1: %\documentclass[12pt,preprint]{aastex}
2:
3: \documentclass{emulateapj}
4: %\topmargin 0.5in
5:
6: \shortauthors{Brammer and van Dokkum}
7: \shorttitle{}
8: \slugcomment{Accepted for publication in ApJL}
9:
10:
11: % Additional private definitions
12:
13: \newcommand{\MSOL}{\mbox{$\:{\rm M}_{\sun}$}}
14: \newcommand{\zphot}{\mbox{$z_{\rm phot}$}}
15: \newcommand{\zspec}{\mbox{$z_{\rm spec}$}}
16: \newcommand{\ang}{\mbox{$\:$\AA}}
17:
18: %%%%%%%%%% % % BEGIN HERE % %%%%%%%%%%
19:
20: \begin{document}
21:
22: \title{The Density and Spectral Energy Distributions of Red Galaxies at $z\sim3.7$}
23:
24: \author{Gabriel B. Brammer\altaffilmark{1} and Pieter G. van
25: Dokkum\altaffilmark{1}} \email{brammer@astro.yale.edu}
26:
27: \altaffiltext{1}{Department of Astronomy, Yale University, New Haven,
28: CT, 06520-8101}
29:
30: \begin{abstract}
31:
32: We use the deep NIR imaging of the FIRES survey to investigate trends with
33: redshift of the properties of galaxies selected to have strong
34: Balmer/4000\ang\ breaks at $2<z<4.5$. Analagous to the $J-K>1.3$ (AB) color
35: criterion designed to select red galaxies at $z>2$, we propose two color
36: criteria, $J-H>0.9$ and $H-K>0.9$, to select red galaxies in two redshift bins
37: at $2<z<3$ and $3<z<4.5$, respectively. From the FIRES catalogs of the HDF-S
38: (4.7 arcmin$^2$) and MS 1054-03 (26.3 arcmin$^2$) fields, we find 18 galaxies
39: with $\left<z_\mathrm{phot}\right>=2.4$ that satisfy $J_s-H>0.9; H<23.4$ and
40: 23 galaxies with $\left<z_\mathrm{phot}\right>=3.7$ that satisfy $H-K_s>0.9;
41: K_s<24.6$, where the flux limits are chosen to match the limiting rest-frame
42: luminosities at the different median redshifts of the two samples. The space
43: densities of the $J_s-H$ and $H-K_s$ samples are $1.5\pm0.5\times10^{-4}$ and
44: $1.2\pm0.4\times10^{-4}\ \mathrm{Mpc}^{-3}$, respectively. The rest-frame
45: $U-B$ colors of galaxies in both samples are similarly red (as expected from
46: the definition of the color criteria), but the rest-frame UV properties are
47: different: galaxies in the higher-redshift $H-K_s$ selected sample have blue
48: NUV-optical colors and UV slopes similar to those of Lyman Break Galaxies,
49: while the $J_s-H$ galaxies are generally red over the entire wavelength range
50: observed. Synthetic template fits indicate that the distinct rest-NUV
51: properties of the two samples are primarily a result of dust: we find
52: $\left<A_V\right>_{JH}=1$ mag and $\left<A_V\right>_{HK}=0.2$ mag. The median
53: stellar mass determined from the template fits decreases by a factor of
54: $\sim5$ from $z=2.4$ to $3.7$, which, coupled with the fact that the space
55: density of such galaxies remains roughly constant, may imply that the stellar
56: mass density in red galaxies decreases by a similar factor over this
57: redshift range.
58:
59: \end{abstract}
60:
61: \keywords{cosmology: observations --- galaxies: evolution --- galaxies: formation}
62:
63: \section{Introduction}
64:
65: Over the last few years, the study of galaxies with red rest-frame optical
66: colors has been extended to ever-increasing redshifts. While the selection of
67: galaxies based on their rest-frame UV emission \citep[e.g. the Lyman break
68: technique;][]{steidel93} has enabled the detailed study of young galaxies at
69: high redshift for some time, it is only with the more recent advent of
70: efficient, large-scale detectors in the near-IR that large numbers of galaxies
71: with redder rest-frame colors have been discovered at early cosmic epochs
72: \citep{franx03,vd03}. Red galaxies with typically very low UV fluxes make up
73: 80\% of the mass contained in the most massive galaxies \citep{vd06} and 25-75\%
74: of the total mass in galaxies at $2<z<3$ \citep{papovich06, marchesini06}, and
75: thus provide critical constraints on theoretical models of galaxy formation and
76: evolution \citep{somerville04, nagamine05}.
77:
78: A proven technique for selecting galaxies with red rest-frame optical colors is
79: the $J_s-K_s>1.3$ criterion \citep[$J-K_\mathrm{Vega}>2.3$;][]{franx03}, which
80: relies on the Balmer/4000\ang\ break redshifted into the $J$ band for redshifts
81: $z>2$. Galaxies selected using this technique comprise a heterogeneous
82: population showing a broad range in dust properties, luminosity-weighted ages,
83: and star-formation rates \citep{forster04,labbe05,papovich06,kriek06b}.
84:
85: In this Letter, we extend the study of red galaxies to redshifts $z>3$, and we
86: compare the number and properties of red galaxies at $z\sim3.5$ to those at
87: $z\sim2.5$ using uniform color selection criteria based on the Balmer/4000\ang\
88: break redshifted into the $H$ and $J_s$ bands, respectively. While there is
89: typically a tail of sources extending beyond $z>3.5$ in $J-K$ selected samples
90: \citep[e.g.][]{forster04}, there have been no systematic studies comparing the
91: numbers and properties of red galaxies at the extremes of the broad $J-K$
92: redshift selection window. We adopt $H_0=70$ km/s, $\Omega_{\mathrm m} = 0.3$,
93: and $\Omega_\Lambda = 0.7$. All magnitudes are given in the AB system.
94:
95:
96: \section{Color Selection of Red Galaxies}
97:
98: The top panel of Fig.~\ref{fig:colors} shows the observed $J_s-K_s$ color of
99: \cite{bc03} template spectra over $0<z<5$. At a given redshift, $J_s-K_s$ is
100: reddest for the oldest template because the strength of the Balmer/4000\ang\
101: break increases as a stellar population ages. Based on the figure,
102: $J_s-K_s>1.3$ should select galaxies at $z>2$ that are either dominated by an
103: evolved stellar population or are highly reddened by dust---the so-called
104: ``distant red galaxies'', or DRGs \citep{franx03}. A number of DRGs have been
105: spectroscopically confirmed \citep{vd03,kriek06,kriek06b} at $z>2$, and large
106: photometric samples of DRGs are found to have $1.5\lesssim\zphot\lesssim3.5$
107: \citep{forster04,papovich06,quadri06}.
108:
109: Because DRGs have a fairly broad redshift distribution, the rest-frame
110: properties of galaxies satisfying the color criterion to a given magnitude
111: limit in a particular selection band change with redshift in three important
112: ways. First, the limiting absolute magnitude in the selection band becomes
113: brighter with increasing redshift for a fixed survey depth, probing sharply
114: decreasing source densities at the bright end of the steep luminosity function.
115: Second, the rest wavelength of the selection band decreases with increasing
116: redshift, probing wavelengths where the scatter in $M/L_\lambda$ is large.
117: Finally, the rest-frame color---essentially the type of galaxy
118: selected---changes with redshift because the Balmer/4000\ang\ break is narrow
119: in wavelength compared to the spacing between the $J_s$ and $K_s$ filters and
120: because the spectral slope is not the same on both sides of the break for galaxies
121: older than $\sim100$ Myr. Together, these effects make it very difficult to
122: compare DRG properties at different redshifts.
123:
124: The bottom panel of Fig.~\ref{fig:colors} demonstrates how splitting the $J-K$
125: criterion into two should divide DRGs into two redshift bins,
126: \begin{eqnarray}
127: J_s-H>0.9\:&:&\:z\gtrsim2 \nonumber \\
128: H-K_s>0.9\:&:&\:z\gtrsim3,
129: \label{eq1}
130: \end{eqnarray}
131:
132: \noindent mitigating the problems described above that prevent direct analysis
133: of the variation of red galaxy properties with redshift. The wavelength
134: baselines of the two NIR colors are similar enough that the same color limit
135: can be used for both criteria to select identical rest-frame powerlaw spectral
136: slopes. The criteria of Equation \ref{eq1} are adopted analagous to
137: $J_s-K_s>1.3$ to select against low-$z$ interlopers, while the selection
138: efficiency of the $J_s-H$ and $H-K_s$ criteria may be further enhanced by the
139: fact that those colors are steeper functions of redshift at the low-$z$
140: selection boundaries than $J_s-K_s$.
141:
142: Alternatively, one could select samples of galaxies based on their rest-frame
143: properties estimated using distances determined from the galaxies' photometric
144: redshifts. We note, however, that the $\zphot$--$\zspec$ calibration is poorly
145: determined at $z>3$, and therefore relying on photometric redshifts alone
146: introduces large uncertainties in the analysis at high redshift. Furthermore,
147: by working in the observed frame, the results can be easily verified by others
148: independent of photometric redshift or analysis techniques.
149:
150: %%%%% Fig 1: Color vs. z %%%%%%%%
151: \begin{figure}
152: %\epsscale{0.6}
153: \plotone{f1.eps}
154: \caption{\textit{Top}: Evolution of $J_s-K_s$ color with redshift. The dashed
155: line shows the DRG criterion of \cite{franx03} designed to select
156: passively-evolving galaxies at $z>2$. \textit{Bottom}: $J_s-H$ (blue) and
157: $H-K_s$ (red) $vs.\ z$. The dashed line indicates the selection criteria used
158: in this Letter, which are analogous to the DRG color selection but that divide
159: the DRG sample into two redshift bins. The shaded bands indicate the range of
160: colors of dust-free \cite{bc03} passively evolving templates with ages between
161: 0.25 and 1.0 Gyr. The thin lines correspond to a template with a constant star
162: formation rate with age 0.1 Gyr and $E(B-V) = 0.5$, showing that moderately
163: reddened starbursts at low redshift are not expected to significantly
164: contaminate the $z>2$ NIR-selected galaxy samples.
165: \label{fig:colors}}
166: \end{figure}
167:
168: \section{Data \label{sec_data}}
169:
170: We use the deep optical+NIR photometry of the FIRES survey \citep{franx00} to
171: select red galaxies at $z>2$ based on the two NIR color criteria described
172: above. Details of the data reduction and $K$-selected source catalogs of the
173: two fields of the survey, HDF-S (4.7 arcmin$^2$) and MS 1054-03 (26.3
174: arcmin$^2$), can be found in \cite{labbe03} and \cite{forster06},
175: respectively. Briefly, the combined catalog of the two fields contains HST
176: $U_{300}B_{450}$ (HDFS-S), $V_{606}I_{814}$ (both fields), ground-based $UBV$
177: (MS 1054-03) and ISAAC-$J_sHK_s$ (both fields) photometry, along with
178: photometric redshifts determined following the procedure described by
179: \cite{rudnick01,rudnick03}. Details of the accuracy of the photometric
180: redshifts in these fields are given in \cite{labbe03} and \cite{forster06}.
181: The combined catalog depth is limited by the shallower MS 1054-03
182: observations. At $K_s=24.6$, the combined catalog is $\sim90\%$ complete and
183: sources in the MS 1054-03 catalog have $\left(S/N\right)_K=5-7$. The FIRES
184: dataset provides a unique opportunity to select red galaxies at $z>3.5$---even
185: the brightest of which should still be quite faint in $K$---since there are no
186: other currently available surveys of comparable depth in all three $JHK$ NIR
187: filters.
188:
189: \section{Densities \label{sec_density}}
190:
191: From the combined FIRES catalog we find 18 sources that satisfy
192: $J_s-H>0.9\:;\:H<23.4$ and 23 sources with $H-K_s>0.9\:;\:K_s<24.6$. The
193: limiting magnitude in $H$ was chosen such that the rest-frame limiting magnitude
194: redward of the Balmer/4000\ang\ break is the same for both samples, and its
195: value was determined based on the ratio of the luminosity distances at the
196: median redshifts of the two samples. The number of selected sources correspond
197: to combined (HDF-S-only) surface densities and Poisson errors of
198: $0.58\pm0.18\;(0.63\pm0.37)$ and $0.74\pm0.19\;(0.42\pm0.30)$ arcmin$^{-2}$ for
199: the $J_s-H$ and $H-K_s$ selected samples, respectively. The photometric
200: redshift distributions of the NIR selected sources are shown in
201: Fig.~\ref{fig:zhist}. The $J_s-H$ sample has $\zphot=2.4\pm0.3$ (1-$\sigma$
202: range) and 16/18 galaxies also satisfy $J_s-K_s>1.3$. For the $H-K_s$ sample,
203: $\zphot=3.9\pm1.0$ and 15/23 galaxies satisfy $J_s-K_s>1.3$.
204:
205: If we consider comoving volumes bounded by tophat redshift distributions of
206: $2<\zphot<3$ and $3<\zphot<4.5$, the area-weighted space densities and
207: associated Poisson errors of the $J_s-H$ and $H-K_s$ samples are
208: $1.5\pm0.5\times10^{-4}$ and $1.2\pm0.4\times10^{-4}\ \mathrm{Mpc}^{-3}$,
209: respectively. Therefore, the space density of a flux-limited sample of red
210: galaxies remains constant within the errors from $\left<z\right>_{\rm
211: JH}\sim2.4$ to $\left<z\right>_{\rm HK}\sim3.7$. We note that the total survey
212: area discussed here is relatively small and that our result is likely subject to
213: cosmic variance, especially in light of recent studies that show that red
214: galaxies at $z>2$ are strongly clustered \citep{daddi03,quadri06}.
215:
216: %%%%%% % Fig 2: redshift histograms %%%%%%%
217: \begin{figure}
218: %\epsscale{0.8}
219: \plotone{f2.eps}
220: \caption{Photometric redshift distribution of galaxies in the FIRES fields with
221: $J_s-H>0.9; H<23.4$ (blue) and $H-K_s>0.9; K_s<24.6$ (red). Although the samples show
222: a small overlap in redshift space, no galaxy meets both selection criteria. For
223: comparison, the dotted line shows the redshift distribution of galaxies satisfying
224: $J_s-K_s>1.3$ and $K_s<24.6$.
225: \label{fig:zhist}}
226: \end{figure}
227:
228: \section{Rest Frame SEDs \label{sec_sed_fits}}
229:
230: Athough there is no appreciable change in the \textit{number} of galaxies
231: selected to have strong Balmer/4000\ang\ breaks from $z\sim2.4$ to $z\sim3.7$,
232: their properties are quite different. Fig.~\ref{fig:seds} shows the spectral
233: energy distributions (SEDs) of the two galaxy samples shifted to the
234: rest-frame. The $J_s-H$ galaxies typically have red rest-frame UV-optical
235: colors and a fairly flat rest-UV spectrum, consistent with previous studies of
236: DRGs in the redshift range $2<z<3$ \citep{forster04}. In contrast, the
237: higher-redshift $H-K_s$ galaxies generally have blue UV-optical colors. The
238: spectral slopes of the two samples through the Balmer/4000\ang\ break are
239: similar for the two samples, confirming that the NIR color criteria proposed
240: here select galaxies with similar $\left(U-B\right)_{\rm rest}$ colors over
241: $2<z<4.5$.
242:
243: We quantify the spectral shapes of the SEDs using the rest-frame UV power-law
244: slope, $F_\lambda\propto\lambda^\beta$ \citep{calzetti94}, measured from a
245: best-fit \cite{bc03} template with an exponentially decaying star formation
246: rate ($\tau=300$ Myr) and solar metallicity. The template fits to the
247: broadband photometry hold the redshift fixed to the catalog \zphot\ values,
248: allow ages between 0.1 Myr and the age of the universe at \zphot\, and allow
249: $A_V=0-3$ mag following the extinction law of \cite{calzetti00}. Corrections
250: for Ly$\alpha$ forest absorption are applied following \cite{madau95}. The
251: distributions of $\beta$ for the two galaxy samples are shown in
252: Fig.~\ref{fig:seds}. The distribution is quite flat for the $J_s-H$
253: sample---similar to the distribution seen for a large sample of massive DRGs
254: by \cite{vd06}. In contrast, the distribution of $\beta$ for the $H-K_s$
255: sample shows a peak at $\beta\sim-2$, values similar to those found by
256: \cite{adelberger00} for UV-selected galaxies and to the sample of massive
257: Lyman break galaxies (LBGs) discussed by \cite{vd06}.
258:
259: Of the 12 galaxies in the $H-K$ sample that have \zphot\ in the range
260: $2.7<z<3.4$ or $3.9<z<4.5$, 10 have synthetic $U_nG{\cal R}I$ colors
261: integrated from the best-fit templates that satisfy the $U$ or $G$ dropout LBG
262: color criteria at those redshifts proposed by \cite{steidel99}. However, 8 of
263: those 10 galaxies with LBG colors have ${\cal R}>25.5$, too faint to be
264: included in typical spectroscopic samples of LBGs.
265:
266: %%%%%%%%%%%%% Fig 3: SEDS %%%%%%%%%
267: \begin{figure}
268: %\epsscale{0.8}
269: \plotone{f3.eps}
270: \caption{Rest-frame SEDs of the $J_s-H$ (blue circles) and $H-K_s$ (red
271: diamonds) galaxy samples limited to $2<\zphot<3$ and $3<\zphot<4.5$,
272: respectively. The SEDs are normalized to the flux at $\lambda_{\rm rest} =
273: 3700$\ang\, determined from a linear interpolation between the two nearest
274: filters in the rest frame. The solid red and blue lines indicate a running
275: median of the 10 neighboring points for the two samples. An unreddened 1 Gyr
276: old $\tau_{300}$ template is shown for reference. The similarity of the SED
277: slopes at the Balmer/4000\ang\ break demonstrates how the two color criteria
278: are matched to select similar rest-frame spectral shapes at two different
279: redshifts. $Inset$: distribution of the synthetic rest-frame UV spectral
280: slope, $\beta$. The thin histograms correspond to the entire samples, while
281: the filled histograms are for sources that fall within the $2<\zphot<3$ and
282: $3<\zphot<4.5$ comparison ranges. With few exceptions, the higher redshift
283: galaxies in the $H-K_s$ sample have UV-optical colors that are significantly
284: bluer than those of the $J_s-H$ selected galaxies at lower redshift.
285: \label{fig:seds}}
286: \end{figure}
287:
288: \section{Discussion}
289:
290: We have shown that we can efficiently select $z\sim3.7$ galaxies in the near-IR
291: with the simple color criterion $H-K>0.9$. The samples described here
292: indicate that the rest-frame UV-optical SEDs of galaxies selected to have
293: strong Balmer/4000\ang\ breaks are significantly different at $z\sim2.4$ and
294: $z\sim3.7$: galaxies in the higher redshift sample have a median NUV/optical
295: flux ratio $2-4$ times greater than that of the galaxies at $z\sim2.4$.
296: Finding evolution in the properties of galaxies over the $\sim1$ Gyr between
297: $z=2.4-3.7$, which at the distant end is only 1.7 Gyr after the Big Bang, is in
298: itself not surprising, as the spectral evolution is rapid at these redshifts
299: for galaxies with a broad range of formation redshifts and subsequent star
300: formation histories. What is interesting is that the galaxies in both samples
301: were selected to have similarly strong Balmer/4000\ang\ breaks, indicating the
302: presence of an evolved stellar population that itself would not be a likely
303: source for the strong NUV component of the $z\sim3.7$ SEDs. A two-burst model
304: whose optical SED is dominated by an evolved stellar population but that also
305: contains a young component that supplies the UV flux could explain the
306: $z\sim3.7$ SEDs. Though two-component model fits are beyond the scope of this
307: Letter, there have been other observations of red galaxies at high-$z$
308: suggesting that composite populations are appropriate, both from SED modelling
309: \citep{yan04} and from observations of distinct UV and optical galaxy
310: morphologies \citep{toft05} of such galaxies.
311:
312: Our $\tau_{300}$ template fits suggest that dust is the primary source of the
313: difference in spectral shape shown in Fig.\ref{fig:seds}. Both samples have a
314: median template age of $\sim1$ Gyr, while the median dust extinction decreases
315: from $A_V=1$ to 0.2 mag going from the $J_s-H$ to the $H-K_s$ sample. In
316: general one expects the dust content of galaxies, along with the metal content,
317: to increase with time, qualitatively consistent with the analysis presented
318: here. It is difficult to quantitatively compare our results to theoretical
319: models of galaxy formation since such models typically rely on ad hoc treatments
320: of dust absorption \citep[e.g.][]{croton06}. Interestingly, our results may be
321: qualitatively consistent with the redshift distribution of
322: submillimeter-selected galaxies. Although still quite uncertain, there is
323: evidence that the number of very dusty, luminous galaxies drops significantly
324: from $z\sim2.4$ to $z\sim3.7$ \citep{chapman05}.
325:
326: The normalization of our population synthesis fits provides a rough estimate of
327: the stellar mass of each galaxy in the sample. With our photometry sampling
328: only the rest-frame UV-optical light of the galaxies in our sample at $z>2$, the
329: stellar mass fits are uncertain to factors of $>2$ due to uncertainties in
330: $\zphot$ and in the IMF and to model degeneracies between age, dust,
331: metallicity, and star formation history. With those caveats in mind, our
332: template fits imply that the median stellar mass of red galaxies decreases by a
333: factor of $\sim5$ from $z\sim2.4$ to $z\sim3.7$. While the uncertainties are
334: large for the masses of individual galaxies, a Mann-Whitney test on the
335: distribution of masses suggests that the difference in the median masses of the
336: two samples is significant at the 99\% confidence level. \cite{erb06} observe a
337: similar trend in the mass of LBGs: the average dynamical and stellar masses of
338: $z\sim3$ LBGs are a factor of 2 smaller than for LBGs at $z\sim2$. We note here
339: that the $H-K$ galaxies are not necessarily direct progenitors of the $J-H$
340: galaxies, just as LBGs at $z=4$ are not necessarily progenitors of LBGs at
341: $z=2$. Many of the $H-K$ galaxies could fade below our magnitude limit after
342: $\sim1$ Gyr; conversely, many of the $J-H$ galaxies could have been much bluer 1
343: Gyr previously.
344:
345: Since the space densities of the $J_s-H$ and $H-K_s$ samples are statistically
346: equivalent, a decrease in the median stellar mass in the higher redshift sample,
347: if real, would indicate a decrease in the stellar mass density of galaxies
348: with evolved stellar populations. This may be consistent with the results of \cite{kriek06b}, who find that the
349: ages of apparently passive galaxies at $z\sim2.3$ (which make up 45\% of their
350: $K$-selected sample) are typically $\lesssim1$ Gyr. It seems unlikely that many
351: of the $z>3$ progenitors of these galaxies were already ``red and dead''.
352: On the other hand, this result may be difficult to reconcile with the existence
353: of a significant population of $M_*>10^{11}\MSOL$ galaxies at $z>6$, as may be
354: implied by the source described by \cite{mobasher05}. Furthermore, the best fit
355: \cite{bc03} $z=6.5$ template of the \cite{mobasher05} source has $\beta\sim2$
356: and would be an outlier even at $z\sim3.7$ compared to the sources in our $H-K$
357: sample.
358:
359: Clearly, much further work is required to fully understand how the properties
360: of red galaxies change with time at $z>2$. The differences shown in
361: Fig.~\ref{fig:seds} are model independent, to the extent that the photometric
362: redshifts approximate the true redshifts, but the results above based on
363: template fits are quite uncertain. The addition of $Spitzer$ photometry would
364: better constrain the stellar mass contained in these galaxies
365: \citep{wuyts06}. The brightest galaxies in the $H-K_s$ sample may be within
366: reach of NIR spectroscopy that could precisely determine redshifts and model
367: stellar populations based on the prominent Balmer/4000\ang\ breaks, as has
368: been done quite effectively for DRGs at $z\sim2.5$ \citep{kriek06}. Finally
369: we note that the UKIDSS Ultra Deep Survey is planned to be $\sim$0.2 mag
370: deeper than our adopted $K_s$ limit over an area $\sim90$ times that of
371: FIRES, which would vastly increase the sample sizes of galaxies selected as
372: described here.
373:
374: \begin{acknowledgements}
375:
376: We thank Ivo Labb{\'e} and Natascha F{\"o}rster Schreiber for providing the FIRES
377: data to the community, Ryan Quadri for discussion and comments, and the referee,
378: Michael Rowan-Robinson, for his suggestions that improved this Letter. Support
379: from National Science Foundation grant NSF CAREER AST 04-49678 is gratefully
380: acknowledged.
381:
382: \end{acknowledgements}
383:
384: \begin{thebibliography}{}
385:
386: \bibitem[Adelberger \& Steidel(2000)]{adelberger00} Adelberger,
387: K.~L., \& Steidel, C.~C.\ 2000, \apj, 544, 218
388:
389: \bibitem[Bruzual \& Charlot(2003)]{bc03} Bruzual, G.~\&~Charlot, S.\ 2003,
390: \mnras, 344, 1000
391:
392: \bibitem[Calzetti et~al.(1994)]{calzetti94} Calzetti, D., Kinney,
393: A.~L., \& Storchi-Bergmann, T.\ 1994, \apj, 429, 582
394:
395: \bibitem[Calzetti et~al.(2000)]{calzetti00} Calzetti, D., Armus,
396: L., Bohlin, R.~C., Kinney, A.~L., Koornneef, J., \& Storchi-Bergmann, T.\
397: 2000, \apj, 533, 682
398:
399: \bibitem[Chapman et~al.(2005)]{chapman05} Chapman, S.~C., Blain,
400: A.~W., Smail, I., \& Ivison, R.~J.\ 2005, \apj, 622, 772
401:
402: \bibitem[Croton et~al.(2006)]{croton06} Croton, D.~J., et~al.\
403: 2006, \mnras, 365, 11
404:
405: \bibitem[Daddi et~al.(2003)]{daddi03} Daddi, E., et~al.\ 2003,
406: \apj, 588, 50
407:
408: \bibitem[Erb et~al.(2006)]{erb06} Erb, D.~K., Steidel, C.~C., Shapley, A.~E.,
409: Pettini, M., Reddy, N.~A., \& Adelberger, K.~L.\ 2006, \apj, 646, 107
410:
411: \bibitem[F\"orster-Schreiber et~al.(2004)]{forster04} F\"orster-Shreiber, N.~M.,
412: et~al.\ 2004, \apj, 616, 40
413:
414: \bibitem[F\"orster Schreiber et~al.(2006)]{forster06} F\"orster Shreiber, N.~M.,
415: et~al.\ 2006, \aj, 131, 1891
416:
417: \bibitem[Franx et~al.(2000)]{franx00} Franx, M., et~al.\ 2000,
418: The Messenger, 99, 20
419:
420: \bibitem[Franx et~al.(2003)]{franx03} Franx, M., et~al.\ 2003, \apjl, 587, L79
421:
422: \bibitem[Kriek et~al.(2006a)]{kriek06} Kriek, M., et~al.\ 2006, \apj, 645, 44
423:
424: \bibitem[Kriek et~al.(2006b)]{kriek06b} Kriek, M., et~al.\ 2006, \apjl, 649, L71
425:
426: \bibitem[Labb{\'e} et~al.(2003)]{labbe03} Labb{\'e}, I., et~al.\ 2003, \aj, 125,
427: 1107
428:
429: \bibitem[Labb{\'e} et~al.(2005)]{labbe05} Labb{\'e}, I., et~al.\ 2005, \apjl,
430: 624, L81
431:
432: \bibitem[Madau(1995)]{madau95} Madau, P. 1995, \apj, 441, 18
433:
434: \bibitem[Marchesini et~al.(2006)]{marchesini06} Marchesini, D., et~al.\ 2006,
435: \apj, submitted (astro-ph/0610484)
436:
437: \bibitem[Mobasher et~al.(2005)]{mobasher05} Mobasher, B., et~al.\ 2005, \apj,
438: 635, 832
439:
440: \bibitem[Nagamine et~al.(2005)]{nagamine05} Nagamine, K., Cen, R., Hernquist,
441: L., Ostriker, J.~P. \& Springel, V.\ 2005, \apj, 618, 23
442:
443: \bibitem[Papovich et~al.(2006)]{papovich06} Papovich, C., et~al.\ 2006, \apj,
444: 640, 92
445:
446: \bibitem[Quadri et~al.(2006)]{quadri06} Quadri, R., et~al.\ 2006, \apj,
447: submitted (astro-ph/0606330)
448:
449: \bibitem[Rudnick et~al.(2001)]{rudnick01} Rudnick, G. et~al.\ 2001, \aj, 122,
450: 2205
451:
452: \bibitem[Rudnick et~al.(2003)]{rudnick03} Rudnick, G. et~al.\ 2003, \apj, 599,
453: 847
454:
455: \bibitem[Somerville et~al.(2004)]{somerville04} Somerville, R.~S. et~al.\ 2004,
456: \apj, 600, L135
457:
458: \bibitem[Steidel \& Hamilton(1993)]{steidel93} Steidel, C.~C. \& Hamilton, D.\
459: 1993, \aj, 105, 2017
460:
461: \bibitem[Steidel et~al.(1999)]{steidel99} Steidel, C.~C., Adelberger, K.~L.,
462: Giavalisco, M., Dickinson, M. \& Pettini, M.\ 1999, \apj,519,1
463:
464: \bibitem[Steidel et~al.(2004)]{steidel04} Steidel, C.~C.,
465: Shapley, A.~E., Pettini, M., Adelberger, K.~L., Erb, D.~K., Reddy, N.~A.,
466: \& Hunt, M.~P.\ 2004, \apj, 604, 534
467:
468: \bibitem[Toft et~al.(2005)]{toft05} Toft, S., van Dokkum, P.,
469: Franx, M., Thompson, R.~I., Illingworth, G.~D., Bouwens, R.~J., \& Kriek,
470: M.\ 2005, \apjl, 624, L9
471:
472: \bibitem[van Dokkum et~al.(2003)]{vd03} van Dokkum, P.~G., et~al.\ 2003, \apjl,
473: 587, L83
474:
475: \bibitem[van Dokkum et~al.(2006)]{vd06} van Dokkum, P.~G., et~al.\ 2006, \apjl,
476: 638, L59
477:
478: \bibitem[Wuyts et~al.(2006)]{wuyts06} Wuyts, S., et~al.\ 2006,
479: \apj, in press (astro-ph/0609548)
480:
481: \bibitem[Yan et~al.(2004)]{yan04} Yan, H., et~al.\ 2004,
482: \apj, 616, 63
483:
484: \end{thebibliography}
485:
486: \end{document}
487: