1: \documentclass[12pt,preprint]{aastex}
2:
3: %%\slugcomment{May 21, 2007}
4:
5: \shorttitle{Near Infrared Background}
6: \shortauthors{Thompson et al.}
7:
8: \begin{document}
9:
10:
11: \title{Evidence for a $Z<8$ Origin of the Source Subtracted Near
12: Infrared Background}
13:
14: \author{Rodger I. Thompson, Daniel Eisenstein, Xiaohui Fan and Marcia Rieke}
15: \affil{Steward Observatory, University of Arizona,
16: Tucson, AZ 85721}
17: \email{rthompson@as.arizona.edu, deisenstein@as.arizona.edu, fan@as.arizona.edu, mrieke@as.arizona.edu}
18:
19: \author{Robert C. Kennicutt\altaffilmark{1}}
20: \affil{Institute of Astronomy, University of Cambridge,
21: Cambridge CB3 OHA UK}
22: \email{robk@ast.cam.ac.uk}
23:
24: \altaffiltext{1}{Steward Observatory, University of Arizona,
25: Tucson, AZ 85721}
26:
27: \begin{abstract}
28:
29: This letter extends our previous fluctuation analysis of the near infrared
30: background at 1.6$\micron$ to the 1.1$\micron$ (F110W) image of the Hubble
31: Ultra Deep field. When all detectable sources are removed the ratio of
32: fluctuation power in the two images is consistent with the ratio expected
33: for faint, $z<8$, sources, and is inconsistent with the expected ratio
34: for galaxies with $z>8$. We also use numerically redshifted model galaxy
35: spectral energy distributions for 50 and 10 million year old galaxies to
36: predict the expected fluctuation power at 3.6$\micron$ and 4.5$\micron$
37: to compare with recent \emph{Spitzer} observations. The predicted
38: fluctuation power for galaxies at z = 0--12 matches the observed
39: \emph{Spitzer} fluctuation power while the predicted power for $z>13$
40: galaxies is much higher than the observed values. As was found in the
41: 1.6$\micron$ (F160W) analysis the fluctuation power in the source subtracted
42: F110W image is two orders of magnitude below the power in the image with all
43: sources present. This leads to the conclusion that the 0.8--1.8$\micron$ near
44: infrared background is due to resolved galaxies in the redshift range $z<8$,
45: with the majority of power in the redshift range of 0.5--1.5.
46:
47: \end{abstract}
48:
49: \keywords{cosmology: observations -- diffuse radiation -- early universe, galaxies:
50: high-redshift}
51:
52: \section{Introduction}
53:
54: In a previous paper \citep{thm07} we addressed the nature of the Near InfraRed
55: Background (NIRB) at 1.6$\micron$ using NICMOS observations in the Hubble Ultra
56: Deep Field (HUDF). The area covered by the NICMOS observations comprises the
57: NICMOS Ultra Deep Field (NUDF) which is smaller than the HUDF. The NIRB at
58: 1.1$\micron$ and 1.6$\micron$ has a total power of 7 nWm$^{-2}$sr$^{-1}$ emitted
59: by resolved galaxies, predominantly in the redshift range z = 0.5--1.5. This is
60: in contrast to the previous results from \citet{mat05} that claimed a peak flux
61: of 70 nWm$^{-2}$sr$^{-1}$ at 1.4$\micron$. The discrepancy arose not in the
62: value of the total measured flux but in the attribution of the flux to the
63: components of zodiacal emission, resolved stars and galaxies and an excess.
64: In \citet{mat05} models were used to determine the first two components and
65: the difference between the modeled components and the total observed flux
66: was attributed to an excess. The \citet{mat05} measurements did not extend
67: shortward of 1.4$\micron$ and the sharp drop in the flux levels between the
68: 1.4$\micron$ value and shorter wavelengths was initially attributed by some
69: authors (eg. \citet{sal03}) as due to the Lyman break in very high redshift
70: galaxies. Our analysis measured the flux from the resolved stars and galaxies
71: and the zodiacal background and found no excess. The difference was in the
72: amount of flux attributed to the zodiacal background. Our measured value was
73: higher than the modeled value used by \citet{mat05} and the difference was
74: equal to the flux attributed to an excess. We therefore concluded that unless
75: there was a extragalactic flux component that was extremely flat, mimicking
76: zodiacal flux, there is no excess and the near infrared background is
77: resolved.
78:
79: \citet{thm07} also addressed the spatial fluctuations in the NUDF field relative
80: to the fluctuations found in deep calibration 2MASS images at the same wavelength
81: by \citet{kas02}. \citet{kas02} subtracted all of the resolved sources from
82: the 2MASS images and found fluctuations in excess of that expected by shot
83: noise at long wavelengths. The excess was attributed to fluctuation power from
84: galaxies at redshifts of 10 and above in several studies (eg. \citet{mag03}).
85: A fluctuation analysis of the NUDF
86: F160W image showed that when sources were removed from the NICMOS image down to
87: the level that were removed from the 2MASS image, the fluctuations were consistent
88: with those found by \citet{kas02}. When sources were removed down to the
89: detection limit of the combined ACS and NICMOS observations in the NUDF the
90: fluctuation power dropped by almost two orders of magnitude. Since all of the
91: detected sources had redshifts between 0 and 8, this meant that the fluctuations
92: found by \citet{kas02} were from galaxies in the redshift range between 0 and 8.
93: Further analysis in \citet{thm07} of images that only retained sources in a given
94: redshift bin determined that the majority of fluctuation power came from galaxies
95: in redshift range between 0.5 and 1.5. The fluctuation spectra from
96: galaxies with redshifts greater than 4 were not detectable above the noise.
97:
98: More recent fluctuation analyses of \emph{Spitzer} IRAC images at 3.6$\micron$
99: and 4.5$\micron$ (\citet{kas05b} and \citet{kas07b}) interpret the residual
100: fluctuations after source subtraction as being due to high redshift, possibly
101: Population III, galaxies \citep{kas07a}. This assertion appears to be based
102: on the assumption that all galaxies below the IRAC detection limit must be faint
103: due to their high redshift. This interpretation has been challenged by
104: \citet{coo07} who did further source subtraction in the field using deep ACS
105: images to identify faint sources missed by the IRAC images. They found
106: the removal of these sources greatly reduced the signal seen by \citet{kas07b}
107: and the signal must be due to sources with redshifts less than 8 that are
108: visible in the ACS images. \citet{kas07a} replied that the excessive amount of
109: area removed from the IRAC images in the source subtraction carried out by
110: \citet{coo07} invalidates their result. Modeling by \citet{sal06} indicates
111: that the observed IRAC fluctuations may be consistent with those created by
112: Pop II stars at redshift greater than or equal to 5. The present study is
113: motivated in part by a desire to investigate whether the observations in
114: the NUDF at both 1.1$\micron$ and 1.6$\micron$ are consistent with the high
115: redshift interpretation of the the IRAC observations by examining the color
116: of the fluctuations and their expected extrapolation to the IRAC bands.
117:
118: \section{Data and Analysis}
119:
120: The location, size and pixel scale of the F110W image is identical to the
121: F160W NUDF image analyzed in \citet{thm07}. Details of the image preparation
122: are given in \citet{thm05}. The basic image production after the processing
123: of the individual images is production of a background image which is the
124: median of all of the individual images, subtraction of the background from
125: the individual images and then combining the images with the drizzle procedure
126: \citep{fru02}. The fluctuation analysis of the F110W image is identical to the
127: analysis techniques used on the F160W image in \citet{thm07} and is not repeated
128: in detail here. Source subtraction is accomplished using the SExtractor
129: \citep{ber96} (SE) pixel map where each pixel that is part of a source is given
130: a value equal to the source ID number and all other pixels have a value of
131: zero. The photometric redshifts derived in \citet{thm06} were used to identify
132: the redshift of each source. Due to the much higher resolution of the NICMOS
133: images relative to the IRAC images, even the all source subtracted image retained
134: 93$\%$ of its pixels. This means that the objections of \citet{kas07a} to
135: the \citet{coo07} analysis do not apply here. All source pixels were set to
136: zero in the subtracted image. There was no attempt to replace them with random
137: noise. The small area of the subtracted sources means that this had no affect
138: on the final fluctuation spectrum. Figure~\ref{fig-fluct} shows the fluctuation
139: power at 1.1$\micron$ and 1.6$\micron$. To determine the redshift distribution of
140: the 1.1$\micron$ fluctuations the fluctuation analysis was also performed on the
141: 1.1$\micron$ image with all sources except those in a given redshift bin removed.
142: The 1.1$\micron$ fluctuation power versus
143: redshift distribution is essentially identical to the 1.6$\micron$ distribution
144: shown in Figure 6 of \citet{thm07}. There is a peak near z=1 and no discernible
145: power above the background for $z>4$.
146:
147: \section{Fluctuation Colors \label{s-fc}}
148:
149: To investigate the nature of the sources of the fluctuation power observed in
150: the all source subtracted NICMOS and IRAC images we calculate
151: the expected fluctuation power color as a function of redshift using galaxy
152: Spectral Energy Distributions (SEDs). At any given angular scale the fluctuation
153: power for a given wavelength band is directly proportional to the power of the
154: galaxies ($\nu I_{\nu}$) in that band. The power ratios of the bands at any
155: redshift are easily calculated from the numerically redshifted model SEDs.
156: The analysis of \citet{thm06} used 7 primary template SEDs labeled 1--7 going
157: from early to late galaxies. The first 5 are observed SEDs and SEDs 6 and
158: 7 are calculated from the models of \citet{bru03}. The properties of these
159: models are given in Table~\ref{tab-temp}. Template 6 is for a 50 million
160: year old galaxy and Template 7 is for a 10 million year old galaxy with the
161: lowest metallicity available in the \citet{bru03} models. Template 7 is
162: similar to a Pop. III SED but redder than a true Pop. III SED (see
163: \citet{sch02} for the expected differences). Although a Pop. III SED may be
164: bluer than template 7, the sharp increase in the 3.6$\micron$ to 1.6$\micron$
165: fluctuation power at high redshift is due to the Lyman break passing through
166: the 1.6$\micron$ and is therefore relatively independent of the SED to the
167: red of the break.
168:
169: \citet{thm06} interpolated
170: between the SEDs in increments of 0.1 to produce a final set of 61 different
171: model SEDs. Figure~\ref{fig-temp} shows the distribution of the template
172: numbers for all sources in the NUDF with redshifts greater than 2.0. The
173: predominance of sources concentrate on templates 6 and 7, therefore, we will take
174: them as the most probable SEDs for sources fainter than our source
175: detection limit. We use these SEDs and the effective response functions
176: in the NICMOS and IRAC filters to calculate the expected ratios of the power at
177: redshifts between 0 and 15. Absorption in the Inter-Galactic Medium is modeled
178: according to the treatment given in \citet{mad96}.
179:
180: The horizontal lines in Figure~\ref{fig-col} show the observed ratios of
181: the fluctuation power in the all source subtracted images using; this analysis,
182: \citet{kas07b} and \citet{thm07}. Since the excess fluctuation power attributed
183: to high redshift galaxies by \citet{kas07b} is at angular scales greater than
184: $30\arcsec$ the horizontal lines show the ratios observed at $100\arcsec$, the
185: largest NICMOS angular scale.
186:
187: The observed 1.1$\micron$ to 1.6$\micron$ fluctuation power ratio is consistent
188: with the lower redshift template 6 predictions, but is incompatible with the
189: calculated ratios at $z>8$ for either of the templates. This is strong
190: evidence that the source subtracted fluctuations in the NICMOS images are
191: due to galaxies at $z<8$. The observed 3.6$\micron$ to 1.6$\micron$ ratio
192: lies between the ratios predicted by templates 6 and 7. Given the good match
193: to the NICMOS flux ratios for template 6 we would expect higher 3.6$\micron$
194: fluctuations than observed. The difference may be due to cosmic variance since
195: the fields are different. It may also be that the much broader \emph{Spitzer}
196: PSFs subtract more background flux than the HST PSFs. For $z>12$ the
197: predicted ratios become incompatible with observations. The 4.5$\micron$ to
198: 3.6$\micron$ ratio is essentially independent of redshift at the redshifts
199: considered and therefore does not contribute any information on the redshift
200: epoch of the fluctuations. The apparent agreement between the predicted and
201: observed ratios for those bands, however, is further evidence that the observed
202: fluctuations are due to faint galaxies at redshifts less than 8.
203:
204: \subsection{Compatibility of the Data Sets}
205:
206: An important question when comparing the NICMOS and IRAC source subtracted remnant
207: fluctuations is whether sources are subtracted to the same level in both data sets.
208: Although \citet{kas07b} does not explicitly state the achieved source subtraction
209: level, \citet{kas07a} states that any remaining sources in the IRAC image must be
210: fainter than 10--20 nJy. For the SEDs used in this study the IRAC limit
211: corresponds to a NICMOS limit of 4.5--9 nJy. From Figure 3b from \citet{thm07}
212: the completeness limit is approximately 20 nJy at 1.6$\micron$, which is 2 to 4
213: times brighter than the IRAC subtraction limit. The sources for subtraction,
214: however, were identified using detections in a combination of the ACS and
215: NICMOS HUDF
216: images, with the ACS images going significantly deeper due to their much longer
217: integration time. Figure 7 in \citet{thm06} shows that the detection limit in
218: the ACS F775W band is approximately 30 in AB magnitudes which is 3.7 nJy. We can
219: use the SEDs discussed in \S\ref{s-fc} to see what IRAC fluxes these correspond
220: to. The results are shown Figure~\ref{fig-cds}. Template SED 6 gives a limit of
221: 20 nJy and the bluer template 7 SED gives a limit of 10 nJy which are compatible
222: with the IRAC subtraction limits given by \citet{kas07a}. The F775W band only
223: removes sources at that limit for redshifts less than 3 but the ACS F850LP long
224: pass filter removes sources at that limit for redshifts up to 6. The two data
225: sets are compatible for all redshifts less than 6. At redshifts between 6 and 8
226: the NICMOS subtraction limit is not as stringent as the IRAC subtraction limit.
227: To account for this difference the horizontal solid line in Figures~\ref{fig-col}a
228: and b would have to be raised by a factor of 2 to 4 but that would still not be
229: compatible with redshifts greater than 12. The redshift distribution of galaxies
230: with NICMOS fluxes less than 20 nJy would have to be significantly different
231: than galaxies above this limit for the z = 6--8 sources to contribute any power
232: to the observed NICMOS fluctuations.
233:
234: \subsection{Signal or Noise?}
235:
236: All of the discussion above is predicated on the assumption that the fluctuations
237: observed after source subtraction are due to faint galaxies below the detection limit
238: and not noise sources from the instrument or telescope. To check for the signature of
239: Gaussian distributed noise in the NICMOS images we populated an image of the same
240: size as the UDF F110W and F160W images with random Gaussian distributed noise with
241: a variance equal to that found in the real images. An identical fluctuation
242: analysis gives the results shown by the triangles
243: in Figure~\ref{fig-fluct}. The random noise is close to the observed fluctuation power
244: at small angular scales, but is far below the observed power at angular scales of
245: $10\arcsec$
246: and larger, the region considered in this analysis. We therefore conclude that
247: the observed fluctuation power is not due to random noise components such as read noise,
248: Poisson photon statistics, or dark current in the NICMOS images.
249:
250: We originally did not include the F110W image in the fluctuation analysis due to the
251: possibility of residual flat fielding errors. The median of the 144 individual images
252: measures any flat field errors very accurately. The F110W mosaic image created by
253: replacing all of the individual actual images with the median image does appear to
254: contain a component of noise due to small flat field errors. It is evident as the
255: higher fluctuation power in Figure~\ref{fig-fluct}a at spatial scales between
256: $10\arcsec$ and $80\arcsec$. A similar signature is not present in the 1.1$\micron$
257: source subtracted fluctuation power. This indicates that the subtraction of the median
258: image from each actual image before the mosaic accurately removes any flat field
259: errors in the regions that have no detectable sources. These are the regions that
260: contribute to the all source subtracted fluctuation power. We therefore exclude flat
261: fielding errors as a source of fluctuation power.
262:
263: If the source subtracted fluctuations are due to faint, low redshift, galaxies then
264: the 1.1$\micron$ and 1.6$\micron$ source subtracted images should be spatially correlated.
265: A simple check for correlation is to subtract the F110W image from the F160W to test
266: whether the fluctuations are reduced as would be expected if they are correlated.
267: Since the exact scaling is not known we subtracted the F110W image with scalings
268: between 0 and 2 in 0.1 increments. The log weighted sum of the fluctuation powers
269: for angular scales greater than $30\arcsec$ versus the scale factor is shown in
270: Figure~\ref{fig-sub}a. There is a broad minimum for a scale factor of 0.6, near
271: the observed ratio of F160W to F110W fluctuation of 0.75 at $100\arcsec$.
272: Figure~\ref{fig-sub}b shows the nested fluctuation spectra. This
273: shows that the subtraction of the scaled F110W image reduces and alters the fluctuation
274: spectrum at larger angular scales. It also indicates that the two images are partially
275: correlated after the subtraction of all known sources as would be expected if fainter
276: undetected sources were present in the images. The lower boundary in Figure~\ref{fig-sub}b
277: is the upper limit on fluctuation power contributed by high redshift, 1.1$\micron$ dropout,
278: galaxies plus any remaining noise in the images. This limit is still above the
279: Gaussian noise spectrum shown in Figure~\ref{fig-fluct}a and b.
280:
281: At the largest scale of $100\arcsec$ the residual fluctuation power in the F160W
282: minus F110W image is 0.2 nWm$^{-2}$sr$^{-1}$. If the residual is due to
283: sources with redshifts less than 10 then the expected power in the IRAC
284: 3.6$\micron$ band is 0.02--0.2 nWm$^{-2}$sr$^{-1}$ depending on which SED
285: template is used. This is consistent with the 0.1 nWm$^{-2}$sr$^{-1}$
286: power observed by \citet{kas07b} at the same spatial scale. If the residual
287: power is due to galaxies at redshifts greater than 8 then the predicted
288: power is in the range of 0.04 to a very large number depending on the SED
289: and redshift. The predicted power starts to exceed the observed flux at a
290: redshift of 11 for template 6 and at a redshift of 12 for template 7. At
291: a redshift of 13.5 the predicted power exceeds the observed 3.6$\micron$
292: power by a factor of 10 for both SEDs. The residual F160W minus F110W power
293: to 3.6$\micron$ power ratio is therefore consistent with emission from
294: galaxies with redshifts less than 8 but cannot definitively rule out
295: that the observed 3.6$\micron$ fluctuation power comes from galaxies
296: in the redshift range between 8 and 13. Redshifts beyond 13, however,
297: are not consistent with the ratio if the difference between the lower
298: limit in Figure~\ref{fig-sub}b and the Gaussian noise is due to real
299: sources. It should be noted, however, that the predominant source of
300: fluctuation power in the observed sources comes from galaxies at
301: redshifts between 0.5 and 1.5. Having the residual F160W minus F110W
302: fluctuation power coming from galaxies with redshifts above 8 would
303: require a remarkable suppression of the luminosity function of the
304: 0.5--1.5 redshift galaxies at low luminosities.
305:
306: A final, but not as compelling, piece of evidence that the residual fluctuations are
307: due to galaxies is that the 1.1$\micron$ to 1.6$\micron$ fluctuation power color is
308: similar to that expected from $z<8$ galaxies. We cannot completely rule out
309: an unknown source of spatial power, which would be the same in both images due to
310: their similar observation pattern and data analysis. However, in the absence of any
311: known source of such spatial power we conclude that the
312: NICMOS source subtracted fluctuations are due to faint, $z<8$, galaxies below the detection limit.
313: These galaxies become evident in the fluctuation analysis because it utilizes $93\%$ of
314: the pixels while the detected source power is contributed by only $7\%$ of the image pixels.
315:
316: Without far more knowledge of the IRAC instrument and data analysis procedures
317: than we currently possess we cannot make the same arguments for the faint galaxy
318: origin of the IRAC fluctuations. We can say that the IRAC fluctuations are at
319: the levels predicted for faint, $z<8$, galaxies from the NICMOS observations.
320: Since all four bands have ratios consistent with a faint, $z<8$, galaxies
321: origin we assume that the fluctuations in all four bands have extragalactic
322: origins. We note that even if the IRAC fluctuations were due to noise, it would
323: still support our conclusions that the IRAC fluctuations are not due to very high
324: redshift galaxies.
325:
326: \section{Conclusions}
327:
328: Our basic conclusion is that the source subtracted fluctuations observed in the two NICMOS bands
329: and the IRAC bands at 3.6$\micron$ and 4.5$\micron$ are consistent with faint,
330: $z<8$, galaxies that are below the individual detection limit in the
331: respective images. We also conclude that the observed ratio of 1.1$\micron$
332: to 1.6$\micron$
333: fluctuations is incompatible with a significant fraction of power from $z>8$
334: galaxies and that the ratio of the IRAC 3.6$\micron$ to NICMOS 1.6$\micron$
335: fluctuation power is incompatible for galaxies at $z>13$. In light of these
336: observations we find no evidence that any of the observed fluctuations require
337: flux from high redshift galaxies. Of course such galaxies may exist, but we
338: conclude that they do not contribute measurable power to the current observations.
339:
340: The argument in \citet{kas07a} for the IRAC fluctuations being due to high
341: redshift galaxies appears to be based on the assumption that any sources
342: fainter than their detection limit must be at high redshift because they are
343: faint. Examination of Figure 3b in \citet{thm07} shows that the z = 0--7
344: galaxy luminosity function shows no break up to a completeness
345: limit of 5 nJy in the F160W band. This implies that there are many more
346: low redshift galaxies fainter than the detection limit, i.e. galaxies that are not
347: faint because they are at $z>8$.
348:
349: A caveat relevant to this study is that if it is postulated that the IRAC
350: fluctuations are due to galaxies with $z>13$ then the predicted
351: 1.6$\micron$ fluctuations would be below our detection level due to the
352: very high ratio of IRAC power to NICMOS power at high redshifts. However if
353: these high-redshift galaxies are the only source of the IRAC fluctuations,
354: then a mechanism must be found to suppress the low redshift IRAC fluctuations
355: predicted from the NICMOS residual fluctuations. At this time we know of no
356: such mechanism.
357:
358:
359: \acknowledgments
360: This article is based on data from observations with the NASA/ESA Hubble Space Telescope
361: obtained at the Space Telescope Science Institute, which is operated by the
362: Association of Universities for Research in Astronomy under NASA contract
363: NAS 5-26555. We would like to thank the anonymous referee for helpful
364: comments in the preparation of this article.
365:
366: \begin{thebibliography}{}
367:
368: \bibitem[Bertin and Arnout(1996)]{ber96}Bertin, E., and Arnouts, S. 1996, \aap,
369: 117, 393.
370:
371: \bibitem[Bruzual \& Charlot(2003)]{bru03} Bruzual, G. \& Charlot, S., 2003,
372: \mnras, 344, 1000
373:
374: \bibitem[Chabrier(2003)]{cha03} Chabrier, G. 2003, \pasp, 115, 763
375:
376:
377: \bibitem[Cooray et al.(2007)]{coo07} Cooray, A., et al. 2007, \apjl, 659, L91
378:
379: \bibitem[Fruchter and Hook(2002)]{fru02} Fruchter, A.S. \& Hook, R.N. 2002, \pasp, 114, 144.
380:
381: \bibitem [Kashlinsky et al.(2002)]{kas02} Kashlinsky, A. et al. 2002, \apjl, 579, L53
382:
383: \bibitem [Kashlinsky et al.(2005)]{kas05b} Kashlinsky, A., Arendt, R. G., Mather, J.,
384: \& Mosley, S. H. 2005, \nat, 438, 45
385:
386: \bibitem[Kashlinsky(2007)]{kas07} Kashlinsky, A. 2007, Astro-ph/0701147
387:
388: \bibitem[Kashlinsky et al.(2007a)]{kas07a} Kashlinsky, A., Arendt, R.G., Mather, J. \&
389: Moseley, S.H. 2007, \apjl, 654, L1, (Kashlinsky et al. 2007a)
390:
391: \bibitem[Kashlinsky et al.(2007b)]{kas07b} Kashlinsky, A., Arendt, R.G., Mather, J. \&
392: Moseley, S.H. 2007, \apjl, 654, L5, (Kashlinsky et al. 2007b).
393:
394: \bibitem[Madau et al. (1996)] {mad96} Madau, P., Ferguson, H.C.,
395: Dickinson, M. E.,Giavalisco, M., Steidel, C. C. \& Fruchter, A.
396: 1996, \mnras, 283, 1388.
397:
398: \bibitem [Magliocchetti, Salvaterra \& Ferrara(2003)]{mag03} Magliocchetti, M.,
399: Salvaterra, R., \& Ferrara, A. 2003, \mnras, 342, L25
400:
401: \bibitem [Matsumoto et al.(2005)]{mat05} Matsumoto, T. et al. 2005, \apj, 626, 31
402:
403: \bibitem [Salvaterra \& Ferrara(2003)]{sal03} Salvaterra, R., Ferrara, A. 2003, \mnras, 339, 973
404:
405: \bibitem[Salvaterra et al.(2006)]{sal06} Salvaterra, R., Magliocchetti, M.,
406: Ferrara, A \& Schneider, R. 2006, \mnras, 368, L6.
407:
408: \bibitem[Schaerer(2002)]{sch02} Schaerer, D. 2002, \aap, 382, 28.
409:
410: \bibitem [Thompson et al.(2005)]{thm05} Thompson, R. I., et al. \aj, 130, 1
411:
412: \bibitem [Thompson et al.(2006)]{thm06} Thompson, R. I. 2006, \apj, 647, 787
413:
414: \bibitem[Thompson et al.(2007)]{thm07} Thompson, R.I., Eisenstein, D., Fan, X.,
415: Rieke, M., \& Kennicutt, R.K. 2007, \apj, 657, 669.
416: \end{thebibliography}
417:
418: \clearpage
419:
420: \begin{figure}
421: \plottwo{f1a.ps}{f1b.eps}
422: \caption{The fluctuation spectra from a) the F110W NUDF image (left) and b) the F160W
423: image (right). The open squares are for no sources subtracted, the open diamonds
424: are for all sources with a F160W AB magnitude of 20 or brighter subtracted.
425: The asterisks are for all sources subtracted and the crosses are for the
426: median background field described in the text. The open triangles are
427: for a Gaussian distributed noise image with a standard deviation equal to
428: that of the images. The dashed line indicates the level of 1.6$\micron$
429: fluctuations found in deep calibration 2MASS fields by \citet{kas02}.
430: \label{fig-fluct}}
431: \end{figure}
432:
433: \clearpage
434:
435: \begin{figure}
436: \plotone{f2.ps}
437: \caption{The histogram of SED template numbers for all sources with redshifts
438: greater than 2.0. The interpolated templates go in increments of 0.1 between
439: the primary template SEDs.
440: \label{fig-temp}}
441: \end{figure}
442:
443: \clearpage
444:
445:
446: \begin{figure}
447: \plottwo{f3a.ps}{f3b.ps}
448: \caption{The predicted ratios of the powers, $\nu I_{\nu}$, in the F110W band
449: to the F160W band, dashed line; the the 3.6$\micron$ \emph{Spitzer} band to
450: F160W band, solid line; and the \emph{Spitzer} 4.5$\micron$ to 3.6$\micron$
451: band, dash dot line; versus redshift. The straight horizontal lines indicate
452: the observed values of the ratios at an angular scale of $100\arcsec$ with the
453: same line style code. a) (left) is for template6 and b) (right) is for
454: template 7, the hottest SED, which is consistent with a possible
455: Population III galaxy.
456: \label{fig-col}}
457: \end{figure}
458:
459: \clearpage
460:
461: \begin{figure}
462: \plotone{f4.ps}
463: \caption{The figure shows the IRAC fluxes in the 3.6$\micron$ and 4.5$\micron$
464: bands that correspond to a subtraction limit of 3.7 nJy in the ACS F775W band
465: as a function of wavelength for the two templates (6 and 7) used in this study.
466: The quoted subtraction level from \citet{kas07b} is 10 to 20 nJy, consistent
467: with subtraction levels in the figure. The sharp rise at approximately z=3
468: is when the Lyman break starts to enter the F775W band.
469: \label{fig-cds}}
470: \end{figure}
471:
472: \clearpage
473:
474: \begin{figure}
475: \plottwo{f5a.ps}{f5b.ps}
476: \caption{a) The left panels shows the $\ln(k)$ weighted sum of fluctuation
477: power for angular scales greater
478: than $30\arcsec$ versus the scaling factor of the F110W image subtracted from the
479: F160W image. The minimum is at a scaling factor of 0.6. b) The right panel shows the nested
480: fluctuation power spectra for the different scaling factors. The lower boundary is the
481: limit for the contribution of high redshift sources plus other sources of spatial noise.
482: \label{fig-sub}}
483: \end{figure}
484:
485: \clearpage
486:
487: \clearpage
488:
489: \begin{deluxetable}{lcccc}
490: \tabletypesize{\scriptsize}
491: \tablecaption{Properties of the template SEDs. Both SEDs are taken from
492: models of \citet{bru03} based on the Padova 1994 models. The IMFs are from
493: \citet{cha03}. \label{tab-temp}}
494: \tablewidth{0pt}
495: \tablehead{
496: \colhead{Template Number} & \colhead{Metallicity} & \colhead{IMF} & \colhead{Low Mass Limit} & \colhead{Upper Mass Limit}
497: }
498: \startdata
499: 6 & 0.004 & Chabrier & 0.1 & 100\\
500: 7 & 0.0001 & Chabrier & 0.1 & 100\\
501: \enddata
502:
503: \end{deluxetable}
504:
505: \end{document}
506: