1: %\documentclass[12pt,preprint]{aastex}
2: \documentclass{emulateapj}
3: \usepackage{apjfonts}
4:
5: % define some shortcuts
6: \newcommand{\Fig}[1]{Figure~\ref{#1}}
7: \newcommand{\Sec}[1]{\S~\ref{#1}}
8: \newcommand{\Tab}[1]{Table~\ref{#1}}
9: \newcommand{\Eqn}[1]{Equation~\ref{#1}}
10: \newcommand{\imag}{\ensuremath{i^\prime}}
11: \newcommand{\zmag}{\ensuremath{z^\prime}}
12: \newcommand{\Msun}{\ensuremath{{\rm M}_\sun}}
13: \newcommand{\Zsun}{\ensuremath{{\rm Z}_\sun}}
14: \newcommand{\pow}[2]{\ensuremath{#1 \times 10^{#2}}}
15: \newcommand{\err}[3]{\ensuremath{#1_{-#2}^{+#3}}}
16: \newcommand{\note}[1]{{\bf \em \{#1\}}}
17:
18: % Running Heads
19: \shorttitle{Ly$\alpha$ Emitting Galaxies at $z = 3.1$}
20: \shortauthors{Lai et al.}
21:
22: \begin{document}
23: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
24: % Front Material
25: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
26:
27: % use these commands to prevent auntomatic capitalization in the title
28: \newcommand{\zee}{z}
29:
30: \title{Spitzer Constraints on the Stellar Populations of Lyman-alpha
31: Emitting Galaxies at $\zee = 3.1$}
32:
33: \author{Kamson Lai\altaffilmark{1}, Jia-Sheng Huang\altaffilmark{1},
34: Giovanni Fazio\altaffilmark{1}, Eric Gawiser\altaffilmark{2,3},
35: Robin Ciardullo\altaffilmark{4}, Maaike Damen\altaffilmark{5},
36: Marijn Franx\altaffilmark{5}, Caryl Gronwall\altaffilmark{4},
37: Ivo Labbe\altaffilmark{6}, Georgios Magdis\altaffilmark{7},
38: Pieter van Dokkum\altaffilmark{2}
39: }
40:
41: \altaffiltext{1}{Harvard-Smithsonian Center for Astrophysics, 60
42: Garden Street, Cambridge, MA 02138, USA; klai@cfa.harvard.edu}
43:
44: \altaffiltext{2}{Yale Astronomy Department and Yale Center for Astronomy and
45: Astrophysics, Yale University, P.O. Box 208121, New Haven, CT 06520,
46: USA}
47:
48: \altaffiltext{3}{Department of Physics and Astronomy, Rutgers
49: University, 136 Frelinghuysen Road, Piscataway, NJ 08854-8019}
50:
51: \altaffiltext{4}{Department of Astronomy \& Astrophysics, The
52: Pennsylvania State University, Davey Lab, University Park, PA 16802,
53: USA}
54:
55: \altaffiltext{5}{Sterrewacht Leiden, Leiden University, Postbus 9513,
56: NL-2300 RA Leiden, Netherlands}
57:
58: \altaffiltext{6}{Carnegie Observatories, 813 Santa Barbara Street,
59: Pasadena, CA 91101, USA}
60:
61: \altaffiltext{7}{Department of Astrophysics, Oxford University, Keble
62: Road, Oxford, OX1 3RH, UK}
63:
64:
65: \begin{abstract}
66: We investigate the stellar populations of a sample of 162 Ly$\alpha$
67: emitting galaxies (LAEs) at $z = 3.1$ in the Extended Chandra Deep
68: Field South, using deep Spitzer IRAC data available from the GOODS and
69: SIMPLE surveys to derive reliable stellar population estimates. We
70: divide the LAEs according to their rest-frame near-IR luminosities
71: into IRAC-detected and IRAC-undetected samples. About 70\% of the
72: LAEs are undetected in 3.6 \micron\ down to $m_{3.6} = 25.2$ AB.
73: Stacking analysis reveals that the average stellar population of the
74: IRAC-undetected sample has an age of $\sim 200$ Myr and a mass of
75: $\sim \pow{3}{8}$ \Msun, consistent with the expectation that LAEs are
76: mostly young and low-mass galaxies. On the other hand, the
77: IRAC-detected LAEs are on average significantly older and more
78: massive, with an average age $\ga 1$ Gyr and mass $\sim 10^{10}$
79: \Msun. Comparing the IRAC colors and magnitudes of the LAEs to $z
80: \sim 3$ Lyman break galaxies (LBGs) shows that the IRAC-detected LAEs
81: lie at the faint blue end of the LBG color-magnitude distribution,
82: suggesting that IRAC-detected LAEs may be the low mass extension of
83: the LBG population. We also present tentative evidence for a small
84: fraction ($\sim 5\%$) of obscured AGN within the LAE sample. Our
85: results suggest that LAEs posses a wide range of ages and masses.
86: Additionally, the presence of evolved stellar populations inside LAEs
87: suggests that the Ly$\alpha$ luminous phase of galaxies may either be
88: a long-lasting or recurring phenomenon.
89: \end{abstract}
90:
91: \keywords{cosmology: observations --- galaxies: evolution ---
92: galaxies: high-redshift --- galaxies: stellar content}
93:
94: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
95: \section{Introduction} \label{Intro}
96: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
97: Lyman-alpha emitting galaxies (LAEs) are important tracers of galaxy
98: formation. The Ly$\alpha$ emission is produced by on-going star
99: formation in the galaxies, and the line emission enables discovery of
100: objects that may be too faint to be seen in the continuum. LAEs
101: therefore offer an opportunity to probe the faint end of galaxy
102: formation at high redshift, and may serve as building blocks of larger
103: galaxies in a hierarchical universe.
104:
105: In recent years, there have been great advancements in narrow-band
106: surveys of LAEs. Large samples of candidate LAEs now exist at
107: redshifts $z = 3 - 7$ \citep[e.g.][]{hu04, malho04, tanig05, shima06,
108: dawso07, muray07, ouchi07}. One outstanding question regarding the
109: LAEs is their stellar population, which is important for understanding
110: the physical nature of the LAEs and their connection with other high
111: redshift galaxy populations such as Lyman break galaxies
112: \citep[LBGs;][]{shapl03, ando06, pente07, stanw07}. A difficulty
113: often encountered when studying the stellar population of
114: high-redshift galaxies is the necessity to have observations at or
115: redward of rest-frame optical in order to constrain the total stellar
116: mass. At $z \ga 3$, this means having either deep near-IR or Spitzer
117: IRAC observations. Much progress has been made recently, however, and
118: several studies have suggested that LAEs are small galaxies with
119: masses $\la 10^9 - 10^{10}$ \Msun\ \citep{gawis06b, lai07, finke07,
120: pirzk07, nilss07}.
121:
122: In this study, we focus on a sample of 162 LAEs at $z = 3.1$
123: discovered in a narrow-band survey of the Extended Chandra Deep Field
124: South (ECDF-S) by \citet{gronw07}. The narrow-band 4990 \AA\ imaging
125: covers an area of 992 arcmin$^2$ and reaches a completeness limit of
126: $\sim \pow{1.5}{-17}$ erg cm$^{-2}$ s$^{-1}$ (equivalent to a
127: Ly$\alpha$ luminosity $L_{{\rm Ly}\alpha} = \pow{1.3}{42}$ erg
128: s$^{-1}$). Spectroscopic results confirm the robustness of the sample
129: (details of the spectroscopy will be presented in \citealt{gawis07}
130: and Lira et al., in preparation). Of the 52 candidates with
131: sufficient S/N in the spectra to yield redshifts, 51 were confirmed to
132: be $z=3.1$ LAEs (the remaining object is a $z=1.60$ AGN with C~III]
133: 1909 \AA\ emission). The most likely contaminants in our sample are
134: $z=0.34$ [O~II] emitters. However, there should be a negligible
135: number of these objects in our sample because by requiring the
136: observer's frame equivalent width to be $> 80$ \AA, we eliminate all
137: but the strongest (and rarest) [O~II] emitters \citep{gronw07}.
138:
139: The narrow-band imaging in ECDF-S is supplemented by optical and
140: near-IR observations from MUSYC \citep{gawis06a}, GOODS
141: \citep{dicki03}, and GEMS \citep{rix04} surveys, and Spitzer
142: observations from GOODS and SIMPLE\footnotemark\ (Damen et al., in
143: preparation). Using the multi-wavelength data available in ECDF-S, we
144: study the rest-frame UV to near-IR properties of this large sample of
145: LAEs. We place special emphasis on the new Spitzer IRAC observations,
146: which sample the rest-frame $0.9 - 2$ \micron\ emission from the $z =
147: 3.1$ LAEs and provide valuable constraints on their stellar population
148: and star formation history.
149: \footnotetext{http://www.astro.yale.edu/dokkum/SIMPLE/}
150:
151: We assume a ${\rm \Lambda}$CDM cosmology with $\Omega_{\rm M} = 0.3$,
152: $\Omega_{\rm \Lambda} = 0.7$, and $h = 0.7$. All magnitudes are in
153: the AB system.
154:
155: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
156: \section{IRAC Observations of $\zee = 3.1$ LAEs} \label{Analysis}
157: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
158: When working with deep IRAC observations, source confusion is often an
159: issue because of the broad IRAC PSF ($\sim 2\arcsec$ FWHM). Since the
160: LAEs are in general quite faint in the IRAC bands, it is important to
161: select a subsample whose IRAC photometry is not significantly
162: contaminated by nearby objects. Beginning with the original sample of
163: 162 LAEs, we select those with no detected neighbors within a
164: 4\arcsec\ radius ($\sim 2\times$ the IRAC FWHM), and additionally
165: reject LAEs with very bright neighbors at $> 4\arcsec$ radius by
166: visual inspection. This results in a high confidence
167: IRAC-uncontaminated sample of 76 LAEs.
168:
169: \begin{figure}
170: %\plotone{f1.eps}
171: \includegraphics[width=\columnwidth]{f1.eps}
172: \caption{ $R - m_{3.6}$ vs.\ $R$ color-magnitude diagram of the
173: IRAC-uncontaminated LAE sample. The dotted line represents the
174: $m_{3.6}=25.2$ flux cut used to define the IRAC-undetected sample
175: (shown as upper limits). Objects that are undetected in both $R$
176: and 3.6 \micron\ (15 in total) are omitted. Note also that there
177: are two objects that are detected in 3.6 \micron\ but not in $R$
178: (shown as lower limits near the right of the plot). Red $R -
179: m_{3.6}>2$ objects are marked by blue squares, with the x-ray
180: detected AGN candidate shown with a filled square. The red
181: five-point stars are results of stacking (in order of increasing
182: $R$) the IRAC-detected and IRAC-undetected samples.
183: \label{Rm36}
184: }
185: \end{figure}
186:
187: We further separate the IRAC-uncontaminated sample into an
188: IRAC-detected sample and a corresponding IRAC-undetected sample. This
189: division separates the rest-frame near-IR luminous objects from the
190: less luminous ones. Of the 76 IRAC-uncontaminated LAEs, 52 are
191: undetected in the IRAC 3.6 \micron\ channel down to a flux limit of
192: 0.3 $\mu$Jy ($m_{3.6}=25.2$). This is nominally the 2$\sigma$
193: detection limit of the survey. However, since the LAE positions are
194: well determined from the narrow band imaging, it is possible to do
195: slightly better with aperture photometry. Aperture corrected fluxes
196: are measured from the IRAC images using 2\farcs5 diameter apertures.
197: The flux cut of 0.3 $\mu$Jy corresponds approximately to a 3$\sigma$
198: detection limit for the individual objects. The initial IRAC-detected
199: sample consists of 24 objects, but 6 objects are rejected as possible
200: AGN or dusty objects (see below), leaving a total of 18 objects in the
201: IRAC-detected sample.
202:
203: IRAC samples rest-frame $0.9-2$ \micron\ at $z \sim 3$ and provides
204: important constraints on the mature stellar population in the LAEs.
205: In particular, the $R-m_{3.6}$ color, shown for the
206: IRAC-uncontaminated sample in \Fig{Rm36}, probes the relative
207: contribution from young and mature stellar populations, and is
208: therefore an age indicator. The dotted line represents the
209: $m_{3.6}=25.2$ flux limit for the IRAC-undetected sample. All objects
210: in this sample will fall as upper limits on the dotted line. The
211: survey limits imply that we are not sensitive to young stellar
212: populations with $R-m_{3.6} < 0$. However, stacking analysis can be
213: used to obtain constraints on the average properties of the
214: IRAC-undetected sample (see next section). The IRAC-detected LAEs
215: typically have red rest-frame UV to near-IR colors, $0 < R-m_{3.6} <
216: 2$, with a median of $R-m_{3.6}=1.2$. The red $R-m_{3.6}$ colors of
217: the IRAC-detected LAEs imply the possible presence of a significant
218: mature stellar population in LAEs, a point we will examine more
219: closely in the following section.
220:
221: \begin{figure}
222: %\plotone{f2.eps}
223: \includegraphics[width=\columnwidth]{f2.eps}
224: \caption{ $m_{3.6}-m_{8.0}$ vs.\ $m_{8.0}$ color-magnitude diagram of
225: the $R-m_{3.6} > 2$ LAEs (blue squares; the x-ray detected object is
226: marked with a filled square), $z \sim 3$ AGN/QSOs (green stars;
227: \citealt{steid03}; Magdis et al., in preparation), and 8 \micron\
228: detected LBGs (black dots; \citealt{rigop06}). The limits obtained
229: by stacking the IRAC-detected sample are marked with the red star.
230: \label{AGN_color}
231: }
232: \end{figure}
233:
234: While the $R-m_{3.6}$ color traces the age of the stellar population,
235: the 3.6 \micron\ magnitude probes the mature stellar mass in the LAEs.
236: Therefore, the IRAC-detected sample, which has 3.6 \micron\ magnitudes
237: ranging from $m_{3.6} \sim 23.5 - 25$, should in general be more
238: massive than the $m_{3.6} > 25.2$ IRAC-undetected sample. The
239: IRAC-detected sample is also in general brighter in $R$, with the
240: majority of IRAC-detected LAEs being brighter than $R \sim 26$.
241:
242: We also explore the possibility of a correlation between Ly$\alpha$
243: luminosity and IRAC flux. The mean Ly$\alpha$ luminosity of the
244: IRAC-detected sample ($\log(L_{{\rm Ly}\alpha}) = 42.47$ erg s$^{-1}$)
245: is about 40\% higher than that of the IRAC-undetected sample
246: ($\log(L_{{\rm Ly}\alpha}) = 42.34$ erg s$^{-1}$). This is only a
247: weak correlation, considering that the average 3.6 \micron\ flux of
248: the IRAC-detected sample is $8\times$ higher than the IRAC-undetected
249: sample (\Fig{sed}). There is also significant overlap in the $L_{{\rm
250: Ly}\alpha}$ distributions of the two samples, in part due to the large
251: ($\sim 0.2$ dex) dispersion in $L_{{\rm Ly}\alpha}$ within the
252: subsamples.
253:
254: It is interesting to note that there are four objects in the sample,
255: corresponding to LAE 1, 11, 117, and 144 in the \citet{gronw07}
256: catalog, with very red colors $R-m_{3.6} > 2$. LAE 1, the brightest
257: Ly$\alpha$ source in the sample, is detected in x-ray and is
258: presumably an AGN \citep{gronw07}. The similar $R-m_{3.6}$ colors of
259: the remaining three x-ray undetected objects suggest that they may
260: possibly be AGN as well. LAE 1, 11, and 117 are spectroscopically
261: confirmed to be at $z \sim 3$. The x-ray detected object (LAE 1)
262: shows C~IV emission in its spectrum, while LAE 11 and 117 show only
263: Ly$\alpha$ emission (Lira et al., in preparation). LAE 144 is
264: observed, but no line was detected in its spectrum. LAE 11 and 117
265: are in the central part of CDF-S where the $0.5-8$ keV (rest-frame
266: $0.1-2$ keV) flux limit of $< \pow{3}{-16}$ erg s$^{-1}$ cm$^{-2}$
267: implies an upper limit of $<\pow{2.5}{43}$ erg s$^{-1}$ for the
268: objects' AGN x-ray luminosities. The four objects have $R-m_{4.5}$
269: colors ranging from $2.4<R-m_{4.5}<4$, consistent with the colors of
270: mid-IR selected AGN \citep{hicko07}. However, the observed
271: $R-m_{4.5}$ colors can also be produced by a dusty stellar continuum
272: with $E(B-V) \sim 0.5$. These four objects are the only 8.0 \micron\
273: detected sources in the IRAC-uncontaminated sample and with $m_{8.0}$
274: ranging from $21 - 22$ they are more than 2 mags brighter than the
275: average IRAC-detected LAE (\Fig{AGN_color}). Two of the four objects
276: have $m_{3.6} - m_{8.0}$ colors consistent with $z \sim 3$ AGN and
277: QSOs (\citealt{steid03}; Magdis et al., in preparation). However,
278: \Fig{AGN_color} shows that the four objects have 8.0 \micron\
279: magnitudes more similar to that of the 8 \micron\ detected LBGs
280: \citep{huang05, rigop06}. There are a total of two x-ray detected
281: objects in the original sample of 162 LAEs (the other x-ray detected
282: object is not in the IRAC-uncontaminated sample), giving an AGN
283: fraction of $\sim 1\%$. The evidence presented here cannot rule out
284: the possibility that the other three red $R-m_{3.6}$ LAEs are dusty
285: LBG-like objects, but if they also turn out to be AGN, then the AGN
286: fraction is $\sim 5\%$.
287:
288: Since the four $R-m_{3.6} > 2$ LAEs have rather different properties
289: than the rest of the LAE population, they are excluded from the
290: conservative IRAC-detected sample. Two more objects are rejected
291: since they are detected in 3.6 \micron\ but not in $R$, which may
292: imply a large amount of dust reddening inconsistent with the rest of
293: the IRAC-detected population. The subsequent analysis will be focused
294: on the conservative IRAC-detected sample of 18 LAEs. However, we will
295: also present results derived using the larger sample of 23
296: IRAC-detected LAEs (the x-ray detected LAE AGN is still excluded).
297:
298:
299: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
300: \section{Stellar Population} \label{SP}
301: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
302: Since the majority of the LAEs in our sample are undetected in the
303: IRAC images, we perform a stacking analysis in order to measure the
304: average properties of the sample. Again, we work only with the sample
305: of 76 IRAC-uncontaminated LAEs. This is especially important when
306: stacking the IRAC-undetected subsample because any flux contributions
307: from bright neighbors could significantly skew the average. Because
308: of crowdedness in the IRAC images, it is also necessary to clean the
309: images before stacking by subtracting detected sources using a PSF
310: fitting routine \citep[StarFinder;][]{diola00}. The optical and
311: near-IR images are stacked in the same manner as the IRAC images,
312: except that the subtraction of detected sources is not necessary in
313: this case. Uncertainties in the stacked photometry are estimated by
314: bootstrapping the original sample, and then repeating the stacking and
315: flux measurement on the bootstrap samples. The uncertainties thus
316: derived take into account both random fluctuations and the intrinsic
317: luminosity dispersion among the LAE population.
318:
319: \begin{figure}
320: %\plotone{f3.eps}
321: \includegraphics[width=\columnwidth]{f3.eps}
322: \caption{ Stacked SEDs and best-fit models of the IRAC-detected
323: (squares, red line) and IRAC-undetected (diamonds, blue line)
324: samples. The models assume constant SFR and solar metallicity.
325: Only the bands with detections in the stacked images (${\rm S/N} >
326: 3$) are included in the plot.
327: \label{sed}
328: }
329: \end{figure}
330:
331: We fit \citet{bc03} stellar population synthesis models with a
332: Salpeter initial mass function to the stacked photometry. The $V$
333: band flux is corrected for the contribution of the Ly$\alpha$ emission
334: before the fits. The best-fit model to the IRAC-undetected sample is
335: shown together with the stacked SED in \Fig{sed}. Assuming a constant
336: SFR and solar metallicity, the IRAC-undetected sample is best fitted
337: by a model with $E(B-V) = \err{0.0}{0.0}{0.1}$, age =
338: \err{160}{110}{140} Myr, and mass = $\err{3}{2}{4}\times 10^8$ \Msun.
339:
340: The average mass of the IRAC-undetected LAEs is smaller than that of
341: typical LBGs by more than a factor of 30. The average SFR derived
342: from the best-fit model is 2 \Msun\ yr$^{-1}$, which is not very large
343: when compared to other populations of galaxies at $z \sim 3$ such as
344: the distant red galaxies (DRGs; \citealt{franx03}) which typically
345: have SFR $\ga 100$ \Msun\ yr$^{-1}$ \citep{forst04}. However, the
346: LAEs have a specific SFR (defined as SFR per unit mass) of \pow{7}{-9}
347: yr$^{-1}$, which is among the highest observed of galaxies at $z \sim
348: 3$. In comparison, LBGs have specific SFR $\sim \pow{3}{-9}$
349: yr$^{-1}$ \citep{shapl01} and DRGs have specific SFR $\sim
350: \pow{1}{-9}$ yr$^{-1}$ \citep{forst04}. This means that the stellar
351: populations in LAEs are increasing in mass at a much faster rate than
352: other populations of galaxies at similar redshifts, suggesting that
353: they are in the early stages of formation.
354:
355: The IRAC-detected sample appears to be significantly different from
356: the IRAC-undetected sample (\Fig{sed}). Again assuming a constant SFR
357: and solar metallicity, we find that the average stellar population of
358: the IRAC-detected LAEs is best fitted by a model with $E(B-V) =
359: \err{0.0}{0.0}{0.1}$, age = $1.6 \pm 0.4$ Gyr, and mass = $9 \pm 3
360: \times 10^9$ \Msun. Compared to the IRAC-undetected sample, the SFR
361: of the IRAC-detected LAEs is slightly higher at 6 \Msun\ yr$^{-1}$.
362: The specific SFR, however, is about $10 \times$ lower at \pow{7}{-10}
363: yr$^{-1}$, similar to LBGs with comparable stellar population ages
364: \citep{shapl01}. We emphasize that although the IRAC-detected and
365: IRAC-undetected samples have distinct properties on average, the
366: individual LAEs exhibit a continuum of properties, as evident in
367: \Fig{Rm36} \citep[see also][]{gronw07}. If we include the 5 red
368: $R-m_{3.6}$ objects originally excluded from our conservative
369: IRAC-detected sample (the x-ray detected LAE AGN is still excluded in
370: this analysis), the best-fit age and mass increase slightly to 2.1 Gyr
371: and \pow{1.8}{10} \Msun, respectively.
372:
373: The best-fit parameters presented above do not depend sensitively on
374: the assumed metallicity. Fitting the data with a 0.2 \Zsun\ model
375: increases the best-fit mass and the age of the IRAC-undetected sample
376: by 30\% and 60\%, respectively, while the effect on the IRAC-detected
377: sample is even smaller at less than 10\%. The best-fit parameters
378: also do not depend sensitively on the assumed star formation history.
379: Comparing constant and exponentially declining SFR models with the
380: same metallicity, we find that the two models in general give best-fit
381: masses and ages that are within $\sim 20\%$ of each other. This is
382: because the best-fit exponential SFR models have mostly long star
383: formation timescales ($\ga 1$ Gyr) and are therefore very similar to a
384: constant SFR model.
385:
386: At $z = 3.1$, AGB stars may contribute significantly to the flux in
387: the IRAC bands, which correspond to rest-frame 0.9 - 2 \micron. In
388: order to examine how the effects of AGB stars would change our
389: conclusions, we perform alternate fits using \citet{maras05} stellar
390: population synthesis models, which include larger contributions from
391: AGB stars than \citet{bc03} models. Fitting exponentially declining
392: SFR models with solar metallicity, we find for the IRAC-undetected
393: sample a best-fit model with age = 100 Myr, star formation timescale =
394: 250 Myr, $E(B-V) = 0$, and mass = \pow{3}{8} \Msun. For the
395: IRAC-detected sample, the best-fit model has age = 700 Myr, star
396: formation timescale = 2 Gyr, $E(B-V) = 0$, and mass = \pow{5}{9}
397: \Msun. The Maraston models in general give lower mass and age
398: estimates, but despite these differences, the distinction between the
399: IRAC-detected and IRAC-undetected populations remains significant.
400:
401:
402: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
403: \section{Discussion} \label{Disc}
404: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
405: There are several previous studies on the stellar population of LAEs
406: at $z \sim 3$ and beyond \citep{gawis06b, finke07, pirzk07, nilss07}.
407: These previous studies found that LAEs have masses ranging from $10^6
408: - 10^9$ \Msun, and ages ranging from $0.005 - 1$ Gyr. Our best-fit
409: average age of 160 Myr and mass of \pow{3}{8} \Msun\ for the
410: IRAC-undetected sample are broadly consistent with these previous
411: studies. In a further analysis of the IRAC-undetected sample
412: presented in this paper, \citet[submitted]{gawis07} fit two-component
413: models and find that the IRAC-undetected LAEs have a total mass of
414: \pow{1}{9} \Msun, with a young stellar component (age $\sim 20$ Myr)
415: accounting for $\sim 20\%$ of the total mass. These results are
416: consistent with ours which are derived using single-component models
417: and should be interpreted as the average properties of the entire
418: stellar population within the LAEs.
419:
420: \begin{figure}
421: %\plotone{f4.eps}
422: \includegraphics[width=\columnwidth]{f4.eps}
423: \caption{ A comparison of the $R-m_{3.6}$ colors and 3.6 \micron\
424: magnitudes between 8 \micron\ detected LBGs (green square), regular
425: LBGs (black dots), and the IRAC-detected LAEs (red stars). The
426: solid line shows the $R<25.5$ selection limit for LBGs. The data
427: for the regular LBGs come from IRAC observations of the
428: \citet{steid03} LBG sample (Magdis et al., in preparation), and the
429: 8 \micron\ LBG data come from combining the samples of Magdis et
430: al.\ and \citet{rigop06}. The mean colors and magnitudes of each
431: sample are plotted as large filled symbols, with error bars showing
432: the dispersion within the samples. We additionally plot the stacked
433: photometry of the IRAC-undetected sample, shown as the red filled
434: red star with the faintest 3.6 \micron\ magnitude.
435: \label{compLBG}
436: }
437: \end{figure}
438:
439: The IRAC-detected LAEs are more luminous than the IRAC-undetected LAEs
440: in both rest-frame UV and near-IR, and represent the massive end of
441: the LAE mass spectrum. Potentially, the IRAC-detected LAEs may
442: provide a link between the LAEs and other galaxy populations. For
443: instance, they may be the $z \sim 3$ analogues to the similarly
444: massive IRAC-detected LAEs found at $z \sim 5.7$ \citep{lai07}. In
445: \Fig{compLBG}, we compare the $R-m_{3.6}$ colors and 3.6 \micron\
446: magnitudes of the LAEs to 8 \micron\ detected LBGs and regular LBGs at
447: $z \sim 3$. The $R-m_{3.6}$ color and 3.6 \micron\ magnitude
448: correlate with stellar mass \citep{rigop06}, with the 8 \micron\ LBGs
449: being the most massive ($\ga 10^{11}$ \Msun), followed by LBGs ($\sim
450: 10^{10} - 10^{11}$ \Msun), IRAC-detected LAEs ($\sim 10^{10}$ \Msun),
451: and finally IRAC-undetected LAEs ($\sim 10^8$ \Msun). The
452: IRAC-detected LAEs occupy the faint and blue end of the LBG
453: color-magnitude distribution, suggesting that they may be the lower
454: mass extension of the LBG population. This interpretation is
455: supported by the fact that the inferred mass of the IRAC-detected LAEs
456: ($\sim 10^{10}$ \Msun) is at the low end of the mass range found for
457: LBGs ($\sim 10^{10} - 10^{11}$ \Msun; \citealt{papov01, shapl01}).
458: Furthermore, most LAEs in the present sample have (observer frame)
459: optical colors that would allow them to be selected as LBGs, although
460: the LAEs tend to be fainter in the rest-frame UV \citep{gronw07,
461: gawis06b}. It has also been observed that $\sim 20\% - 25\%$ of LBGs
462: at $z \sim 3$ exhibit Ly$\alpha$ emission strong enough to be narrow
463: band excess objects \citep{steid00, shapl03}. There is thus
464: accumulating evidence suggesting that the IRAC-detected LAEs may be
465: the bridge connecting the LAE and LBG populations.
466:
467: Using the stellar mass estimates derived in the previous section and
468: assuming a survey volume of \pow{1.1}{5} Mpc$^3$ \citep{gronw07}, we
469: find stellar mass densities of $0.3 \pm 0.3 \times 10^6$ and $3 \pm 1
470: \times 10^6$ \Msun\ Mpc$^{-3}$ for the IRAC-undetected and
471: IRAC-detected populations, respectively. The errors in the stellar
472: mass densities include uncertainties in the mass from the fits and
473: uncertainties in the number density coming from Poisson fluctuations
474: and a $\sim 20\%$ cosmic variance given the observed LAE bias of $\sim
475: 1.7$ \citep[submitted]{gawis07}. We should stress that the above
476: stellar mass densities only account for LAEs with Ly$\alpha$
477: luminosities above our survey completeness limit of \pow{1.3}{42} erg
478: s$^{-1}$. The IRAC-undetected LAEs account for about $9 \pm 10\%$ of
479: the total stellar mass in LAEs, even though they make up 2/3 of the
480: total by number. Compared to LBGs and DRGs at $z \sim 3$, which have
481: stellar mass densities of $\sim \pow{1}{7}$ and \pow{7}{6} \Msun\
482: Mpc$^{-3}$ respectively \citep{grazi07}, the stellar mass contained in
483: LAEs is smaller, but not insignificant. However, it is important to
484: keep in mind that there may be substantial overlap between the LAEs
485: and LBGs. A better understanding of the overlap between these two
486: populations is necessary before a direct comparison of the stellar
487: mass densities can be made.
488:
489: The present results show that the LAEs posses a wide range of masses
490: and ages, from the massive and evolved IRAC-detected LAEs to the young
491: and small IRAC-undetected LAEs. The range of photometric properties
492: shown in the LAE sample suggests that LAEs exhibit a continuum of
493: properties between these two extremes. The presence of both young and
494: evolved stellar populations within the overall LAE population implies
495: that the Ly$\alpha$ luminous phase of galaxies may last $\ga 1$ Gyr,
496: or that the Ly$\alpha$ luminous phase is recurring. Interestingly,
497: \citet{shapl01} found that LBGs with best-fit stellar population ages
498: $\ga 1$ Gyr also show strong Ly$\alpha$ emission in their spectra.
499: The authors suggest that vigorous past star formation has destroyed
500: and/or expelled the dust inside the galaxies, allowing the Ly$\alpha$
501: photons to escape. One characteristic of these evolved Ly$\alpha$
502: emitting LBGs is that they tend to have more quiescent SFR than the
503: rest of the population. The IRAC-detected LAEs, with their evolved
504: stellar population and low specific SFR, are consistent with this
505: scenario. If the LAE phase is recurring, then the young age of the
506: IRAC-undetected LAEs implies that their number and stellar mass
507: densities can be as much as a factor $\sim 10$ higher (very roughly
508: the ratio of the age of the universe to the stellar age). Similarly
509: young stellar populations ($\la 100$ Myr) have also been found at
510: higher redshifts \citep{finke07, pirzk07, verma07}. The stochastic
511: nature of galaxies with short lifetimes suggests that there may be a
512: related population of undetected pre or post-starburst galaxies that
513: may contribute significantly to the stellar mass and star formation
514: rate densities at $z \ga 3$.
515:
516: \acknowledgements
517:
518: We thank the referee for comments that helped improve the paper. KL
519: thanks Ryan Hickox for helpful discussions. This work is based in
520: part on archival data obtained with the Spitzer Space Telescope, which
521: is operated by the Jet Propulsion Laboratory, California Institute of
522: Technology under a contract with NASA. Support for this work was
523: provided by NASA through an award issued by JPL/Caltech.
524:
525:
526: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
527: \bibliographystyle{apj}
528: \bibliography{ms}
529: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
530:
531: \end{document}
532:
533: