650, 812
676: \bibitem[]{706} Zheng, Z., et al. 1999, AJ, 117
677: \bibitem[]{707} Zibetti, S. \& Ferguson, A. M. N., 2004, MNRAS, 352, L6
678: \bibitem[]{708} Zibetti, S., White, S. D. M \& Brinkmann, J. 2004, MNRAS, 347, 556
679:
680: \end{thebibliography}
681:
682: \clearpage
683:
684: \begin{deluxetable}{ccc}
685: \tablecaption{Surface density of the $S/N>5$ source net excess in Par1 as a function of magnitude}
686: \tablewidth{0pt}
687: \tablehead{
688: \colhead{magnitude} &
689: \colhead{$\Sigma$(\acsi) (arcmin$^{-2})$} &
690: \colhead{$\Sigma$(\acsz) (arcmin$^{-2})$}
691: }
692: \startdata
693:
694: 25.55 & 0.51 & 1.35 \\
695: 25.65 & 1.96 & 0.93 \\
696: 25.75 & 2.52 & 2.62 \\
697: 25.85 & 5.13 & 2.51 \\
698: 25.95 & 5.54 & 4.67 \\
699: 26.05 & 7.11 & 6.99 \\
700: 26.15 & 9.60 & 13.13 \\
701: 26.25 & 14.58 & 15.86 \\
702: 26.35 & 18.93 & 19.76 \\
703: 26.45 & 29.74 & 29.16 \\
704: 26.55 & 34.35 & 30.55 \\
705: 26.65 & 36.93 & 24.61 \\
706: 26.75 & 38.54 & 18.89 \\
707: 26.85 & 42.51 & 12.57 \\
708: 26.95 & 43.04 & 8.78 \\
709: 27.05 & 47.17 & 7.04 \\
710: 27.15 & 45.84 & 8.32 \\
711: 27.25 & 43.59 & 2.24 \\
712: 27.35 & 36.64 & --- \\
713: 27.45 & 39.49 & --- \\
714: 27.55 & 37.16 & --- \\
715: 27.65 & 26.10 & --- \\
716: 27.75 & 24.45 & --- \\
717: 27.85 & 18.44 & --- \\
718: 27.95 & 11.06 & --- \\
719: 28.05 & 9.66 & --- \\
720: 28.15 & 0.58 & --- \\
721:
722: \enddata
723:
724: \tablenotetext{1.} {The net excess is as shown in the insets in Fig. 7, which is
725: calculated using the count in the HUDF as the reference.}
726:
727: \end{deluxetable}
728:
729: \clearpage
730: \begin{figure}
731: \plotone{f1.eps}
732: \caption{The locations of the fields studied in this work are shown on top of
733: the image from the Digital Sky Survey. Both
734: fields are several arcminutes away from the interacting galaxy pair,
735: M60 \& NGC4647, which are members of the Virgo Cluster. Par1 was taken in
736: parallel mode by the ACS when the WFPC2 was observing NGC4647, while Par2 was
737: taken by the ACS in parallel when NICMOS was observing two regions in Par1.
738: }
739: \end{figure}
740:
741: \clearpage
742: \begin{figure}
743: \plotone{f2.eps}
744: \caption{The left panel shows the locations of the $i$-dropouts
745: in Par1 as red dots on top of the \acsz-band mosaic of this field. The yellow
746: dots in the right panel show the locations of the $i$-dropouts that are
747: detected the NIC3 mosaic, which is the image in the background (note that
748: the objects that are seemingly close to the edge are actually well within the
749: region of good coverage).
750: }
751: \end{figure}
752:
753: \clearpage
754: \begin{figure}
755: \plotone{f3.eps}
756: \caption{The image stamps of one $i$-dropout in our sample. The stamps are
757: 6$^{''}$ on a side, and the circles in the middle are 0.2$^{''}$ in radius.
758: }
759: \end{figure}
760:
761: \clearpage
762: \begin{figure}
763: \plotone{f4.eps}
764: \caption{The optical-to-IR colors of the $i$-dropouts in Par1 (left) are
765: compared with those of the $i$-dropouts in the HUDF (right). The $i$-dropouts
766: are the red points with error bars, while the field objects are shown as the
767: black dots. M-, L- and T-type brown dwarf stars are the green, blue and brown
768: asterisks, respectively. The colors of E/S0 galaxy at $z=1$--3 are shown
769: as the yellow crosses. Two model galaxies at $5.5\leq z\leq 6.5$, constructed
770: using the models of Bruzual \& Charlot (2003) of different star formation
771: histories, are shown as the open magenta squares. The $i$-dropouts in Par1
772: occupy a similar region in the color space as do those in the HUDF.
773: }
774: \end{figure}
775:
776: \clearpage
777: \begin{figure}
778: \plotone{f5.eps}
779: \caption{Comparison of the \acsz\, vs. \acsi$-$\acsz\, diagrams between the
780: parallel fields and the HUDF. The objects plotted in the left panel are
781: limited to the sources with $S/N>5$, while those in the right panel are limited
782: to the sources with $S/N>3$. The dotted lines show the \acsi$-$\acsz$=1.3$
783: color threshold above which the $i$-dropouts are selected. A unique feature
784: in the two parallel fields is the large excess objects at the faint end, which
785: is not seen in the HUDF. Due to its shallower depth, Par2 does not show this
786: feature very clearly if only the sources at $S/N>5$ are included; the excess
787: becomes prominent when sources to $S/N=3$ are also included.
788: }
789: \end{figure}
790:
791: \clearpage
792: \begin{figure}
793: \plotone{f6.eps}
794: \caption{Comparison of the field source surface densities in the two parallel
795: fields (solid and dashed black curves) and in the HUDF (blue dotted curves) at
796: two threshold levels.
797: The \acsi-based catalogs are used for the statistics in \acsi-band (top), while
798: the \acsz-based catalogs are used for the statistics in \acsz-band (bottom).
799: The left panel shows the statistics based on sources with $S/N>3$, while the
800: right panel shows the statistics based on sources with $S/N>5$. The object
801: excess in the two parallel fields is obvious. The surface densities of
802: $i$-dropouts (all have $S/N>5$ in \acsz) in Par1, Par2 and HUDF are also shown
803: as the magenta, cyan and red curves, respectively.
804: }
805: \end{figure}
806:
807: \clearpage
808: \begin{figure}
809: \plotone{f7.eps}
810: \caption{Field source excess of Par1 as a function of magnitude. The solid
811: histograms are the surface densities in Par1, and the dotted histograms are
812: those in the HUDF. The \acsi-based catalogs are used for the figure in the
813: left panel, while the \acsz-based catalogs are used for the one in the right.
814: Only sources at $S/N>5$ are included for the statistics. The insets show the
815: net excess using the HUDF as the reference. The excess extends to $\sim$ 28.2
816: and 27.2 mag in \acsi\, and \acsz, respectively. The average surface
817: brightness of the net excess corresponds to 26.6 and 27.2 mag/arcsec$^2$ in
818: \acsi, and \acsz\, respectively. The total magnitudes of the net excess
819: integrated over Par1 are \acsi$=$15.0 and \acsz$=$15.7 mag.
820: }
821: \end{figure}
822:
823: \clearpage
824: \begin{figure}
825: \plotone{f8.eps}
826: \caption{The spatial location of the excess field population has a clear
827: gradient towards the direction of the interacting galaxy pair M60/NGC4647.
828: This figure plots the positions of the \acsz-based sources with $S/N>5$ and
829: \acsz$>26.0$ mag, which is the brightness level where the excess starts to
830: be significant. North is up and East is to left. M60/NGC4647 is to the south
831: of the field (see also Fig. 1). Note that the gap in the middle is caused
832: by the gap between the two CCD chips of the ACS/WFC.
833: }
834: \end{figure}
835:
836: \clearpage
837: \begin{figure}
838: \plotone{f9.eps}
839: \caption{The excess population in the parallel fields is red in color. The
840: solid histograms show the \acsi$-$\acsz\, color distributions of Par1 sources
841: that are within the magnitude ranges where the excess occurs, while the dotted
842: histograms show the distribution of the HUDF sources within the same magnitude
843: ranges. The insets show the distributions of the net excess using the HUDF as
844: the reference. The \acsi-based catalogs are used in the left panel, and the
845: \acsz-based catalogs are used in the right panel. The red wing of the excess
846: population in \acsz\, continuously extends to the regime where the $i$-dropouts
847: are selected.
848: }
849: \end{figure}
850:
851: \clearpage
852: \begin{figure}
853: \plotone{f10.eps}
854: \caption{While the $i$-dropouts in the two parallel fields might well be
855: part of the excess stellar population related to M60/NGC4647 instead of
856: galaxies at $z\approx 6$, we have not yet been able to identify the type of
857: stars that can create the colors of the $i$-dropouts. The most promising
858: candidates are AGB stars, which have the necessary luminosities and are red
859: in colors. However, AGB stars still cannot fully explain the colors of the
860: $i$-dropouts. To illustrate this, the same color-color diagrams as in the left
861: panel of Fig. 4 are shown here again, and the colors of AGB stars (cyan
862: squares) from the models of Lancon \& Mouchine (2002) are superposed. To avoid
863: confusion, only the models that satisfy \acsi$-$\acsz$>1.3$ mag are superposed
864: in the right panel.
865: }
866: \end{figure}
867:
868: \end{document}
869: