1: %\documentclass[12pt,preprint]{aastex}

2: \documentclass{emulateapj}

3: \usepackage{natbib}

4: %\usepackage[dvips]{color}

5: 

6: \begin{document}

7: 

8: \title{The H$\alpha$-based Star Formation Rate Density of the Universe

9:   at z=0.84.}

10: 

11: \author{V\'{i}ctor Villar, Jes\'us Gallego, Pablo G. P\'erez-Gonz\'alez, Sergio Pascual}

12: 

13: \affil{Universidad Complutense de Madrid, E-28040 Madrid, Spain}

14: \email{viv,jgm,pgperez,spr@astrax.fis.ucm.es}

15: 

16: \author{Kai Noeske, David C. Koo} 

17: 

18: \affil{Lick Observatory, University of California, Santa Cruz, CA

19: 95064}

20: \email{koo,kai@ucolick.org}

21: \and

22: 

23: 

24: \author{Guillermo Barro, Jaime Zamorano} 

25: 

26: \affil{Universidad Complutense de Madrid, E-28040 Madrid, Spain}

27: \email{gbc,jaz@astrax.fis.ucm.es}

28: \begin{abstract}

29: We present the results of an H$\alpha$ {\it near-infrared} narrow-band

30: survey searching for star-forming galaxies at redshift $z=0.84$. This

31: work is an extension of our previous narrow-band studies in the

32: optical at lower redshifts. After removal of stars and redshift

33: interlopers (using spectroscopic and photometric redshifts), we build

34: a complete sample of 165 H$\alpha$ emitters in the Extended Groth

35: strip and GOODS-N fields with

36: L(H$\alpha$)$>$10$^{41}$~erg~s$^{-1}$. We compute the H$\alpha$

37: luminosity function at $z=0.84$ after corrections for [NII] flux

38: contamination, extinction, systematic errors, and incompleteness. Our

39: sources present an average dust extinction of

40: A(H$\alpha$)=1.5~mag. Adopting H$\alpha$ as a surrogate for the

41: instantaneous star formation rate (SFR), we measure a

42: extinction-corrected SFR density of $0.17^{+0.03}_{-0.03}$ M$_{\odot}$

43: yr$^{-1}$ Mpc$^{-3}$. Combining this result to our prior measurements

44: at z$=$0.02, 0.24, and 0.40, we derive an H$\alpha$-based evolution of

45: the SFR density proportional to (1+z)$^\beta$ with $\beta=3.8\pm0.5$.

46: This evolution is consistent with that derived by other authors using

47: different SFR tracers.

48: \end{abstract}

49: 

50: \keywords{galaxies: evolution -- galaxies: high redshift -- galaxies: starburst}

51: 

52: \section{Introduction}

53: The cosmic Star Formation Rate (SFR) density evolution of the Universe

54: is an important constraint on galaxy formation and evolution

55: models. Deep redshift surveys have proved that star-formation activity

56: substantially increases with redshift from z$=$0 to $z\simeq1$

57: \citep[see][for a review]{fer00}. This behavior has been reproduced by

58: current galaxy evolution theories

59: \citep[see the review by][]{Bau06}.

60: 

61: Several tracers can be used to obtain SFRs at different redshifts:

62: ultraviolet (UV) continuum, nebular lines such as [O{\sc

63: ii}]$\lambda$3727 or H$\alpha$, total infrared (TIR), or radio

64: continuum luminosities. For a summary of SFR density measurements, see

65: \cite{Hop04} and \cite{Hop06}. In the redshift regime from z$\sim$1 to

66: $z=0$, the observational data in the Far-IR and UV is now much more

67: robust with the results from {\it Spitzer} \citep{PG05} and those from

68: GALEX \citep{Arn05,Schimi05} and the VVDS \citep{tresse2007}.

69: 

70: Focusing on the H$\alpha$ SFR tracer, the local SFR density was first

71: measured by \citet[][see also \citealt{PG03}]{Gallego95} using the UCM

72: Survey \citep{Zam94,Zam96}. Similar values at z$=$0 have also been

73: obtained more recently by the SDSS \citep{Brinchmann04} and SINGG

74: \citep{Hanish06} projects. At z=0.24, \cite{Tresse98} and

75: \cite{P01} obtained similar SFR densities for a sample of CFRS

76: galaxies and a sample selected using a narrow band technique like

77: ours. Recently, \cite{Shi07} also used the narrow band technique at

78: this redshift, reaching fainter luminosities. \cite{Jones01} used

79: their narrow band counts obtained with a tunable filter to study the

80: redshift range z$=$0.0$-$0.4. \cite{gla04} also used a tunable filter

81: to detect a total of eight emission-line galaxies in the Hubble Deep

82: Field, three of them being H$\alpha$ emitters at $z=0.40$. At

83: $z\sim$1, \cite{gla99} obtained a pioneering result from near-IR

84: spectroscopy of eight CFRS galaxies in the 0.79$<$z$<$1.1 redshift

85: range. Their results were completed by \cite{Tresse02}, who obtained

86: near-IR spectroscopy with VLT for 30 galaxies with redshifts

87: 0.5$<$z$<$1.1. \cite{Do06} have recently obtained an average H$\alpha$

88: luminosity for 38 galaxies at 0.77$<$z$<$1 by stacking near-IR spectra

89: where the H$\alpha$ emission was not individually detected (for most

90: of the targets). Aperture and luminosity bias corrections are needed

91: to compare SFR densities from such slit spectroscopy studies with

92: other data. Slit-less spectroscopy from HST data for galaxies in the

93: 0.7$<$z$<$1.9 redshift range allowed the analysis of \cite{Yan99} and

94: \cite{Hop00} for 33 and 37 emission-line galaxies, respectively. \cite{Ly07} used

95: several narrow band filters to study emission-line galaxies at

96: different redshifts and through different emission-lines.  Finally,

97: \cite{Red07} have estimated the SFRd at z$=$2--3 based on UV and FIR 

98: luminosity functions, also predicting the H$\alpha$ luminosity

99: function from them.

100: 

101: Significant discrepancies have been found when comparing the values

102: obtained from different studies and tracers, due to dust extinction,

103: metallicity and different spatial origins of the emission. The

104: H$\alpha$ luminosity is an excellent tracer of the SFR

105: \citep{Ken98,Charlot01}. It is essential when computing the SFR of a

106: galaxy from its optical spectrum \citep{Mous06}. The H$\alpha$

107: luminosity shares with the UV and TIR emissions the dependence to the

108: Initial Mass Function (IMF).  H$\alpha$-based SFRs are affected by

109: obscuration but are not very sensitive to metallicity. The TIR is not

110: affected by dust attenuation, but it may miss the unobscured star

111: formation which may be an important fraction of the total in certain

112: galaxies \citep{PG06}. There are also large uncertainties linked to

113: the estimation of the TIR emission (from 8 to 1000~$\mu$m) from

114: monochromatic measurements (e.g., the 24~$\mu$m flux). In addition,

115: other sources different from the recent star formation (old stellar

116: populations, AGNs) contribute to the dust heating in unknown

117: (difficult to quantify) amounts. The UV luminosity not only traces the

118: current SFR but also relatively old stellar populations \citep{Cal05}

119: and is heavily affected by obscuration.  As shown by \cite{Bell03},

120: obscuration corrected H$\alpha$ is consistent, within a factor of 2,

121: with the summed SFRs estimated using the UV and TIR

122: luminosities. Consequently, H$\alpha$ observations of galaxy samples

123: with UV and TIR data provides an invaluable tool to understand the

124: evolution of the SFR and the role of obscuration in the determination

125: of global SFR for galaxies.

126: 

127: Our group measured the SFR density locally \citep{Gallego95} using a

128: sample of H$\alpha$-selected galaxies from the objective-prism UCM

129: Survey \citep{Zam94,Zam96}. We then extended this measurement to

130: $z\simeq0.24$ \citep{P01,P05} and also $z\simeq0.4$ \citep{P05}, the

131: maximum redshift for which H$\alpha$ can be reached with CCDs. To

132: select the H$\alpha$ emitters, we successfully used our own optical

133: narrow band filters tuned to the wavelength of the redshifted

134: H$\alpha$ line. The goal of this paper is to extend our previous work

135: to z=0.84 using a narrow band filter centered at 1.20$\mu$m.

136: 

137: This paper is structured as follows. In Section~2, we present the

138: data, the observations, and reduction process. In Section~3, we

139: describe the different steps to select the final sample, including our

140: simulations to analyze the sample biases. In Section~4, we describe

141: the procedure to obtain the H$\alpha$ fluxes for each galaxy. In

142: Section~5, we present the H$\alpha$ luminosity functions (corrected

143: and uncorrected for extinction) and the star formation rate

144: density. Finally, we summarize our results and conclusions in

145: Section~6.

146: 

147: A concordance cosmology is assumed throughout this paper with

148: $H_{\rm0}=70$ km s$^{-1}$ Mpc$^{-1}$, $\Omega_M=0.3$,

149: $\Omega_\Lambda=0.7$ (Lahav \& Liddle 2006). With this cosmology,

150: 1$\arcsec$ at z=0.84 corresponds to 7.63 kpc, the typical surveyed

151: volume for a 1\% narrow-band filter is $\sim$130000 Mpc$^{3}$

152: $^{\square^{-1}}$ and the Universe is 6.44 Gyr old.

153: 

154: \section{Data}

155: 

156: \subsection{Observations}

157: 

158: This work is based on deep near-infrared imaging obtained with a

159: broad- and a narrow-band filters. The narrow-band filter is the

160: J-continuum ($Jc$) centered at 1.20$\mu$m, corresponding to H$\alpha$

161: at z=0.84. The broad-band filter is used to determine an approximate

162: continuum level near the H$\alpha$ emission-line.

163: 

164: The survey was carried out with the near-infrared camera

165: OMEGA-2000\footnote{http://www.mpia-hd.mpg.de/IRCAM/O2000/index.html}

166: on the 3.5m telescope at the Calar Alto Observatory (Almer\'{\i}a,

167: Spain) with the $J$ and narrow-band filters. OMEGA-2000 is equipped

168: with a 2k$\times$2k Hawaii-2 detector with 18$\mu$m pixels

169: (0$\farcs$45 on the sky, 15$\arcmin \times$15$\arcmin$ field of view).

170: Three different 15$\arcmin \times$15$\arcmin$ pointings were obtained,

171: two in the Extended Groth Strip (EGS) and another one in the

172: GOODS-North field, in April 2005 and May 2006. The characteristics of

173: these 3 pointings are shown in Table~\ref{tab_fields}.

174: 

175: Each field was observed with a dithering pattern consisting of 20

176: different positions with typical relative offsets of

177: 20$\arcsec$-30$\arcsec$. After that sequence, the telescope starts

178: another observation block at the initial position (slightly offseted

179: to remove artifacts). With this combination of patterns, the telescope

180: visits 400 different positions without repeating anyone. For the

181: $J$-band, we co-added 15 images of 10 seconds each, for a total of 150

182: seconds at each position. For the narrow-band filter ($Jc$), 5 images

183: of 30 seconds were coadded at each position, for a total of 150

184: seconds as well. The total average exposure time per pixel was

185: $\sim$7.2~ks in the J band and $\sim$18~ks in the $Jc$ filter.

186: 

187: \begin{deluxetable*}{cccccccccc}

188: \tabletypesize{\scriptsize}

189: \tablecaption{Observed fields}

190: \tablehead{

191: \colhead{Field} & \colhead{$\alpha$ (J2000)} & \colhead{$\delta$ (J2000)} & \colhead{area}& \colhead{t$_{exp}$ NB} & \colhead{t$_{exp}$ BB} & \colhead{FHWM($\arcsec$) NB} & \colhead{FWHM($\arcsec$) BB}& \colhead{m$_{lim}$ NB} & \colhead{m$_{lim}$ BB}\\

192: \colhead{(1)} & \colhead{(2)} & \colhead{(3)} & \colhead{(4)}& \colhead{(5)} & \colhead{(6)} & \colhead{(7)} & \colhead{(8)}& \colhead{(9)} & \colhead{(10)}

193: }

194: \startdata

195: Groth2  &14 17 31 &+52 28 11&0.0468 &17850&7200 &1.1&0.9 & 20.54 & 22.30\\

196: Groth3  &14 18 14 &+52 42 15&0.0648 &18000&7200 &0.9&1.1 & 20.99 & 22.43\\

197: GOODS-N &12 36 40 &+62 12 16&0.0622 &20300&9000 &1.0&0.9 & 20.83 & 21.89\\

198: \enddata

199: \tablecomments{(1) Field name (2) RA (J2000) (3) DEC (J2000) (4) Area (square degrees) (5) narrow-band exposure time (s) (6) broad-band exposure time (s) (7) narrow-band FWHM ($\arcsec$) (8) broad-band FWHM ($\arcsec$) (9) narrow-band limiting J$_{VEGA}$ magnitude (3$\sigma$) (10) broad-band limiting J$_{VEGA}$ magnitude (3$\sigma$)}

200: \label{tab_fields}

201: \end{deluxetable*}

202: 

203: 

204: 

205: 

206: 

207: \subsection{Reduction}

208: We used a combination of the IRAF package {\sc XDIMSUM} and our own

209: dedicated software for the reduction of the data. In a first

210: iteration, the dithered images were dark subtracted and combined

211: without shifting to produce a master flat-field. Pixels marked as

212: cosmetic defects were not used in this computation. After the flat

213: field correction, the sky was subtracted. In this first iteration, we

214: used the median value of each image as the sky value. At this point,

215: we checked the photometry and seeing for each individual image,

216: discarding those images that presented the worst seeing or with low

217: object signal due to the presence of clouds or low transparency. We

218: then combined the remaining images to produce a final mosaic. In the

219: combination, for each final pixel in the image, we rejected pixel

220: values from the individual images that exceeded by 3$\sigma$ the mean

221: signal. This allowed us to get rid of cosmetic defects and cosmic

222: rays. With this first final image, we produced an object mask by

223: detecting all the sources with Sextractor

224: \citep{bertin96}. In a second iteration, we repeated the same process

225: except that we changed the method to construct the flat-field and the

226: sky images. This time, for the flat-field construction we combined the

227: science frames rejecting, in addition to cosmetic defects, object

228: pixels. The sky subtraction was performed with {\sc XDIMSUM}, taking

229: previous and subsequent images to compute the sky for each individual

230: image. Object pixels and cosmetic defects were also excluded in the

231: sky construction. In a third iteration, we normalized the science

232: images dividing by the sky images before creating the flat-field. The

233: new flat-field is then not affected by the shape of the sky. The rest

234: of the process is the same as in the second iteration, obtaining the

235: final science image.

236: 

237: The observing runs were not fully photometric and we had to use bright

238: 2MASS stars to do the photometric calibration. We introduced a

239: color-term because our $J$-filter was not exactly the same as the

240: 2MASS $J$ filter. However, the color term was very small in most

241: cases, with ($J$-$J$$_{2MASS}<$0.08). We estimated that zeropoint

242: errors were lower than $\sim$0.15~mag. The narrow-band filter was

243: calibrated using the $J$-band as reference, assuming that the mean

244: color for the bright objects was zero.

245: 

246: \subsection{Additional data}

247: 

248: In order to estimate photometric redshifts and extinctions, we have

249: also used complementary datasets in both the GOODS-N and the EGS

250: fields. For the GOODS-N, we have used optical and NIR data spanning

251: from the U- to the $HK_{s}$-bands

252: \citep[UBVRIzHK$_{s}$, ][]{capak04} and our own K$_{s}$ imaging

253: data (Barro et al., in preparation). {\it Spitzer} IRAC data and MIPS

254: 24~$\mu$m images were also used, jointly with GALEX observations in

255: the far ultraviolet (150~nm; FUV) and near ultraviolet (230~nm; NUV)

256: bands. In addition, we also used the {\em bviz} HST ACS imaging

257: covering the whole field.

258: 

259: For the EGS, we used the multi-wavelength dataset published by the

260: All-wavelength Extended Groth Strip International Survey (AEGIS, see

261: \citealt{davis07} for a detailed description). These data consist in

262: CFHT {\em ugriz} imaging, CFHT {\em BRI} \citep{coil04} observations,

263: {\em vi} HST-ACS data, {\it Spitzer} IRAC and MIPS images, and GALEX

264: FUV and NUV observations.

265: 

266: There is also a wealth of publicly available spectroscopic redshifts

267: in both fields. For the EGS, the DEEP2 Galaxy Redshift Survey

268: \citep{faber03} obtained over 15,000 redshifts in the whole EGS. In

269: the GOODS-N field, spectroscopy is available for $\sim$1,500 sources

270: \citep{Wir04,Cow04,Red06}.

271: 

272: 

273: 

274: \section{Sample selection}

275: \subsection{Color-magnitude diagram}

276: \label{col-mag}

277: Emission-line objects were selected by their excess flux when

278: comparing the narrow band and the broad band images. The candidates

279: were selected as those showing a clear flux excess. The criterion used

280: was:

281: 

282: \begin{equation}

283: (m_{BB} - m_{NB}) > \mu(m_{BB} - m_{NB}) + n_{\sigma} \; \sigma(m_{BB} - m_{NB})

284: \end{equation}

285: 

286: \noindent where $m_{BB}$ is the apparent magnitude in the broad-band filter, $m_{NB}$

287: is the apparent magnitude in the narrow-band filter, $\mu$ is an

288: offset parameter, i.e. the average deviation from the zero color,

289: $\sigma$ is the standard deviation of the color distribution, and

290: $n_{\sigma}$ is the level of significance. The offset parameter and

291: standard deviation can be expressed as a function of the narrow-band

292: magnitude, and they can be calculated directly from the distribution

293: of objects. Thus, we have a certain curve, dependent on the

294: narrow-band magnitude, above which objects are selected as

295: emission-line candidates. In Figure~\ref{cm-diag}, we show the

296: color-magnitude diagram with the selection curve for one of our

297: fields.

298: 

299: \begin{figure}

300: \includegraphics[width=6cm, angle=-90]{f1.eps}

301: \caption{\label{cm-diag} Color-magnitude diagram for the Groth3 field. 

302: Emission-line candidates are represented with black circles, and the

303: rest of the objects are represented with grey circles. The

304: 2.5-$\sigma$ selection curve is also plotted. Fluxes are measured

305: within a four pixel diameter aperture.}

306: \end{figure}

307: 

308: 

309: Fluxes in each band were measured within fixed circular apertures of

310: different sizes. Thus, our measurements have the same spatial origin

311: avoiding the mix of light from different regions in extended

312: galaxies. The selection of the candidates was carried out using

313: different apertures, typically ranging from the PSF FWHM to five times

314: this quantity, for a total of 9-10 apertures. 

315: 

316: The main goal in the selection process is to efficiently select

317: emission-line objects avoiding (redshift and non-emission-line)

318: interlopers. Taking advantage of the large spectroscopic surveys in

319: both the EGS and GOODS-N fields, we studied the the level of

320: significance and range of apertures that yielded better results.

321: 

322: In order to study the best level of significance, we created several

323: selection curves with values of $n_{\sigma}$ ranging from 1.5 to 3.0

324: in steps of 0.25. Each selection curve defines a sample of emitting

325: candidates. We then obtained spectroscopic redshifts for each sample

326: by cross-matching our selected samples with spectroscopic

327: catalogs. The search radius was set to 1$\arcsec$. The objects with

328: spectroscopic redshift can be divided into those selected by an

329: emission-line (i.e. selected by H$\alpha$, [O{\sc

330: iii}]$\lambda\lambda$5007,4959 or [O{\sc ii}]$\lambda$3727 line flux),

331: and those not selected by any of these emission-lines. The fraction of

332: the former objects over the total tell us how accurately we are

333: selecting genuine emission-line galaxies. The final goal is to select

334: the maximum number of objects without losing accuracy. Using the

335: lowest significance level, we obtain a sample of candidates with the

336: largest number of objects, but many of them could be redshift

337: interlopers. Assuming that every object in the redshift range is an

338: emission-line object, we can measure the fraction of objects recovered

339: over the total in the lowest significance level. The best level of

340: significance will be a compromise between accuracy and number of

341: selected objects. In Figure~\ref{check_sigma}, we show the results for

342: each level of significance. We demonstrate here that a level of

343: significance $n_{\sigma}$=2.5 is a good compromise between the number

344: of selected emission-line objects and the accuracy of the selection.

345: 

346: 

347: 

348: To include as many different line emitters as possible, it is

349: necessary to use several apertures. The smallest apertures are more

350: adequate for the detection of small, low luminosity emission-line

351: objects, since the corresponding fluxes are less affected by the sky

352: noise. This is also the case for bright objects with high nuclear star

353: formation. On the other hand, large, low surface brightness objects

354: with extended star formation are better selected with the larger

355: apertures. Large apertures measurements are very noisy for small

356: objects, so apertures significantly larger than the object were not

357: considered. Figure~\ref{check_sigma} shows the fraction of objects

358: selected in each aperture over the total number obtained by taking

359: into account all the apertures. In addition, we represent the accuracy

360: at each aperture. If we select emission-line candidates in a

361: 1.8$\arcsec$ (4 pixels) diameter aperture, we recover $\sim$70\% of

362: the objects in the final sample using all apertures. Thus, we are

363: losing $\sim$30\% of the objects if we only use one aperture, even if

364: it is the one that selects the highest number of objects. The accuracy

365: in each aperture remains constant, even for the larger apertures where

366: the sky noise could severely affect small objects fluxes. The reason

367: is that we reject those objects selected in apertures much larger than

368: its size.

369: \begin{figure}

370: \includegraphics[width=7.5cm,angle=-90]{f2a.eps}

371: \includegraphics[width=7.5cm,angle=-90]{f2b.eps}

372: \caption{\label{check_sigma} Left: Analysis of different significance

373: levels $n_{\sigma}$. Open circles represent the fraction of

374: emission-line objects in the total selected sample at each

375: $n_{\sigma}$. Filled circles represent the fraction of emission-line

376: objects selected at each $n_{\sigma}$ level, considering the total as

377: the number of objects selected at the lowest significance level,

378: i.e. $n_{\sigma}$=1.5. Right: Selection results in different

379: apertures. Open circles: fraction of total number of objects selected

380: at each aperture over the total number obtained with all the

381: apertures. Filled circles: fraction of confirmed emission-line

382: objects.}

383: \end{figure}

384: 

385: \subsection{Star-galaxy segregation}

386: 

387: After selecting the candidates to be an emission-line at z$=$0.84, we

388: must determine if the source is a star or a galaxy. The discrimination

389: between stars and galaxies was carried out using eight different

390: criteria. The main criterium was the {\tt STELLARITY} parameter given

391: by Sextractor in each optical and NIR band were the object was

392: detected. Every object presenting an average value of the {\tt

393: STELLARITY} parameter higher than 0.95 was classified as a star.

394: 

395: In addition we used the following color criteria based on IRAC and NIR

396: magnitudes \citep{Eisen04,RR05}: a) $[3.6] - [8.0] > -2$ and

397: $[3.6]-[8.0] < -1$ and $[8.0] < 20.$, or $[3.6]-[4.5] > -1$ and $[3.6]

398: - [4.5] < -0.5$ and $[4.5] < 19.5$; b) $[5.8] - [8.0] > -1$, $[5.8] -

399: [4.5] < -0.2$ and $[8.0] < 20.$; c) $I - [8.0] < -1$ or $I - [3.6] <

400: 1$ and $[3.6] < 18.$ or $I - [8.0] < -1$ and $[3.6] - [8.0] < - 1$; d)

401: $B - I > 2 × (I - [3.6]) + 0.070;$ e) $J - K + 0.956 < 0.5$; and f)

402: $[3.6]_{3\arcsec} - 0.460 - [3.6]_{auto} > -0.25$ and $[3.6] < 15.$

403: and $[3.6]_{3\arcsec} -0.460-[3.6]_{auto} < 0.2$, or $[3.6]_{3\arcsec}

404: - [3.6]_{auto} < - 0.25$, where [band]$_{3\arcsec}$ is the magnitude

405: in a 3$\arcsec$ diameter aperture, and [band]$_{auto}$ is the mag auto

406: magnitude given by sextractor (an estimation of the integrated

407: magnitude). The BzK criterion $(z-K)_{AB} <0.3\times (B-z)_{AB}-0.5$

408: \citep{daddi04} was also used.

409: 

410: Only 4 objects where classified as stars in the total sample of 243

411: candidates. Half of them were selected in the GOODS-N field and the

412: other half in the field Groth2. This represents 1.6\% of the whole

413: sample, a clearly negligible fraction.

414: 

415: \subsection{Photometric redshifts}

416: 

417: Once the stellar objects have been removed from the sample, we tried

418: to get rid of the objects outside the redshift range we are

419: studying. We showed in Section~\ref{col-mag} that we can have two

420: types of redshift interlopers in our sample: a) those selected by

421: other emission-lines; and b) those selected due to noise or strange

422: spectral features. Spectroscopic redshifts with enough quality were

423: available for 98 out of 239 objects (241 if we include the

424: stars). This means that there are 141 objects (59\% of the entire

425: sample) without spectroscopic data. Estimating photometric redshifts

426: (despite their relatively high uncertainties compared to spectroscopic

427: values) for these objects is important to get a highly complete and

428: reliable sample of galaxies at z$\sim$0.84.

429: 

430: \begin{figure}

431: \includegraphics[width=7.5cm]{f3a.eps}

432: \includegraphics[width=7.5cm]{f3b.eps}

433: \caption{\label{photo-z} Comparison between spectroscopic and photometric 

434: redshift for sources selected in the $J$-band in the GOODS-N (left

435: panel) and Groth fields (right panel). Black points are objects whose

436: spectroscopic redshift quality is very good. Grey points have

437: spectroscopic redshifts with low reliability flags. Note that,

438: although it seems to be a lot of dispersion in the EGS field, 86\% of

439: the objects with reliable spectroscopic redshift fall within

440: $\sigma_{z}/(1+z)<$0.1 (dark lines in the figure), and 95\% fall

441: within $\sigma_{z}/(1+z)<$0.2.}

442: \end{figure}

443: 

444: 

445: We obtained photometric redshifts for our sources using the same

446: method presented in \cite{PG05} and appendix B of

447: \cite{PG07}. First, we measured consistent (aperture matched) photometry 

448: in each band where the object was detected. Then, a set of templates

449: (built with stellar population and dust emission models) was

450: redshifted (in steps of $\Delta z=0.1$) and convolved with the

451: observed filters. A $\chi^{2}$ minimization algorithm was used to

452: estimate the most probable photometric redshift for each object. A

453: preliminary step determining the 1.6$\mu$m bump feature helped to

454: constrain the final solution. An additional constraint was imposed to

455: the template that best fitted the data points: it had to be younger

456: than the age of the universe at the given photometric redshift. The

457: photometric redshift probability distribution was built with the best

458: $\chi^{2}$ values for each redshift. This probability distribution was

459: very useful because some objects had two or even more peaks, making

460: them compatible with different redshifts.

461: 

462: \begin{figure}

463: \includegraphics[width=8.5cm]{f4.eps}

464: \caption{\label{hist_photz08} Photo-redshift histogram for all 

465: galaxies in our three fields with reliable spectroscopic redshift

466: within our redshift range. Some of these objects have several peaks in

467: the probability distribution that move them to the central Gaussian

468: distribution.  }

469: \end{figure}

470: 

471: 

472: 

473: We estimated photometric redshifts for all the objects with

474: spectroscopic redshifts detected in the $J$-band

475: images. Figure~\ref{photo-z} shows the comparison between

476: spectroscopic and photometric redshifts for the GOODS-N and Groth

477: fields. In the first panel, we show the comparison for the 1430

478: $J$-band sources with available spectroscopy in GOODS-N. Although

479: there are some sources that lie quite far from the one-to-one

480: relation, most of them have a photometric redshift in good agreement

481: with the spectroscopic value. There is no evidence of a significant

482: systematic error, given that the average difference $\delta

483: z=z_{spec}-z_{photo}$ is 0.011, 90\% of

484: the objects with reliable spectroscopic redshift fall within

485: $\sigma_{z}/(1+z)<$0.1, and 97\% fall

486: within $\sigma_{z}/(1+z)<$0.2.

487: 

488: The results for the EGS fields are shown in the second panel, with a

489: total of 3810 sources. In this case, the quality of the

490: photo-redshifts is very similar to that achieved in GOODS-N. The

491: average difference between redshifts is $\delta z$=-0.011, 86\% of

492: the objects with reliable spectroscopic redshift fall within

493: $\sigma_{z}/(1+z)<$0.1, and 95\% fall

494: within $\sigma_{z}/(1+z)<$0.2.

495: 

496: 

497: Due to the typical photometric redshift uncertainties, the objects

498: with spectroscopic redshift in the redshift range of interest (the one

499: corresponding to the H$\alpha$ emission for our NB filter) are spread

500: over a much wider photo-redshift range. Figure~\ref{hist_photz08}

501: shows the histogram of photometric redshifts for the galaxies with

502: spectroscopic redshifts within our filter's range. Most of the sources

503: are close to the expected spectroscopic value of z$=$0.84 with some

504: outliers. However, some of these outliers presents peaks in the

505: probability distribution that shift them close to the spectroscopic

506: redshift. The mean (median) photo-redshift value of the distribution

507: plotted in Figure~\ref{hist_photz08} is z$_\mathrm{photo}$$=$0.822

508: (0.820), whereas z$_\mathrm{spec}$$=$0.839 (0.838). The difference is

509: $\delta z$$=$0.017 (0.018), and the standard deviation for the

510: photometric redshifts is 0.16.

511: 

512: 

513: 

514: Objects with a measured spectroscopic redshift outside the range

515: covered by the filter were removed from the sample. Sources with no

516: spectroscopic redshift and photometric redshift

517: z$_\mathrm{photo}$$<$0.5 or z$_\mathrm{photo}$$>$1.1 were also

518: discarded. Note that we checked the photo-z probability distribution

519: for each of these sources and, in the cases where there was a peak at

520: 0.5$<$z$_\mathrm{photo}$$<$1.1, it was introduced again in the

521: sample. Finally, eight galaxies without spectroscopic nor photometric

522: redshift were kept in the final sample.

523: 

524: After the removal of stars and redshift interlopers, the final sample

525: of H$\alpha$ emitters at z$\sim$0.84 has 165 objects: 51 in Groth3, 56

526: in Groth2 and 58 in GOODS-N. Of these, 165 galaxies, 79 (58\%) are

527: confirmed spectroscopically. Table~\ref{tab_objects} lists the objects

528: in the final sample. The [O{\sc ii}]$\lambda$3727 and [O{\sc

529: iii}]$\lambda\lambda$4959,5007 emitters will be analyzed elsewhere.

530: 

531: 

532: 

533: \subsection{Survey detection limits and completeness}

534: \label{incomplete}

535: The narrow-band technique to select emission-line galaxies at

536: different redshifts has been extensively used over the last

537: years. However, most of the times, the line flux detection limit is

538: not consistently determined. The problem is that two different images

539: are used and that each line flux could come from different

540: combinations of narrow and broad band fluxes (i.e., galaxies could

541: cover a wide range in equivalent widths).

542: 

543: In this work, we decided to tackle this problem performing simulations

544: of the selection and measurement processes, in order to determine: a)

545: the completeness of our selection; b) corrections for incompleteness

546: in the luminosity function; and c) systematic errors that could lead

547: to erroneous line flux measurements.

548: 

549: The method consisted in introducing a well known sample of fake

550: galaxies in the science images and, working exactly in the same way as

551: we do with the real images, check whether or not we recover the

552: original properties of the fake sample.

553: 

554: \begin{figure}

555: \includegraphics[width=8.5cm]{f5.eps}

556: \caption{\label{color_sim} Color- magnitude selection diagram of simulated 

557: objects in the Groth3 field. The dark-grey points have

558: f$_l$$=$10$^{-16}$~erg~s$^{-1}$~cm$^{-2}$ and the light-grey

559: ones have

560: f$_l$$=$1.5$\times$10$^{-16}$~erg~s$^{-1}$~cm$^{-2}$. All

561: magnitudes were measured in circular apertures for 4 pixels

562: diameter. The solid line is the selection curve for this field and

563: aperture size, with only objects above it considered as candidates

564: (for this aperture).}

565: \end{figure}

566: 

567: The analysis of the HST morphology of our sample (Villar et al, in

568: preparation) shows that most of our galaxies are disky, and that a

569: significant fraction of the global star formation (typically less than

570: 50\%, with a mean value of 30\%) is distributed in several star

571: forming regions (being the mean number 5) covering the whole

572: galaxy. We used this information to model disks with star formation

573: distributed in five star forming regions, randomly distributed, within

574: the galaxy's half light radius. This produces models of galaxies with

575: highly concentrated as well as more extended and diffuse star

576: formation. We used a exponential law for the disks, limiting the

577: models to three different half light radius and three different

578: inclinations. The star forming regions were modeled with Gaussian

579: profiles and a half light radius of 600pc, which is the average radius

580: we found in the morphological study. In Table~\ref{tab_prop}, we give

581: the range of parameters covered by the models. The fake galaxy images

582: were constructed using {\tt GALFIT} \citep{Peng02}, convolving the

583: model with the field's PSF. No additional noise was added to the

584: models because the main source of noise for faint objects in our

585: images was the sky background level. For each combination of

586: parameters, we inserted 200 fake galaxies in the science image. We

587: then carried out the detection of candidates in the standard way.

588: 

589: \begin{deluxetable}{lr}

590: \tablecaption{Simulations: range of parameters}

591: \tablehead{

592: \colhead{Physical property} &\colhead{Range}

593: }

594: \startdata

595: Log Line flux(erg s$^{-1}$ cm$^{-2}$)& -15.1 -- -16.5\\

596: H$\alpha$ equivalent width (\AA)& 10 -- 2500\\

597: Effective radius (kpc) & 2.5, 5.0, 7.5\\

598: Inclination ($^\circ$)&0, 45, 70\\

599: 

600: \enddata

601: \label{tab_prop}

602: \end{deluxetable}

603: 

604: 

605: For a certain line flux, we have different narrow-band and broad-band

606: fluxes. The magnitudes for each band are given by:

607: \begin{eqnarray}

608: m_{NB}=C-2.5 \log (f_c+f_l/\Delta_{NB})\\

609: m_{BB}=C-2.5 \log (f_c+f_l/\Delta_{BB}) 

610: \end{eqnarray}

611: 

612: \noindent where $C$ is the zero-point, $f_c$ is the continuum flux 

613: per wavelength unit, $f_l$ is the line flux and $\Delta_{NB}$ and

614: $\Delta_{BB}$ are the narrow- and broad-band filter effective

615: widths. Objects with the same line flux present different narrow-band

616: magnitudes with a maximum value given by $m_{NB}=C-2.5 \log

617: (f_l/\Delta_{NB})$.  In Figure~\ref{color_sim}, the different

618: locations in the color-magnitude diagram are shown for objects with

619: the same flux level simulated in the Groth3 field. The fluxes were

620: measured in apertures of 4 pixels in diameter and the corresponding

621: selection curve is also shown. We can see that the color, i.e., the

622: equivalent width, increases with narrow-band magnitude. The reader

623: should note that low equivalent width objects are not selected by our

624: method. However, these sources also present bright $J$-magnitudes and

625: they are not relatively very numerous, so the completeness will not be

626: seriously affected. The fluxes for the faintest sources are recovered

627: with less accuracy and the dispersion becomes larger, preventing the

628: selection of the whole fraction of objects.

629: 

630: \begin{figure*}

631: \epsscale{.10}

632: \includegraphics[width=8.0cm]{f6a.eps}

633: \includegraphics[width=9.0cm]{f6b.eps}

634: \includegraphics[width=8.0cm]{f6c.eps}

635: \includegraphics[width=9.0cm]{f6d.eps}

636: \includegraphics[width=8.0cm]{f6e.eps}

637: \includegraphics[width=9.0cm]{f6f.eps}

638: \caption{\label{simu1} Completeness and line flux accuracy for each field surveyed. 

639: From top to bottom: Groth2, Groth3, GOODS-N. Left: Completeness versus

640: line flux. Different symbols represent different half-light radius of

641: the simulated sources: circles for 2.5 kpc, triangles for 5 kpc and

642: squares for 7.5 kpc (assuming z=0.84). Right: Recovered flux versus

643: input line flux. Symbols are the same as in the left panels. }

644: \end{figure*}

645: 

646: In addition, Figure~\ref{color_sim} shows that there is an upper limit

647: to the color that decreases with narrow-band magnitude. The

648: explanation is that high equivalent widths imply very faint fluxes in

649: the broad-band image. Consequently, these objects would not be

650: selected by our technique, since we need a simultaneous two band

651: detection to assure the existence of the source. It is possible,

652: however, to use the narrow-band image to detect all objects and then

653: measure at the same position in the broad-band image. We did not apply

654: this method because the alignment of the images was not good enough

655: throughout the whole image, producing wrong centered apertures in the

656: broad-band images in some regions, which lead to incorrect results. A

657: refinement of the alignment process could not be done without

658: substantial transformation of the images, which could alter the final

659: results. We conclude that we may have missed faint galaxies with large

660: equivalent width values undetected in the $J$-band. In order to

661: analyze this systematic detection effect, we checked the images

662: looking for objects only detected in the narrow-band images. We could

663: not find any reliable candidate. Moreover, the maximum observed

664: equivalent width measured in our sample is 1077\AA, with the rest of

665: the sample below 600\AA. This is in good agreement with the results

666: found in \cite{Gallego95}, \cite{Tresse02}, and \cite{P05}. A deep

667: narrow band survey looking for Ly$\alpha$ emitters at redshift z=8.8

668: \citep{wil05} did not find any population of high EW(H$\alpha$)

669: emitters. However, this survey was carried out over a small

670: area. Our survey covers a much wider area and confirm these previous

671: results.

672: 

673: Figure~\ref{simu1} (left) shows the fraction of selected objects for

674: the different surveyed fields. Each panel correspond to a different

675: field and shows the completeness for different half light radius. The

676: completeness curve shows a smoother decline than that we can found in

677: magnitude completeness studies. The reason is that in the narrow-band

678: technique, two different magnitudes are involved and for each line

679: flux we span a wide range in broad- and narrow-band magnitudes (see

680: Figure~\ref{color_sim}). Another important issue is the effect of

681: increasing the half light radius. The completeness drops from

682: $\sim$80\% to $\sim$50\% when we move from r$_{eff}$=2.5 kpc to

683: r$_{eff}$=7.5 kpc in the GOODS-N field at $\log$(l$_{f}$)$\sim$-15.65,

684: and would move to lower fractions for higher half light

685: radii. However, this is not a major concern in our case, since 85\% of

686: galaxies present half light radius lower than 7.5 kpc, with all of

687: them except one below 10 kpc (Villar et al., in preparation).

688: 

689: 

690: 

691: The simulations allowed us to check the reliability of the measured

692: fluxes.  For the objects that satisfy the selection criteria, there is

693: a good agreement between the mean recovered value and the mean

694: simulated flux (see Figure~\ref{simu1}, right panel), even for the

695: faintest line fluxes. The comparison between individual objects in

696: each line flux bin give us a better estimation of the error than that

697: determined with photometric errors. Error bars in figure~\ref{simu1}

698: show the standard deviation in the recovered line flux, computed as

699: the standard deviation of the absolute difference between recovered

700: and simulated line fluxes. The errors clearly increase as we move to

701: fainter line fluxes, ranging from a 10\% relative error for the

702: brightest objects to a 60\% for the faintest ones, although they keep

703: below 30\% up to f(H$\alpha$+[NII])$=$5$\times$10$^{-17}$ erg s$^{-1}$

704: cm$^{-2}$.

705: 

706: \section{H$\alpha$ luminosities for z$\sim$0.84 objects}

707: \subsection{Line flux estimation}

708: 

709: Emission-line fluxes were computed using:

710: \begin{equation}

711: f_{\mathrm{l}} =  \Delta_{\mathrm{NB}} \left(f_{\mathrm{NB}}-f_{\mathrm{BB}}\right)\frac{1}{1-\epsilon}

712: \end{equation}

713: 

714: \noindent where $\ensuremath{f_{\mathrm{NB}}}$ and $\ensuremath{f_{\mathrm{BB}}}$ 

715: are total fluxes in the narrow- and broad-bands, $f_{\mathrm{l}}$ is

716: the line flux (including [NII]$\lambda\lambda$6548,6584),

717: $\Delta_{\mathrm{NB}}$ is the width of the narrow-band filter computed

718: following the procedure specified by \cite{P07}, and $\epsilon$ is the

719: ratio of the widths of the narrow- and broad-band filters.

720: 

721: To estimate the integrated emission-line flux of each galaxy, we used

722: the whole set of apertures. The flux grows with aperture diameter

723: until the end of the emission region or the sky are reached. Since

724: galaxies present a variety angular sizes, apertures of different sizes

725: must be used. For small objects, the maximum flux will be reached in a

726: small aperture, whereas for large objects it will be reached in larger

727: apertures. We visually checked the emission-line fluxes for each

728: aperture in each galaxy to select the more reliable integrated

729: emission-line flux.

730: 

731: H$\alpha$ luminosities were computed from the line fluxes. The

732: underlying stellar absorption for H$\alpha$ has a negligible effect

733: when compared with errors from photometry, so no correction was added

734: \citep[see][]{nakamura04}. Nitrogen contamination to the narrow-band

735: flux was removed following the approach in \cite{P07}. In that work,

736: the shape of the narrow-band filter is considered when computing the

737: average [NII] contribution to the measured flux, assuming a certain

738: $I([NII]\lambda6584)/I(H\alpha)$ value. These authors assumed an

739: average ratio $I([NII]\lambda6584)/I(H\alpha)$=0.32, the mean value

740: obtained for the UCM Survey sample \citep{Gallego97} and the galaxies

741: in the Sloan Digital Sky Survey Release 4 \citep[SDSS

742: DR4][]{adel06}. In our case, we have used the SDSS DR4 to study the

743: dependence of $I([NII]\lambda6584)/I(H\alpha)$ with the equivalent

744: width of H$\alpha$ plus the [NII] contribution

745: [$EW(H\alpha+[NII])$]. Figure~\ref{nitro_fig} shows $\log

746: (I([NII]\lambda6584)/I(H\alpha))$ versus $\log

747: (EW(H\alpha+[NII]\lambda6584))$. There is a clear trend of decreasing

748: $I[NII]\lambda6584/I(H\alpha)$ as we move to higher equivalent widths,

749: which can be explained due to a metallicity decrease. The circles

750: represents the mean values obtained from the SDSS sample. The

751: dispersion remains at $\sim$0.4 dex for equivalent widths below $\log

752: (EW(H\alpha + [NII]\lambda6584))$=2. For higher equivalent widths, the

753: dispersion increases up to $\sim$1 dex. This relation was used to

754: estimate the [NII] contribution to the emission-line flux measured in

755: the narrow-band images for each of our sources, obtaining a mean

756: (median) value of $I[NII]\lambda6584/I(H\alpha)$=0.26 (0.27), ranging

757: from 0.04 to 0.4.

758: 

759: \begin{figure}

760: %\epsscale{.10}

761: \includegraphics[width=9.cm]{f7.eps}

762: \caption{\label{nitro_fig} Ratio $I[NII]\lambda6584/I(H\alpha)$ 

763: as a function of EW(H$\alpha + [NII]\lambda6584)$ for the Sloan Digital Sky

764: Survey (the number of galaxies is represented in gray scale) is represented in gray scale and the UCM survey (inverted

765: triangles). The mean values for the SDSS are represented as circles.}

766: \end{figure}

767: 

768: The $I[NII]\lambda6584/I(H\alpha)$ may evolve with redshift as galaxy populations could be very different in the past than local ones. However, changing this ratio by a factor of 2 implies $\sim$20\%-30\% variation in the H$\alpha$ line fluxes, which is of the order of the errors.

769: 

770: 

771: 

772: \subsection{Reddening correction}

773: \label{extinction}

774: 

775: Following \cite{buat05}, we used the ratio ($F_{dust}/F_{FUV}$) to

776: compute the extinction in the ultraviolet. Dust emission is given by

777: L(8-1000$\mu$m) and can be estimated for the objects detected by MIPS,

778: which traces the rest-frame continuum at 13$\mu$m in our redshift

779: regime. To carry out this estimation, we first subtracted the stellar

780: emission predicted by the stellar population templates (obtained in

781: the photo-z determination) from the fluxes at rest-frame wavelengths

782: redder than $\sim$4$\mu$m to obtain the pure emission of the

783: dust. Then, we fitted this emission with \cite{cha01} dust emission

784: models. The model that best matched the observed dust emission colors

785: [observed F(24)/F(8)] was selected, and we computed the TIR luminosity

786: L(8-1000$\mu$m) from this model \citep[for more details, see][]{PG07}.

787: The stellar population template, convolved with the FUV filter

788: transmission curve, also give us the FUV rest-frame flux. With the

789: $F_{dust}/F_{FUV}$ ratio, we compute A(FUV) and then the extinction in

790: H$\alpha$ applying the Calzetti extinction law \citep{cal00}, assuming

791: that the attenuation of the stellar emission is 0.44 times the

792: attenuation of the nebular emission. This law was empirically obtained

793: from local starburst with star formation rates of up to a few tens

794: M$_{\odot}$ yr$^{-1}$, very similar to our galaxies.

795: 

796: However, 86 objects were not detected at 24$\mu$m, not allowing us to

797: obtain the dust flux. In this case, we approached the problem from the

798: ultraviolet side. The slope in the ultraviolet is another tracer of

799: the dust obscuration and it is correlated with $F_{dust}/F_{FUV}$

800: ratio, as \cite{Meu95} found for starburst galaxies. More recently,

801: \cite{Gil06} showed that the (FUV - NUV) color, which relates to the

802: UV slope \citep[see][]{kong04}, is also correlated with the

803: $F_{dust}/F_{FUV}$ ratio. The FUV - NUV color was computed convolving

804: the best stellar population template with the appropriate filter

805: transmission curves. Then, we estimated the $F_{dust}/F_{FUV}$ ratio

806: using the GALEX Ultraviolet Atlas of Nearby Galaxies. Each source in

807: our sample was assigned the mean GALEX atlas $F_{dust}/F_{FUV}$ at the

808: same (FUV -NUV). Figure~\ref{galex_fig} shows $F_{dust}/F_{FUV}$

809: vs. (FUV -NUV) for our 79 objects with MIPS detections. Late-type

810: galaxies in the GALEX atlas have also been represented. Our sample

811: follows, with higher dispersion, the general trend of nearby galaxies,

812: although they are, in general, redder than the local sample,

813: indicating that the extinctions are higher than those of the GALEX

814: atlas sample. This also indicates that there is little evolution of this relation with redshift, not having a significant effect on our results. Moreover, considering the objects with extinctions available by both methods, we obtain similar mean extinctions: A(H$\alpha$)=1.67~mag and A(H$\alpha$)=1.87~mag using the UV slope and the infrared excess respectively.

815: 

816: The mean extinction in our sample is A(H$\alpha$)=1.48~mag, a value

817: $\sim$0.5~mag higher than the mean values obtained for the SDSS

818: \citep{Brinchmann04} and UCM \citep{Gallego95} samples. This

819: implies an increase in the typical extinction of star-forming galaxies

820: with redshift of 0.5~mag from the local Universe to

821: z=0.84. 

822: 

823:  \cite{tresse2007} found that the dust obscuration at 1500\AA\

824: was A(FUV)=2 mag from z=0.4 to z=2, decreasing to $\sim$0.9--1~mag for

825: z$<$0.4. Our sample has a mean value of A(FUV)=2.15, in good agreement

826: with these authors. They argue that the decrease in extinction at low

827: redshift is due to the change of the dominant galaxy population. They

828: show that the emission from early-type galaxies starts to dominate in

829: the $B$-band below z$<$0.5, and they make the assumption that in the

830: FUV the early-type population will still dominate. Therefore, as the

831: dust content in early-type galaxies is much lower than in late-types,

832: the amount of extinction will decrease as we move to lower redshifts.

833: 

834: \begin{figure}

835: \includegraphics[angle=-90,width=8.5cm]{f8.eps}

836: \caption{\label{galex_fig} Dust flux to FUV flux ratio (irx) 

837: as a function of FUV - NUV color, i.e. the UV slope. The crosses are

838: the late-type galaxies in the GALEX Nearby Galaxy Atlas

839: \citep{Gil06}. Circles are objects in our sample that are detected at 24~$\mu$m

840: by MIPS. Rhombus show the mean values and dispersion for our sample. }

841: \end{figure}

842: 

843: However, the difference we find when comparing the mean extinction of

844: our sample with that of local samples of star-forming galaxies cannot

845: be explained with that argument. \cite{Brinchmann04} showed that only

846: 12\% of the SFR density comes from galaxies with D4000$>$1.8, and only

847: 2\% from galaxies with D4000$>$2. Thus, only a very small fraction of

848: the star formation can be located in old systems with very poor dust

849: content. Moreover, \cite{Brinchmann04} pointed out that these systems

850: with high D4000 are probably spiral systems with significant

851: bulges. In addition, \cite{PG01} and

852: \cite{Vit96} did not find any elliptical galaxy

853: in the UCM sample and only 7\% of lenticular objects. So, taking into

854: account the previous discussion and the fact that our sample is

855: dominated by disks, thus sharing the morphology of the SDSS and UCM

856: samples, the higher extinction in our sample has to be caused by an

857: increase in the dust content in the galaxies that host the star

858: formation.

859: 

860: 

861: \section{The H$\alpha$ luminosity function at z=0.84}

862: \label{sec_lf}

863: \subsection{The observed H$\alpha$ luminosity function}

864: \label{ha_lf}

865: The H$\alpha$ luminosity function was calculated applying the

866: V/V$_{max}$ method \citep{Sch68}:

867: 

868: \begin{eqnarray}

869: \label{lumfunc}

870: \phi (\log L_i)=\frac{1}{\Delta \log L}\sum_{j} \frac{1}{V(z)_j}\\

871: %\nonumber \text{with} | \log L_j - \log L_i | < \frac{\Delta \log L}{2}

872: \end{eqnarray}

873: 

874: \noindent where $L_i$ is the central luminosity in the bin $i$ and $V(z)_j$

875: is the maximum volume in which object $j$ can be detected.

876: 

877: To properly compute the volumes defined by the narrow-band filter, we

878: followed the procedure described in section 5.3 of \cite{P07}. These

879: authors consider the volume in which an object would be detected in a

880: narrow-band survey based on its position in the color-magnitude

881: diagram, expanding the method to cope with several lines inside the

882: narrow-band filter. The effect of the Nitrogen lines become important

883: when the filter's transmittance falls and the H$\alpha$ line is

884: detected there. In that case, one of the Nitrogen lines would be in

885: the high transmittance region of the filter, increasing the total flux

886: and, hence, the detection probability. To include this contribution,

887: we considered the width of the filter affected by the Nitrogen lines

888: as in Equation (34) of \cite{P07}. If we consider and average volume

889: determined only by the narrow-band filter's FWHM, we would be

890: overestimating the surveyed volume by $\sim$20\% ($\sim$18\%) on

891: average (median), leading to a similar underestimation (i.e., a

892: systematic error) of the LF points. It is also important to take into

893: account the Nitrogen lines in the volume determination for each

894: individual object. If not considered, volumes are subestimated by

895: $\sim$40\% ($\sim$20\%) on average (median).

896: 

897: The H$\alpha$ luminosity function with no extinction correction is

898: shown in Figure~\ref{lf_nored}. The best fit to a \cite{Sch76}

899: function yields the following parameters:

900: 

901: \begin{eqnarray*}

902: \nonumber \phi^*=10^{-1.74\pm0.11} {\rm Mpc}^{-3}\\

903: \nonumber L^*=10^{41.69\pm0.07} {\rm erg}\ {\rm s}^{-1} \\

904: \end{eqnarray*}

905: 

906: We fixed $\alpha$=-1.35 (based on \citealt{Tresse98} and

907: \citealt{Shi07}) as our LF did not reach faint enough luminosities

908: to accurately determine it.

909: 

910: Errors were obtained from simulations. We computed a large number ($\sim$1000) of LFs, randomly changing the line flux for each object within a gaussian distribution, with $\sigma$ determined by the object line flux error. The final errors in the LF are the standard deviations of the distributions obtained from the simulations. We apply this same method for the errors in the Schechter fit. We did simulations varying the LF within the errors distributions, obtaining distributions the Schechter parameters. The final errors in these parameters are the standard deviations of these distributions.

911: 

912: 

913: Figure~\ref{lf_nored} depicts the expected distribution of observed line

914: fluxes for the H$\alpha$ line (since the Nitrogen correction was

915: already applied). To correct for incompleteness, we computed the

916: fraction of galaxies detected and selected at a certain line flux

917: level (Section~\ref{incomplete}), what we call the completeness

918: fraction. Then, we assumed that this fraction was the probability for

919: a galaxy with these properties to be detected and selected in our

920: sample. Thus, for each selected galaxy, we would expect the inverse of

921: the completeness factor to be the real number density of

922: galaxies. This is equivalent to multiplying each source's detection

923: volume by its completeness factor. Thus, in the LF computation, we

924: multiplied the detection volume of each galaxy by the completeness

925: factor. The LF corrected for incompleteness is shown in

926: Figure~\ref{lf_nored}. The correction is more severe as we move

927: towards fainter luminosity bins. It is very strong in the faintest

928: bin, but still it is most probably underestimated.

929: 

930: \begin{figure}

931: \includegraphics[width=8.5cm]{f9.eps}

932: \caption{\label{lf_nored} H$\alpha$ luminosity function not 

933: corrected for extinction (solid circles) with the best fit to a

934: Schechter function (thick line). Open circles represent the derived LF

935: before applying the completeness correction. for comparison

936: \cite{Tresse02} LF (dot-dashed line) and \cite{Hop00} LF (dashed line)

937: not corrected for extinction are also shown. }

938: \end{figure}

939: 

940: The LFs (not corrected for extinction) published by \cite{Tresse02}

941: and \cite{Hop00} are also shown in Figure~\ref{lf_nored}

942: \citep[converted to the cosmology used in this work,

943: see][]{Hop04}. \cite{Tresse02} observed a sample of galaxies at

944: z$\sim$0.7 selected from the Canada-France redshift survey with

945: EW([O{\sc ii}]$\lambda$3727])$\geq$12\AA.  Our LF is very similar to

946: Tresse's, although ours extends to higher luminosities. Also, our LF

947: presents a higher density at the faint end. This could be due to the

948: fact that \cite{Tresse02} applied a global completeness correction

949: independent of the line flux. This was the best they could do, since

950: they could not select the objects directly by their H$\alpha$

951: equivalent width or flux, but by [O{\sc ii}]$\lambda$3727 equivalent

952: width. A global completeness correction make the whole LF move to

953: higher densities whereas a flux dependent completeness correction

954: change the shape of the LF, raising the faint region. There is also

955: another caveat in their selection: not all the H$\alpha$ emitters show

956: [O{\sc ii}]$\lambda$3727\AA\ emission. \cite{Yan06} showed that in the

957: SDSS survey, 20\% of all emission-line galaxies with an H$\alpha$

958: detection have no [O{\sc ii}]$\lambda$3727\AA\ emission. In this

959: sense, if targets are selected in a spectroscopic survey using the

960: oxygen line equivalent width, a considerable fraction of an H$\alpha$

961: selected sample would not be detected. On the other hand, $\sim$30\% of

962: emission-line galaxies show oxygen emission with very low H$\alpha$

963: emission (22\%) or no emission at all (8\%).

964: 

965: \cite{Yan99} and \cite{Hop00} used the slitless spectroscopy technique 

966: to study emission-line galaxies at z$\sim$1.

967: \cite{Yan99} selected 33 emitters at 0.75$\leq$ z $\leq$ 1.9. Their 

968: data was not deep enough to constrain $\alpha$, so they assumed

969: $\alpha$=-1.35. \cite{Hop00} extended the study adding their deeper

970: data to that of \cite{Yan99}. The LF was similar to that of

971: \cite{Yan99}, although steeper. There is a huge discrepancy between 

972: our LF and theirs in the bright end of the LF, as they found many more

973: brighter objects. \cite{Tresse02} pointed out that, to some extent, it

974: could be an effect related to the Nitrogen correction. Indeed, the

975: slitless spectroscopy did not allow a proper deblending of the

976: H$\alpha$ line from the [NII]$\lambda$6548,6584 lines. However, we

977: have the same problem and we estimate a lower density of high

978: luminosity objects as well. Two explanations are possible: a

979: change in the shape of the LF at higher redshifts or field to field

980: variantions. The redshift range surveyed in

981: \cite{Yan99} and \cite{Hop00} is much larger than ours,

982: reaching higher redshifts (0.75$\leq$z$\leq$1.9). Star-forming

983: galaxies at z$\geq$1.4 could be very different from those at

984: z=0.8. For example, $\sim$30\% of the H$\alpha$ emitters at z$\sim$2

985: studied by \cite{Erb06} have $\log$(L$_{H_{\alpha}})>42.5$, whereas

986: our whole sample have lower luminosities. On the other hand, in the

987: volume surveyed by these authors there could be a high density region

988: due to cosmic variance. Probably, both effects are playing a role in

989: the comparison.

990: 

991: 

992: \subsection{The reddening corrected H$\alpha$ luminosity function}

993: \label{cor_lumfunc}

994: 

995: Two major effects are affecting our sample: extinction and field to field variance. The extinction correction was applied to each individual object and was explained in Section~\ref{extinction}. Field to field variance implies galaxy density changes depending on the observed field. Within our three surveyed fields, we notice significant field to field variations. Figure~\ref{f2f} show the different LFs computed for each field. The Groth2 and GOODS-N fields show an overdensity over the Groth3 field. If we limit the comparison to the bins log(L$_{H\alpha}$)=$\lbrace 41.5, 41.9 \rbrace$, which are less affected by low number statistics, the density of objects is $\sim$2.3 and $\sim$1.7 times higher in Groth2 and GOODS-N than in the Groth3 field, respectively. \cite{Taka07} reported a similar variation among the COSMOS and Subaru Deep (SDF) fields for their [O{\sc ii}]$\lambda$3727 emitters. In addittion, we notice that none of our fields could be representative of the mean density of star forming galaxies in the Universe at this redshift.

996: 

997: 

998: 

999: 

1000: In order to correct for the field to field variance effect we use photometric redshift and spectroscopic redshifts to estimate: a) the mean density of our fields with respect to other fields and b) the relative density of galaxies within our redshift range over this mean density.

1001: 

1002: We used photometric redshifts (\citealt{PG07}, P\'erez-Gonz\'alez et

1003: al., in prep.) for the EGS, GOODS-N, Chandra Deep Field South (CDFS)

1004: and Lockman Hole (LH) to estimate the mean density in a redshift range

1005: centered at our redshift, ranging from z$=$0.75 to z$=$0.9. We used a

1006: redshift range wider than the narrow-band filter redshift range

1007: because photometric redshifts do not work properly in such a small

1008: range. We found that there was an overdensity of galaxies in both EGS

1009: and GOODS-N fields. The overdensity factor was $\sim$1.05 and

1010: $\sim$1.16 for GOODS-N and EGS respectively. This first estimation

1011: tell us that we are observing fields with a higher density of galaxies than the mean density in the redshift range 0.75 $<$ z $<$ 0.9. However, at the small range covered by our narrow-band filter densiites could be very different. Fortunately, spectroscopic

1012: redshift surveys are precise enough to reveal the structure in

1013: redshift ranges as small as our. Then, we measured the density ratio

1014: of objects with reliable spectroscopic redshift within our redshift

1015: range over those within 0.75 $<$ z $<$ 0.9. For GOODS-N we found that

1016: this factor was $\sim$1.9, which translates to $\sim$2.0 when we take

1017: into account the density factor for GOODS-N over the mean density. For

1018: Groth2 and Groth3 fields we first measured the density factors between

1019: these fields and the whole EGS field for the redshift range 0.75 $<$ z

1020: $<$ 0.9. We obtained $\sim$1.07 and $\sim$0.71 for Groth2 and Groth3

1021: respectively, showing that could be high variations from field to

1022: field. Then, we measured the ratios between the galaxies within our

1023: redshift range and the galaxies in the wider redshift range, obtaining

1024: $\sim$2.27 and $\sim$1.30, which become $\sim$2.43 and $\sim$0.93 for

1025: Groth2 and Groth3 fields respectively when compared to the whole

1026: EGS. Finally, applying the overdensity factor of the EGS, we

1027: obtained the final factors: $\sim$2.8 and $\sim$1.08 for Groth2 and

1028: Groth3 fields respectively. 

1029: 

1030: \begin{figure}

1031: \includegraphics[width=8.5cm]{f10.eps}

1032: \caption{\label{f2f} Derived LFs for the Groth2 (squares), 

1033: Groth3 (triangles) and GOODS-N fields. The more populated points in

1034: the central region shows a density ratio of $\sim$2.3 and $\sim$1.7

1035: for Groth2 and GOODS-N fields over Groth3 field.}

1036: \end{figure}

1037: 

1038: 

1039: We applied the same method to compute the luminosity function once we

1040: applied the extinction and field to field variance correction.  The

1041: resulting best fit to a Schechter function gives:

1042: 

1043: \begin{eqnarray*}

1044: \nonumber \phi^*=10^{-2.76\pm0.32} {\rm Mpc}^{-3}\\

1045: \nonumber L^*=10^{42.97\pm0.27} {\rm erg}\ {\rm s}^{-1} \\

1046: \nonumber \alpha=-1.34\pm0.18\\

1047: \end{eqnarray*}

1048: 

1049: Note that this time we also fitted the faint-end slope. In the fitting

1050: process, we discarded the faintest and the brightest bins. The

1051: faintest bin was clearly affected by incompleteness. The brightest bin

1052: fell off the general shape of the best Schechter fit. Moreover, it

1053: contains only one object, which could be there due to a wrong

1054: estimation of the reddening or a photo-z outlier.  Figure~\ref{lf_red}

1055: shows the extinction-corrected LF derived in this work as well as

1056: \cite{Tresse02} and \cite{Hop00} corrected for extinction

1057: LFs. \cite{Tresse02} applied an overall extinction correction

1058: A$_{V}$=1 mag obtained from the CFRS sample, except for two galaxies

1059: where high quality spectra were available and f(H$\beta$) and

1060: f(H$\delta$) could be measured. \cite{Hop00} did not attempt the

1061: extinction correction though we can apply the typical correction

1062: A(H$\alpha$)=1 mag \citep[see ][ and references therein]{P07} for this

1063: kind of surveys.

1064: 

1065: The change in the shape of the LF after correcting for extinction and

1066: field to field variance is evident. The typical H$\alpha$ luminosity

1067: has increased more than 1 dex, from $\log$ L$^*$(H$\alpha$)=41.69 to

1068: $\log$ L$^*$(H$\alpha$)=42.97, and the density $\phi^{*}$ has

1069: decreased from -1.74 to -2.76, although this is explained in part because no density correction was applied to the observed LF. Now we can see a clear difference between \cite{Tresse02} LF and this work in the bright regime.

1070: 

1071: In order to check if the bright regime may be affected by errors in extinction, we repeated the procces to obtain the Schecter parameters, including a typical error in extinction of 0.3 mag. The effect on the Schecter parameteres was found negligible.

1072: 

1073: Applying an individual

1074: extinction to each object modifies the whole shape of the LF, because

1075: the objects with highest corrected H$\alpha$ luminosities present high

1076: extinctions. Note that most of the previously published H$\alpha$ LFs

1077: assume an average extinction. However, for the total integrated SFRd,

1078: we obtain very similar results with both approaches (see

1079: section~\ref{sfrd}).

1080: 

1081: \begin{figure}

1082: \includegraphics[width=8.5cm]{f11.eps}

1083: \caption{\label{lf_red} H$\alpha$ luminosity function corrected 

1084: for extinction (solid circles) with the best fit to a Schechter

1085: function (thick line). Open circles represent the derived LF before

1086: applying the completeness correction. \cite{Gallego95} local LF

1087: (dotted line) is also shown. \cite{Tresse02} LF (dot-dashed line) and

1088: \cite{Hop00} LF (dashed line) corrected for extinction are also

1089: shown. No correction was applied to the \cite{Hop00} LF originally so

1090: we applied the typical A(H$\alpha$)=1 mag. }

1091: \end{figure}

1092: 

1093: Now the shape of our LF is very similar to that of \cite{Hop00},

1094: although they still present a higher density at faint

1095: luminosities. However, we have applied a global extinction correction

1096: to their LF, so we might expect a change in shape and an increase in

1097: luminosity if we make a careful extinction correction. Moreover, we

1098: obtain a mean extinction A(H$\alpha$)=1.48 in our sample and we expect

1099: even higher attenuation as we move to higher redshifts, so probably

1100: their LF would move to higher H$\alpha$ luminosities.

1101: 

1102: 

1103: 

1104: 

1105: 

1106: 

1107: 

1108: 

1109: 

1110: 

1111: \subsection{The H$\alpha$-based cosmic star formation rate density}

1112: \label{sfrd}

1113: 

1114: Once we have the LF, we can compute the H$\alpha$ luminosity density through:

1115: 

1116: \begin{equation}

1117:  \rho _L (H\alpha)=\phi^*\space L^*\space \Gamma (2+\alpha)

1118: \end{equation}

1119: 

1120: where $\phi ^*$, $L^*$ and $\alpha$ are the parameters obtained in the

1121: Schechter fitting to the LF.

1122: 

1123: We convert this luminosity density to star formation rate density

1124: through the \cite{Ken98} calibration.

1125: 

1126: %  Recently \cite{Pfla07} showed

1127: % that such a linear relation is not valid for low H$\alpha$

1128: % luminosities, becoming non-linear below $\sim$10$^{39}{\rm erg}\ {\rm

1129: % s}^{-1}$. This is not a major concern for our results, since our

1130: % luminosities are a few orders of magnitude above that regime.

1131:  We find

1132: that the inferred extinction-, field to field variance-corrected star

1133: formation rate density is $\dot\rho_*$=0.17$^{+0.03}_{-0.03}$

1134: M$_{\odot}$ yr$^{-1}$ Mpc$^{-3}$.

1135: 

1136: As a consistency check, we checked that the observed and extinction corrected SFR densities differ by the mean extinction correction. We integrated the observed LF and applied the mean extinction correction. However, the observed LF was also affected by field to field variance so we applied a mean density correction (see section~\ref{cor_lumfunc}). The SFRd obtained in this case is $\dot\rho_*$=0.19$^{+0.03}_{-0.03}$ M$_{\odot}$ yr$^{-1}$ Mpc$^{-3}$ , in good agreement with the previous value. This shows that although a mean extinction correction may not be the appropriate method to obtain real shape of the LF, it is enough to accurately determine the luminosity density or SFRd.

1137: 

1138: This value has not been corrected for AGN contribution as the effects are very small and other authors have not corrected their values either. The AGN contamination is a very difficult to solve problem, and a detailed analysis is out of the scope of this paper. We have tried to quantify how many of our galaxies harbor a luminous AGN by cross-correlating our sample with X-ray catalogs. We looked for X-ray detections in the Chandra 2Ms X-ray point source catalog \citep{Alex03} in GOODS-N. We found 4 X-ray detections out of 58 candidates, within a 2$\arcsec$ search radius. The amount of H$\alpha$ flux concentrated in these sources is 10\% of the total flux in the whole GOODS-N sample, whereas their contribution to the number of galaxies is 8\% (4/58). This result is in good agreement with \cite{Do06} who found an AGN upper limit contribution of 9.5\% to the flux density. \cite{Gallego95} found higher values for the UCM local sample: 10\% in number and 15\% in flux density. In any case, it is important to notice that, although X-ray emission primarily come from the AGN, H$\alpha$ emission could come from a mixture of star forming processes and AGN activity. Hence, the fraction of H$\alpha$ flux concentrated in the X-ray detected sources is an upper limit to the H$\alpha$ flux coming from AGN activity. These X-ray catalogs could be missing very obscured AGN. We have checked the MIR SED of all our objects and none of them would qualify as a power-law galaxy (i.e., a heavily extincted AGN; see, e.g., \citealt{Alo06}). Still, even for the X-ray emitters, it would be impossible to quantify (with the data in our hands) whether the AGN or star formation dominate the H $\alpha$ (or MIR) emission.

1139: 

1140: 

1141: We compare our result with other SFRd measured via H$\alpha$ line flux

1142: in figure~\ref{sfrd_fig}. We took the different values from

1143: \cite{Hop04} except values at z=0.24 and z=0.4 that were taken from

1144: \cite{P05} and the result at z$\sim$0.82 by \cite{Do06}. The

1145: \cite{Pet01} value, obtained via H$\beta$, is also shown. In the

1146: \cite{Hop04} compilation, all SFRd values were corrected for

1147: extinction using a SFR dependent obscuration when the LF was available

1148: and the correction by the original authors (if any) was overall and

1149: SFR independent. If the LF was not available and no correction was

1150: made by the original authors, a mean obscuration correction of

1151: A(H$\alpha$)=1 mag was applied.

1152: 

1153: \begin{figure}

1154: \includegraphics[width=8.5cm]{f12.eps}

1155: \caption{\label{sfrd_fig} Evolution of the star formation rate 

1156: density with redshift (scale is given by log (1+z)) for estimations

1157: based on H$\alpha$ measurements. The dark pentagon is this work

1158: result. Other H$\alpha$ measurements come from \cite{Gallego95} and

1159: \cite{PG03} (filled circles), \cite{Sul00} (triangle), \cite{P05}

1160: (empty squares), \cite{Tresse98} and \cite{Tresse02} (filled squares),

1161: \cite{Do06} (inverted filled triangle), \cite{gla99} (inverted

1162: triangle), \cite{Yan99} and \cite{Hop00} (empty circles), \cite{Mor00}

1163: (filled triangle), and \cite{Pet01} (star).}

1164: \end{figure}

1165: 

1166: 

1167: The closest values in redshift are those by \cite{Tresse02},

1168: \cite{Do06}, and \cite{gla99}. Our result is systematically higher,

1169: about a factor of $\sim$1.5. However in the figure we can see that the

1170: difference between \cite{Tresse02} and this work could be an evolution

1171: effect, as they follow the general trend in redshift (evolution will

1172: be discussed in more detail in Section~\ref{sfrd_evo}). The other two

1173: points fall off the general trend. \cite{Do06} value was corrected for

1174: incompleteness by a factor of $\sim$3 due to the inherent difficulty

1175: of multi-object fiber spectroscopy observations. Only 9 out of 38

1176: galaxies observed were clearly detected ($\geq$5$\sigma$) and the

1177: others were stacked in order to get some information about H$\alpha$

1178: low luminosity objects. They also had to apply aperture corrections,

1179: with a mean value of 2.4 but with some individual values above 4. In

1180: spite of all the efforts they put in to correct for incompleteness and

1181: flux loss, they could still be missing an important fraction of flux

1182: density. Another bias that could have affected their result is that

1183: the selection was made in the {\em R} band, which samples

1184: $\sim$3600\AA\ rest-frame (i.e. the {\em U} band), taking the

1185: precaution to select only targets with identified emission-lines. The

1186: {\em U}-band, although a good tracer of star forming galaxies

1187: \citep{Mous06}, is not a direct tracer as can be the rest-frame UV. In

1188: addition, as pointed out in Section~\ref{ha_lf}, 20\% of objects with

1189: no [O{\sc ii}]$\lambda$3727 emission have H$\alpha$ emission in the

1190: SDSS, so they could be missing a fraction of H$\alpha$

1191: emitters. \cite{gla99} SFRd is $\sim$2 times lower than our

1192: value. However they only detected 8 galaxies with H$\alpha$ in

1193: emission, which could severely affect their results.

1194: 

1195: \subsection{Star formation rate density evolution}

1196: \label{sfrd_evo}

1197: It is obvious from figure~\ref{sfrd_fig} that a decrease in SFRd has

1198: occurred from z$\sim$1 to the local Universe, being the latter

1199: $\sim$10 times less active forming stars. It is common to parametrize

1200: the evolution of the SFR density with a power law:

1201: $\dot\rho_{*}\propto (1+z)^{\beta}$.

1202: 

1203: Combining just the H$\alpha$-based SFR densities obtained by our group

1204: at $z=0.02$, $z=0.24$, $z=0.40$ and $z=0.84$ for H$\alpha$-selected

1205: samples, we obtain an evolution of the cosmic SFR density

1206: $\propto(1+z)^\beta$ where $\beta=3.8\pm0.5$. The fitted power law is

1207: shown in figure~\ref{sfrd_evo_fig}. This $\beta$ value is similar to

1208: the one estimated by \cite{Tresse02} using H$\alpha$ observations, and

1209: by

1210: \cite{PG05} for a thermal IR-selected sample. However, \cite{Tresse02} 

1211: value was calculated for an Einstein-de Sitter cosmology with H$_{\rm

1212: 0}$=50 km$^{-1}$. The cosmology change softens this value to $\approx$

1213: 3.5 \citep{Do06}. Thus, we find a slightly higher value but still

1214: compatible within errors.  Our value is also comparable to that of

1215: \cite{Hop04} who used data obtained with multiple star formation

1216: tracers and obtained $\beta$=3.19$\pm$0.26 for a luminosity dependent

1217: obscuration correction.

1218: 

1219: \begin{figure}

1220: \includegraphics[width=9cm]{f13.eps}

1221: \caption{\label{sfrd_evo_fig} Evolution of SFRd measured by the UCM 

1222: through the H$\alpha$ line (\citealp{Gallego95}, \citealp{PG03},

1223: \citealp{P05} and this work). The line represents the best fit to a

1224: law $\propto$(1+z)$^\beta$ with $\beta$=3.8.}

1225: \end{figure}

1226: 

1227: More interesting is to compare the evolution of the SFRd obtained

1228: through different estimators. In Figure~\ref{sfrd_evo_mw_fig} we plot

1229: the SFRd history obtained in H$\alpha$, IR \citep[from ][]{PG05} and

1230: UV \citep[from ][]{Schimi05}.  H$\alpha$-based SFRd values are

1231: corrected for reddening while the UV values are not. \cite{PG05}

1232: obtained $\beta$=3.98$\pm$0.22 for a IR selected sample and

1233: \cite{Schimi05} obtained $\beta$=2.5$\pm$0.7, both up to z=1. Our

1234: H$\alpha$ measurement agrees quite well with that obtained with the IR

1235: sample. The UV slope is significantly lower, which could be caused by

1236: an evolution in the extinction properties with redshift, in the case

1237: the populations selected with each method were mainly the same.

1238: 

1239: 

1240: 

1241: There is another interesting question that arises when comparing

1242: populations selected with different observables: are we selecting the

1243: same objects or are there substantial differences?

1244: 

1245: To answer this question, we have considered the galaxies detected by

1246: MIPS and having a reliable spectroscopic redshift. There are 11, 2 and

1247: 18 objects in Groth2, Groth3 and GOODS-N respectively that are within

1248: our filter redshift range, are detected by MIPS, but not selected in

1249: our survey. This imply a 17\%, 8\% and 47\% of the total 24$\mu$m flux

1250: in spectroscopically confirmed MIPS galaxies (at z$\sim$0.84),

1251: corresponding the lowest fraction to our deepest field and the high

1252: fraction to our shallowest filed. Thus, the extinction in these

1253: objects could make their H$\alpha$ flux fall below our detection

1254: limits, being worst for the shallower fields. However, the fraction in

1255: GOODS-N is still quite high to be explained by the different field

1256: depths. The explanation comes from the different extinction for the

1257: objects in these fields. Whereas for the Groth fields we have a mean

1258: $\overline{A}(H\alpha)$=1.75 mag, somewhat higher than the mean value

1259: for the whole sample, for the GOODS-N field this value is

1260: $\overline{A}(H\alpha)$=3 mag.

1261: 

1262: On the other hand, we are missing 11, 2 and 22 objects detected in the

1263: GALEX NUV-band which is very close to rest-frame FUV (a good estimator

1264: of the SFR). Most of these objects are missed because they fall below

1265: our detection limit. GALEX reach smaller SFRs, but only in the case of

1266: low attenuation. However, some objects show UV emission corresponding

1267: to a star formation rate that could be selected by our method. These

1268: missed objects could also be post starburst (that over predict current

1269: star formation) although there is no H$\alpha$ emission.

1270: 

1271: \begin{figure}

1272: \includegraphics[width=8.5cm]{f14.eps}

1273: \caption{\label{sfrd_evo_mw_fig} Evolution of the star formation rate 

1274: density with redshift for estimations based on H$\alpha$, IR and FUV

1275: measurements. The dark pentagon is this work result. Other H$\alpha$

1276: measurements (open circles) come from \cite{Gallego95}, \cite{PG03},

1277: \cite{Sul00}, \cite{P05}, \cite{Tresse98}, \cite{Tresse02},

1278: \cite{Do06}, \cite{gla99}, \cite{Yan99}, \cite{Hop00}, \cite{Mor00}

1279: and \cite{Pet01}. IR measurements \citep{PG05} are represented as open

1280: triangles and FUV estimations \citep{Schimi05} as open squares. The

1281: derived evolution law is represented as a thick line for H$\alpha$, a

1282: dashed line for UV and a dot-dashed line for IR. }

1283: \end{figure}

1284: 

1285: The opposite case is also present. We detect 23 (41\%), 32 (63\%) and

1286: 20 (34\%) objects (including those with only photometric redshift)

1287: that do not show MIPS 24$\mu$m emission. The mean star formation rate

1288: for these objects is 2.8, 2.2 and 2.6 M$_{\odot}$ yr$^{-1}$ with mean

1289: extinctions of H$\alpha$ of 1.0, 0.9 and 1.3 mag, thus having a mean

1290: corrected star formation rate of 7.0, 5.0 and 8.6 M$_{\odot}$

1291: yr$^{-1}$. These values are below the 80\% completeness limit of the

1292: MIPS instrument in these fields: 83 $\mu$Jy which corresponds to

1293: $\sim$10 M$_{\odot}$ yr$^{-1}$ \citep{PG05}. When considering the UV

1294: emission there are 37(66\%), 28(55\%) and 26(44\%) objects not

1295: detected in the GALEX NUV band. The mean H$\alpha$ star formation rate

1296: for these objects is 2.7, 2.5 and 3.45 M$_{\odot}$ yr$^{-1}$ with the

1297: following mean extinctions: 1.5, 1.5 and 1.6 magnitudes in

1298: H$\alpha$. If we translate these SFRs to observed SFRs in the UV we

1299: obtain 1.4, 1.34 and 1.7 M$_{\odot}$ yr$^{-1}$. These low SFRs are

1300: similar to the detection limit for the GALEX NUV-band ($\sim$1.5

1301: M$_{\odot}$ yr$^{-1}$) at z=0.84 based on the analysis of the GALEX

1302: catalog.

1303: 

1304: As we have shown, the most significant loss is the FIR emitters

1305: because they have high SFRs, but lie below our detection limit due to

1306: the presence of dust. We notice that although we are losing a fraction

1307: of the FIR objects our completeness correction is also recovering a

1308: fraction of them. The UV objects not recovered in our sample are very

1309: faint and contribute to the low luminosity regime of the luminosity

1310: function. On the other hand, FIR and UV surveys miss a significant

1311: fraction of objects. In the case of FIR it is worth to notice that,

1312: although we are missing a fraction of star forming galaxies, we obtain

1313: a very similar SFRd value to that of \cite{PG05} (even higher). Hence, the

1314: objects not detected by MIPS with lower SFRs are playing a more

1315: important role than that estimated by \cite{PG05}. We conclude that

1316: our work is complementary to FIR and UV surveys, as it goes fainter

1317: than FIR detection limits and is not as affected by extinction as the

1318: UV.

1319: 

1320: \section{Summary and conclusions}

1321: Using an H$\alpha$ selected sample of star-forming galaxies we have

1322: estimated the H$\alpha$ luminosity function for the Universe at

1323: $z=0.84$. This work is the continuation of previous surveys where our

1324: group used the H$\alpha$ emission to select representative samples of

1325: star-forming galaxies at intermediate redshifts.  We argue that, since

1326: the H$\alpha$ emission provides a good estimate of the instantaneous

1327: star formation, the galaxies have been selected in a homogeneous way

1328: up to z$\sim$1 by their current SFR. Therefore, we can use the

1329: H$\alpha$ luminosity function to determine the ``current SFR

1330: function'' describing the number of star-forming galaxies as a

1331: function of their SFR. Integrating over all H$\alpha$ luminosities (or

1332: SFRs) we determine the current SFR density for galaxies.

1333: 

1334: A total of 165 objects have been selected as H$\alpha$ emitters using

1335: the narrow band technique. We have tested the reliability of our

1336: emission-line candidates in three different ways: 1) analyzing the use

1337: of photometric apertures of different sizes; 2) carrying out a

1338: star-galaxy segregation; and 3) estimating photometric redshifts.

1339: 

1340: Line luminosities have been corrected for Nitrogen contribution and

1341: dust reddening. To correct for Nitrogen flux contamination, we used

1342: the SDSS sample to estimate the mean [NII] to H$\alpha$ flux ratio for

1343: a given EW(H$\alpha$+[NII]). For the dust reddening correction, we

1344: proceeded in several steps: for the objects with MIPS detection, we

1345: estimated the total infrared luminosity and a synthetic FUV flux from

1346: templates and then compute the IRX ratio, which is related to the

1347: extinction in the FUV band. If the object was not detected in MIPS we

1348: used the UV slope, given by FUV-NUV \citep{Gil06}. We found a mean

1349: extinction for the whole sample A(H$\alpha$)=1.45 mag, ranging from 0

1350: to 4.16.

1351: 

1352: We performed simulations to determine the limiting flux and

1353: completeness corrections. The limiting fluxes vary from field to field

1354: from 8$\times10^{-17}$ erg\,s$^{-1}$\,cm$^{-2}$ to 14 $\times

1355: 10^{-17}$ erg\,s$^{-1}$\,cm$^{-2}$ for a 70\% completeness level. The

1356: completeness correction was applied to the computation of the

1357: extinction not corrected and corrected LF.

1358: 

1359: We computed the observed LF not corrected for extinction, obtaining

1360: the following parameters when fitting to a Schechter function:

1361: $\phi^*$=$10^{-1.74\pm0.11}$ Mpc$^{-3}$, L$^*=10^{41.69\pm0.07}$ erg

1362: s$^{-1}$. We fixed the low-luminosity slope $\alpha$=-1.35 because our

1363: LF was not deep enough. Our LF has higher density than that of

1364: \cite{Tresse02}, that could be explained by an evolutionary effect due

1365: to the different mean redshifts explored or by the selection method in

1366: each case: directly by H$\alpha$ in this work whereas they had to use

1367: the $I$-band and spectroscopic redshifts. \cite{Yan99} and

1368: \cite{Hop00} LFs extend to higher luminosities than ours. A possible

1369: explanation could be that these authors surveyed a higher redshift

1370: range (up to z$\sim$1.9), where star forming galaxies properties could

1371: be significantly different than at z$\sim$0.84.

1372: 

1373: The LF corrected for extinction and field to field variance yielded:

1374: $\alpha$=-1.34$\pm$0.18, $\phi^*$=10$^{-2.76\pm0.32}$ Mpc$^{-3}$ and

1375: $L^*$=10$^{42.97\pm0.27}$ erg s$^{-1}$. The LF extends now to

1376: similar luminosities than \cite{Hop00} LF, although, as in the

1377: original work no extinction correction was applied, we applied a mean

1378: correction A(H$\alpha$)=1. However, this mean correction could lead to

1379: subestimate L$^*$, as the highest attenuated sources wouldn't move to

1380: their actual high luminosities. On the other hand, we found a mean

1381: attenuation for our sample A(H$\alpha$)=1.45 whereas we have applied

1382: the typical mean correction for the \cite{Hop00} LF.

1383: 

1384: Analyzing each field independently and compared to the mean density of

1385: galaxies we found that there is an overabundance factor of $\sim$2.8,

1386: $\sim$1.08 and $\sim$2.0 for Groth2, Groth3 and GOODS-N fields

1387: respectively.

1388: 

1389: The SFRd derived from the extinction and field to field variance

1390: corrected LF is $\dot\rho_{*}$=0.17$^{+0.03}_{-0.03}$ M$_{\odot}$

1391: yr$^{-1}$ Mpc$^{-3}$. The strong increase from z=0. to z$\sim$1 found

1392: in other surveys is confirmed. Combining just the H$\alpha$-based SFR

1393: densities obtained by our group from $z=0.02$, $z=0.24$, $z=0.40$ and

1394: $z=0.84$ H$\alpha$-selected samples, we obtain an evolution of the

1395: cosmic SFR density $\propto(1+z)^\beta$ where $\beta=3.8\pm0.5$. This

1396: $\beta$ value is similar to the one estimated by \cite{Tresse02} and

1397: by \cite{PG05} for thermal IR-selected samples.

1398: 

1399: The H$\alpha$ approach is complementary to FIR and UV surveys as it

1400: reach fainter SFRs than FIR surveys and is less affected by extinction

1401: than UV surveys. The fraction of objects detected in FIR not detected

1402: by H$\alpha$ is around $\sim$15\% unless very high extincted objects

1403: are present, as in the case of GOODS-N.

1404: 

1405: \acknowledgments

1406: We thank Nicolas Cardiel and Armando Gil de Paz for advice and useful comments.

1407: V.\ Villar acknowledges the receipt of a \emph{Formaci\'on de

1408: Personal Investigador} fellowship from

1409: the Spanish Ministerio de Educaci\'on y Ciencia.

1410: This work was supported in part by the Spanish \emph{Plan Nacional

1411: de Astronom\'{\i}a y Astrof\'{\i}sica} under grant AYA2003-01676 and AYA2006-02358.

1412: This work was supported in part by a generous grant of the \emph{New del Amo

1413: Foundation Program}.

1414: 

1415: {\it Facilities:} \facility{CAO:3.5m} \facility{Keck:II}

1416: \bibliographystyle{apj}

1417: \bibliography{referencias_paper,apj-jour}

1418: 

1419: \clearpage

1420: \setcounter{table}{2}

1421: \begin{deluxetable}{ccccrrcrr}

1422: \tabletypesize{\scriptsize}

1423: \tablecaption{Observed sample}

1424: \tablehead{

1425: \colhead{ID} & \colhead{$\alpha$ (J2000)} & \colhead{$\delta$ (J2000)} & \colhead{J}& \colhead{f$_{H_{\alpha}}$} & \colhead{EW(H$\alpha$)} & \colhead{A(H$\alpha$)} & \colhead{SFR$_{obs}$} & \colhead{SFR$_{cor}$} \\

1426: \colhead{(1)} & \colhead{(2)} & \colhead{(3)} & \colhead{(4)}& \colhead{(5)} & \colhead{(6)} & \colhead{(7)} & \colhead{(8)}& \colhead{(9)}

1427: }

1428: \startdata

1429: f2\_481&214.39872&52.36067&20.22$\pm$0.08&23.6$\pm$4.7&67&0.45&4.6&6.9\\

1430: f2\_566&214.49003&52.36594&21.62$\pm$0.12&15.4$\pm$3.9&209&1.81&3.2&16.9\\

1431: f2\_592&214.28326&52.36750&21.88$\pm$0.16&14.1$\pm$4.4&241&1.60&3.0&13.0\\

1432: f2\_688&214.26028&52.37098&21.76$\pm$0.14&14.1$\pm$4.4&121&3.04&2.9&48.4\\

1433: f2\_959&214.44781&52.37737&19.15$\pm$0.03&6.8$\pm$3.4&17&1.97&1.2&7.4\\

1434: f2\_997&214.43842&52.38060&21.20$\pm$0.09&14.2$\pm$4.0&132&1.04&2.8&7.3\\

1435: f2\_1238&214.40850&52.38895&19.95$\pm$0.05&20.9$\pm$5.6&34&2.02&3.9&24.9\\

1436: f2\_1461&214.21682&52.39630&18.95$\pm$0.02&53.8$\pm$7.8&33&2.08&10.0&68.1\\

1437: f2\_1463&214.40130&52.39658&22.26$\pm$0.27&11.1$\pm$4.1&282&1.06&2.4&6.4\\

1438: f2\_1650&214.49389&52.40403&21.12$\pm$0.09&25.6$\pm$6.0&111&1.25&5.3&16.6\\

1439: f2\_1694&214.36583&52.40625&21.12$\pm$0.12&14.8$\pm$3.8&179&1.41&3.0&11.0\\

1440: f2\_1822&214.20626&52.41180&22.48$\pm$0.25&9.2$\pm$3.7&273&0.07&2.0&2.1\\

1441: f2\_2091&214.50282&52.41996&21.82$\pm$0.19&30.1$\pm$4.3&513&0.08&7.2&7.7\\

1442: f2\_2121&214.51324&52.42082&21.70$\pm$0.17&12.8$\pm$4.1&231&1.65&2.7&12.2\\

1443: f2\_2219&214.25459&52.42487&21.40$\pm$0.11&20.0$\pm$4.6&115&0.72&4.1&8.0\\

1444: f2\_2362&214.50548&52.42970&21.57$\pm$0.12&16.0$\pm$4.2&114&1.12&3.3&9.2\\

1445: f2\_2522&214.46666&52.43515&23.48$\pm$0.32&10.9$\pm$3.2&611&0.23&2.7&3.3\\

1446: f2\_2539&214.50109&52.43483&20.83$\pm$0.08&8.1$\pm$4.3&32&3.33&1.5&32.4\\

1447: f2\_2556&214.30817&52.43651&22.47$\pm$0.21&17.1$\pm$4.3&450&0.07&4.0&4.3\\

1448: f2\_2598&214.49855&52.43669&20.71$\pm$0.07&18.7$\pm$5.6&59&0.92&3.6&8.3\\

1449: f2\_2668&214.31872&52.43925&20.84$\pm$0.13&22.4$\pm$5.1&189&1.34&4.6&15.6\\

1450: f2\_2706&214.50345&52.44056&20.67$\pm$0.09&40.4$\pm$7.0&223&1.04&8.4&22.0\\

1451: f2\_2993&214.23302&52.44979&19.86$\pm$0.06&20.6$\pm$5.8&35&1.75&3.8&19.3\\

1452: f2\_3093&214.24402&52.45767&22.55$\pm$0.29&9.3$\pm$3.7&275&0.45&2.0&3.0\\

1453: f2\_3166&214.21455&52.45995&21.72$\pm$0.21&16.0$\pm$5.2&292&0.84&3.5&7.6\\

1454: f2\_3194&214.24633&52.45995&21.16$\pm$0.12&10.0$\pm$2.9&74&2.46&2.0&18.9\\

1455: f2\_3345&214.25875&52.46516&20.67$\pm$0.07&16.1$\pm$4.2&52&2.75&3.1&38.7\\

1456: f2\_3358&214.24843&52.46479&20.64$\pm$0.09&22.4$\pm$5.4&74&2.31&4.4&36.9\\

1457: f2\_3458&214.20323&52.46636&19.19$\pm$0.03&27.5$\pm$6.6&24&1.47&5.0&19.5\\

1458: f2\_3460&214.53003&52.46774&19.95$\pm$0.05&18.8$\pm$5.2&37&1.43&3.5&13.1\\

1459: f2\_3507&214.21475&52.47060&19.48$\pm$0.04&20.0$\pm$5.1&26&1.98&3.7&22.8\\

1460: f2\_3614&214.41869&52.47512&22.75$\pm$0.26&7.7$\pm$2.5&342&2.24&1.7&13.7\\

1461: f2\_3726&214.36825&52.47954&22.64$\pm$0.33&8.3$\pm$3.1&340&0.34&1.9&2.6\\

1462: f2\_3742&214.32370&52.47898&21.12$\pm$0.11&10.4$\pm$3.4&71&2.54&2.0&20.9\\

1463: f2\_3786&214.42755&52.47950&19.44$\pm$0.04&60.5$\pm$7.4&65&2.16&11.7&86.1\\

1464: f2\_3992&214.28128&52.48944&20.03$\pm$0.05&25.8$\pm$6.3&43&1.27&4.9&15.6\\

1465: f2\_4054&214.19562&52.49363&22.31$\pm$0.22&8.9$\pm$3.2&287&0.04&1.9&2.0\\

1466: f2\_4179&214.45653&52.49805&23.14$\pm$0.32&12.0$\pm$3.9&549&1.41&2.9&10.6\\

1467: f2\_4323&214.25305&52.50370&21.13$\pm$0.13&9.7$\pm$3.4&77&2.34&1.9&16.5\\

1468: f2\_4422&214.53702&52.50786&23.18$\pm$0.32&20.1$\pm$3.4&1077&0.38&5.1&7.2\\

1469: f2\_4692&214.23025&52.51940&21.68$\pm$0.14&6.1$\pm$2.8&129&1.02&1.2&3.0\\

1470: f2\_4713&214.29729&52.51900&20.15$\pm$0.07&25.5$\pm$5.2&63&0.85&4.9&10.8\\

1471: f2\_4769&214.28798&52.52265&23.26$\pm$0.28&8.9$\pm$3.2&505&3.74&2.1&65.8\\

1472: f2\_4937&214.30355&52.52710&19.77$\pm$0.05&12.6$\pm$4.1&27&2.76&2.3&29.2\\

1473: f2\_4986&214.43021&52.52997&22.48$\pm$0.21&11.4$\pm$3.5&190&1.16&2.6&7.5\\

1474: f2\_5405&214.53927&52.54503&20.83$\pm$0.10&18.1$\pm$5.6&69&1.12&3.5&9.9\\

1475: f2\_5408&214.53529&52.54495&20.28$\pm$0.06&79.5$\pm$6.7&167&1.02&17.5&45.1\\

1476: f2\_5459&214.47510&52.54793&22.12$\pm$0.18&7.7$\pm$4.7&150&0.75&1.5&3.1\\

1477: f2\_5508&214.28716&52.56664&21.12$\pm$0.10&35.9$\pm$6.1&154&0.63&7.8&14.0\\

1478: f2\_5584&214.30300&52.56433&21.20$\pm$0.10&20.0$\pm$5.2&95&1.26&4.0&12.8\\

1479: f2\_5594&214.55351&52.56617&22.04$\pm$0.25&31.9$\pm$8.8&646&0.44&7.8&11.7\\

1480: f2\_5639&214.28714&52.56724&21.16$\pm$0.10&11.2$\pm$3.8&68&1.04&2.2&5.6\\

1481: f2\_5959&214.29914&52.55195&21.79$\pm$0.20&13.8$\pm$4.2&136&3.04&2.9&48.2\\

1482: f2\_5993&214.55324&52.54776&19.99$\pm$0.06&18.7$\pm$9.8&42&2.88&3.5&50.0\\

1483: f2\_7462&214.28432&52.56822&20.50$\pm$0.06&27.1$\pm$5.0&77&1.26&5.3&17.1\\

1484: f3\_530&214.35434&52.58357&21.47$\pm$0.12&27.0$\pm$6.2&288&0.68&5.9&11.0\\

1485: f3\_578&214.54496&52.58483&22.35$\pm$0.18&12.0$\pm$4.3&280&1.31&2.6&8.7\\

1486: f3\_863&214.60952&52.58966&19.46$\pm$0.03&17.5$\pm$4.8&20&1.34&3.2&10.9\\

1487: f3\_1282&214.66531&52.60264&19.19$\pm$0.02&29.3$\pm$6.3&25&1.10&5.3&14.8\\

1488: f3\_1316&214.51123&52.60716&21.41$\pm$0.09&16.3$\pm$3.9&94&4.13&3.3&147.8\\

1489: f3\_1344&214.69324&52.60773&23.53$\pm$0.31&11.1$\pm$3.3&599&1.25&2.7&8.6\\

1490: f3\_1390&214.37272&52.60940&21.53$\pm$0.10&15.2$\pm$3.7&99&1.93&3.1&18.1\\

1491: f3\_2440&214.71246&52.64082&20.89$\pm$0.07&13.8$\pm$4.0&98&1.04&2.6&6.9\\

1492: f3\_2588&214.40606&52.64745&21.81$\pm$0.12&10.2$\pm$3.6&158&1.06&2.0&5.4\\

1493: f3\_2634&214.65503&52.64757&20.41$\pm$0.05&23.0$\pm$5.4&98&2.02&4.4&28.2\\

1494: f3\_2635&214.70055&52.64851&22.29$\pm$0.14&10.5$\pm$3.2&245&1.00&2.2&5.6\\

1495: f3\_2694&214.40516&52.65156&21.87$\pm$0.17&9.4$\pm$2.8&246&0.58&2.0&3.4\\

1496: f3\_2919&214.66029&52.65804&21.60$\pm$0.12&7.4$\pm$2.7&123&1.14&1.4&4.1\\

1497: f3\_3041&214.75493&52.66071&21.29$\pm$0.11&10.9$\pm$4.6&103&0.08&2.1&2.2\\

1498: f3\_3112&214.42338&52.66452&21.34$\pm$0.12&4.7$\pm$2.2&111&0.92&0.9&2.1\\

1499: f3\_3417&214.57087&52.67417&20.32$\pm$0.04&16.2$\pm$3.7&41&1.86&3.0&16.9\\

1500: f3\_3525&214.41201&52.67866&21.26$\pm$0.10&8.6$\pm$2.9&136&1.06&1.7&4.5\\

1501: \enddata

1502: \tablecomments{(1) Identification (2) RA (J2000) (3) DEC (J2000) (4) J-magnitude (Vega) (5) Line flux (10$^{-17} erg s^{-1} cm^{-2}$ (6) Restframe equivalent width (\AA) (7) Extinction in magnitudes (8) SFR not corrected for extinction (M$_{\odot}$ yr$^{-1}$) (9) SFR corrected for extinction (M$_{\odot}$ yr$^{-1}$).}

1503: \label{tab_objects}

1504: \end{deluxetable}

1505: 

1506: \setcounter{table}{2}

1507: \begin{deluxetable}{ccccrrcrr}

1508: \tabletypesize{\scriptsize}

1509: \tablecaption{Observed sample}

1510: \tablehead{

1511: \colhead{ID} & \colhead{$\alpha$ (J2000)} & \colhead{$\delta$ (J2000)} & \colhead{J}& \colhead{f$_{H_{\alpha}}$} & \colhead{EW(H$\alpha$)} & \colhead{A(H$\alpha$)} & \colhead{SFR$_{obs}$} & \colhead{SFR$_{cor}$} \\

1512: \colhead{(1)} & \colhead{(2)} & \colhead{(3)} & \colhead{(4)}& \colhead{(5)} & \colhead{(6)} & \colhead{(7)} & \colhead{(8)}& \colhead{(9)}

1513: }

1514: \startdata

1515: 

1516: f3\_3675&214.42497&52.68249&19.15$\pm$0.02&28.2$\pm$6.7&22&1.29&5.1&16.7\\

1517: f3\_3694&214.40237&52.68428&20.77$\pm$0.06&20.6$\pm$4.4&69&0.66&4.0&7.3\\

1518: f3\_3972&214.54982&52.69515&21.85$\pm$0.15&11.8$\pm$3.6&186&0.08&2.4&2.6\\

1519: f3\_4116&214.42933&52.70046&22.28$\pm$0.18&8.3$\pm$2.8&184&1.54&1.7&6.9\\

1520: f3\_4119&214.75000&52.69880&20.75$\pm$0.07&12.6$\pm$4.3&79&2.20&2.4&18.1\\

1521: f3\_4133&214.56011&52.70091&22.28$\pm$0.17&9.9$\pm$3.1&248&0.08&2.1&2.3\\

1522: f3\_4222&214.48923&52.70217&19.91$\pm$0.04&22.2$\pm$5.6&36&0.52&4.2&6.7\\

1523: f3\_4368&214.42725&52.70791&20.10$\pm$0.04&21.4$\pm$4.2&46&3.75&4.0&128.2\\

1524: f3\_4659&214.34934&52.71719&20.37$\pm$0.06&6.5$\pm$6.3&15&4.04&1.1&47.0\\

1525: f3\_4741&214.41970&52.72037&22.27$\pm$0.15&20.5$\pm$4.2&468&0.08&4.8&5.2\\

1526: f3\_4858&214.37258&52.72191&19.58$\pm$0.03&24.2$\pm$5.1&30&4.17&4.5&207.6\\

1527: f3\_4891&214.40727&52.72368&19.53$\pm$0.02&25.2$\pm$5.1&50&1.25&4.6&14.6\\

1528: f3\_4901&214.54208&52.72290&20.72$\pm$0.08&6.9$\pm$3.0&27&1.96&1.3&7.7\\

1529: f3\_4916&214.63473&52.72382&20.13$\pm$0.04&13.1$\pm$3.9&27&2.38&2.4&21.5\\

1530: f3\_4927&214.36864&52.72580&21.69$\pm$0.12&9.1$\pm$3.3&163&0.07&1.8&1.9\\

1531: f3\_4955&214.40295&52.72560&19.63$\pm$0.03&20.5$\pm$5.2&25&3.58&3.8&101.4\\

1532: f3\_5080&214.60435&52.73046&20.65$\pm$0.08&8.6$\pm$3.0&79&0.04&1.6&1.7\\

1533: f3\_5362&214.71217&52.74022&21.54$\pm$0.16&11.6$\pm$4.0&185&1.31&2.4&7.9\\

1534: f3\_5603&214.56435&52.74962&22.61$\pm$0.28&6.7$\pm$2.5&319&0.95&1.5&3.6\\

1535: f3\_5746&214.75617&52.75273&19.48$\pm$0.03&30.6$\pm$6.4&30&1.65&5.7&25.9\\

1536: f3\_5781&214.64338&52.75524&21.76$\pm$0.12&14.7$\pm$4.0&117&1.34&3.1&10.5\\

1537: f3\_5785&214.50875&52.75358&18.97$\pm$0.02&7.3$\pm$3.2&17&2.33&1.3&11.0\\

1538: f3\_5808&214.48128&52.75600&21.25$\pm$0.09&11.7$\pm$2.9&159&0.84&2.3&5.0\\

1539: f3\_5857&214.67298&52.75791&21.91$\pm$0.13&6.0$\pm$2.4&127&1.00&1.2&2.9\\

1540: f3\_5893&214.75849&52.75765&20.74$\pm$0.08&6.1$\pm$2.7&50&1.14&1.2&3.3\\

1541: f3\_6178&214.38307&52.77068&22.23$\pm$0.14&12.1$\pm$3.4&288&0.64&2.6&4.7\\

1542: f3\_6456&214.36018&52.82435&22.15$\pm$0.15&10.2$\pm$3.7&244&0.38&2.2&3.1\\

1543: f3\_6483&214.39506&52.82217&21.80$\pm$0.19&4.1$\pm$2.2&146&1.41&0.8&3.0\\

1544: f3\_6553&214.46315&52.82027&22.48$\pm$0.25&8.7$\pm$3.2&285&1.25&1.9&6.0\\

1545: f3\_7108&214.44095&52.80083&21.42$\pm$0.13&6.4$\pm$2.5&127&0.04&1.2&1.3\\

1546: f3\_7402&214.51934&52.79224&20.14$\pm$0.06&4.8$\pm$2.2&89&2.06&0.9&6.1\\

1547: f3\_7493&214.47743&52.79039&20.76$\pm$0.06&7.4$\pm$4.1&46&0.30&1.4&1.8\\

1548: f3\_7702&214.39626&52.78315&19.75$\pm$0.04&33.5$\pm$6.4&83&1.65&6.3&29.0\\

1549: g1\_375&189.12078&62.08836&21.86$\pm$0.30&6.8$\pm$3.5&67&1.12&1.3&3.7\\

1550: g1\_456&189.24120&62.09161&22.12$\pm$0.29&5.2$\pm$2.4&196&1.19&1.1&3.2\\

1551: g1\_465&189.19594&62.09143&20.99$\pm$0.12&15.5$\pm$4.1&130&1.95&3.0&18.1\\

1552: g1\_496&188.92963&62.09167&19.12$\pm$0.05&38.9$\pm$9.1&29&2.05&7.2&47.6\\

1553: g1\_516&189.42234&62.09258&20.77$\pm$0.12&12.8$\pm$4.7&54&1.33&2.4&8.3\\

1554: g1\_713&189.22507&62.10214&20.36$\pm$0.12&12.4$\pm$3.2&47&2.65&2.3&26.9\\

1555: g1\_908&189.21466&62.11220&19.97$\pm$0.08&35.5$\pm$5.1&115&2.31&6.8&57.2\\

1556: g1\_1034&189.30751&62.11791&20.88$\pm$0.12&15.0$\pm$3.5&66&0.58&2.9&5.0\\

1557: g1\_1082&189.14916&62.12032&21.44$\pm$0.26&6.8$\pm$2.4&124&1.37&1.3&4.7\\

1558: g1\_1159&189.15771&62.12317&20.66$\pm$0.12&14.6$\pm$3.1&64&1.27&2.8&9.1\\

1559: g1\_1516&189.00670&62.13943&20.59$\pm$0.13&14.1$\pm$3.9&50&2.01&2.7&17.0\\

1560: g1\_1665&189.12674&62.14533&20.87$\pm$0.13&21.4$\pm$3.9&86&1.58&4.2&18.2\\

1561: g1\_1735&189.12719&62.14752&20.34$\pm$0.11&27.2$\pm$4.8&79&3.31&5.4&113.1\\

1562: g1\_1995&189.06608&62.15986&20.93$\pm$0.13&23.8$\pm$4.6&178&0.79&4.8&9.9\\

1563: g1\_2073&189.39735&62.16152&20.71$\pm$0.12&13.4$\pm$4.3&66&3.53&2.6&67.0\\

1564: g1\_2141&189.16036&62.16473&18.85$\pm$0.04&39.5$\pm$6.6&22&1.14&7.2&20.5\\

1565: g1\_2198&189.11342&62.16728&20.39$\pm$0.12&18.2$\pm$3.9&105&0.48&3.5&5.4\\

1566: g1\_2205&189.35772&62.16781&20.94$\pm$0.15&17.2$\pm$5.0&171&1.88&3.4&19.5\\

1567: g1\_2339&189.23896&62.17391&20.90$\pm$0.16&5.9$\pm$2.1&78&1.43&1.1&4.1\\

1568: g1\_2387&188.95030&62.17764&22.01$\pm$0.21&12.0$\pm$3.2&233&0.79&2.5&5.2\\

1569: g1\_2450&188.94491&62.18024&20.56$\pm$0.13&12.1$\pm$3.6&43&2.75&2.3&28.5\\

1570: g1\_2537&188.93729&62.18332&21.46$\pm$0.17&16.2$\pm$3.3&134&1.74&3.4&17.0\\

1571: g1\_2669&189.00505&62.19133&21.77$\pm$0.20&10.1$\pm$2.8&170&2.36&2.0&17.7\\

1572: g1\_2815&189.15753&62.19707&20.64$\pm$0.12&12.8$\pm$4.0&85&2.29&2.4&20.0\\

1573: g1\_2827&189.15960&62.19750&20.44$\pm$0.11&21.0$\pm$4.6&54&3.05&4.0&66.4\\

1574: g1\_2882&189.01208&62.20036&20.63$\pm$0.11&17.9$\pm$4.0&58&0.87&3.4&7.7\\

1575: g1\_3000&189.42950&62.20486&20.38$\pm$0.10&15.7$\pm$8.3&66&1.46&2.9&11.2\\

1576: g1\_3068&188.97302&62.20893&20.06$\pm$0.09&21.7$\pm$4.6&44&0.98&4.1&10.1\\

1577: g1\_3178&189.16642&62.21390&19.55$\pm$0.06&17.8$\pm$3.5&30&2.41&3.3&30.3\\

1578: g1\_3205&189.28483&62.21460&20.72$\pm$0.14&18.0$\pm$4.6&59&1.56&3.5&14.6\\

1579: g1\_3339&189.13999&62.22222&19.91$\pm$0.07&19.3$\pm$4.0&32&2.25&3.6&28.5\\

1580: g1\_3400&189.30463&62.22606&20.96$\pm$0.18&20.0$\pm$3.8&181&0.79&4.0&8.3\\

1581: g1\_3531&188.95300&62.23167&20.57$\pm$0.12&23.4$\pm$3.7&164&0.34&4.7&6.4\\

1582: g1\_3581&189.39394&62.23232&19.00$\pm$0.05&33.9$\pm$10.0&26&1.88&6.2&35.2\\

1583: g1\_3630 & 189.40551 & 62.23646 & 20.31$\pm$0.12 & 9.8  $\pm$  6.3 & 27 & 1.06 & 1.8 & 4.8\\

1584: g1\_3655&189.27626&62.25499&20.23$\pm$0.15&34.2$\pm$4.7&180&0.71&6.9&13.2\\

1585: g1\_3693&189.27757&62.25355&21.69$\pm$0.30&7.8$\pm$3.6&70&1.00&1.5&3.8\\

1586: g1\_3694&189.28492&62.25408&20.29$\pm$0.09&50.3$\pm$4.9&222&0.98&10.5&25.8\\

1587: \enddata

1588: \tablecomments{(1) Identification (2) RA (J2000) (3) DEC (J2000) (4) J-magnitude (Vega) (5) Line flux (10$^{-17} erg s^{-1} cm^{-2}$ (6) Restframe equivalent width (\AA) (7) Extinction in magnitudes (8) SFR not corrected for extinction (M$_{\odot}$ yr$^{-1}$) (9) SFR corrected for extinction (M$_{\odot}$ yr$^{-1}$).}

1589: \label{tab_objects}

1590: \end{deluxetable}

1591: 

1592: \setcounter{table}{2}

1593: \begin{deluxetable}{ccccrrcrr}

1594: \tabletypesize{\scriptsize}

1595: \tablecaption{Observed sample}

1596: \tablehead{

1597: \colhead{ID} & \colhead{$\alpha$ (J2000)} & \colhead{$\delta$ (J2000)} & \colhead{J}& \colhead{f$_{H_{\alpha}}$} & \colhead{EW(H$\alpha$)} & \colhead{A(H$\alpha$)} & \colhead{SFR$_{obs}$} & \colhead{SFR$_{cor}$} \\

1598: \colhead{(1)} & \colhead{(2)} & \colhead{(3)} & \colhead{(4)}& \colhead{(5)} & \colhead{(6)} & \colhead{(7)} & \colhead{(8)}& \colhead{(9)}

1599: }

1600: \startdata

1601: 

1602: g1\_3756&189.36488&62.24973&21.58$\pm$0.22&14.7$\pm$5.2&261&1.63&3.1&14.1\\

1603: g1\_3779&189.37307&62.24834&21.18$\pm$0.16&16.1$\pm$6.1&94&0.31&3.2&4.3\\

1604: g1\_4171&189.02571&62.32623&22.01$\pm$0.22&3.5$\pm$2.5&40&1.28&0.7&2.2\\

1605: g1\_4268&189.27285&62.33080&20.67$\pm$0.22&36.1$\pm$5.9&278&2.19&7.8&59.0\\

1606: g1\_4284&189.27174&62.33083&20.41$\pm$0.19&9.8$\pm$3.7&80&0.08&1.9&2.0\\

1607: g1\_4398&189.33542&62.32533&21.85$\pm$0.24&11.8$\pm$5.0&219&2.23&2.5&19.1\\

1608: g1\_4423&189.39726&62.32167&19.42$\pm$0.06&36.7$\pm$7.8&42&2.21&6.9&52.8\\

1609: g1\_4558&189.35632&62.31392&19.22$\pm$0.05&51.9$\pm$8.4&51&2.22&9.9&76.7\\

1610: g1\_4806&189.39811&62.30147&18.99$\pm$0.05&32.3$\pm$10.6&23&0.94&5.9&14.0\\

1611: g1\_4832&189.32014&62.30673&20.18$\pm$0.12&32.9$\pm$5.1&179&2.26&6.6&53.3\\

1612: g1\_4908&189.42394&62.29461&21.93$\pm$0.21&26.4$\pm$6.7&272&2.63&6.3&70.8\\

1613: g1\_5183&189.34373&62.28091&20.86$\pm$0.15&16.6$\pm$4.7&177&1.04&3.3&8.7\\

1614: g1\_5226&189.17678&62.27920&21.12$\pm$0.19&9.7$\pm$2.8&64&1.95&1.9&11.2\\

1615: g1\_5276&189.33590&62.27489&20.48$\pm$0.13&27.6$\pm$7.5&77&2.05&5.4&35.9\\

1616: \enddata

1617: \tablecomments{(1) Identification (2) RA (J2000) (3) DEC (J2000) (4) J-magnitude (Vega) (5) Line flux (10$^{-17} erg s^{-1} cm^{-2}$ (6) Restframe equivalent width (\AA) (7) Extinction in magnitudes (8) SFR not corrected for extinction (M$_{\odot}$ yr$^{-1}$) (9) SFR corrected for extinction (M$_{\odot}$ yr$^{-1}$).}

1618: \label{tab_objects}

1619: \end{deluxetable}

1620: 

1621: \end{document}

1622: 

1623: