1: \documentclass[12pt, preprint]{aastex}
2: %\usepackage{graphicx}
3: %\usepackage[figuresright]{rotating}
4:
5: %\def\lsimeq{{_<\atop^{\sim}}}
6: %\def\gsimeq{{_>\atop^{\sim}}}
7: \slugcomment{Submitted to ApJ}
8: \shorttitle{The VLA-COSMOS Survey}
9: \shortauthors{Bondi et al.}
10:
11: \begin{document}
12:
13: \title{The VLA-COSMOS Survey: \\
14: III. Further Catalog Analysis and the Radio Source Counts}
15:
16: \author{M. Bondi\altaffilmark{1}, P. Ciliegi\altaffilmark{2},
17: E. Schinnerer\altaffilmark{3}, V. Smol\v{c}i\'{c}\altaffilmark{3},
18: K. Jahnke\altaffilmark{3}, C. Carilli\altaffilmark{4},
19: G. Zamorani\altaffilmark{2}}
20:
21: \altaffiltext{1}{INAF - Istituto di Radioastronomia,
22: Via Gobetti 101, I-40129 Bologna, Italy}
23: \altaffiltext{2}{INAF - Osservatorio Astronomico di Bologna, Via Ranzani 1,
24: I-40127, Bologna, Italy}
25: \altaffiltext{3}{Max-Planck-Institut f\"ur Astronomie, K\"onigsthul 17, D-69117
26: Heidelberg, Germany}
27: \altaffiltext{4}{National Radio Astronomy Observatory, P.O. Box O, Socorro,
28: NM 87801-0387, U.S.A.}
29:
30: \begin{abstract}
31: The VLA-COSMOS Large Project has imaged the 2 deg$^2$ COSMOS field with a
32: resolution of 1.5 arcsec and a sensitivity of about 11 $\mu$Jy (1$\sigma$)
33: yielding to a catalog of $\sim 3600$ radio sources. In this paper
34: we present a further analysis of the VLA-COSMOS Large Project catalog
35: of radio sources
36: aimed to: 1) quantify and correct for the effect of bandwidth smearing in the
37: catalog, 2) determine the incompleteness produced by
38: the noise bias and the resolution bias in the new catalog and
39: 3) derive the radio source counts at 1.4 GHz.
40:
41: The effect of bandwidth smearing on the radio sources in the catalog
42: was quantified
43: comparing the peak and total flux densities in the final mosaic and in
44: each of the individual pointings where the source was closest to the center
45: of the field.
46: We find that the peak flux densities in the original VLA-COSMOS Large Project
47: catalog have to be divided by a factor about 0.8 or 0.9, depending on the
48: distance from the mosaic center.
49:
50: The completeness of the radio catalog has been
51: tested using samples of simulated radio sources with different angular
52: size distributions.
53: These simulated sources have been added to the radio image
54: and recovered using the same techniques used to produce the radio
55: catalog. The fraction of missed sources as a function of the total flux
56: density is a direct measure of the incompleteness.
57:
58: Finally, we derived the radio source counts down to 60 $\mu$Jy
59: with unprecedented good statistics. Comparison to the findings of
60: other surveys shows good agreement in the flux density range 0.06-1 mJy
61: confirming the upturn at $\sim 0.5$ mJy and a possible decline of the
62: source counts below $\sim 0.1$ mJy.
63:
64: \end{abstract}
65: \keywords{Surveys -- Radio continuum: galaxies -- Methods: data analysis}
66:
67: \section{Introduction}
68:
69: In recent years, we have experienced a renaissance of deep radio surveys,
70: usually associated with multi-waveband observational programmes mainly
71: designed to
72: probe the formation and evolution of galaxies and Super Massive Black Holes
73: (SMBH).
74: These surveys, mostly carried out at 1.4 GHz, have enabled the study of
75: sub-mJy and $\mu$Jy radio source
76: populations (hereafter $\mu$Jy population) and in particular
77: the determination, with robust statistics, of the radio source counts down to
78: a few tens
79: of $\mu$Jy \citep[e.g.][]{Hopk98,Rich00,Bond03,Cili03,Hopk03,Seym04,Huyn05,
80: Pran06,Foma06,Simp06,Ivis07,Bond07}.
81:
82: While it is clear that the $\mu$Jy population is a mixture of different
83: classes of objects (radio loud and radio quiet AGN, starburst galaxies,
84: spirals), the exact contribution of these different classes to
85: the radio source counts is still not very well
86: established and often dependent on the optical limit of the spectroscopic
87: follow-up \citep{Benn93,Hamm95,Geor99,Grup99,Pran01,Afon05,Foma06,Simp06}.
88: The role played by the cosmological evolution of the different classes of
89: objects is even more uncertain.
90: Recently, \citet{Smol06} have developed a new classification method to
91: separate star forming galaxies and AGN based on
92: photometric rest-frame colors for local galaxy samples.
93: The application of this classification
94: method to the VLA-COSMOS sources \citep{Smol08a} shows that the radio
95: population above $\sim$ 50
96: $\mu$Jy is a mixture of roughly 30--40\% of star forming galaxies
97: and 50--60\% of AGN galaxies, with a minor contribution ($\sim$ 10\%) of QSOs.
98:
99: The international COSMOS (Cosmic Evolution) survey \citep{Scov07a}
100: is designed to probe the correlated evolution of galaxies, star
101: formation, AGN up to high redshift. The survey includes
102: state of the art imaging data {covering the entire wavelength range} from the
103: X-rays to the radio domain
104: \citep{Hasi07,Zamo07,Tani07,Scov07b,Capa07,Sand07,Bert07,Schi04,Schi07}
105: supported by extensive optical
106: spectroscopic campaigns using mainly the VLT/VIMOS and the
107: Magellan/IMACS
108: instruments \citep{Lill07,Trum07}.
109:
110: The radio observations of the COSMOS field were done with the VLA in A
111: and C configuration and are described in Schinnerer et al.
112: (2004, 2007). The whole 2 deg$^2$ field was imaged
113: in mosaic mode with a resolution of 1.5 arcsec.
114: The sensitivity is uniform in the inner
115: 1 deg$^2$ with an average rms value of 10.5 $\mu$Jy/beam and it increases
116: almost regularly to the outer regions of the 2 deg$^2$ field.
117: From the mosaic a catalog of radio sources, the VLA-COSMOS Large Project
118: Catalog, was
119: extracted. The catalog lists $\sim 3600$ radio sources selected
120: above the local $4.5\sigma$ threshold over an area of 2 deg$^2$.
121: A full description of the observations, data reduction and catalog
122: extraction is given in Schinnerer et al. (2007, hereafter S07).
123:
124: In this paper we present a further analysis on the VLA-COSMOS Large Project
125: Catalog
126: aimed to quantify the impact of bandwidth smearing (Section 2) and the survey
127: completeness (Section 3). Finally, we derive the radio source counts and
128: compare them with those derived from similar surveys (Section 4).
129:
130: A more detailed interpretation of the radio data is left to
131: other papers which are currently in preparation. In particular, the optical
132: identifications of the VLA-COSMOS radio sources are
133: presented by \citet{Cili08}, the fraction of star-forming galaxies and AGN
134: in the sub-mJy population of the VLA-COSMOS field is discussed in
135: \citet{Smol08a} and the dust un-biased cosmic star formation history
136: is derived from the 1.4 GHz radio data in \citet{Smol08b}.
137:
138: \section{Bandwidth smearing correction}
139: Bandwidth smearing (or chromatic aberration) affects all synthesis
140: observations made with a finite width of the receiver channels.
141: Imaging sources at large distances from the phase center can yield to radial
142: smearing, increasing the source size and reducing the peak flux density,
143: while the integrated flux density is conserved. The image smearing is
144: proportional to the fractional bandwidth (the bandwidth divided by the central
145: frequency of observation) and to the distance from the phase center in units
146: of the synthesized beam \citep[e.g.][]{Thom99}.
147: For single pointing observation the bandwidth
148: smearing simply increases with the radial distance from the
149: phase centre. For a mosaic image, produced by multi-pointing observations,
150: the contribution from all the pointings has to be taken into account and the
151: resulting radial pattern will be much more complicated depending also
152: on the adopted spacing pattern of the individual pointings.
153:
154: \placefigure{bw_fig1}
155:
156: In order to quantify the effect of bandwidth smearing on the VLA-COSMOS
157: observations we ran the same procedure that produced the radio catalog on
158: each of the 23 individual pointings (see S07).
159: For the strongest sources we compared their peak and total flux densities in
160: the final mosaic with the corresponding peak and total flux densities in the
161: pointing where the sources are closest to the field center.
162: The total flux density of each source in the mosaic and in the individual
163: pointings are in very good agreement (the median value of the ratio is 1.03
164: with an r.m.s. dispersion of 0.04)
165: confirming that the total flux density is properly recovered.
166: On the other hand,
167: as expected for observations affected by bandwidth smearing, the peak flux
168: densities are underestimated in the final mosaic with respect to the peak
169: fluxes
170: in the individual pointings where the sources are closest to the field center.
171: To quantify this effect
172: we selected only sources which are within a
173: radius $R_{\rm min}\le 5\arcmin$ from the field center in, at least, one
174: individual pointing. The peak flux density of a source
175: within $5\arcmin$
176: from the field center is almost unaffected (less than 5\% for the
177: VLA-COSMOS observations)
178: by bandwidth smearing and can be compared
179: with the peak flux in the mosaic.
180: The image of the source in the mosaic is
181: the combination of all the images in the individual pointings where the
182: source is within the cut-off radius ($16.8\arcmin$, see S07).
183:
184:
185: In Figure~\ref{bw_fig1} we show the peak flux ratios for compact sources
186: (angular sizes less than 4$\arcsec$) with peak fluxes
187: brighter than 0.2 mJy/beam and $R_{\rm min}\le 5\arcmin$ as a function of
188: the radial distance from the center in
189: the final mosaic.
190:
191: \placefigure{pointings}
192:
193: The ratio is about $0.8\pm 0.05$ for sources in the inner region of the
194: mosaic, within 30$\arcmin$ from the mosaic center.
195: In the region between 30$\arcmin$ and 45$\arcmin$, the ratio is $0.9\pm
196: 0.03$, implying a smaller
197: correction in this area. This is due to the fact that the number of
198: overlapping pointings is smaller at larger radii (
199: see Fig.~\ref{pointings}).
200: Within these two regions
201: the effect appears to be independent from the position in the mosaic considering the
202: dispersion of the points. Therefore, we adopted a constant bandwidth
203: smearing correction factor of 0.8 for the sources within $30\arcmin$ from the
204: mosaic center and a constant value of 0.9 for those between $30\arcmin$ and
205: $45\arcmin$.
206:
207: We performed this analysis in the inner 1 deg$^2$ ($1\times 1$ deg.
208: square region) of the final mosaic
209: since this region has the deepest and most uniform sensitivity and it
210: will be used to derive the source counts in the following Section.
211: Beyond $45\arcmin$ from the mosaic center this correction is no more valid as
212: each source in the mosaic receives a
213: contribution from only one pointing and therefore
214: the bandwidth smearing correction will
215: be a function of the distance from the center of the individual pointing
216: where each source is located, rather than of the distance from the mosaic
217: center.
218:
219: \placefigure{res_unres}
220:
221: Consequently, to derive the radio source counts in the inner 1 deg$^2$
222: we divided all the
223: peak flux densities by 0.8 or 0.9 depending on the distance from the
224: center of the mosaic and derived the new
225: resolved/unresolved criterion following the procedure described in S07.
226: Since the ratio between total and peak fluxes is a direct measure of the
227: extent of a radio source, we used it to discriminate between resolved or
228: extended sources (i.e. larger than the beam) and unresolved sources.
229:
230: Figure~\ref{res_unres} shows the
231: ratio of the total integrated flux density $S_{\rm total}$ and the corrected
232: peak flux density $S_{\rm peak}$ as
233: function of the signal to noise ratio. To select the resolved sources we have
234: fitted a lower envelope in Fig.~\ref{res_unres} which contains $\simeq 95\%$
235: of the sources with $S_{\rm total} < S_{\rm peak}$
236: and mirrored it above the $S_{\rm total}/ S_{\rm peak}=1$ line.
237: The criterion of 95\% of sources contained in the lower envelope of
238: Fig.~\ref{res_unres} is slightly different from that used in S07, where a
239: value of 99\% was adopted. We reckon that the value adopted here is the best
240: compromise to properly classify the sources as resolved or unresolved
241: because it fully characterises the shape of the distribution of sources with
242: $S_{\rm total} < S_{\rm peak}$ and, at the same time, it excludes the most
243: extreme values. This envelope is described by the equation
244:
245: \begin{equation}
246: S_{\rm total}/S_{\rm peak}= 1+ [100/(S_{\rm peak}/{\it rms})^{2.4}]
247: \end{equation}
248:
249:
250: It is important to note that in Fig.~\ref{res_unres} there is no systematic
251: offset from the
252: $S_{\rm total}/ S_{\rm peak}=1$ line at high SNRs. This is an a
253: posteriori confirmation
254: that we adopted an appropriate bandwith smearing correction.
255: Considering only sources with SNR$\ge 5$ in the inner 1 deg$^2$,
256: we obtain 484 resolved sources, for which the total flux is given by
257: the total flux of the Gaussian fitting, and 1208
258: unresolved ones, for which the total flux is set equal to the peak flux.
259:
260: It is worth noting a few points regarding the effects of the correction
261: on the VLA-COSMOS Large Project Catalog:
262: \begin{itemize}
263: \item The correction has been determined only out to a radius of 45 arcmin
264: from the mosaic center and should be safely adopted only in this region.
265:
266: \item
267: The catalog is still selected on the
268: old, pre-bandwidth smearing correction, peak flux densities or signal-to-noise
269: ratio,
270: and we applied the bandwidth smearing correction only to the sources
271: already in the catalog.
272: In the remaining we will refer to the signal-to-noise ratio as the
273: original one, calculated with the uncorrected peak flux density.
274:
275: \item
276: The correction affects the resolved/unresolved classification of a source
277: and therefore also the total flux density and size for the sources, previously
278: classified as resolved, that are found unresolved after correcting for the
279: smearing. This affects about 500 sources, almost one third of the total
280: number of sources in the inner 1 deg$^2$ with
281: SNR $\ge 5$.
282:
283: %\item
284: %For the sources which are still resolved after the correction, the source
285: %size derived with Gaussian fitting in the catalogue are
286: %overestimated because of
287: %bandwidth smearing. Corrected source sizes can be obtained by multiplying the
288: %dimensions by the appropriate bandwidth smearing correction factor.
289:
290: \end{itemize}
291:
292:
293: \section{Survey completeness}
294:
295: \subsection{The visibility area}
296: The VLA-COSMOS Large Project Catalog contains $\sim 3600$ radio sources
297: extracted from a 2 deg$^2$
298: area. As shown in S07, the noise is not uniform over the whole area
299: and increases in the outer regions. The central 1 deg$^2$ area has
300: the lowest and most uniform noise (mean rms 10.5 $\mu$Jy) and therefore
301: it is very well-suited
302: to derive the source counts. Extending the analysis to the whole area and
303: catalog
304: would increase the number of objects at higher flux densities
305: ($> 0.1-0.2$ mJy), but it would not add much to the accuracy of the source counts at
306: the lowest flux densities, which is the main focus of this paper.
307: On the contrary, the correction factors that would need to be applied to take
308: into account the loss of the faintest sources in the outer and noiser region
309: would add more systematic uncertainties at the faint end of the source counts.
310: Furthermore, in our following analysis we consider only
311: sources with signal to noise ratio greater than or equal to 5 (the VLA-COSMOS
312: Large
313: Project Catalog has been selected with a signal to noise ratio threshold of
314: 4.5) and without a flag for possible sidelobe residuals,
315: in order to minimize contamination from spurious sources.
316: Given these constraints, we have a catalog of
317: 1692 sources with SNR $\ge 5.0$ within the inner 1 deg$^2$ region.
318:
319: \placefigure{visi_fig}
320:
321: The visibility area of the inner 1 deg$^2$ of the VLA-COSMOS survey as a
322: function of the rms noise is shown in Fig.~\ref{visi_fig}.
323: This is the area over which a source with given peak flux density
324: (5 times the noise in our case) can be detected.
325: Since the noise in this region is very smooth the visibility area increases
326: very rapidly and becomes flat above an r.m.s. noise of $\simeq 13$
327: $\mu$Jy/beam. The same curve, computed over the entire 2 deg$^2$, becomes
328: flat only above an r.m.s. noise of $\simeq 35$ $\mu$Jy/beam (see Fig.~13 in
329: S07).
330: This plot has been derived using the original peak flux
331: densities and not those corrected for the bandwidth smearing since the
332: catalog is still selected on the basis of the uncorrected peak flux.
333:
334: \subsection{The resolution bias and the intrinsic angular size distribution
335: of sub-mJy radio sources}
336: Besides the visibility area of each source, other correction factors need to
337: be considered to estimate the completeness of the radio catalog and to
338: derive the intrinsic source counts: the
339: errors introduced by the fitting routines, the noise bias, and the resolution
340: bias can all be modelled using simulated samples of radio sources.
341: In particular,
342: the resolution bias can be quite important since the beam of the
343: VLA-COSMOS observations is $1.5\arcsec \times 1.4\arcsec$
344: (e. g. about 15 times less
345: in area than for similar surveys carried out at the same frequency).
346: Since the completeness of the radio catalog is defined in terms of the
347: peak flux, while the source counts are derived as a function of the
348: total flux density, corrections need to be applied to take into account
349: resolved sources with peak fluxes below the catalog threshold and
350: total flux densities above the nominal limit of the survey.
351:
352: Our strategy to estimate the correction factor has been to simulate samples of
353: radio sources, with a realistic flux density and angular size distributions,
354: to insert these sources in the mosaic image and to recover these mock sources
355: using the same technique used to produce the catalog. From the comparison
356: between the number of sources added to the image and those effectively
357: recovered we can estimate the correction factors for different flux density
358: bins.
359: The correct choice of the flux density distribution is rather
360: straightforward, as a broken power law derived from the observed
361: distribution is a reasonable assumption, consistent with the extrapolation from
362: shallower surveys.
363: The choice of the angular size distribution for the simulated radio sources is
364: more uncertain.
365: While there is good evidence that the median of the intrinsic source size
366: at 1 mJy is about $2\arcsec$ and decreases with
367: decreasing flux density at sub-mJy level, it is not known how exactly
368: this happens and what is
369: the best analytical relationship to model this behaviour.
370: As we will show later, this is an important parameter in deriving the
371: completeness of the VLA-COSMOS catalog.
372: In the past different scaling relationships between the median angular size,
373: $\theta_{\rm med}$, and the total flux density, $S$, have been used, e.g.
374: $\theta_{\rm med}\propto S^m$ with $m=0.3$ \citep{Wind90} or
375: $m=0.5$ \citep{Rich00}.
376:
377: More recently, new surveys have provided some results on the
378: angular size distributions of the $\mu$Jy radio sources.
379: \citet{Foma06} obtained a
380: complete sample of 289 radio sources with flux density $\ga$ 40
381: $\mu$Jy and found that
382: 64\% of objects are unresolved with sizes less than $1.2$ arcsec.
383: MERLIN observations at higher angular resolution ($\sim 0.2\arcsec$) provided
384: consistent results for
385: a sample of 92 radio sources, 80 of which with flux density
386: between 40 $\mu$Jy and 200 $\mu$Jy \citep{Muxl05}.
387: The angular size distribution derived from the latter sample of
388: objects, which are all resolved, shows a dominant narrow gaussian component
389: centered at $\simeq 0''$.7 and a tail
390: of sources ($\sim 20\%$ of the sample) with angular sizes between 2$''$
391: and 4$''$.
392: We derived the median of the angular size distribution for the sources
393: observed by Muxlow et al. (2005) in the flux density range 40--90 $\mu$Jy.
394: The 55 sources in this flux density range have a median angular size of
395: $0\arcsec.7$ (with a r.m.s. dispersion of 0$\arcsec$.4) and a median flux
396: density of 54 $\mu$Jy. For comparison, \citet{Bond03} found that the
397: median angular size for sources in the flux density range
398: 0.4-1.0 mJy is $1\arcsec.8$ with a median flux density of 0.56 mJy.
399: The value of $m$ consistent with these median angular sizes at different
400: flux densities is $m=0.4$. It is worth noting that this value for $m$ has
401: to be considered as a lower limit
402: since the angular sizes given by \citet{Muxl05} are the largest angular sizes
403: while the angular sizes given in \citet{Bond03} are the FWHM.
404:
405: %We derived the median of the largest angular size distribution for the
406: %sources observed by Muxlow et al. (2005) taking into account the upper limits
407: %and we obtained a value of 0$''$.8.
408: These results provide strong evidences that radio sources with flux density
409: around 100 $\mu$Jy have typically sub-arcsecond sizes even if the quality of
410: the data (e.g. large number of upper limits in the angular sizes derived in
411: Bondi et al. 2003) and the statistics (the above result on $m$ is derived from
412: less than 100 sources) do not allow one to derive a firm estimate of $m$.
413:
414: Given these uncertainties, we decided to use a complementary method to
415: derive the intrinsic angular size distribution, simulating mock samples of
416: radio sources with different angular size
417: distributions following a general power law $\theta\propto S^m$
418: with the exponent $m$ ranging from 0.2 to 0.5, and normalized to the integral
419: angular size distribution derived from the VVDS survey in \citet{Bond03}.
420:
421: We constructed a simulated sample of radio sources containing about 3100
422: radio sources down to a total flux density
423: level of 30 $\mu$Jy. This allows us to count also sources with flux density
424: below
425: the limit which, because of positive noise fluctuations, might have a measured
426: peak flux density above the threshold.
427: Then, all the simulated sources were randomly injected in the inner 1
428: deg$^2$ of the VLA-COSMOS mosaic and subsequently recovered using the same
429: procedures adopted for the real sources and binned in flux
430: density intervals.
431: This was repeated for 3 different samples, and for four different
432: angular size distributions (with $m$ equal to 0.2, 0.3, 0.4, and 0.5)
433: yielding a total of 12
434: simulated samples with more than 35,000 simulated radio sources.
435: Finally, from the comparison between the number of simulated sources
436: detected in each bin and the number of sources in the simulated input
437: sample in the same flux density bin we derived the correction
438: factors for each value of the exponent $m$.
439: These correction factors, listed in Table~\ref{corr_5sigma},
440: account for the visibility area, noise bias, fitting errors
441: and resolution bias.
442:
443: \placetable{corr_5sigma}
444:
445: From Table~\ref{corr_5sigma} it is evident that
446: for the brightest sources the correction factors are $\la 10\%$ and
447: the differences among the correction factors for the various angular size
448: distributions are negligible. On the other hand,
449: below $0.20$ mJy the correction factors become quite large with
450: significant differences for the various angular size distributions.
451: Therefore, it is important to infer the intrinsic angular size
452: distribution in order to reconstruct the properly corrected source counts.
453: Another interesting feature of the correction factors which is worth noting
454: is that the maximum is not at the lowest flux density bin but around 0.12 mJy.
455: This is due to a combination of the resolution bias and the capability to
456: classify a radio source as resolved, which depends on the signal-to-noise
457: ratio. A source with a given total flux density and size will
458: be catalogued in one of the following three classes:
459: 1) undetected, if the peak flux is below the sample threshold,
460: 2) detected with the correct total flux density if the signal to noise ratio
461: is high enough to allow for proper deconvolution, or,
462: 3) detected with a lower total flux density
463: if the signal to noise ratio is not high enough to recognize the source as
464: resolved. The sources in the last class are
465: assigned a total flux density lower than the intrinsic one and
466: equal to the peak flux. A significant fraction of the sources with intrinsic
467: total flux densities $\la 0.15$ mJy is redistributed to lower total flux
468: densities since the signal-to-noise ratio is not sufficient to allow a proper
469: deconvolution.
470:
471: \placetable{median_sim}
472:
473: To constrain the intrinsic angular size distribution we performed the
474: following test.
475: We derived the median major axis of the simulated samples
476: of radio sources for each of the tested power laws ($m=0.2-0.5$) in the
477: flux density range 0.25-0.4 mJy.
478: This flux density range was chosen as a compromise to minimise the
479: number of unresolved sources (and therefore upper limits on source size)
480: and allow for a
481: statistically significant large sample of sources reaching as close as
482: possible to the lowest fluxes.
483: We calculated both the median major axis of the original
484: injected sources (row 1 in Tab.~\ref{median_sim}) and the median major axis
485: of the simulated sources recovered by the fitting routine (row 2 in
486: Tab.~\ref{median_sim}). These latter values were obtained considering also the
487: upper limits derived for the sources fitted as unresolved
488: ($\la$ 15\% of the total number of sources in this flux
489: density range).
490:
491: Table~\ref{median_sim} shows that,
492: in the flux density range 0.25-0.4 mJy,
493: although the measured median major axis of the simulated sources is
494: somewhat larger than that actually injected, the differences are not
495: statistically significant for any given
496: value of $m$.
497: On the other hand, different values of $m$ yield to
498: significantly different median major axis.
499:
500: Therefore, the value of the median major axis in this flux density range for
501: the real sources could be used to roughly discriminate between different
502: intrinsic angular size distributions. Unfortunately, the measured sizes
503: of the radio sources are affected by bandwidth smearing and they can not be
504: used for such a test. However, as we have seen in Section 2, the
505: total flux densities are properly determined and the ratio between the
506: total and the corrected peak flux densities can be used as an estimate of the
507: area covered by the radio sources. Then we made the assumption that
508: radio sources are extended in only one direction and we derived the major axis
509: from the ratio between the total and the corrected peak flux densities.
510: This assumption can be justified since extended radio emission is
511: generally associated to elongated features (e.g. radio jets in AGN or spiral
512: arms in star-forming galaxies), and when this is not true the derived source
513: sizes will be an upper limit of the real ones.
514: In this way we obtained a median value of the major axis
515: for the observed sources in the flux density range 0.25-0.4 mJy of
516: $0\arcsec.99\pm 0\arcsec.09$.
517: From the comparison with the values listed in Tab.~\ref{median_sim}, we
518: conclude that the observed median major axis
519: is more consistent with that obtained using simulated samples of
520: radio sources following
521: an angular size distribution with $m=0.5$ or $m=0.4$.
522:
523: With both methods, the comparison between the angular size distribution
524: obtained by \citet{Muxl05} and \citet{Bond03} in different flux density
525: intervals and the comparison between the estimate of the true angular sizes
526: and those derived from the simulated samples, we obtained consistent results
527: which exclude values of $m=0.2$ and $m=0.3$ and suggest $m\simeq 0.5$. In the
528: following analysis we apply the correction factors obtained for $m=0.5$
529: to the radio source counts.
530:
531: \section{The radio source counts}
532:
533: We have derived the radio source counts at 1.4 GHz in the inner 1 deg$^2$
534: region of the COSMOS field
535: from the VLA-COSMOS catalog, corrected for the effect of bandwidth smearing
536: and with a signal-to-noise threshold of 5.
537: The source counts up to $\simeq 1$ mJy are presented in
538: Table~\ref{counts_tab}. For each flux
539: density bin we give the mean flux density, the observed
540: number of sources, the corresponding differential source
541: density d$N$/d$S$ (in sr$^{-1}$Jy$^{-1}$), the normalised differential counts
542: (d$N$/d$S$)$S^{2.5}$ (in sr$^{-1}$Jy$^{1.5}$) with their estimated Poissonian
543: errors, the incompleteness correction factors from
544: Table~\ref{corr_5sigma} for $m=0.5$, and the cumulative number of effective
545: sources after applying the corrections. The correction factors above
546: 0.2 mJy are $\le 10\%$. Such variations are roughly consistent with
547: the poissonian error of the source counts and for this reason the source
548: counts above 0.2 mJy are assigned a correction factor of 1.
549: The mean flux density in each bin was calculated as the
550: geometric mean of the flux density extrema.
551:
552: \placetable{counts_tab}
553:
554: In Fig.~\ref{counts1} and Fig.~\ref{counts2} we show
555: the raw and the corrected source counts derived from
556: the VLA-COSMOS survey compared to those from other surveys at 1.4 GHz:
557: the VLA-VVDS \citep{Bond03}, the ATCA-HDF\_S
558: \citep{Huyn05}, the VLA-HDF\_N \citep{Rich00}, the PDF survey
559: \citep{Hopk03}, the SSA 13 field \citep{Foma06} and the FIRST survey
560: \citep{Whit97}.
561: The full range of source counts up to $\simeq 100$ mJy is shown in
562: Fig.~~\ref{counts1}, and a blow-up of the sub-mJy region is shown in
563: Fig.\ref{counts2}. In both figures the raw source counts from the VLA-COSMOS
564: survey are shown with empty circles and those corrected for incompleteness
565: with filled ones.
566: The solid line in Fig.~\ref{counts1} is a linear least-squares
567: sixth-order polynomial fit obtained using the VLA-COSMOS source counts
568: above 0.06 mJy supplemented with the FIRST source counts above 2.5 mJy and
569: up to 1 Jy. The resulting polynomial fit is given by
570:
571: \begin{equation}
572: \log[(dN/dS)/(S^{-2.5})]=\sum_{i=0}^6 a_i[\log(S/mJy)]^i
573: \end{equation}
574:
575: with $a_0=0.805$, $a_1=0.493$, $a_2=0.564$, $a_3=-0.129$, $a_4=-0.195$,
576: $a_5=0.110$, and $a_6=-0.017$.
577: The residuals from the polynomial fit have an rms of about 0.06 in the
578: logarithm of the normalized counts.
579: As discussed by \citet{Hopk03} the sixth-order
580: polynomial is necessary to follow the different curvatures shown by the source
581: counts.
582: In Fig.~\ref{counts1} we also show for comparison with a dashed line
583: the sixth-order polynomial fit obtained by \citet{Hopk03} from the Phoenix
584: Deep Survey source counts. The two fits are in reasonably good agreement with
585: our fit slightly higher for flux density below 0.2 mJy.
586:
587:
588: \placefigure{counts1}
589: \placefigure{counts2}
590:
591: It is evident that in some cases, the field-to-field differences in the radio
592: source counts below 0.5 mJy are larger than the combined errors.
593: It is likely
594: that these differences (at least partly) can be ascribed
595: to instrumental defects or different recipes to account for the incompleteness.
596: In particular, the assumption of the intrinsic angular size
597: distribution of the $\mu$Jy radio sources in high resolution radio surveys is
598: certainly a factor, as we have shown in the previous Section. Nevertheless,
599: the most extreme cases (e.g. the SSA 13 and HDF North fields) should reflect
600: real cosmic variance \citep{Foma06}.
601:
602: There is a particularly good agreement between the source counts derived
603: from the VLA-COSMOS, after correction, and the HDF-S fields. This is
604: reassuring for the
605: resolution bias correction we applied since the HDF-S field was observed with
606: a resolution of about $5\arcsec$, and for this reason
607: is much less affected by the precise
608: shape of the intrinsic angular size distribution.
609:
610: It is worth noting that while the VLA-COSMOS survey is not as deep as
611: the SSA13 and HDF-N surveys, it is, by far, the one with the most robust
612: statistics. At a flux density level of $\sim 65$ $\mu$Jy the VLA-COSMOS
613: survey counts almost 400 sources while typical numbers for other surveys,
614: in this flux density range, are between 50 and 90 sources.
615:
616: A possible drop off in the radio source counts below $\sim 100$ $\mu$Jy has
617: been already detected and discussed by other surveys
618: imaging various fields with different sensitivity and resolution
619: \citep[e.g.]{Hopk03, Huyn05, Foma06}.
620: Although, in every case, the drop off relies on the faintest bins it should
621: not be immediately discounted as incompleteness in the source counts as
622: it is accepted that the flat region of the normalised source counts below
623: $\sim 0.3$ mJy can not continue indefinetely \citep{Wind93}.
624: From the VLA-COSMOS radio source counts there is some indication that the
625: density of sources is beginning to decrease below 100-150 $\mu$Jy producing
626: a downturn in the observed radio source counts, even if the result is
627: strongly dependent on the fidelity of the first flux density bin.
628: It is important to note that if incompleteness is responsible for the drop
629: off observed in the first flux density bin of Fig.~\ref{counts2}, it would
630: mean that we are missing about 130 sources in the range 0.06-0.0735 mJy, which
631: is extremely unlikely. So, while incompleteness at the lowest flux
632: density levels can still be present at some level, it can not explain by
633: itself the drop off observed in the radio source counts. Thus, this decline
634: can provide important limits when modeling the
635: populations contributing to the radio source counts.
636:
637: As we have explained in the previous section, the intrinsic angular size
638: distribution adopted to derive the incompleteness correction factor has a
639: large impact in this flux density range, but adopting distributions with
640: $m< 0.5$ would only enhance the drop off below $\simeq 0.15$ mJy.
641:
642: Also the uncertainties in the bandwidth smearing correction could, in
643: principle, significantly
644: modify the radio source counts and the observed drop off observed at lower
645: flux density. We have tested
646: this assuming two extreme cases, a constant correction over the whole
647: area of 0.8 and 0.9 respectively. As we can expect,
648: the first case would produce a very small
649: difference, well within the Poissonian error boxes, since we already applied
650: the 0.8 correction over almost 80\% of the 1 deg$^2$. On the other hand,
651: adopting a correction
652: factor of 0.9 over the whole area, would significantly lower only the radio
653: source counts in the first three bins, increasing the slope of the drop off
654: region. As we have shown in Section 2, the assumption of a constant correction
655: over the whole 1 deg$^2$ is not corroborated by the real data. The data
656: require a mixture of the two correction factors and with the previous test we
657: have just verified that the drop off in the source counts below 0.15 mJy can
658: not be produced by the applied smearing corrections.
659:
660: We can conclude that given the large statistics and the fact that the
661: incompleteness factors
662: have been carefully investigated and applied, the decrease in the
663: source counts below $\simeq 0.10-0.15$ mJy observed in the VLA-COSMOS radio
664: source counts should be considered real.
665:
666: \section{Summary \& conclusions}
667:
668: In this paper we have presented a further analysis of the VLA-COSMOS Large
669: Project Catalog
670: with the goals of: 1) quantifying and correcting for the bandwidth smearing
671: effect which affects the VLA-COSMOS radio mosaic and catalog;
672: 2) determining the incompleteness of the VLA-COSMOS
673: radio catalog, through extensive use of Monte-Carlo simulations
674: taking into account different intrinsic angular size distributions,
675: in order to
676: quantify the effects of noise bias and resolution bias; 3) deriving the radio
677: source counts at 1.4 GHz in the VLA-COSMOS field.
678:
679: The bandwidth smearing correction factors have been determined from the
680: comparison of peak and total flux densities in the final mosaic and
681: in individual pointings where the source is within $5\arcmin$ from the center.
682: We have found that a two-value correction factor is a good approximation:
683: out to a radius of $30\arcmin$ from the center,
684: the peak flux densities in the VLA-COSMOS catalog must be
685: divided by a factor 0.8, while in the region between $30\arcmin$ and
686: $45\arcmin$ from the
687: center by a factor 0.9. Inside these two regions the correction
688: factor is rather constant, within the dispersion of the data, with a sharp
689: transition from 0.8 to 0.9 at a radial distance of $30\arcmin$ reflecting
690: the change in overlapping pointings.
691: Consequently, a new classification in resolved and unresolved sources has been
692: derived. A number of sources,
693: originally classified as resolved, turned out to be unresolved
694: after the bandwidth smearing correction was applied. For these sources
695: also the total flux density was corrected.
696:
697: We simulated several samples of radio sources with a flux density
698: distribution compatible with the observed one and with different
699: angular size distributions.
700: The angular size distribution was modelled using $\theta\propto S^m$
701: with $m=0.2,0.3,0.4,0.5$.
702: These mock sources were added to the radio mosaic
703: and recoverd using the same procedure that yielded to the radio catalog.
704: In such a way we were able to quantify the effects of the combined noise and
705: resolution bias affecting the completeness of the radio catalog.
706: A comparison of the median angular size at a flux density level
707: not significantly affected by the resolution bias for the simulated samples of
708: sources and the observed one, strongly suggest a general distribution of
709: radio source sizes following $\theta\propto S^{0.5}$.
710:
711: Finally, we derived the radio source counts in the VLA-COSMOS field (both
712: uncorrected and corrected for the incompleteness). The source counts extend
713: down to
714: 60 $\mu$Jy and are consistent with those derived by other surveys but with
715: much more robust statistics. In particular, the drop off of the source counts
716: below $\sim 100$ $\mu$Jy has to be considered real and not due to
717: incompleteness.
718:
719:
720:
721: \begin{acknowledgements}
722: The National Radio Astronomy Observatory (NRAO) is operated by
723: Associated Universities, Inc., under cooperative agreement with the
724: National Science Foundation. KJ acknowledges support by the German DFG
725: under grant SCHI~536/3-1.
726: \end{acknowledgements}
727:
728: \begin{thebibliography}{}
729:
730: \bibitem[Afonso et al.(2005)]{Afon05}
731: Afonso, J., Georgakakis, A., Almeida, C., Hopkins, A. M., Cram, L. E.
732: Mobasher, B., Sullivan, M. 2005, ApJ, 624, 135
733:
734: \bibitem[Benn et al.(1993)]{Benn93}
735: Benn, C.R., Rowan-Robinson, M., McMahon, R. G., Broadhurst, T.J.,
736: \& Lawrence, A. 1993, MNRAS, 263, 98
737:
738: \bibitem[Bertoldi et al.(2007)]{Bert07}
739: Bertoldi, F., et al. 2007, ApJS, 172, 132
740:
741: \bibitem[Bondi et al.(2003)]{Bond03}
742: Bondi, M., et al. 2003, A\&A, 403, 857
743:
744: \bibitem[Bondi et al.(2007)]{Bond07}
745: Bondi, M., et al. 2007, A\&A, 463, 519
746:
747: \bibitem[Capak et al.(2007)]{Capa07}
748: Capak, P., et al. 2007, ApJS, 172, 99
749:
750: \bibitem[Ciliegi et al.(2003)]{Cili03}
751: Ciliegi, P., Zamorani, G., Hasinger, G., Lehmann, I., Szokoly, G., Wilson, G.
752: 2003, A\&A, 398, 901
753:
754: \bibitem[Ciliegi et al.(2008)]{Cili08}
755: Ciliegi, P., et al. 2008, in preparation
756:
757: \bibitem[Fomalont et al.(2006)]{Foma06}
758: Fomalont, E.B., Kellerman, K.I., Cowie, L.L., Capak, P., Barger, A. J.,
759: Partridge, R. B., Windhorst, R. A., Richards, E. A. 2006, ApJS, 167, 103
760:
761: \bibitem[Georga\-ka\-kis et al.(1999)]{Geor99}
762: Georgakakis, A., Mobasher, B., Cram, L., Hopkins, A., Lidman, C.,
763: Rowan-Robinson, M. 1999, MNRAS, 306, 708
764:
765: \bibitem[Gruppioni et al.(1999)]{Grup99}
766: Gruppioni, C., Mignoli, M., \& Zamorani, G. 1999a, MNRAS, 304, 1999
767:
768: \bibitem[Hammer et al.(1995)]{Hamm95}
769: Hammer, F., Crampton, D., Lilly, S.J., Le Fevre, O., \& Kenet, T.
770: 1995, MNRAS, 276, 1085
771:
772: \bibitem[Hasinger et al.(2007)]{Hasi07}
773: Hasinger, G., et al. 2007, ApJS, 172, 29
774:
775: \bibitem[Hopkins et al.(1998)]{Hopk98}
776: Hopkins, A. M., Mobasher, B., Cram, L.E., \& Rowan-Robinson, M. 1998,
777: MNRAS, 296, 839
778:
779: \bibitem[Hopkins et al.(2003)]{Hopk03}
780: Hopkins, A. M., Afonso, J., Chan, B., Cram, L. E., Georgakakis, A.,
781: Mobasher, B. 2003, AJ, 125, 465
782:
783: \bibitem[Huynh et al.(2005)]{Huyn05}
784: Huynh, M.T., Jackson, C.A., Norris, R.P., Prandoni, I. 2005, AJ, 130, 1373
785:
786: \bibitem[Ivison et al.(2007)]{Ivis07}
787: Ivison, R.J., et al. 2007, ApJ, 660, 771
788:
789: \bibitem[Lilly et al.(2007)]{Lill07}
790: Lilly, S.J., et al. 2007, ApJS, 172, 70
791:
792: \bibitem[Muxlow et al.(2005)]{Muxl05}
793: Muxlow, T.W.B., et al. 2005, MNRAS,
794: 358, 1159
795:
796: \bibitem[Prandoni et al.(2001)]{Pran01}
797: Prandoni, I., Gregorini, L., Parma, P., de Ruiter, H. R., Vettolani, G.,
798: Zanichelli, A., Wieringa, M. H., Ekers, R. D. 2001 A\&A, 369, 787
799:
800: \bibitem[Prandoni et al.(2006)]{Pran06}
801: Prandoni, I., Parma, P., Wieringa, M.H., de Ruiter, H. R., Gregorini, L.,
802: Mignano, A., Vettolani, G., Ekers, R. D. 2006 A\&A, 457, 517
803:
804: \bibitem[Richards(2000)]{Rich00}
805: Richards, E.A. 2000, ApJ, 533, 611
806:
807: \bibitem[Sanders et al.(2007)]{Sand07}
808: Sanders, D.B., et al. 2007, ApJS, 172, 86
809:
810: \bibitem[Schinnerer et al.(2004)]{Schi04}
811: Schinnerer, E., et al. 2004, AJ, 128, 1974
812:
813: \bibitem[Schinnerer et al.(2007)]{Schi07}
814: Schinnerer, E., et al. 2007, ApJS, 172, 46 (S07)
815:
816: \bibitem[Scoville et al.(2007a)]{Scov07a}
817: Scoville, N., et al. 2007, ApJS, 172, 1
818:
819: \bibitem[Scoville et al.(2007b)]{Scov07b}
820: Scoville, N., et al. 2007, ApJS, 172, 38
821:
822: \bibitem[Seymour, McHardy \& Gunn(2004)]{Seym04}
823: Seymour, N., McHardy, I.M., Gunn, K.F. 2004, MNRAS, 352, 131
824:
825: \bibitem[Simpson et al.(2006)]{Simp06}
826: Simpson, et al. 2006, MNRAS, 372, 741
827:
828: \bibitem[Smol\v{c}i\'{c} et al.(2006)]{Smol06}
829: Smol\v{c}i\'{c}, V., et al. 2006, MNRAS, 371, 121
830:
831: \bibitem[Smol\v{c}i\'{c} et al.(2008a)]{Smol08a}
832: Smol\v{c}i\'{c}, V., et al. 2008a, ApJ in press
833:
834: \bibitem[Smol\v{c}i\'{c} et al.(2008b)]{Smol08b}
835: Smol\v{c}i\'{c}, V., et al. 2008b, in preparation
836:
837: \bibitem[Taniguchi et al.(2007)]{Tani07}
838: Taniguchi, Y., et al. 2007, ApJS, 172, 9
839:
840: \bibitem[Thompson(1999)]{Thom99}
841: Thompson, A.R., 1999 in ASP Conf. Ser. 180: Synthesis Imaging in Radio
842: Astronomy II, ed. G. B. Taylor, C. L. Carilli, \& R. A. Perley
843:
844: \bibitem[Trump et al.(2007)]{Trum07}
845: Trump, J.R., et al. 2007, ApJS, 172, 383
846:
847: \bibitem[White et al.(1997)]{Whit97}
848: White, R.L., Becker, R.H., Helfand, D.J., Gregg, M.D. 1997, ApJ, 475 479
849:
850: \bibitem[Windhorst et al.(1990)]{Wind90}
851: Windhorst, R.A., Mathis, D., \& Neuschaefer, L. 1990, ASP Conf. Ser.,
852: Evolution of the universe of galaxies, p.389
853:
854: \bibitem[Windhorst et al.(1993)]{Wind93}
855: Windhorst, R.A., Fomalont, E.B., Partridge, R.B., Lowenthal, J.D. 1993, ApJ,
856: 405, 498
857:
858: \bibitem[Zamojski et al.(2007)]{Zamo07}
859: Zamojski, M.A., et al. 2007, ApJS, 172, 468
860:
861: \end{thebibliography}
862:
863: \begin{table}
864: \centering
865: \caption{Correction factors for different angular size distributions}
866: \label{corr_5sigma}
867: \begin{tabular}{cccccccc}
868: \hline
869: Flux bin & $m=0.2$ & $m=0.3$ & $m=0.4$ & $m=0.5$ \\
870: (mJy) & & & & \\
871: \hline
872: 0.0600-0.0735 & 1.29 & 1.19 & 1.06 & 0.99 \\
873: 0.0735-0.0900 & 1.59 & 1.35 & 1.23 & 1.16 \\
874: 0.0900-0.1103 & 2.12 & 1.81 & 1.53 & 1.36 \\
875: 0.1103-0.1351 & 2.34 & 2.12 & 1.81 & 1.60 \\
876: 0.1351-0.1655 & 1.60 & 1.46 & 1.32 & 1.29 \\
877: 0.1655-0.2028 & 1.24 & 1.21 & 1.21 & 1.15 \\
878: 0.2028-0.2484 & 1.05 & 1.03 & 1.00 & 0.99 \\
879: 0.2484-0.3043 & 1.03 & 1.02 & 0.95 & 0.93 \\
880: 0.3043-0.4564 & 1.10 & 1.08 & 1.09 & 1.07 \\
881: 0.4564-0.6846 & 1.11 & 1.10 & 1.10 & 1.09 \\
882: 0.6846-1.0270 & 0.98 & 0.98 & 0.97 & 0.98 \\
883: $\ge 1.0270$ & 1.07 & 1.07 & 1.08 & 1.09 \\
884: \hline
885: \end{tabular}
886: \end{table}
887:
888: \begin{table}
889: \centering
890: \caption{Median major axis in the flux density range 0.25-0.4 mJy for the
891: input and output simulated samples}
892: \label{median_sim}
893: \begin{tabular}{cccccccc}
894: \hline
895: Sample & $m=0.2$ & $m=0.3$ & $m=0.4$ & $m=0.5$ \\
896: &$\theta_{maj,med}$&$\theta_{maj,med}$&$\theta_{maj,med}$&$\theta_{maj,med}$\\
897: \hline
898: Sim. Input &$1.32\pm 0.05$&$1.17\pm 0.04$&$1.03\pm 0.04$&$0.91\pm 0.04$\\
899: Sim. Output &$1.36\pm 0.06$&$1.24\pm 0.06$&$1.11\pm 0.05$&$1.00\pm 0.05$\\
900: \hline
901: \end{tabular}
902: \end{table}
903:
904:
905: \begin{table}
906: \centering
907: \caption{The 1.4 GHz radio source counts}
908: \label{counts_tab}
909: \begin{tabular}{cccccccc}
910: \hline
911: $S$ & $<S>$ & $N$& d$N$/d$S$& (d$N$/d$S$)$S^{2.5}$ & C & N$_{c}$($>S$)\\
912: mJy & mJy & &sr$^{-1}$Jy$^{-1}$& sr$^{-1}$Jy$^{1.5}$& & deg$^{-2}$\\
913: \hline
914: 0.0600-0.0735 & 0.066 & 380 & $9.24\times 10^{10}$& $3.32\pm 0.17$ & 0.99
915: & $1669\pm 41$\\
916: 0.0735-0.0900 & 0.081 & 324 & $6.43\times 10^{10}$& $3.84\pm 0.21$ & 1.16
917: & $1515\pm 39$\\
918: 0.0900-0.1103 & 0.100 & 203 & $3.29\times 10^{10}$& $3.26\pm 0.23$ & 1.36
919: & $1336\pm 37$\\
920: 0.1103-0.1351 & 0.122 & 142 & $1.88\times 10^{10}$& $3.09\pm 0.26$ & 1.60
921: & $1246\pm 35$\\
922: 0.1351-0.1655 & 0.150 & 135 & $1.46\times 10^{10}$& $3.99\pm 0.34$ & 1.29
923: & $~822\pm 29$\\
924: 0.1655-0.2028 & 0.183 & 109 & $9.61\times 10^{9}$ & $4.36\pm 0.42$ & 1.15
925: & $~577\pm 24$\\
926: 0.2028-0.2484 & 0.224 & ~74 & $5.33\times 10^{9}$ & $4.02\pm 0.47$ & 1.00
927: & $~393\pm 20$\\
928: 0.2484-0.3043 & 0.275 & ~72 & $4.23\times 10^{9}$ & $5.30\pm 0.63$ & 1.00
929: & $~319\pm 18$\\
930: 0.3043-0.3728 & 0.337 & ~45 & $2.16\times 10^{9}$ & $4.49\pm 0.67$ & 1.00
931: & $~247\pm 16$\\
932: 0.3728-0.4566 & 0.413 & ~33 & $1.29\times 10^{9}$ & $4.47\pm 0.78$ & 1.00
933: & $~202\pm 14$\\
934: 0.4566-0.5593 & 0.505 & ~28 & $8.95\times 10^{8}$ & $5.14\pm 0.97$ & 1.00
935: & $~169\pm 13$\\
936: 0.5593-0.6851 & 0.619 & ~24 & $6.26\times 10^{8}$ & $5.97\pm 1.28$ & 1.00
937: & $~141\pm 12$\\
938: 0.6851-0.8393 & 0.758 & ~12 & $2.56\times 10^{8}$ & $4.05\pm 1.17$ & 1.00
939: & $~117\pm 11$\\
940: 0.8393-1.0282 & 0.929 & ~12 & $2.09\times 10^{8}$ & $5.49\pm 1.58$ & 1.00
941: & $~105\pm 10$\\
942: \hline
943: \end{tabular}
944: \end{table}
945:
946: \clearpage
947:
948: \begin{figure}
949: \plotone{f1.eps}
950: \caption[]
951: {Ratio between the peak flux densities in the
952: final mosaic and in the individual pointing where the source is within
953: $5\arcmin$ from the center versus
954: the radial distance from the center in the final mosaic.
955: Only sources with peak flux density greater than 0.2 mJy/beam and sizes
956: smaller than 4$\arcsec$ are plotted. The straight lines at
957: S(mosaic)/S(single pointing)= 0.8 and 0.9 for R $<30\arcmin$ and
958: R $>30\arcmin$ show the adopted bandwidth smearing correction factors.}
959: \label{bw_fig1}
960: \end{figure}
961:
962: \begin{figure}
963: \plotone{f2.eps}
964: \caption[]
965: {Layout of the 23 pointings for the VLA-COSMOS observations. The two circles
966: have a radius of $30\arcmin$ and $45\arcmin$.
967: }
968: \label{pointings}
969: \end{figure}
970:
971: \begin{figure}
972: \plotone{f3.eps}
973: \caption[]{Ratio of the total flux to the corrected peak flux (see text for
974: details) as a function of the
975: signal-to-noise ratio. The solid lines show the upper and lower envelopes of
976: the flux ratio distribution containing the sources considered unresolved.
977: Open symbols show the sources considered extended.
978: This Figure should be compared to Fig. 15 in S07. }
979: \label{res_unres}
980: \end{figure}
981:
982: \begin{figure}
983: \plotone{f4.eps}
984: \caption[]{Visibility area of the inner 1 deg$^2$ of the COSMOS survey.}
985: \label{visi_fig}
986: \end{figure}
987:
988: \begin{figure}
989: \plotone{f5.eps}
990: \caption[]{Radio source counts at 1.4 GHz from the VLA-COSMOS survey (dots)
991: and from other surveys. Empty circles show the radio counts not corrected for
992: incompleteness, filled circles the corrected ones using $m=0.5$.
993: The VLA-COSMOS source counts
994: are shown along with those obtained by other deep surveys (see text). The
995: solid line is least-squares sixth-order polynomial fit obtained using the
996: VLA-COSMOS and the FIRST source counts. The dashed line is the fit obtained
997: by \citet{Hopk03}.}
998: \label{counts1}
999: \end{figure}
1000:
1001: \begin{figure}
1002: \plotone{f6.eps}
1003: \caption[]{Blow-up of the radio source counts in the sub-mJy region.
1004: Symbols are the same as in Fig.~\ref{counts1}.}
1005: \label{counts2}
1006: \end{figure}
1007:
1008: \end{document}
1009:
1010: