1: %\documentclass[12pt,preprint]{aastex}
2: \documentclass{aastex}
3: %\usepackage{emulateapj5}
4: \usepackage{epsfig}
5: \usepackage{graphicx}
6: \usepackage{latexsym,multirow}
7: %\usepackage{slashbox}
8: \usepackage{latexsym,multirow}
9: \usepackage{slashbox}
10: \newcommand{\asec} {\mbox{$^{\prime \prime} $} }
11: \newcommand{\amin} {\mbox{$^{\prime}$}}
12: \newcommand{\etal}{{\it et al.}}
13: %\newcommand{\EWalpha}{EW$\alpha$}
14: \newcommand{\Halpha}{H$\alpha$}
15: \newcommand{\bq}{\begin{equation}}
16: \newcommand{\eq}{\end{equation}}
17: \newcommand{\msun}{\mbox{M$_{\odot}$}}
18: \newcommand{\hmsun}{\mbox{$h^{-1}$M$_{\odot}$}}
19: \newcommand{\zsun}{\mbox{Z$_{\odot}$}}
20: \newcommand{\lsun}{\mbox{L$_{\odot}$}}
21:
22: \shorttitle{Stellar Mass and Star Formation Rate}
23: \shortauthors{Mobasher {\it B. Mobasher et al}}
24:
25: \begin{document}
26:
27: \title{Relation Between Stellar Mass and Star Formation Activity in Galaxies$^1$}
28: \author{Bahram Mobasher\altaffilmark{2};
29: Tomas Dahlen \altaffilmark{3};
30: Andrew Hopkins \altaffilmark{4};
31: Nick Z. Scoville \altaffilmark{5};
32: Peter Capak \altaffilmark{5};
33: R. Michael Rich \altaffilmark{6};
34: David B. Sanders \altaffilmark{7};
35: Eva Schinnerer \altaffilmark{8};
36: Olivier Ilbert \altaffilmark{7};
37: Mara Salvato \altaffilmark{5};
38: Kartik Sheth\altaffilmark{9}
39: }
40:
41: \altaffiltext{1}{Based on observations with the NASA/ESA {\em
42: Hubble Space Telescope}, obtained at the Space Telescope Science
43: Institute, which is operated by AURA Inc, under NASA contract NAS
44: 5-26555; also based on data collected at : the Subaru Telescope, which is operat
45: ed by
46: the National Astronomical Observatory of Japan; the XMM-Newton, an ESA science m
47: ission with
48: instruments and contributions directly funded by ESA Member States and NASA; the European Southern Observatory under Large Program 175.A-0839, Chile; Kitt Peak
49: National Observatory, Cerro Tololo Inter-American
50: Observatory, and the National Optical Astronomy Observatory, which are
51: operated by the Association of Universities for Research in Astronomy, Inc.
52: (AURA) under cooperative agreement with the National Science Foundation;
53: the National Radio Astronomy Observatory which is a facility of the National Sci
54: ence
55: Foundation operated under cooperative agreement by Associated Universities, Inc
56: ;
57: and and the Canada-France-Hawaii Telescope with MegaPrime/MegaCam operated as a
58: joint project by the CFHT Corporation, CEA/DAPNIA, the National Research
59: Council of Canada, the Canadian Astronomy Data Centre, the Centre National
60: de la Recherche Scientifique de France, TERAPIX and the University of
61: Hawaii.}
62: %
63: \altaffiltext{2}{Department of Physics and Astronomy, University of California, Riverside, CA 92521 USA}
64: \altaffiltext{3}{Space Telescope Science Institute, 3700 San Martin
65: Drive, Baltimore, MD 21218 USA}
66: \altaffiltext{4} {School of Physics, University of Sydney, NSW 2006, Australia}
67: \altaffiltext{5}{California Institute of Technology, MC 105-24, 1200 East
68: California Boulevard, Pasadena, CA 91125 USA}
69: \altaffiltext{6}{Department of Physics and Astronomy, University of California, Los Angles, CA 90095 USA}
70: \altaffiltext{7}{Institute for Astronomy, 2680 Woodlawn Dr., University of
71: Hawaii, Honolulu, Hawaii, 96822 USA}
72: \altaffiltext{8}{Max Planck Institute for Astronomy, Konigstuhl 17,
73: 69117 Heidelberg, Germany}
74: \altaffiltext{9}{Spitzer Science Center, California Institute of Technology, Pasadena, CA 91125 USA}
75:
76: \begin{abstract}
77:
78: For a mass-selected sample
79: of 66544 galaxies with photometric redshifts ($z_{phot}$)
80: from the Cosmic Evolution Survey (COSMOS), we examine
81: the evolution of star formation activity as a function of stellar mass
82: in galaxies.
83: We estimate the cosmic star formation rates (SFR) over the range
84: $ 0.2 < z_{phot} < 1.2$, using the rest-frame 2800\,\AA \, flux
85: (corrected for
86: extinction). We find the mean SFR to be a strong function of the galactic
87: stellar mass
88: at any given redshift, with massive systems (log$(M/M_\odot) > 10.5$)
89: contributing
90: less (by a factor of $\sim$ 5) to the total star formation rate density
91: (SFRD).
92:
93: Combining data from the COSMOS and Gemini Deep Deep Survey (GDDS),
94: we extend the $SFRD-z$ relation as a function of stellar mass to $z
95: \sim 2$.
96: For massive galaxies, we find a steep increase in the $SFRD-z$ relation
97: to $z\sim 2$; for the less massive systems, the SFRD which also
98: increases from z $= 0$ to
99: 1, levels off at $z\sim 1$. This implies that the massive systems have had
100: their major star formation activity at earlier epochs ($z> 2$) than
101: the lower mass galaxies.
102:
103: We study changes in the SFRDs as a function of both redshift and
104: stellar mass for galaxies of different spectral types. We find that the
105: slope of the {\it SFRD-z} relation for different spectral type of galaxies
106: is a strong function of their stellar mass. For low and intermediate mass
107: systems, the main contribution to the cosmic SFRD comes from the star-forming
108: galaxies while, for more massive systems, the evolved galaxies are the
109: most dominant population.
110:
111:
112: \end {abstract}
113:
114: \keywords{galaxies: evolution --- galaxies: starburst: galaxies}
115:
116: \section {Introduction}
117:
118: Among the most outstanding issues in observational astronomy today is
119: understanding the physics of the star formation process and the parameters
120: responsible for it. The cosmic evolution of the star formation rate (SFR),
121: for example, provides clues towards the assembly of mass and the development
122: of
123: the Hubble sequence, while parameters which govern the star formation activity
124: are directly responsible for the evolution of galaxies. Despite significant
125: progress on this subject in recent years, there are still a number of
126: open questions concerning the star formation process in galaxies.
127:
128: The main problems in studying star formation
129: in galaxies can be divided into two general groups: first,
130: uncertainties due to dust extinction and variation in properties of different
131: star formation diagnostics \citep{Cra:98,Hop:01,Sul:01}; second, lack
132: of knowledge about the parameters which control star formation activity
133: in galaxies (Ellison et al 2007; Li et al 2007; Kaufmann et al 2007; Gomez et al 2005). There have been
134: major advances in resolving the problems in the first group by comparing
135: star formation rate (SFR) estimates from different diagnostics
136: \citep{Cra:98,Hop:03,Hop:04,dad:07}, investigating the cause
137: of their
138: difference \citep{Sul:01} and exploring the role of dust extinction
139: in SFR measurements \citep{Cal:07,Sul:01}. Due to severe observational
140: constraints, however, there has been little progress
141: in addressing problems in the second group.
142:
143: \citet{Cow:96} first studied dependence of the specific SFR (SFR per
144: unit mass) with redshift. They found the star forming galaxies to have
145: relatively brighter absolute K-band luminosities ($M_K$) at higher
146: redshifts ($z\sim 1$) compared to those at lower-$z$. As the $M_K$
147: is analogous to the integrated stellar mass of galaxies, they conclude
148: that higher mass systems to mainly contribute to the star formation rate
149: density (SFRD) at higher redshifts with the main contribution shifting
150: to lower mass systems at lower-$z$, a phenomenon they called ``downsizing".
151: This was confirmed in later studies
152: of nearby galaxies \citep{Bri:04,Hea:04} and
153: more distant systems \citep{Jun:05,Zhe:07,Scob:07}.
154: Observed changes in ``downsizing" as a function of
155: redshift reveals that, the most massive galaxies
156: form first, with galaxy formation proceeding later on to smaller
157: systems \citep{Jun:05}.
158:
159: The correlation between stellar mass and SFR in galaxies as a function
160: of redshift
161: provides insights into the integrated SFR over
162: the history of galaxies and hence, a picture of galaxy evolution over
163: cosmic time \citep{Pap:06,Dic:04}. Moreover, study of
164: these relations for different morphological types of galaxies elucidates
165: the origin of the Hubble sequence (Abraham et al 2007; Scarlata et al 2007).
166:
167: Here, we study
168: the relation between SFR with stellar mass and redshift, using a
169: homogeneous, complete and unbiased sample of galaxies covering a
170: wide range in all these parameters. We use the extensive multiwavelength
171: data from the Cosmic
172: Evolution Survey- COSMOS (Scoville et al 2007a, Mobasher et al 2007;
173: Capak et al 2007a).
174: The combination of the depth, area and wavelength coverage in COSMOS is
175: extremely beneficial for such investigations and complements
176: existing deeper but smaller area surveys
177: \citep{Jun:05,Gla:05,Gia:04}. Results from this study are extensively used
178: in Scoville et al (2007b) to explore the influence of local density enhancement
179: on star formation activity in galaxies.
180:
181: Throughout this paper we assume $H_0 =70$ km\,s$^{-1}$\,Mpc$^{-1}$,
182: $\Omega_\Lambda = 0.70$ and $\Omega_M =0.30$. Magnitudes are given
183: in the AB system.
184:
185: \section {Observations, Photometric Redshifts and Stellar Masses}
186:
187: We use the photometric data obtained from our imaging
188: surveys of the COSMOS field. This includes data in
189: U (CFHT), BV{\it riz} (Subaru) and K$_s$ (Kitt Peak/CTIO) bands,
190: covering the entire 2 sq.\ deg.\ area of the COSMOS field.
191: Details of the photometric observations and data reduction are
192: presented in \citet{Cap:07a} and Taniguchi et al (2007). For the
193: present study, we use a subset of the {\it i}-band selected catalog,
194: extended over the entire COSMOS field and complete
195: to $i < 25$ mag. To this magnitude limit, we derive accurate
196: photometric redshifts and spectral types (Mobasher et al 2007)
197: and can reliably estimate the SFRs. This covers a significantly larger area
198: with a depth similar to previous studies (see the compilation by
199: Hopkins and Beacom 2006).
200:
201: Photometric redshifts for galaxies in COSMOS are calculated using
202: the template fitting method incorporating priors, as described in
203: Mobasher et al. (2007).
204: For each object, we derive the photometric redshift, the redshift probability
205: distribution and the best-fitting spectral type. The spectral types cover
206: E, Sbc, Scd and Im templates (\citeauthor{Col:80} \citeyear{Col:80},
207: extended to the UV and NIR by \citeauthor{Bol:00} \citeyear{Bol:00})
208: in addition to two starburst templates \citep{Kin:96}.
209: We also consider extinction as an independent variable in the SED fits.
210:
211: It is possible that by considering extinction as a free parameter
212: in the fit, we increase the likelihood of degeneracy in the final
213: result (i.e. redshift, dust extinction and presence of the old population
214: all make the SEDs redder). This is examined in Mobasher et al (2007) by
215: studying the accuracy of photometric redshifts by comparing
216: the photometric ($z_{phot}$) and spectroscopic ($z_{spec}$) redshifts
217: for a sample of galaxies in the COSMOS with available such data,
218: expressed by the ratio:
219: $\Delta_z\equiv\langle|z_{\rm phot}-z_{\rm spec}|/(1+z_{\rm spec})\rangle$.
220: The smallest {\it rms} scatter between the estimated photometric
221: and available spectroscopic redshifts is obtained when
222: treating the extinction as a free parameter (see Table 4 and Figure 5 in
223: Mobasher et al 2007). We find $\Delta_z=0.03 $
224: with a small fraction ($\sim$2.5\%) of outliers, defined as $\Delta_z>0.3$.
225: Since Mobasher et al (2007) use a sub-sample of the COSMOS galaxies with
226: available spectroscopy to calibrate photometric redshifts, it is possible that
227: the photometric redshift sample is biased against fainter galaxies, for which
228: spectroscopic data do not exist in COSMOS. To minimize this,
229: we constrain the sample to galaxies with $z < 1.10$ as these have
230: the most reliable photometric redshifts and least
231: number of outliers (Mobasher et al 2007).
232:
233: We estimate the stellar mass corresponding to each galaxy, using the
234: photometric redshift, the rest-frame V-band absolute magnitudes ($M_V$)
235: and $M/L_V $ ratio dependent on the spectral type estimated
236: from the best fit SED
237: for each galaxy, as described in Mobasher et al (2007).
238:
239: \section{Star Formation Rates from UV Luminosities}
240:
241: The SFRs associated with individual galaxies are
242: estimated using rest-frame 2800\,\AA\ fluxes.
243: We estimate this for each galaxy, using
244: photometric data covering UBVr{\em izK} bands. This is done by
245: interpolating between the two bands that straddle rest-frame
246: 2800\,\AA\, wavelength, making use of the shape of the best-fit
247: SED when doing the interpolation. The exception is at low
248: redshifts ($0.20 < z < 0.34$), where
249: we use the observed U-band magnitude, extrapolated to 2800\,\AA\,
250: using the best-fit SED.
251: Having the observed apparent flux, K-correction and photometric redshifts
252: obtained from the best-fit SEDs, we then estimate the absolute flux in
253: rest-frame $\lambda=$2800\,\AA. A
254: detailed description of this technique is presented in \citet{Dah:07}.
255:
256: The UV luminosity at 2800\,\AA\ is mainly produced
257: by short-lived O and B stars and is therefore closely related to the
258: ongoing star formation activity.
259: Here, we follow the approach in \citet{Dah:07} to relate the UV flux
260: to the SFR. This involves using predictions from stellar synthesis models
261: and assuming parametric forms for the past star formation history to mimic
262: the evolution of the SFR with redshift, as summarized below. We
263: derive the conversion factor between the UV flux
264: and the ongoing SFR, using the stellar population synthesis code GALEXEV
265: \citep{BC:03}. We assume solar metallicity and a modified Salpeter IMF
266: spanning
267: $0.1 < M/M_\odot < 100$, but having a turnover at low mass \citep{BG:03}.
268: To model the dependence of SFR on redshift, we use a parametric fit to
269: the data from the Great Observatories Origins Deep Survey (GOODS),
270: presented in \citet{Gia:04} and parameterized in \citet{Str:04}.
271: Convolving the past SFR history with the results from the stellar
272: synthesis models, we derive a redshift dependent conversion factor,
273: $k_\lambda (z)$, between the ongoing SFR and UV flux as
274: \begin{equation}
275: L_{2800}=k_{2800} (z)\frac{{\rm SFR}}{M_{\odot}~{\rm yr}^{-1}}{\rm ergs}~{\rm
276: s}^{-1}
277: ~{\rm Hz}^{-1}
278: \end{equation}
279: The dependence of $k_{2800} (z)$~ on redshift is shown in Figure~\ref{fig1}.
280: Note that we use the same volume averaged past SFR history to calcluate the
281: conversion factor for all galaxies. Therefore, galaxies that have
282: SFR histories
283: significantly different from the average, may be assigned a SFR that differs
284: from the actual value. Furthermore, galaxies of different stellar
285: masses are characterized by different star formation histories (Panter et
286: al 2007), introducing a mass-dependent bias into the SFR estimates for
287: individual galaxies. We do not correct the SFR calibration here for this
288: effect, as this is highly model-dependent and introduces more free parameters
289: into the analysis. Moreover, any such correction would complicate the
290: interpretation of a mass-SFR-redshift relation discussed in the following
291: sections. We show the redshift
292: dependence of the conversion factor in Figure 1, illustrating only a
293: weak dependence with a change of only 2\% over the redshift
294: interval considered here. This suggests that the dependence on the assumed
295: SFR history is weak when deriving the SFR.
296: While this is an acknowledged source of uncertainty affecting calibration
297: of the SFR for any individual galaxy, it has no effect on the estimates
298: of the total SFR density (SFRD). This is because the SFRD is calculated by
299: integrating over UV luminosity functions estimated from the data, after
300: scaling the UV luminosity appropriately to SFR using the redshift dependent
301: conversion shown in Figure 1. We will return to the discussion of the
302: accuracy of the SFR calibration again in section 5.1, when we compare
303: our estimated SFRDs with those derived independently and compiled from
304: literature.
305:
306: Accurate estimate of dust extinction is essential in correcting the SFRs measured from rest-frame UV (2800 \AA) flux. Using the extinction values estimated for invidual galaxies through the SED fits, and assuming a Calzetti extinction law \citep{Cal:00}, we derive the reddening, $E_{B-V}$, for each galaxy. To examine the accuracy of the $E_{B-V}$ values in this study, we explore their distribution for different spectral types of galaxies. As expected, we find the median extinction ($E_{B-V}$) values to increase in galaxies from old and evolved to young and star-forming spectral types. Furthermore, we estimate the dust corrected SFRDs as presented in section 5.1, with different extinction corrections: the median $E_{B-V}$ values for each spectral type, individual extinction corrections for each galaxy and a constant $E_{B-V}$ value for all the galaxies. We find that the final results do not depend on the specific prescription for extinction correction.
307:
308: \section{Sample Selection and Completeness}
309:
310: The sample considered in the present study is selected
311: in {\em i}-band. This is the filter with the
312: deepest COSMOS photometry and provides a
313: sample most like a mass-selected survey (at least to $z\sim 0.5$).
314: Samples of star forming galaxies, selected based on emission lines
315: ([OII], H$\alpha$), broad-band (i.e. UV, U- band), far-infrared or radio
316: flux, can be very biased, since they select only a particular class of
317: galaxies. This is the main reason for selecting a sample
318: resembling a stellar mass-selected survey over much of its
319: redshift range. The final sample has:
320:
321: \begin {itemize}
322:
323: \item $i_{AB} < 25$: the majority of the objects fainter
324: than this limit escape detection at some wavelengths (especially in the
325: near-infrared), resulting in poor wavelength coverage of the SEDs and
326: thus, uncertain photometric redshifts, SFRs and stellar masses.
327:
328: \item $0.20 < z < 1.10$. The lower-limit for the photometric redshift
329: is adopted to minimize extrapolation to rest-frame 2800~\AA~
330: for low redshift galaxies, while the upper-limit is selected to avoid
331: objects with uncertain photometric redshifts (Mobasher et al 2007).
332:
333: \item $9.5 < \log(M/M_{\odot}) < 11.5$: this allows
334: a sample complete in terms of stellar mass. (Figure~\ref{fig2} shows the
335: distribution of stellar mass as a function of redshift). We find that galaxies with redder SEDs (i.e. those with spectral types resembling the early-type systems), are highly incomplete to $\log(M/M_{\odot}) \sim 9.5$ while this is the appropriate completness limit for galaxies with bluer SEDs (corresponding to later-type systems).
336:
337:
338: \item 0 $M_\odot$/yr $ < SFR_{2800} < 100\ ~ $M$_\odot$/yr: this will exclude
339: sources with excessive (and probably uncertain) SFRs, which could dominate
340: the SFR densities, and AGNs wrongly classified as star-forming
341: galaxies due to their extreme UV flux.
342: Figure~\ref{fig3} shows the SFR distribution as a function of redshift.
343: Very few galaxies are found to have $ SFR_{2800} > 100 \ M_\odot yr^{-1}$ and therefore, the results are not sensitive to the upper limit for the SFRs.
344:
345: \item $M_V < -19$: to allow the selection of galaxies from the same part of
346: the luminosity function at different redshifts and minimize luminosity (stellar
347: mass) dependent biases. This criterion excludes only an additional 61
348: galaxies on top of the above criteria, predominantly located in the
349: middle two redshift bins.
350:
351: \end{itemize}
352:
353: After applying the above selection criteria, we have a total
354: of 66544 galaxies in our sample. Following Scoville et al (2007b), we divide
355: the sample into the following redshift bins: $0.20 < z < 0.43$ (5594 galaxies);
356: $0.43 < z < 0.65$ (9608); $0.65 < z < 0.88$ (22374) and $0.88 < z < 1.10$
357: (28968). In terms of stellar mass, our combined sample
358: is complete to $M\sim 3\times 10^9$ M$_\odot$ over the redshift range
359: $0.20 < z < 1.1$.
360:
361: The fraction of the AGNs in the final sample is $< 1\%$ (a total of 1865
362: sources. M. Salvato -private communication), as identified by their X-ray
363: flux. We remove all the identified AGNs from the sample. There is a
364: possibility that the final sample is still contaminated by a small number
365: of AGNs. However, while massive galaxies
366: at $z\sim 2$ could host obscure AGNs (Daddi et al 2007), the fraction of
367: galaxies at $z < 1$ with strong nuclear activity is small.
368: It is also likely that we miss a population of highly
369: dusty star-forming galaxies with masses above our stellar mass
370: completeness limit. Our estimated global SFRs are therefore likely to be
371: lower limits, although our corrections for dust obscuration to the rest-frame UV flux and the technique used to estimate the total SFRDs are
372: expected to account for the majority of such missing systems.
373:
374: \section{Results}
375:
376: \subsection{Measurement of the Star Formation Densities}
377:
378: The aim of this study is to explore the effect of stellar mass on the
379: star formation activity in galaxies and to follow its behavior as a function
380: of look-back time. Previous such studies have either been too shallow
381: (i.e. 2dF and SDSS- Gomez et al 2005) or, covered only small volumes
382: with very few sources
383: \citep{Jun:05}, therefore, suffering from cosmic variance.
384: The present study, based on the COSMOS field, uses data in a wide (2 sq.\ deg)
385: area to medium depth, providing a highly homogeneous sample in the
386: range $ 0.20 < z < 1.10$, with well-known selection bias.
387:
388: We estimate the star formation rate densities (SFRD) in four redshift
389: intervals: $0.20 < z < 0.43$; $0.43 < z < 0.65$; $0.65 < z < 0.88$ and
390: $0.88 < z < 1.10$. In each redshift interval the SFRD
391: is estimated in two ways: (1) by summing over the extinction corrected SFRs
392: for galaxies above the stellar mass completeness limit in each
393: redshift interval and normalizing them to the volume corresponding to
394: that redshift slice; (2) by integrating the rest-frame UV luminosity function
395: of galaxies (corrected for extinction) in the above redshift intervals,
396: measuring the UV luminosity densities
397: and converting them to SFRDs. The SFRDs from the two methods are listed
398: in Table 1, with the evolution of the SFRD
399: with redshift, derived from the COSMOS galaxies, presented
400: in Figure~\ref{fig4}. This is
401: compared with the SFRDs estimated independently from other studies, as
402: compiled in \citep{HB:06}. All the data in Figure 4 are scaled
403: to a modified Salpeter IMF
404: described by \citet{BG:03}.
405: Errorbars are measured
406: assuming Poisson statistics. The SFRDs per redshift bin estimated from
407: the sum of the contributions from individual galaxies is lower
408: than those derived from the rest-frame UV LF (Table 1 and Figure 4).
409: This is due to the absence
410: of sources fainter than the flux limit of the survey from this method.
411: The agreement between the SFRDs estimated here using the extinction corrected
412: rest-frame UV LF, and those in the literature
413: over the redshift range $0.20 < z < 1.10$ is very good,
414: given that the SFRDs are measured from different diagnostics, different
415: surveys and are based on different prescriptions for extinction correction.
416: The agreement here lends support to our estimated values for
417: redshift, SFR and extinction. Because of their selection wavelengths, however,
418: all these surveys are biased against very dusty, extremely star-forming
419: galaxies \citep{Afo:03}, implying the estimated SFRDs in Figure~\ref{fig4}
420: are likely lower limits. We further discuss this and its implications in section 6.
421:
422: \subsection{Relation between the SFRDs and Stellar Mass}
423:
424: The main result of this paper is presented in Figure~\ref{fig5}, where we
425: find changes in the SFRDs as a function of redshift and stellar
426: mass. We divide the sample into three mass intervals:
427: $9.5 < \log(M/M_{\odot}) < 10.0$, $10.0 < \log(M/M_{\odot}) < 10.5$ and
428: $10.5 < \log(M/M_{\odot}) < 11.5$ in order to have a sufficient number of
429: galaxies in each stellar mass interval. The points corresponding to
430: the lower mass bins are likely to be lower limits due to incompletness.
431: Errorbars correspond to Poisson statistics, and are often smaller than
432: the symbol size. We find that for galaxies in all
433: mass intervals, there is a clear increase in the SFRDs with redshift, changing
434: by a factor of 2 to 4 over the redshift range considered here. Moreover,
435: at any given redshift, more massive galaxies ($\log(M/M_{\odot}) > 10.5$)
436: contribute less (by a factor of 5) to the total SFRD. This is because
437: they have already gone through their intensive star formation phase
438: and have built up their mass. Results
439: from Figure~\ref{fig5} also imply that, while massive galaxies
440: continue to undergo modest
441: star formation activity to $z\sim 1$, they have
442: acquired most of their mass before this redshift (i.e. $z>1$).
443: The results here are consistent with the ``downsizing"
444: picture of galaxy formation, where the SFR changes from high-mass
445: to low-mass systems (\cite{Cow:96,Jun:05}).
446:
447: The mass-dependent evolution of the SFRDs show progressively
448: smaller contributions to the total SFRD (including all masses)
449: as the SFRD declines to low-$z$. The fraction of the total
450: SFRD, however, is more or less constant over the redshift range
451: probed for each mass bin (i.e. at any given mass bin, the ratio of
452: the SFRD in that bin to the total SFRD does not change over the
453: redshift range considered here). This result differs significantly
454: from that found by Seymour et al (2008) for the evolution
455: in the high-luminosity end of the luminosity function,
456: where very strong mass-dependent evolution is seen for
457: this fractional contribution to the total SFRD. This result
458: is likely to be due to the different region of the luminosity
459: function sampled by different selection effects arising
460: at different wavelengths. Furthermore, the radio selected sample used in
461: Seymour et al (2008) is mainly
462: dominated by galaxies from the high-end of the luminosity function while
463: the optically selected sample here is more typically dominated
464: by $L^*$ galaxies.
465:
466: Using a sample of galaxies detected at 24 $\mu$m wavelength, Zheng et al (2007)
467: studied the evolution of the SFRD with redshift as a function of stellar mass.
468: Their sample covers the same redshift range as the present study and uses a
469: ``dust-free" measure of the SFR by combining the UV and total infrared flux
470: (8-1000 $\mu$m) for individual galaxies.
471: Therefore, by comparing our results with those in Zheng et al (2007),
472: one could quantify possible dust-induced biases in Figure 5, despite the very
473: different selection criteria used for these samples. The results from
474: the two studies are compared in Figure 6, where they are divided into the same
475: mass intervals. Over the redshift range covered by these studies, the rate
476: of change of the SFRDs with redshift (i.e. the slope of the SFRD-$z$ relations)
477: is the same, independent of the stellar mass. For the intermediate mass
478: galaxies with
479: $10^{10.19} < M/M_\odot < 10^{11.19}$, there is excellent agreement
480: between the the two studies. However, for the lowest mass bin at
481: $10^{9.5} < M/M_\odot < 10^{10.19}$, the SFRD estimated here is significantly
482: higher (by a factor of 5) at all redshifts. This discrepancy is not caused by
483: dust extinction in our estimate of the SFRD compared to Zheng et al (2007),
484: as this would have led to a relatively smaller SFRD
485: (with respect to Zheng et al) for our sample.
486: This is likely a consequence of both incompleteness and selection
487: effects in the lowest mass bin of Zheng et al (2007). Their method of
488: infrared (24\,$\mu$m) stacking, based on the optically selected sample
489: of COMBO-17 sources in this bin, is biased against optically faint
490: low-mass galaxies. These are exactly those systems that are likely (by
491: virtue of dust obscuration) to be, on average, brighter in the infrared
492: wavelengths than that measured by the stacking result from the optically
493: detected sources.
494: This causes a bias, leading to an underestimation in the inferred SFRD
495: for their lowest mass bin. Moreover, the inclusion of low-mass red
496: galaxies in the stacking analysis (as done by Zheng et al 2007) would
497: serve to reduce the infered average SFR, adding another bias towards
498: underestimating the true SFRD. A detailed investigation of the extent of
499: the bias in the low mass bin of Zheng et al (2007) study would be
500: facilitated by discriminating between blue (star-forming) and red
501: (non-star-forming) populations. Given the above discussion, it is
502: likely that their result for the lowest mass bin is an underestimate.
503: The disagreement seen in the lowest mass bin in Figure 6 is therefore
504: not surprising. To summarize, while for the intermediate mass galaxies
505: the SFRDs at different redshifts are in excellent agreement, the observed
506: discrepancy for the lowest mass galaxies is likely caused by incompletness
507: and selection bias in the Zheng et al sample. Correcting for these effects
508: brings the Zheng et al's result into better agreement with that in the present
509: study. For the highest mass bin ($M > 10^{11.15}$ M$_\odot$), there is
510: serious incompletness in our sample, which explains the relatively lower
511: SFRDs measured for galaxies with higher stellar mass from the present study.
512:
513: We find the number density evolution in the present sample to
514: depend on the stellar mass of galaxies.
515: The impact of this number density evolution on the SFRD can be
516: interpreted by considering the ratio of the SFRD
517: ($\dot{\rho}_*$) to the number density
518: of star forming galaxies ($\rho_N$), $\dot{\rho}_* / \rho_N$, defined as
519: the ``characteristic star formation
520: rate" (cSFR), as a function of mass and redshift.
521: The cSFR remains
522: flat for our low-mass galaxies, at a level of $\approx
523: 1\,M_{\odot}$\,yr$^{-1}$, while dropping by more than an order of
524: magnitude for high mass systems from $\approx 2\,M_{\odot}$\,yr$^{-1}$
525: at $z\approx 1$ to $\approx 0.1\,M_{\odot}$\,yr$^{-1}$ at
526: $z\approx 0.3$.
527: Intermediate masses show an intermediate level
528: of evolution, with their cSFRs comparable to the high-mass objects
529: at high-$z$, and to low-mass objects at low-$z$. This behavior exhibits
530: the characteristics of ``downsizing" in galaxy evolution in the sense that
531: the cSFR decreases with redshift faster for massive galaxies
532: than for low-mass systems.
533:
534: Using spectroscopic observations of galaxies over the range $1 < z < 2$,
535: selected from Gemini Deep Deep Survey (GDDS), \citet{Jun:05} studied
536: changes in the SFRDs with redshift as a function of stellar mass of
537: galaxies. This provides
538: a natural extension of the COSMOS study to $z\sim 2$. To study the behavior of
539: this relation over the redshift range covered by the combined COSMOS and GDDS,
540: we divide our sample into the same mass bins as those in \citet{Jun:05} and
541: compare the results in Figure~\ref{fig7}. For the most massive galaxies
542: we find the trend in SFRD with redshift to continue to
543: $z\sim 2$, however for the less massive systems it flattens around
544: $z\sim 1.1$. This implies that downsizing has been effective at $z > 1$,
545: and that the massive galaxies were formed before $z\sim 2$ after going
546: through a period of intensive star formation activity. Results
547: in Figures 5 and 7
548: confirm that, at any given redshift out to $z\sim 2$, massive galaxies
549: contribute less to the total SFRDs than objects with smaller stellar mass.
550: However, although the sample here is selected to be close to
551: Juneau et al's galaxies (in terms of the range in their stellar mass),
552: differences in the selection criteria between the two samples is likely to
553: affect the above result.
554:
555: \subsection{Evolution of the SFRD as a Function of Stellar Mass and
556: Galaxy Type}
557:
558: The coverage of the COSMOS field by the Advanced Camera for Surveys (ACS),
559: on-board the Hubble Space Telescope (HST), provides high resolution imaging
560: of galaxies and measurement of their morphologies. Since the
561: coverage is in only one HST-ACS band, it is not possible
562: to measure morphologies for galaxies at the same rest-frame wavelength in
563: all the 4 redshift bins studied here. We therefore use the spectral types
564: as a proxy to rest-frame ACS morphologies. We have already shown that our
565: estimated spectral types closely agree with the rest-frame morphologies
566: in COSMOS (Capak et al 2007b). Nevertheless, one needs to be cautious that the spectral types and morphologies of galaxies are only loosely related. In the following discussion, these terms are used synonymously. Moreover, relations involving photometric redshifts, spectral types and stellar masses of galaxies should be interpreted carefully as these parameters are not independently derived here.
567:
568: In Figure~\ref{fig8}, we present the evolution of the SFRD with redshift as
569: a function of both stellar mass and spectral type of galaxies.
570: The spectral types are divided into those with redder SEDs (early types),
571: intermediare color SEDs (spirals) and bluer SEDs (late-types and starbursts)-
572: (Mobasher et al 2007).
573: As with any type classification, there will be some level of misclassification
574: or systems that do not fall easily into the defined spectral types.
575: For the following discussion this limitation must be kept in mind.
576: In particular, there is some ambiguity regarding irregular galaxies,
577: which may in some instances fall either in the ``spiral" or ``starburst"
578: classes as all these are represented by SEDs similar to those of star-forming
579: galaxies. Nevertheless,
580: we find a clear trend for all the spectral types in Figure 8. Due to serious incompleteness in galaxies with redder SEDs at $\log(M/M_\odot)\sim 9.5$, the relation for these galaxies is not shown in Figure 8.
581: For the low
582: and intermediate mass galaxies,
583: starbursts dominate the star formation activity at high redshifts with
584: a comparable contribution from spirals at lower redshifts.
585: Also, for the lowest mass systems ($ 9.5 < log(M/M_\odot) < 10$),
586: we find a steep increase with redshift in the SFRD for starbursts, compared
587: to that for the spirals which remain unchanged with redshift. However,
588: for the intermediate mass systems ($ 10 < log(M/M_\odot) < 10.5$), the
589: slope of the SFRD-$z$ relation increases for both spirals and starbursts,
590: with the spirals making a relatively higher contribution to the total
591: SFRD at lower redshifts.
592: For higher mass galaxies, the sample is dominated by early-types, with
593: the spirals showing a relatively steeper change in their
594: SFRD with redshift compared to earlier types.
595: This is consistent with the expectation from
596: the downsizing scenario that the most massive systems are dominated
597: by early-type old galaxies, which have undergone intense star formation activity
598: at higher redshifts ($z\approx 1$). An interesting feature in Figure 7
599: is that the slope of the SFRD-$z$
600: relation for different spectral types of galaxies are strong functions
601: of their stellar mass.
602:
603: Perhaps the most revealing population in this diagram are the starbursts.
604: These are a dominant form of star formation in galaxies at $z\gtrsim 1$,
605: but their prevalance at lower redshifts declines progressively with
606: decreasing redshift. This is reflected in the sharp decrease from
607: $z\sim1$ to $z\sim 0.20$ of the
608: starburst contribution to the SFRD with redshift for low and intermediate
609: mass systems. The fact that most definitions of a starburst favor
610: low-mass star forming galaxies can be seen in
611: the relatively large contribution of this population in the lowest-mass
612: bin, and progressively less towards higher masses. In the high-mass bin,
613: there are too few starbursts observed to reliably fit luminosity functions,
614: and the lower limits shown reflect the sum over the detected population only.
615: Overall, starbursts account for the majority of the star formation in low- to
616: mid-mass systems at higher redshifts ($z\gtrsim 0.7$), but at lower redshifts
617: the dominant population shifts to the spiral types in these mass ranges.
618: The lower the galaxy mass, the lower the redshift to which the starbursts
619: remain dominant, consistent with the low-mass irregular starburst galaxies
620: seen in the local universe (Figure 3 in Kennicutt 1998). This is a
621: consequence of the change of slope of SFRD-$z$ relations found in
622: different stellar mass intervals.
623:
624: The spiral types are the dominant contributors at low redshift and low and
625: intermediate mass, although they make a contribution similar to the
626: early types
627: in the highest mass bin. The evolution of the SFRD for spirals is flatter than
628: any other classes, which is due to a transformation
629: from starburst to spiral spectral types with decreasing redshift. We should
630: note that, particularly for early types, the lowest mass systems
631: are expected to form their stars very late in most hierarchical models.
632:
633: \section {Discussion}
634:
635: A number of studies in recent years have explored changes in SFRD
636: with redshift as a function of the stellar mass of galaxies
637: \citep{Gla:05,Jun:05,Zhe:07}.
638: Here we investigate this relation using a large and homogeneous sample
639: of galaxies with well-known selection biases, accurate photometric redshifts,
640: SFR estimates and stellar mass measurements.
641:
642: We find a strong dependence of the SFRD on stellar mass in the range
643: $0.20 < z < 1.10$, with higher mass
644: galaxies ($M > 10^{10.5}$ M$_\odot$) contributing less (by an order of
645: magnitude) to the global SFRD than lower mass systems at any given redshift.
646: However, the rate of evolution of the SFRD with redshift is the same,
647: regardless of the stellar mass of galaxies, in agreement with the
648: independent study by Zheng et al (2007), in which the effect of
649: dust obscuration in selection of star forming galaxies and measurement
650: of their SFRs is taken into account.
651:
652: Using the GDDS with spectroscopic redshifts in the range $1 < z < 2$,
653: \citet{Jun:05} studied this relation from $z\sim 1$ to $z\sim 2$. The combined
654: COSMOS and GDDS data show that for high mass galaxies
655: ($11.06 < log (M/M_\odot) < 11.76 $), the upward trend in the SFRD continues
656: to $z\sim 2$, while for
657: lower mass systems, after a steep increase with redshift, the SFRD flattens
658: at $z > 1$. This implies that massive galaxies produced most of their stars
659: by $z\sim 2$ (when the Universe was $\sim 3.5\,$Gyr old), with the
660: less massive
661: galaxies hosting efficient star formation activity only after $z\sim1$.
662: These results confirm that the stellar mass of galaxies regulates the relative
663: contribution of a galaxy to the global SFRD, consistent with the ``downsizing''
664: scenario for the formation of galaxies. Moreover, we find that
665: galaxy ``downsizing'' was already occuring
666: at $z\sim 2$ and has continued to the present epoch.
667:
668: The estimated SFRD in this study is likely to be a lower limit due to
669: selection at optical bands biasing against dusty star-forming galaxies,
670: which could significantly contribute to the global SFRD.
671: Using a near-IR selected sample with $K_s < 21.5$,
672: mimicking a mass selected sample complete to $M \sim 10^{10}$ M$_\odot$,
673: Caputi et al (2006) estimated the space density of Luminous Infrared
674: Galaxies (LIRGs) and Ultra-Luminous Infrared Galaxies (ULIRGs) in the range
675: $0.5 < z < 1$, similar to the redshift range covered in the present study.
676: They found that 24\% of galaxies with
677: $10^{11} $M$_\odot$ $< M < 2.5\times 10^{11}$ M$_\odot$
678: are either LIRGs or ULIRGs (excluding AGNs), with a negligible fraction
679: of LIRGs and ULIRGs to have masses in excess of
680: $2.5 \times 10^{11}$ M$_\odot$. Furthermore,
681: using the MIPS observations in
682: the GOODS-S field, Caputi et al (2005) show a very small fraction of LIRGs or
683: ULIRGs at $0.4 < z < 1$ to have $M < 5\times 10^{10}$ M$_\odot$. Indeed,
684: ULIRGs of any stellar mass are found to be very rare below $z \sim 0.5$
685: \citep{Flo:99,Capu:06,LeF:05}.
686: The conclusion from these studies is that, the fraction of massive ($M > 10^{11}$ M$_\odot$) LIRGs and ULIRGs at $z < 1$ is very small, implying that these
687: objects do not significantly contribute to the total SFRD over this redshift
688: range. Since the majority of the LIRGs and ULIRGs at $0.5 < z < 1$
689: are associated with intermediate mass systems, the absence of these galaxies
690: from our sample is likely to lead to an underestimation of the contribution
691: from these objects (i.e. intermediate mass systems) to the global SFRD.
692: Moreover, \citep{LeF:05} found that at $z > 0.6$, dusty starbursts with $SFR > 10$ M$_\odot$ yr$^{-1}$ dominate the cosmic star formation density, with their contribution increasing from 50\% at $z\sim0$ to 80\% at $z\sim1$ (i.e. only 20\% of the SFRD at $z\sim1$ is due to the UV bright galaxies).
693: Any selection bias against these galaxies may possibly compromise
694: our estimate of the SFRD in the range $0.6 < z < 1.2$. We believe that
695: a possible bias against such galaxies does not affect our
696: results here. By estimating the SFRDs through integrating the rest-frame
697: UV(2800 \AA) luminosity functions (corrected for dust
698: obscuration) over their entire magnitude range, we incorporate a
699: contribution from galaxies at the faint-end of the UV luminosity function,
700: where heavily obscured starbursts would lie.
701: The excellent agreement between our estimated SFRDs from COSMOS and those
702: based on the sample compiled by Hopkins and Beacom (2006) in Figure 4
703: supports the fact that the bias against dusty starburst galaxies does
704: not significantly affect the results in the present study.
705: We note that Hopkins and Beacom (2006) estimate the total contribution to
706: the SFRD from their compilation by combining the SFRD inferred from the
707: UV and far-infrared luminosities. Nevertheless, the limitations imposed
708: by selection criteria and any biases so introduced remains a major source
709: of uncertainty in any such study.
710:
711: Using a sample of 24 $\mu$m selected galaxies in GOODS-S, Daddi et al (2007)
712: found a roughly linear relation between the stellar mass and SFR densities.
713: For a given mass, the SFRD at $z\sim 2$ was larger by factors of
714: $\sim 4$ and $\sim 30$ relative to the star formaing galaxies
715: at $z\sim 1$ and $z\sim 0$ respectively. This is roughly consistent
716: with our results in the two highest mass bins in Figure 6. In the range
717: $1 < z < 2$, the observed 24 $\mu$m wavelength corresponds to rest-frame
718: $8-12\ \mu$m bands, sampling the PAH features which are most sensitive
719: to the star forming galaxies. The agreement between or results and
720: Daddi et al (2007) further implies that the potential selection bias against
721: dusty star forming galaxies in our sample is not significantly affecting
722: the estimated SFR densities.
723:
724: By dividing our sample into three spectral types; ellipticals, spirals and
725: starbursts, we study the evolution (with redshift) of the SFRD-Mass
726: relation for each type. A striking feature, presented in Figure 7, is
727: that the slope of the SFRD$-z$ relations, found
728: for different spectral types of galaxies are strong functions of the
729: stellar mass. We find the massive galaxies, which are
730: dominated by early-types, to significantly contribute to the total SFRD
731: at earlier epochs ($z\sim 1$). This is expected as more massive galaxies
732: are able to support higher rates of star formation than lower mass systems,
733: all else being equal.
734: For the intermediate mass systems,
735: the contribution from early-type galaxies to the global SFRD is minimal, with
736: no significant change with redshift. The relation for lower mass systems,
737: which are dominated by spirals and starbursts, show a progressive build up
738: of the stellar mass (i.e. increasing SFRD) with cosmic time.
739: The results
740: here imply that the stellar mass of galaxies plays a crucial role in
741: determining their star formation activity and spectral types and governs
742: the contribution of individual galaxies to the total SFRD.
743:
744: \section {Summary and Conclusion}
745:
746: Using multiwavelength data from the COSMOS survey we studied the dependence
747: of the cosmic star formation density on redshift, stellar mass and
748: spectral types. We explored the main parameters which govern the star formation
749: process in galaxies. Our main results are summarized as follows:
750:
751: \begin {itemize}
752:
753: \item There is a strong dependence of the SFRD on stellar mass of galaxies,
754: with the most massive syatems (log$(M/M_\odot) > 10.5$)
755: contributing least (by a factor of $\sim$5) to the cosmic SFRD
756: at any given redshift.
757:
758: \item Combining data from the COSMOS and GDDS, we extend the $SFRD-z-$mass
759: relation to $z\sim 2$. For high mass galaxies, we find a steep increase
760: in this relation till $z\sim 2$. This implies that the massive galaxies
761: seen today,
762: went through intensive star formation activity at $z>1$, to generate their
763: current stellar mass. For the less massive systems, the $SFRD-z$
764: relation flattens at $z\sim 1$, indicating that these systems are currently
765: undergoing build up of their mass, with most of their stellar mass
766: formed at lower redshifts ($z < 1$).
767:
768: \item We study dependence of the SFRD$-z-$mass relation on the spectral
769: type of galaxies. We find that the slope of the SFRD$-z$ relation
770: for different spectral types is a strong function of
771: their stellar mass. For low and intermediate mass systems, the main
772: contribution to the cosmic SFR comes from the star-forming
773: galaxies while, for more massive galaxies, older and redder galaxies
774: are the most
775: dominant population, with their contribution to the global SFRD
776: increasing with redshift.
777:
778: \end{itemize}
779:
780: \begin{thebibliography}{}
781: \bibitem[Abraham et al.(2007)]{Abr:07}
782: Abraham, R., et al. 2007, in ``Galaxy Evolution across the Hubble Time,"
783: Proceedings IAU Symposium No. 235, eds F. Combes \& J. Palou
784: \bibitem[Afonso et al.(2003)]{Afo:03}
785: Afonso, J., Hopkins, A., Mobasher, B., Almeida, C. 2003, \apj, 597, 269
786: \bibitem[Baldry \& Glazebrook(2003)]{BG:03}
787: Baldry, I. K., Glazebrook, K. 2003, \apj, 593, 258
788: \bibitem[Bell \& de Jong(2001)]{BdJ:01}
789: Bell, E. F., de Jong, R. S. 2001, \apj, 550, 212
790: \bibitem[Bolzonella et al.(2000)]{Bol:00}
791: Bolzonella, M., Miralles, J.-M., Pell{\'o}, R. 2000, \aap, 363, 476
792: \bibitem[Brinchmann et al.(2004)]{Bri:04}
793: Brinchmann, J., Charlot, S., White, S. D. M., Tremonti, C.,
794: Kauffmann, G., Heckman, T., Brinkmann, J. 2004, \mnras, 351, 1151
795: \bibitem[Bruzual \& Charlot(2003)]{BC:03}
796: Bruzual, G., Charlot, S. 2003, \mnras, 344, 1000
797: \bibitem[Calzetti et al.(2000)]{Cal:00}
798: Calzetti, D., Armus, L., Bohlin, R. C., Kinney, A. L., Koornneef, J.,
799: Storchi-Bergmann, T. 2000, \apj, 533, 682
800: \bibitem[Calzetti et al.(2007)]{Cal:07}
801: Calzetti, D. 2007, in ``Pathways through an Eclectic Universe", ASP
802: Conference Series, 2007, eds. J.H. Knapen, T.J. Mahoney, and A. Vazdekis
803: (arXiv:0707.0467)
804: \bibitem[Capak et al.(2007a)]{Cap:07a}
805: Capak, P., et al. 2007, \apjs, 172, 99
806: \bibitem[Capak et al.(2007b)]{Cap:07b}
807: Capak, P., et al. 2007, \apjs, 172, 284
808: \bibitem[Caputi et al (2006)]{Capu:06}
809: Caputi, K. I.; McLure, R. J.; Dunlop, J. S.; Cirasuolo, M.; Schael, A. M. 2006, Mon. Not. R. Astr. Soc. 366, 609
810: \bibitem[Coleman et al.(1980)]{Col:80}
811: Coleman, G. D., Wu, C.-C., Weedman, D. W. 1980, \apjs, 43, 393
812: \bibitem[Cowie et al.(1996)]{Cow:96}
813: Cowie, L. L., Songaila, A., Hu, E. M., Cohen, J. G. 1996, \aj, 112, 839
814: \bibitem[Cram et al.(1998)]{Cra:98}
815: Cram, L., Hopkins, A., Mobasher, B., \& Rowan-Robinson, M.
816: 1998, \apj, 507, 155
817: \bibitem[Daddi et al (2007)]{dad:07}
818: Daddi, E. et al astro-ph 07052831
819: \bibitem[Dahlen et al.(2007)]{Dah:07}
820: Dahlen, T., Mobasher, B., Dickinson, M., Ferguson, H. C.,
821: Giavalisco, M., Kretchmer, C., Ravindranath, S. 2007, \apj, 654, 172
822: \bibitem[Dickinson et al.(2004)]{Dic:04}
823: Dickinson, M., et al. 2004, \apj, 600, L99
824: \bibitem[Ellison et al (2007)]{ell:07}
825: Ellison et al astro-ph 07114833
826: \bibitem[Flores et al (1999)]{Flo:99}
827: Flores, H., et al. 1999, \apj, 517, 148
828: \bibitem[Giavalisco et al.(2004)]{Gia:04}
829: Giavalisco, M., et al. 2004, \apjl, 600, L103
830: \bibitem[Glazebrook et al.(2005)]{Gla:05}
831: Glazebrook, K., et al. 2005, in ``Maps of the Cosmos," Proceedings
832: of IAU Symposium No. 216, eds. M. Colless, L. Staveley-Smith, R. Stathakis
833: \bibitem[G{\'o}mez et al.(2003)]{Gom:03}
834: G{\'o}mez, P. L., et al. 2003, \apj, 584, 210
835: \bibitem[Heavens et al.(2004)]{Hea:04}
836: Heavens, A., Panter, B., Jimenez, R., Dunlop, J. 2004, Nature, 428, 625
837: \bibitem[Hopkins(2004)]{Hop:04}
838: Hopkins, A. M. 2004, \apj, 615, 209
839: \bibitem[Hopkins \& Beacom(2006)]{HB:06}
840: Hopkins, A. M., \& Beacom, J. F. 2006, \apj, 651, 142
841: \bibitem[Hopkins et al.(2003)]{Hop:03}
842: Hopkins, A. M., et al. 2003, \apj, 599, 971
843: \bibitem[Hopkins et al.(2001)]{Hop:01}
844: Hopkins, A. M., Connolly, A. J., Haarsma, D. B., Cram, L. E.
845: 2001, \aj, 122, 288
846: \bibitem[Juneau et al.(2005)]{Jun:05}
847: Juneau, S., et al. 2005, \apjl, 619, L135
848: \bibitem[Kaufmann et al.(2007)]{kauf:07}
849: Kaufmann, T., Wheeler, C., Bullock, J. S. astro-ph 7060210
850: \bibitem[Kinney et al.(1996)]{Kin:96}
851: Kinney, A. L., Calzetti, D., Bohlin, R. C., McQuade, K.,
852: Storchi-Bergmann, T., Schmitt, H. R. 1996, \apj, 467, 38
853: \bibitem[Le Floc'h et al.(2005)]{LeF:05}
854: Le Floc'h, E., et al. 2005, \apj, 632, 169
855: \bibitem[Lewis et al.(2002)]{Lew:02}
856: Lewis, I., et al. 2002, \mnras, 334, 673
857: \bibitem[Li et al.(2007)]{Li:07}
858: Li, C., Kauffmann, G., Heckman, T., Jing, Y. P.,
859: White, S. D. M. astro-ph 07113792
860: \bibitem[Madau et al.(1996)]{Mad:96}
861: Madau, P., Ferguson, H. C., Dickinson, M. E., Giavalisco, M.,
862: Steidel, C. C., Fruchter, A. 1996, \mnras, 283, 1388
863: \bibitem[Mobasher et al. (2007)]{mob:07}
864: Mobasher et al. 2007 ApJ Suppl. 172, 117
865: \bibitem[Panter et al. (2007)]{pan:07}
866: Panter, B., Jimenez, R., Heavens, A. \& Charlot, S. 2007 Mon. Not. R. Astr. Soc 378, 1550
867: \bibitem[Papovich et al.(2006)]{Pap:06}
868: Papovich, C., et al. 2006, \apj, 640, 92
869: \bibitem[Salpeter(1955)]{Sal:55}
870: Salpeter, E. E. 1955, \apj, 121, 161
871: \bibitem[Scoville et al. (2007a)]{Scoa:07}
872: Scoville, N. Z et al 2007a ApJ Suppl. 172, 1
873: \bibitem[Scoville et al. (2007b)]{Scob:07}
874: Scoville, N. Z et al 2007b ApJ Suppl. 172, 150
875: \bibitem[Seymour et al.(2008)]{Sey:08}
876: Seymour, N., et al., 2008, MNRAS, 386, 1695
877: \bibitem[Strolger et al.(2004)]{Str:04}
878: Strolger, L.-G., et al. 2004, \apj, 613, 200
879: \bibitem[Sullivan et al.(2001)]{Sul:01}
880: Sullivan, M., Mobasher, B., Chan, B., Cram, L., Ellis, R., Treyer, M.,
881: Hopkins, A. 2001, \apj, 558, 72
882: \bibitem[Zheng et al.(2007)]{Zhe:07}
883: Zheng, X. Z., Bell, E. F., Papovich, C., Wolf, C., Meisenheimer, K.,
884: Rix, H.-W., Rieke, G. H., Somerville, R. 2007, \apj, 661 L41
885: \end{thebibliography}
886:
887: \begin{table*}
888:
889: \caption[]{SFRDs estimated from the sum of the SFRDs for individual galaxies and from integrating the rest-frame Luminosity functions at four redshift intervals considered here}
890: \begin{tabular}{ccc}
891: & & \\
892: z & $log (SFRD) $ M$_\odot$/yr/Mpc$^3$ & $ log (SFRD)$ M$_\odot$/yr/Mpc$^3$ \\
893: & LF & simple sum \\
894: & & \\
895: $0.315 \pm 0.115$ & $-1.4286 \pm 0.0058$ & $-1.5550 \pm 0.0058$\\
896: $0.540 \pm 0.110$ & $-1.2733 \pm 0.0045$ & $-1.4196 \pm 0.0045$\\
897: $0.765 \pm 0.115$ & $-1.0192 \pm 0.0029$ & $-1.2273 \pm 0.0029$\\
898: $0.990 \pm 0.110$ & $-0.9656 \pm 0.0026$ & $-1.1033 \pm 0.0026$\\
899:
900: \end{tabular} \label{table1}
901: \end{table*}
902:
903:
904:
905: \begin{figure}
906: %\epsscale{0.8}
907: \includegraphics[angle=0,scale=1]{fig1.eps}
908: %\plotone{phot_spec.eps}
909: \caption{Plot of the conversion factor, $k_{2800}$, between UV luminosity
910: (at 2800~\AA) and ongoing SFR as a function of redshift.
911: $L_{2800} = k_{2800} (z) \, {SFR \over M_\odot yr^{-1}}$
912: erg s$^{-1}$Hz$^{-1}$. The
913: star formation history used is taken from Strolger et al. (2004), while
914: stellar synthesis models are taken from \citet{BC:03}, assuming
915: solar metallicity and a Salpeter IMF. The conversion factor is given in
916: units of $10^{27}$.
917: \label{fig1}}
918:
919: \end{figure}
920:
921: \begin{figure}
922: %\epsscale{0.8}
923: \includegraphics[angle=-90,width=15cm]{fig2.ps}
924: %\plotone{phot_spec.eps}
925: \caption{Distribution of the stellar mass of galaxies in different redshift intervals. The redshift bins are adopted following Scoville et al (2007b).
926: The mass completeness limits for the COSMOS are estimated and used to select
927: a survey complete in stellar mass.
928: \label{fig2}}
929: \end{figure}
930:
931: \begin{figure}
932: %\epsscale{0.8}
933: \includegraphics[angle=-90,width=15cm]{fig3.ps}
934: %\plotone{phot_spec.eps}
935: \caption{Distribution of the estimated SFRs (corrected for extinction) for
936: COSMOS galaxies in different redshift intervals.
937: \label{fig3}}
938:
939: \end{figure}
940:
941: \begin{figure}
942: %\epsscale{0.8}
943: \includegraphics[angle=-90,width=15cm]{fig4.ps}
944: %\plotone{phot_spec.eps}
945: \caption{The extinction-corrected SFR densities, $SFRD_{2800}$, {\it vs.}
946: redshift, derived from COSMOS galaxies are compared with other
947: independent estimates of the SFRD. Total SFRDs are estimated
948: from integrating the rest-frame UV luminosity function (filled circles)
949: or are estimated by summing over the SFRs for individual galaxies
950: (Open circles). Vertical errorbars correspond to
951: Poisson counting statistics, and are all smaller than the symbol size
952: for the COSMOS galaxies (blue symbols). The filled black symbols are the
953: compilation of data from literature (Hopkins \& Beacom 2006),
954: reduced to the same cosmology used here.
955: \label{fig4}}
956: \end{figure}
957:
958: \begin{figure}
959: %\epsscale{0.8}
960: \includegraphics[angle=-90,width=15cm]{fig5.ps}
961: %\plotone{phot_spec.eps}
962: \caption{Extinction corrected $SFR_{2800}$ density from COSMOS is plotted
963: against redshift in different stellar mass intervals. Errorbars are again
964: estimated assuming Poisson statistics, and typically smaller than the
965: symbol size.
966: \label{fig5}}
967: \end{figure}
968:
969: \begin{figure}
970: %\epsscale{0.8}
971: \includegraphics[angle=-90,width=15cm]{fig6.ps}
972: %\plotone{phot_spec.eps}
973: \caption{The SFRDs from the COSMOS (this study)- (filled dots) are compared with those from Zheng et al (2007)- (stars). Different colors indicate different
974: mass intervals.
975: The two samples are binned in the same stellar mass intervals. The SFRDs from
976: Zheng et al (2007) are estimated using the combined UV and total infrared
977: (8-1000 $\mu$m) flux of galaxies. Errorbars correspond to poisson statistics.
978: \label{fig6}}
979: \end{figure}
980:
981:
982: \begin{figure}
983: %\epsscale{0.8}
984: \includegraphics[angle=-90,width=15cm]{fig7.ps}
985: %\plotone{phot_spec.eps}
986: \caption{Changes in SFRD with redshift in different stellar mass intervals,
987: extended
988: to $z\sim 2$ using the \citet{Jun:05} estimates from GDDS (asterix). The SFRDs
989: from COSMOS are re-binned to the same mass intervals as those in
990: the \citet{Jun:05} sample. The data from GDDS is reduced to the
991: same cosmology and IMF as for the COSMOS.
992: \label{fig7}}
993: \end{figure}
994:
995: \begin{figure}
996: %\epsscale{0.8}
997: \includegraphics[angle=-90,width=15cm]{fig8.ps}
998: %\plotone{phot_spec.eps}
999: \caption{SFRD as a function of stellar mass and galaxy type.
1000: Blue: Starburst; Green: Spiral (Sa-Sd); Red: Early (E/S0);
1001: Black: all types.
1002: \label{fig8}}
1003: \end{figure}
1004:
1005: \end{document}
1006: