1: %% The command below calls the preprint style

2: %% which will produce a one-column, single-spaced document.

3: %% Examples of commands for other substyles follow. Use

4: %% whichever is most appropriate for your purposes.

5: %%

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

7: 

8: %% manuscript produces a one-column, double-spaced document:

9: 

10: %\documentclass[manuscript]{aastex}

11: 

12: %% preprint2 produces a double-column, single-spaced document:

13: 

14: \documentclass[preprint2]{aastex}

15: 

16: 

17: %

18: % new commands

19: %

20: \newcommand\KAB{K_{\rm AB}}

21: \newcommand\Msun{M_\odot}

22: \newcommand\Ha{{\rm H}\alpha}

23: \newcommand\Hb{{\rm H}\beta}

24: %

25: %

26: 

27: %\usepackage{lscape}

28: 

29: %% Sometimes a paper's abstract is too long to fit on the

30: %% title page in preprint2 mode. When that is the case,

31: %% use the longabstract style option.

32: 

33: %% \documentclass[preprint2,longabstract]{aastex}

34: 

35: %% You can insert a short comment on the title page using the command below.

36: 

37: %%\slugcomment{accepted for publication in}

38: 

39: %% If you wish, you may supply running head information, although

40: %% this information may be modified by the editorial offices.

41: %% The left head contains a list of authors,

42: %% usually a maximum of three (otherwise use et al.).  The right

43: %% head is a modified title of up to roughly 44 characters.

44: %% Running heads will not print in the manuscript style.

45: 

46: %% \shorttitle{}

47: %% \shortauthors{}

48: 

49: %% This is the end of the preamble.  Indicate the beginning of the

50: %% paper itself with \begin{document}.

51: 

52: \begin{document}

53: 

54: \title{

55: Star Formation Rates and Metallicities of $K$-selected Star Forming

56: Galaxies at $z \sim 2$\altaffilmark{1}

57: }

58: 

59: \author{

60: Masao Hayashi \altaffilmark{2},

61: Kentaro Motohara \altaffilmark{3},

62: Kazuhiro Shimasaku \altaffilmark{4},

63: Masato Onodera \altaffilmark{5,8},

64: Yuka Katsuno Uchimoto \altaffilmark{3},

65: Nobunari Kashikawa \altaffilmark{6},

66: Makiko Yoshida \altaffilmark{2},

67: Sadanori Okamura \altaffilmark{4},

68: Chun Ly \altaffilmark{7},

69: and

70: Matthew A. Malkan \altaffilmark{7}

71: }

72: 

73: \email{hayashi@astron.s.u-tokyo.ac.jp}

74: 

75: \altaffiltext{1}{

76: Based on data collected at Subaru Telescope, which is operated by

77: the National Astronomical Observatory of Japan.

78: Use of the UKIRT 3.8-m telescope for the observations is supported by NAOJ.}

79: \altaffiltext{2}{

80: Department of Astronomy, Graduate School of Science,

81: University of Tokyo, Tokyo 113-0033, Japan}

82: \altaffiltext{3}{

83: Institute of Astronomy, Graduate School of Science, University of

84: Tokyo, Mitaka, Tokyo 181-0015, Japan}

85: \altaffiltext{4}{

86: Department of Astronomy and Research Center for the Early Universe,

87: Graduate School of Science,

88: University of Tokyo, Tokyo 113-0033, Japan}

89: \altaffiltext{5}{

90: Department of Astronomy, Yonsei University, Sinchon-dong 134,

91: Seodaemun-gu, Seoul, Korea}

92: \altaffiltext{6}{

93: Optical and Infrared Astronomy Division, National Astronomical

94: Observatory, Mitaka, Tokyo 181-8588, Japan}

95: \altaffiltext{7}{

96: Department of Physics and Astronomy, University of California at Los

97: Angeles, P. O. Box 951547,  Los Angeles, CA 90095-1547, USA}

98: \altaffiltext{8}{

99: Service d'Astrophysique, CEA Saclay, Orme des Merisiers, 91191

100: Gif-sur-Yvette Cedex, France}

101: 

102: 

103: \begin{abstract}

104:  We present spectroscopy of 15 star-forming BzK galaxies

105:  (sBzKs) with $\KAB \lesssim 23$ in the Subaru Deep Field, for which

106:  $\Ha$ and some other emission lines are detected in 0.9 to 2.3

107:  $\micron$ spectra with a resolution of $R$=500. 

108:  %

109:  Using $\Ha$ luminosities, we obtain star formation rates (SFRs), and

110:  then specific SFRs (SSFRs) dividing SFRs by stellar masses, 

111:  which are derived from SED fitting to $BVRi'z'K$ photometry. 

112:  It is found that sBzKs with higher stellar masses have larger SFRs. 

113:  A negative correlation is seen between stellar mass and SSFR,  

114:  which is consistent with the previous results for $z \sim 2$

115:  galaxies. This implies that a larger growth of stellar mass occurs in

116:  less massive galaxies.  

117:  %

118:  In addition, gas-phase oxygen abundances, 12+log(O/H), are

119:  derived from the ratio of [\ion{N}{2}]($\lambda6584$) to $\Ha$ using

120:  the N2 index method. We have found a correlation between stellar mass

121:  and oxygen abundance in the sense that more massive sBzKs tend to be

122:  more metal rich, which is qualitatively consistent with the relation

123:  for UV-selected $z \sim 2$ galaxies. However, the metallicity of the

124:  sBzKs is $\sim 0.2$ dex higher than that of UV-selected galaxies with

125:  similar stellar masses, which is significant considering the small 

126:  uncertainties.   

127:  The sBzKs in our sample have redder $R-K$ colors than the UV-selected

128:  galaxies.  

129:  %

130:  This galaxy color-dependence in the oxygen abundance may be caused by

131:  older or dustier galaxies having higher metallicities at $z \sim 2$.

132:  %

133: \end{abstract}

134: 

135: \keywords{ galaxies: abundances --- galaxies: high-redshift  ---

136: galaxies: starburst --- galaxies: formation 

137: }

138: 

139: \section{Introduction}

140: Recent studies suggest that the era of $z \sim 2$ is a turning point

141: in galaxy formation and evolution. The cosmic star formation

142: rate (SFR) begins to drop at $z \sim 1-2$ from a flat plateau at higher

143: redshifts \citep{dic03,fon03,ly07}. Significant evolution of the Hubble

144: sequence occurred at $z \sim 1-2$ \citep{kaj01}.   

145: It is also found that the number density of QSOs has a peak at 

146: $z \sim 2$ \citep{ric06}. These facts suggest that dramatic changes in

147: the galaxy population occurred at $z \sim 2$, which are important to 

148: study in further detail. 

149: 

150: Galaxy properties are determined from spectral

151: characteristics such as emission lines, absorption lines and strong

152: continuum breaks.  

153: At $z \sim 2$, the strongest of these features fall in the

154: near-ultraviolet (NUV) and near-infrared (NIR) wavebands.  

155: NUV and NIR observations are thus essential to reveal detailed

156: properties of $z \sim 2$ galaxies, but these observations encounter many 

157: difficulties, including poor sky transparency except for some

158: atmospheric windows, and strong OH airglow emission lines and thermal emission

159: from the atmosphere.

160: This is why the era of $z \sim 2$ had been called a

161: {\lq}redshift desert{\rq} until recently.  

162: However, recent advances in wide-field imaging in the NUV and NIR, along

163: with multi-object spectrographs, are beginning to fill in this

164: {\lq}desert{\rq} (e.g., Subaru Deep Field (SDF) survey, GOODS-North and

165: South, UKIDSS, GMASS, and MUSYC 

166: \citep{mccarthy99,red06b,erb06c,hay07,dad07,gra07,lan07,hal08,kri08,ly08}).  

167: 

168: A photometric selection technique using $B-z$ and $z-K$ colors has been

169: proposed to choose an unbiased sample of $z \sim 2$ galaxies (BzK

170: galaxies: \citet{dad04}). Using these colors to find galaxies with

171: a Balmer break between $z$ and $K$, the selected

172: galaxies are then classified into star forming galaxies (sBzKs) and

173: passively evolving galaxies (pBzKs). 

174: Thanks to wide and deep imaging surveys in the NIR as well as the optical, 

175: large new BzK samples are enabling us to measure their statistical

176: properties, such as clustering strength \citep{kon06,hay07,bla08}. 

177: The multiwavelength data also show that sBzKs are very actively

178: forming stars\citep{dan06,dad07}. On the other hand, 

179: due to lack of spectroscopic observations, there are relatively few

180: accurately known redshifts, which are needed   

181: to derive accurate stellar masses and star formation rates

182: from SED fitting to the multiwavelength data. Also, their physical

183: properties such as metallicity are still not well-known.

184: 

185: More spectroscopy has recently become available for 

186: $z \sim 2$ galaxies selected by their UV colors.

187: \citet{erb06a,erb06b} presented a mass-metallicity ($M$-$Z$) relation

188: and SFRs for UV-selected galaxies at $z \sim 2$ using 114 NIR spectra. 

189: They found a $M$-$Z$ correlation at $z \sim 2$ as

190: there is in the local universe, with the more massive galaxies being

191: more metal rich. Although the correlation has a similar slope to the

192: local one, it is shifted to lower metallicity by $\sim 0.3$ dex.

193: They have also found that SFRs derived from $\Ha$ luminosities are in

194: good agreement with those from de-reddened rest-UV luminosities, with an

195: average dust-corrected SFR of $\sim 30 \Msun {\rm yr^{-1}}$. 

196: 

197: However, the rest-UV color selection misses about half of $z \sim 2$

198: galaxies, which have a large amount of dust extinction \citep{kon06}. 

199: The spectroscopic properties of this dusty population have yet not been

200: clearly revealed. 

201: $K$-limited samples are better suited than UV-limited ones to

202: investigate the relations between stellar mass and other fundamental

203: properties at $z \sim 2$, because they approximately correspond to

204: samples limited by stellar mass, without excluding galaxies with large

205: dust extinction.  

206: \citet{kri08} carried out a NIR spectroscopic survey for $K$-bright

207: galaxies at $z \sim 2.3$. They suggest that studies with broadband

208: photometric data alone may overestimate the number of massive galaxies

209: at $z \sim 2$, and underestimate the evolution of the stellar mass

210: density, again indicating the importance of spectroscopy.

211: 

212: In this paper, we present NIR spectroscopy for 44 BzK galaxies in the

213: SDF, and their spectroscopic properties.  

214: The structure of this paper is as follows.

215: The photometric and spectroscopic data are described in \S

216: \ref{sec;data}. In \S \ref{sec;analysis}, we examine the properties of

217: the detected emission lines, including spectroscopic redshifts and

218: fluxes. We then investigate active galactic nuclei (AGN) 

219: contamination in our spectroscopic sample in \S \ref{sec;agn}.

220: In \S \ref{sec;discussions}, physical properties are derived from the

221: emission lines, and then discussed.   

222: A summary is given in \S \ref{sec;conclusion}. 

223: Throughout this paper, magnitudes are in the AB system, 

224: and we adopt cosmological parameters of $h=0.7$, 

225: $\Omega_{m0}=0.3$ and $\Omega_{\Lambda 0}=0.7$. 

226: 

227: 

228: \section{Data}

229: \label{sec;data}

230: 

231: \subsection{BzK galaxies in the SDF}

232: \label{sec;BzKs}

233: Here we briefly describe the imaging data and sample selection of BzKs

234: in the SDF, which are almost the same as those in \citet{hay07}. 

235: The SDF is a large ``blank'' field centered on ($13^{\rm h} 24^{\rm m} 38\fs9$,

236: $+27\arcdeg29\arcmin25\farcs9$; J2000), a rectangle of 

237: $29\farcm7 \times 36\farcm7$, with

238: deep multiwavelength data from the NUV to

239: mid-IR (MIR) wavebands. Among these data, the optical and the NIR data

240: are mainly used in this study. The optical data, $B$, $V$, $R$, $i'$,

241: $z'$, NB816, and NB921, were obtained with the Suprime-Cam on the Subaru

242: telescope for the Subaru Deep Field Project \citep{kas04}, and then the

243: NIR data, $J$ and $K$, were obtained with the wide field camera (WFCAM)

244: on the United Kingdom Infra-Red Telescope (UKIRT) (Motohara et al. in

245: preparation).   

246: The NIR data available to this study were only what was obtained on 2005

247: April 14 and 15, which limited the overlapping area with

248: the optical data to about $40\%$ of the whole SDF \citep{hay07}. 

249: The $3\sigma$ limiting magnitudes in a $2\arcsec$-diameter aperture

250: are 28.45 magnitude for $B$, 26.62 for $z'$, and 24.05 for $K$. The

251: optical images have been convolved to a seeing size of $1\farcs14$,

252: which is the value for $K$. 

253: 

254: MIR data, 3.6 -- 8.0 $\micron$, were obtained with the InfraRed Array

255: Camera (IRAC) on the Spitzer space telescope. Since the point spread

256: functions in the MIR data are much broader than those in optical and

257: NIR data, some galaxy images are blended with nearby objects. 

258: For these, we had to remove the contaminating flux contributions from 

259: close objects in order to do accurate photometry. 

260: To do this we use GALFIT \citep{pen02} to

261: deblend images with adjacent objects (Hayashi et al. in

262: preparation), and accurately measure the MIR fluxes of objects. 

263: The $3\sigma$ limiting magnitudes are 23.60 for $3.6 \micron$, 23.32 for

264: $4.5 \micron$, 21.65 for $5.8 \micron$, and 20.99 for $8.0 \micron$,

265: which are measured with $4.4\arcsec$, $4.4\arcsec$, $5.0\arcsec$, and

266: $6.4\arcsec$ diameter apertures, respectively.    

267: 

268: We applied the $BzK$ color selection proposed by \citet{dad04} to the

269: $B$, $z'$, and $K$ data in the 180 arcmin$^2$ of the SDF which have good 

270: detection uniformity in all three passbands. In selecting BzK galaxies,

271: deep optical data are crucial, because objects without $B$ or $z'$

272: detections can take arbitrary positions on the $BzK$ color diagram, with

273: almost all of them left unclassified. We selected the BzKs with

274: great care for the optically undetected objects (See also

275: \citet{hay07}). We obtained samples of 1092 sBzKs and 56 pBzKs with 

276: $K < 23.2$, corresponding to the $5\sigma$ limiting total magnitude.

277: 

278: 

279: \subsection{Spectroscopic observation}

280: \label{sec;observation}

281: NIR spectroscopy of the BzKs in the SDF was carried out with the

282: Multi-Object InfraRed Camera and Spectrograph \citep[MOIRCS;][]{ich06}

283: on the Subaru telescope on 2007 May 3 and 4. MOIRCS has a wide field

284: of view of $4\arcmin \times 7\arcmin$, and its $zJ500$ and $HK500$

285: grisms provide spectra with a resolution of $R$=500, covering wavelengths

286: of 0.9 -- 1.8 $\micron$ and 1.3 -- 2.5 $\micron$, respectively.

287: The dispersions of the grisms are 5.57 and 7.72 \AA/pixel,

288: respectively. 

289: 

290: Forty-four spectroscopic targets were selected from the samples of 1092 sBzKs

291: and 56 pBzKs. We gave priority to $K<22.0$ BzKs in our samples for 

292: slitmask design, and then $K>22.0$ BzKs were allocated to the blank

293: remaining slits. We used two slitmasks to cover the 44 BzKs, that is, each

294: slitmask covers 22 BzKs.   

295: Each slit has a width of $0\farcs7$ with a length of 10 -- 12

296: $\arcsec$ according to the space between the slits.

297: Among the 44 galaxies, 40 are classified as sBzKs and 4 as pBzKs. 

298: Half of them are fainter than $K=22.0$. 

299: Figure \ref{fig;color_target} shows the colors of the 44 BzKs. 

300: 

301: We observed them consecutively at two positions (A and B) on the slit,

302: using the $zJ500$ and $HK500$ grisms for one slitmask per night 

303: with single exposure times of 600 or 900 sec. 

304: The total on-source integration time was then 2 -- 2.6 hours for both

305: grisms.   

306: During the two nights, the weather was good, and the seeing was as

307: good as $\sim 0\farcs4$ in $K$, except for the end of May 4, when it was

308: $\gtrsim 1\farcs0$.  

309: We also observed A0 stars with $V \sim 9.0$ magnitude at different air

310: masses with both grisms on each night to correct for the telluric 

311: absorption and the instrumental efficiency. Before and after the

312: observations, dome flats with lamps on and off, and Th-Ar lamp

313: data were obtained for wavelength calibration. 

314: 

315: \begin{figure}[tb]

316:  \begin{center}

317:   \plotone{f1.eps}

318:   \caption{The colors of 44 targets. The filled circles are BzKs with

319:   $K<22.0$, and the filled triangles are those with $K>22.0$. BzKs with

320:   emission lines detected are squares.  

321:            }

322:   \label{fig;color_target}

323:  \end{center}

324: \end{figure}

325: 

326: \setcounter{footnote}{8}

327: 

328: \subsection{Data Reduction}

329: \label{sec;reduction}

330: Data reduction is done with standard IRAF%

331: %

332: \footnote{IRAF is

333: distributed by the National Optical Astronomy Observatories, which are

334: operated by the Association of Universities for 

335: Research in Astronomy, Inc., under cooperative agreement with the

336: National Science Foundation.}

337: %

338: procedures.

339: First, we create A -- B frames from successive frames observed at

340: the two positions. The A -- B frames are then divided by a flatfield image, 

341: and bad pixels and cosmic rays are removed. 

342: The flat image is created by taking difference between the dome

343: flat frames with the lamps on and off, and normalizing it. 

344: Next, distortion is corrected using calibration data files provided by

345: the MOIRCS instrument team.    

346: Wavelength calibration is done with only the OH airglow lines for the

347: $HK$ data, since there are numbers of the strong OH lines in the

348: wavelengths longer than $1.1 \micron$. On the other hand, for the $zJ$

349: data, Th-Ar data are also used for wavelength calibration at

350: wavelengths shorter than $1.1 \micron$.  

351: The wavelength uncertainties in the strong lines are less than 15\AA.

352:   

353: Residual sky subtraction is then carried out, since only the A -- B

354: procedure may not completely remove the sky background due to its

355: time variation.   

356: The telluric absorption and the instrumental efficiency are corrected

357: using the spectra of A0 stars and the stellar spectral library given by

358: \citet{pic98}. 

359: Finally, all the frames for each galaxy are added and 1D spectra are

360: extracted combining 5 pixels along a slit.

361: 

362: Flux calibration of the spectra is done using the

363: photometry from the imaging data in $z'$ and $K$.   

364: The $1\sigma$ error of the spectrum is estimated from the standard

365: deviation in the sky region of the 2D spectrum.

366: 

367: 

368: \section{Measurement of emission lines}

369: \label{sec;analysis}

370: \subsection{Detection of emission lines}

371: We have detected $\Ha$ and some other emission lines in 15 of the 40

372: sBzKs, 9 of which have $K>22.0$.   

373: Since there are many strong OH airglow lines in the NIR region, we have

374: first confirmed all lines in the 2D spectra by visual inspection.

375: Among the confirmed lines, then, we regard lines with a peak flux larger

376: than the $3\sigma$ sky noise as a detection using 1D spectra with a

377: resolution of $R$=500, that is, unbinned spectra. 

378: For the 4 pBzKs, no spectra with enough S/N to detect any Balmer/4000\AA

379: \ break were obtained due to the insufficient exposure times. In what

380: follows, we examine and discuss spectroscopic properties of the 15 sBzKs 

381: with emission lines detected. Figures \ref{fig;spectra-a} --

382: \ref{fig;spectra-c} show the binned 1D spectra of the lines, and Tables

383: \ref{table;photo} and \ref{table;spec} show properties of the 15 sBzKs.  

384: 

385: \begin{figure*}[t]

386:  \begin{center}

387:   \epsscale{1.4}

388:   \plotone{f2.eps}

389:   \caption{The spectra of $\Ha$ and [\ion{N}{2}] emission lines we have

390:   detected. These spectra have a binned dispersion of $\sim 16$ \AA.

391:   The solid lines show the spectra and the shaded regions in gray are

392:   the $1\sigma$ sky noise. We regard lines with a peak flux larger than 

393:   the $3\sigma$ sky noise as a detection using unbinned 1D spectra.

394:   Labeled redshifts are derived from $\Ha$ lines.

395:            }

396:   \epsscale{1.0}

397:   \label{fig;spectra-a}

398:  \end{center}

399: \end{figure*}

400: 

401: To measure spectroscopic redshifts and fluxes of the emission lines,

402: Gaussian profiles with a constant continuum are fitted, 

403: where the free parameters are amplitude, line width, redshift, and

404: continuum level.    

405: The $1\sigma$ error of the spectrum is used as a sigma in the

406: calculation of $\chi^2$. 

407: Because $\Ha$ ($\lambda$6563) and [\ion{N}{2}]

408: ($\lambda\lambda$6548,6584) are blended, they are simultaneously fitted,

409: assuming the same line width and redshift, and a 6584/6548 ratio of

410: 3-to-1.  

411: The $1\sigma$ error of each parameter in the fitting is estimated as the

412: range where the increase in $\chi^2$ from the best-fit value is less

413: than the appropriate value for the number of degrees of freedom

414: (see \S14.5 of Numerical recipes by \citet{pre}).

415: 

416: \begin{figure}[bth]

417:  \begin{center}

418:   \epsscale{1.1}

419:   \plotone{f3.eps}

420:   \caption{Same as Fig. \ref{fig;spectra-a}, but for $\Hb$ and [\ion{O}{3}].

421:            }

422:   \epsscale{1.0}

423:   \label{fig;spectra-b}

424:  \end{center}

425: %

426:  \begin{center}

427:   \epsscale{0.7}

428:   \plotone{f4.eps}

429:   \caption{Same as Fig. \ref{fig;spectra-a}, but for [\ion{O}{2}].

430:            }

431:   \epsscale{1.0}

432:   \label{fig;spectra-c}

433: \end{center}

434: \end{figure}

435: 

436: 

437: \subsection{Spectroscopic redshifts}

438: \label{sec;spec-zflux}

439: Spectroscopic redshifts are derived from the $\Ha$ lines using the

440: central wavelength of the fitted Gaussian. For four sBzKs, which are

441: ID21360, ID21667, ID22602, and ID27677, only one emission line is

442: detected, but we regard it as $\Ha$ for the following reasons:  

443: (1) The redshift of ID21360 is also confirmed to be 1.174 by optical

444: spectroscopy with the Hectospec on the Multiple Mirror Telescope

445: \citep{ly08}.   

446: Therefore, the line is certainly $\Ha$.

447: (2) For the other three objects, it is possible that the detected lines, 

448: seen in the $HK$ grism spectra, are from transitions other than $\Ha$.

449: If the lines are $\Ha$, the redshifts should be 1.75, 1.87, and 1.71,

450: respectively. 

451: On the other hand, if the lines are [\ion{O}{3}]($\lambda$5007), those

452: would be 2.60, 2.76, and 2.55, respectively.

453: The transition of $\Hb$ would result in even higher redshifts.

454: Then, because the BzK technique isolates selecting star forming galaxies

455: at $1.4 \lesssim z \lesssim 2.5$ by detecting the Balmer break between

456: the $z$-band and $K$, $\Ha$ is the most likely identification of lines

457: in the $HK$ region. 

458: (3) In addition, we note that almost all the derived spectroscopic

459: redshifts are consistent with photometric redshifts.   

460: The photometric redshifts are obtained by fitting template SEDs to

461: $BVRi'z'K$ photometric data using the Hyper$z$ code (version 1.1)

462: \citep{bol00}. 

463: The used set of template SEDs is what is provided in the Hyper$z$

464: package, which consists of the observed spectra based on the CWW

465: templates \citep{cww80}. 

466: The standard deviation of $\sigma(z_{\rm spec} - z_{\rm phot})$ is 0.61 

467: for the 12 sBzKs with the secure spectroscopic redshfits. The

468: photometric redshifts for ID21667 and ID22602 are 1.47 and 1.46,

469: respectively, supporting our identification of their lines as $\Ha$. 

470: ID27677 would have a photometric redshift of 2.46, suggesting the

471: possibility that the line is $\Hb$ or [\ion{O}{3}]. 

472: Since the flux ratio of [\ion{O}{3}] to $\Hb$ can widely range from 0.1

473: to 10 \citep{vei87}, it is possible that only one of the two lines is

474: detected. 

475: Note that [\ion{O}{3}]($\lambda$4959) would not be detected due to the

476: inadequate S/N even if the identification of [\ion{O}{3}]($\lambda$5007) 

477: is correct. These considerations suggest that all the three of

478: $\Ha$, $\Hb$, and [\ion{O}{3}] are candidates for the single line of

479: ID27677, and that no secure conclusion is available for the line

480: identification.

481: Therefore, adopting the consideration (2), we assume that the line for

482: ID27677 is also most likely to be $\Ha$. Even if the line is not $\Ha$,

483: our conclusions discussed below would remain unchanged.

484: 

485: Since the spectral resolution is $R$=500, uncertainties in spectroscopic

486: redshifts are estimated to be

487: $\Delta z = \Delta\lambda_{\rm obs}(\Ha)/6563 \sim 0.002$. 

488:   

489: Figure \ref{fig;Nz} shows the redshift distribution of the sBzKs with

490: spectroscopic redshift, indicating a somewhat unexpected feature.

491: The distribution does not peak at $z \sim 2$, and no object lies at 

492: $z > 2$, contrary to the fact that $BzK$ color selection

493: is proposed to select galaxies at $1.4 < z < 2.5$. 

494: In fact, \citet{dad07} have reported that the redshift distributions of

495: sBzKs with spectroscopic and photometric redshifts both peak at 

496: $z \sim 2$, and that this selection technique works well to

497: select $z \sim 2$ galaxies, which is also supported by other results

498: with photometric redshifts \citep{gra07,bla08}.

499: The discrepancy of our result may be due to a flux-limited bias,

500: or the fact that the strongest emission line, H$\alpha$, is shifted

501: into a wavelength region of high sky background for $z > 2$, as

502: discussed in the next section.

503: The opposite problem also occurs, namely that some

504: galaxies at $z < 1.4$ are selected, which are also seen in Fig. 2 of

505: \citet{dad07}. These lower-$z$ objects are near the boundary of $BzK$

506: selection criterion, and lie 2.5 -- 2.7 sigma from the color selection

507: threshold. This result suggests that a fraction of star-forming galaxies 

508: at lower redshifts ($z < 1.4$) can meet the sBzK criterion.   

509: 

510: 

511: \begin{figure}[tb]

512:  \begin{center}

513:   \plotone{f5.eps}

514:   \caption{The redshift distribution of the sBzKs with spectroscopic

515:   redshift derived from the $\Ha$ lines. 

516:   Errors are $\Delta z \sim 0.002$.

517:            }

518:   \label{fig;Nz}

519:  \end{center}

520: \end{figure}

521: 

522: 

523: \subsection{Objects without detected emission lines}

524: Among the sBzK targets, no emission line is detected for 25 sBzKs.

525: It is important to know the cause of the non-detections.

526: Judging from the distribution of these sBzKs without a line detection on

527: Figure \ref{fig;color_target}, it is possible that they have large dust

528: extinction.  Applying equation (4) of \citet{dad04} to the $B-z$ color,

529: we find that sBzKs with $E(B-V)$ larger than $\sim 0.5$--which

530: corresponds to Av=2.0 magnitude--have no lines detected. This may be one

531: of the reasons for the non-detection of lines. 

532: Since there are many strong OH airglow lines in the NIR region, it

533: is also possible that an emission line falls on a strong OH line. 

534: In addition, lines are redshifted into the $K$-band region in the case

535: of $z>2$ sBzKs. As described in section \ref{sec;spec-zflux}, the S/N in

536: the region may be too low for lines to be detected.    

537: In fact, there were no lines detected at a wavelength longer than

538: $2.0 \micron$. 

539: Finally, it is also possible that only sBzKs with higher SFRs have

540: strong enough lines to be detected in this $K$-limited sample.

541: The lowest $\Ha$ flux we detected is 

542: 4.2 $\times 10^{-17}$ ergs s$^{-1}$ cm$^{-2}$. 

543: 

544: 

545: 

546: \section{AGN contamination}

547: \label{sec;agn}

548: Before considering the spectroscopic properties of sBzKs, we should

549: account for possible AGN contamination in our sample. 

550: We use two methods to identify AGN candidates. 

551: One is the BPT emission-line diagnostic diagram \citep{bpt81}, and the

552: other is a power-law shaped SED from optical to MIR wavelengths. 

553: 

554: Figure \ref{fig;BPT} is the emission-line diagnostic diagram showing the

555: [\ion{O}{3}]($\lambda$5007)/$\Hb$ vs. [\ion{N}{2}]($\lambda$6584)/$\Ha$

556: ratios. The only three sBzKs--ID26821, ID26840, and ID31569--have all

557: the required lines detected. Undetected line fluxes for the other

558: galaxies are estimated as follows. 

559: We derive $\Hb$ fluxes from the $\Ha$ fluxes, $E(B-V)$ described

560: in \S 5.1, and the assumed intrinsic Balmer decrement of $\Ha/\Hb=2.86$.   

561: The fluxes of [\ion{O}{3}] and [\ion{N}{2}] are the $1\sigma$ upper

562: limits derived from the sky noise at the wavelengths where each line

563: should be seen, assuming the same line width as $\Ha$.

564: As discussed in \citet{kew02} and \citet{erb06a}, the flux ratio of

565: [\ion{N}{2}]/$\Ha$ is enhanced by shock excitation (in LINERs), or by

566: the hard continuum which photo-ionizes the Narrow Line Region (NLR) of

567: all Seyfert nuclei. 

568: Figure 7 in \citet{kew02} shows that the ratio of [\ion{N}{2}]/$\Ha$ is

569: always less than 0.63 in purely star-forming galaxies.  

570: Among the 15 spectroscopically confirmed sBzKs in our sample, four--

571: ID21193, ID22424, ID25528, and ID31340--have [\ion{N}{2}]/$\Ha$ greater

572: than 0.63, which may imply a significant contribution from either a

573: LINER or Seyfert NLR component.

574: In particular, ID25528 shows the high [\ion{N}{2}]/$\Ha$ ratio of 1.2,

575: suggesting that it is an AGN. 

576: On the other hand, we cannot strongly assert that the other three

577: galaxies are LINERs or Seyferts due to the uncertainties of their line

578: ratios.

579: Also, as described below, the SEDs suggest that their continuum is

580: dominated by stellar emission, and that the AGN contribution is

581: small. If they are type II AGNs, the strongest observed contribution to

582: the spectrum from the active nucleus would be high-ionization emission

583: lines, not continuum.   

584: Similarly, one galaxy--ID 31569--shows a high enough

585: [\ion{O}{3}]($\lambda$5007)/$\Hb$ ratio of 4.9 to be a possible Seyfert  

586: 2 galaxy. However, the weak [\ion{N}{2}] line suggests that this is

587: instead a metal-poor star-forming galaxy. 

588: 

589: 

590: \begin{figure*}[tb]

591:  \begin{center}

592:   \plotone{f6.eps}

593:   \caption{The emission-line diagnostic diagram showing the

594:   [\ion{O}{3}]($\lambda$5007)/$\Hb$

595:   vs. [\ion{N}{2}]($\lambda$6584)/$\Ha$ ratios. The solid line is the

596:   boundary separating star forming galaxies (left side) and AGNs (right

597:   side) \citep{kau03}. 

598:   Circles show objects identified as star forming galaxies, while a

599:   square and triangles as an AGN and possible AGN candidates,

600:   respectively.  

601:   The undetected fluxes of [\ion{O}{3}] and [\ion{N}{2}] are

602:   estimated from the $1\sigma$ upper limits, which are shown by

603:   arrows. The $\Hb$ fluxes are inferred from the $\Ha$ flux and $E(B-V)$

604:   on the assumption of an intrinsic Balmer decrement of

605:   $\Ha/\Hb=2.86$. The error bars are shown with $1\sigma$. 

606:   }

607:   \label{fig;BPT}

608:  \end{center}

609: \end{figure*}

610: 

611: 

612: These ratios of forbidden to Balmer emission lines do not however

613: provide a definitive search for AGN. This is because a strong

614: contribution from the Broad Line Region (BLR) in Seyfert 1 nuclei

615: produces a low [\ion{N}{2}]/$\Ha$ ratio which would not be noticeably

616: different from those of \ion{H}{2} regions. 

617: Our spectral resolution of 600 km s$^{-1}$ should be sufficient

618: to detect most broad wing of $\Ha$, but no such wing is found.

619: Furthermore, only one of the $\Ha$ lines--ID 26821--satisfies the

620: criteria of \citet{hicks02} set for an Seyfert 1:

621: EW $>$ 100\AA \ and 

622: $\log L(\Ha,{\rm observed})[{\rm ergs\ s^{-1}}] \ge 42.5$ 

623: (Table \ref{table;spec}).  

624: This implies that there is a contribution from a Seyfert BLR component 

625: in ID26821 sBzK. However, as shown in \S \ref{sec;sfruv}, the SFR

626: derived from $\Ha$ luminosity is in agreement with that from UV

627: luminosity within a factor of 2, which may indicate that the

628: contribution from a BLR to the $\Ha$ line is not significant.

629: 

630: 

631: We have also applied another test for Seyfert 1 nuclei in our sample. 

632: We checked the long-wavelength SEDs of the 15 sBzKs using the photometry

633: from $3.6 - 8.0 \mu m$ taken with IRAC on the Spitzer space telescope.  

634: Most of the sBzKs have a strong bump at a rest wavelength of 1.6 \micron

635: \ in their SEDs, indicating that their SEDs are dominated by stellar

636: emission \citep{saw02,alo07}. 

637: AGN-dominated galaxies, especially those with broad emission lines

638: (Seyfert 1's) have a power-law SED at rest wavelengths longer than

639: 1 \micron \ \citep{mal83,spin95}.

640: In our sample, no such clear power-law SEDs are seen, but three

641: sBzKs have power-law-like SEDs: ID21973, ID26821, and ID27677 

642: (Figure \ref{fig;powerlaw}). 

643: These might perhaps contain "Seyfert 1.9" or "Narrow Line Seyfert 1"

644: nuclei. However, intense star formation also produces substantial

645: near-IR emission from heated dust, which could mimic such an SED. 

646: Figure \ref{fig;powerlaw} shows these three power-law-like SEDs with SED

647: models of starbursts overplotted \citep[SBURT:][]{takagi03}. 

648: These model SEDs have no contribution of AGN.

649: 

650: With all these considerations, we classify ID25528 and ID26821 as

651: AGN. Although the AGN contribution to the spectra, particularly the continuum,

652: may not be significant, the two galaxies are excluded from the

653: discussions on emission lines in the following section. 

654: We then classify the three sBzKs of ID21193, ID22424, and  ID31340 as AGN

655: candidates judging from the [\ion{N}{2}]/$\Ha$ ratios.

656: However, since it is possible that the three AGN candidates are star

657: forming galaxies, we consider them as well as the other ten sBzKs

658: classified as star forming galaxies in the following discussions.

659: 

660: \citet{kon06} have estimated that about 25\% of sBzKs are AGNs using

661: XMM-$Newton$ X-ray data, which would imply that 3 -- 4 sBzKs in our

662: sample could be AGN. However, the sample of \citet{kon06} has only

663: $K$-bright BzKs. According to \citet{alo07}, the characteristic stellar

664: mass of the host galaxy of AGN increases with redshift, and at 

665: $z \sim 2$ it is $\sim (1-3) \times 10^{11}\Msun$.   

666: In our sample, almost all galaxies are less than $10^{11}\Msun$, which

667: indicates that the fraction of AGN can be lower than 25\%.

668: The NICMOS imaging of \citet{colbert05} also indicates that fewer than a

669: quarter of these less massive galaxies should host an AGN detectable in

670: X-rays. 

671: The AGN fraction of 2/15 (13\%) in our sample is consistent

672: with the number of AGN expected from other studies.

673: If the three candidates are actually AGNs, the fraction increases to 33\%, 

674: which would still be consistent with other studies. 

675: 

676: 

677: \begin{figure}[tb]

678:  \begin{center}

679:   \plotone{f7.eps}

680:   \caption{The long-wavelength power-law-like SEDs of ID21973, ID26821,

681:   and ID27677 sBzKs, where model SEDs of starbursts (without any AGN

682:   contribution) from \citet{takagi03} are overplotted.  

683:            }

684:   \label{fig;powerlaw}

685:  \end{center}

686: \end{figure}

687: 

688: 

689: \section{Properties of star-forming BzK galaxies}

690: \label{sec;discussions}

691: Emission line information enables us to estimate SFR

692: and metallicity.  

693: In the local universe, more massive galaxies tend to have lower SFRs and

694: higher metallicities \citep[e.g.,][]{gal05}. 

695: It is important to examine whether a similar relation exists among these

696: properties for high redshift galaxies.

697: 

698: In this section, we first describe the properties obtained from

699: multiwavelength photometric data by SED fitting method. 

700: Then, spectroscopic properties are discussed, along with the

701: photometric properties. 

702: 

703: 

704: \subsection{Properties from SED fitting}

705: \label{sec;sedfit}

706: Properties such as stellar mass and dust extinction, $E(B-V)$, are

707: obtained from fitting synthetic model spectra of continuous bursts

708: \citep{bru03} to the $BVRi'z'K$ photometry, assuming a standard \citet{sal55}

709: initial mass function (IMF), solar abundance, and the empirical dust

710: extinction of \citet{cal00}.   

711: These are shown in Table \ref{table;photo}.

712: Stellar mass is one of the most important parameters of galaxy properties

713: and the most robust one obtained from SED fitting.

714: Although sBzKs are well-fitted by the continuous burst model, fits using

715: the $\tau$ model, where the SFR declines exponentially with time, 

716: would make the stellar masses slightly smaller than those with the

717: continuous burst model.   

718: Note that SEDs of the simple stellar population (SSP) model are

719: irrelevant in this work, since sBzKs are still star-forming at 

720: $z \sim 2$.   

721: It is also noted that if a \citet{cha03} IMF is used, stellar masses

722: (and SFRs) are smaller by a factor of 1.8 than those with a Salpeter IMF. 

723: Thus, stellar masses are determined within a factor of $\sim 2.5$.

724: Fitted $E(B-V)$s are mainly used to calibrate the dust extinction of

725: emission line fluxes in section \ref{sec;sfr}. As discussed in section

726: \ref{sec;agn}, almost all SEDs of the sBzKs are dominated by stellar

727: emission with a clear 1.6$\mu m$ bump. Therefore, any AGN contribution

728: to the SED is small enough that parameters obtained from the SED fitting

729: reflect properties of the stellar populations. 

730: 

731: 

732: \subsection{Star Formation Rates}

733: \label{sec;sfr}

734: \subsubsection{SFR derived from H$\alpha$ and UV continuum}

735: \label{sec;sfrha}

736: SFRs for sBzKs are derived from the luminosity of $\Ha$ using the equation

737: given in \citet{ken98};  

738: \begin{equation}

739: {\rm SFR}_{\Ha}[\Msun {\rm yr^{-1}}] = 7.9 \times 10^{-42} L_{\Ha}

740: [{\rm ergs\ s^{-1}}],

741: \end{equation}

742: assuming solar abundance, Salpeter IMF with mass limits of 0.1 and 100

743: $\Msun$, and a constant SFR.

744: $\Ha$ luminosities are corrected for the dust extinction by 

745: equation (3) in \citet{cal00}, 

746: $E(B-V) = 0.44 E_{\rm gas}(B-V)$.

747: The amount of nebular extinction is then estimated from 

748: $E_{\rm gas}(B-V)$ using the extinction law given in \citet{cal00}. 

749: The median amount of extinction of $\Ha$ is 2.83 magnitude. 

750: 

751: The amount of extinction at $\Ha$ can be directly estimated 

752: from the apparent $\Ha/\Hb$ ratio. However, we do not apply 

753: this method for our sample, because only three galaxies have 

754: detection of both lines and even for these three the amount 

755: of extinction is not strongly constrained due to a large 

756: uncertainty in the $\Hb$ flux.

757: 

758: The rest-UV luminosity is also used as a tracer of SFR. 

759: We derive SFR from the continuum at 1500\AA, $L_{1500}$, which

760: is predicted by the best-fit model to the $BVRi'z'K$ photometry.

761: $L_{1500}$ is corrected for the dust extinction with Calzetti extinction

762: law and converted into a SFR using the equation given in \citet{dad04};

763: \begin{equation}

764: {\rm SFR}_{\rm UV}[\Msun {\rm yr^{-1}}] = \frac{L_{1500}[{\rm ergs\

765:  s^{-1}\ Hz^{-1}}]}{8.85} \times 10^{-27}.  

766: \end{equation}

767: 

768: \subsubsection{Comparison between ${\rm SFR}_{\Ha}$ and ${\rm SFR}_{\rm UV}$}  

769: \label{sec;sfruv}

770: Figure \ref{fig;sfr_ha_uv} compares the SFR from the rest-UV

771: luminosity with that from $\Ha$. 

772: A correlation is seen between the two SFRs, but ${\rm SFR}_{\Ha}$ tends

773: to be larger than ${\rm SFR}_{\rm UV}$ by a factor of three. 

774: The median ${\rm SFR}_{\Ha}$ and ${\rm SFR}_{\rm UV}$ are 182 and

775: 61 $\Msun {\rm yr^{-1}}$, respectively.  

776: Also, this discrepancy between SFR measured by $\Ha$ and from shorter

777: wavelength observations is quite similar to what \citet{hicks02} found

778: in their sample of IR-selected galaxies.

779: 

780: On the other hand, \citet{erb06b} report that ${\rm SFR}_{\Ha}$ for

781: UV-selected $z\sim 2$ galaxies is in good agreement with 

782: ${\rm SFR}_{\rm UV}$, which is inconsistent with our result.

783: However, their observed $\Ha$ luminosities are similar to ours. That is,

784: the median $\Ha$ luminosities of our sample and that of \citet{erb06b}

785: are 2.0 and 2.2 times $10^{42}$ ergs s$^{-1}$, respectively. 

786: This disagreement on resulting SFRs with \citet{erb06b} may be due to 

787: different corrections for dust extinction at $\Ha$. They 

788: used the color excess for the stellar continuum to de-redden

789: $\Ha$, that is, $E_{\rm gas}(B-V)=E(B-V)$, which implies that

790: their amount of dust correction for $\Ha$ luminosity is systematically

791: lower than ours by 1.6 magnitude on average.  

792: This leads to a difference in derived SFRs of a factor of 4.4.

793: Using the same method as \citet{erb06b} for de-reddening $\Ha$,

794: our correlation between ${\rm SFR}_{\Ha}$ and ${\rm SFR}_{\rm UV}$

795: becomes consistent with that of \citet{erb06b}. 

796: In this case, the median ${\rm SFR}_{\Ha}$ is 50 $\Msun {\rm yr^{-1}}$.

797: 

798: \citet{dad04} report that most color excesses of 24 sBzKs in the

799: K20/GOODS area, which are derived from $U$ to $K$ SEDs, fall within a

800: range of $0.2 \lesssim E(B-V) \lesssim 0.5$, which is consistent with

801: our result (Table \ref{table;photo}). 

802: The median $E(B-V)$ in our sample is 0.38.

803: This implies that our estimation of $E(B-V)$ by SED fitting is reliable.

804: 

805: This difference between ${\rm SFR}_{\Ha}$ and ${\rm SFR}_{\rm UV}$

806: implies two possibilities. One is that a Salpeter IMF is unsuitable, if

807: the equation of \citet{cal00}, $E(B-V) = 0.44 E_{\rm gas}(B-V)$, 

808: is assumed to be valid for star-forming galaxies at $z \sim 2$. 

809: A more top-heavy IMF than Salpeter would account for the difference,

810: since nebular emission results from ionizing photons shortward of the

811: Lyman limit. 

812: The other is that the equation of \citet{cal00} cannot be applied to 

813: $z \sim 2$ galaxies. 

814: In any case, in what follows we use ${\rm SFR}_{\Ha}$ for sBzKs in

815: the comparison with their stellar masses.  

816: Even if ${\rm SFR}_{\rm UV}$ were used instead, the discussion in the next

817: section would remain unchanged.

818: 

819: \begin{figure}[tb]

820:  \begin{center}

821:   \plotone{f8.eps}

822:   \caption{Comparison between ${\rm SFR}_{\Ha}$ and 

823:   ${\rm SFR}_{\rm UV}$. 

824:   ${\rm SFR}_{\Ha}$s are derived from $\Ha$ luminosities, and then 

825:   ${\rm SFR}_{\rm UV}$s are from the luminosities at 1500 \AA \  estimated

826:   from the best-fitted SEDs. 

827:   Both SFRs are corrected for dust extinction.

828:   The dotted line shows that ${\rm SFR}_{\rm UV}$ is equal to 

829:   ${\rm SFR}_{\Ha}$. 

830:            }

831:   \label{fig;sfr_ha_uv}

832:  \end{center}

833: \end{figure}

834: 

835: 

836: \subsubsection{Relation to stellar mass}

837: \label{sec;RtoMstar}

838: We examine the relation between ${\rm SFR}_{\Ha}$ and stellar mass.

839: Figure \ref{fig;mstar_sfr} suggests that sBzKs with higher stellar

840: masses have larger SFRs, which is qualitatively consistent with

841: \citet{dad07}. 

842: However, the slope of our relation is shallower than theirs.

843: One cause for this disagreement in the slope may be a difference between

844: ${\rm SFR}_{\Ha}$ and ${\rm SFR}_{\rm UV}$. 

845: Figure \ref{fig;sfr_ha_uv} suggests the difference between these two

846: estimates of SFR is larger at lower SFRs.

847: Since less massive sBzKs have lower SFRs, this difference results in

848: shallower slope of the correlation between stellar mass and 

849: ${\rm SFR}_{\Ha}$. 

850: Other cause may be a difference in the method used to derive stellar

851: masses. \citet{dad07} have used the empirical relation between the

852: stellar mass and $K$ given in \citet{dad04}, which is calibrated based

853: on $z-K$ color. 

854: We compare the stellar masses derived from a SED-fitting with

855: those from the relation of \citet{dad04}, finding that the difference

856: between the two estimates of stellar mass is larger at smaller stellar

857: mass. The stellar masses from the relation are larger by a factor of 2

858: -- 3 in $\sim 10^{10} \Msun$. 

859: Because this relation is derived using bright $K<22$ BzKs,

860: the stellar masses of faint BzKs may be overestimated compared with

861: those from SED fitting, resulting in the shallower correlation

862: between stellar mass and ${\rm SFR}_{\Ha}$.

863: 

864: It is also possible that a flux-limited bias, discussed in \S\S 3.2 and

865: 3.3, produces this disagreement.  

866: If galaxies with $\left<E(B-V)\right>=0.38$ and 

867: $F(\Ha)=4.2\times10^{-17} {\rm ergs\ s^{-1}\ cm^{-2}}$, which is the

868: lowest $\Ha$ flux we have detected, are at $\left<z\right>\sim1.7$, the

869: SFR would be $\sim87\Msun {\rm yr^{-1}}$. This suggests that it is

870: likely that we cannot detect any emission lines for galaxies with SFR

871: $\lesssim$ 87 $\Msun {\rm yr^{-1}}$. 

872: Moreover, it is found that ${\rm SFR}_{\Ha}$s is three times larger

873: than ${\rm SFR}_{\rm UV}$s on average. Combining this with the

874: relation between stellar mass and ${\rm SFR}_{\rm UV}$ given in Fig. 17

875: of \citet{dad07}, we find that the limiting SFR of 87 

876: $\Msun {\rm yr^{-1}}$ corresponds to a stellar mass of 

877: $\sim 1\times10^{10}\Msun$.  

878: These estimates indicate that it is possible that no galaxies with

879: $\lesssim 1\times10^{10}\Msun$ and $\lesssim 87 \Msun {\rm yr^{-1}}$ are

880: detected due to flux-limited bias. 

881: However, even if Figure \ref{fig;mstar_sfr} is limited to a region with

882: SFRs higher than the estimated detection limit, a mild correlation

883: between stellar mass and SFR can be seen.

884: If there are less massive galaxies with lower SFRs, as found in

885: \citet{dad07}, the correlation would be steeper.

886: 

887: 

888: \citet{erb06b} have also reported a similar correlation between SFR and

889: stellar mass for UV-selected $z \sim 2$ galaxies, but their correlation

890: is shifted to much lower SFR than ours. 

891: This is mainly due to differences in the assumed IMF as well as dust

892: correction, as discussed in section \ref{sec;sfruv}. 

893: In \citet{erb06b}, a Chabrier IMF is used to estimate stellar masses and

894: SFRs. SFRs derived with a Chabrier IMF are smaller than those with 

895: a Salpeter IMF by a factor of 1.8. 

896: These differences make our SFR 7.9 times larger than that of

897: \citet{erb06b}.

898: After correcting these systematic differences, the comparison of the

899: results between sBzKs and UV-selected galaxies indicates that the former  

900: tend to have slightly higher SFRs than UV-selected galaxies with the

901: similar stellar masses. This reflects the intrinsic difference

902: between the the dust amount in the two populations--only

903: sBzKs with higher SFRs may be easily detected, due to the larger dust

904: extinction than in UV-selected galaxies.

905: 

906: The positive correlation between SFR and stellar mass may be attributed

907: to a larger amount of gas in more massive galaxies, as suggested by

908: \citet{dad07}. 

909: 

910: \begin{figure}[tb]

911:  \begin{center}

912:   \plotone{f9.eps}

913:   \caption{The relation between ${\rm SFR}_{\Ha}$ and stellar

914:   mass. Stellar masses are derived from SED fitting to $BVRi'z'K$

915:   photometry.    

916:   The dotted line shows the limiting SFR of 

917:   $87 \Msun {\rm yr^{-1}}$ estimated from the lowest detected flux on

918:   the assumption that the galaxy is at $<z>=1.7$ and suffer a typical

919:   color excess of 0.38. 

920:            }

921:   \label{fig;mstar_sfr}

922:  \end{center}

923: \end{figure}

924: 

925: \subsubsection{Specific SFR}

926: \label{sec;ssfr}

927: We derived specific SFRs (SSFRs) from ${\rm SFR}_{\Ha}$s and stellar

928: Masses.  SSFR is an indicator of the time scale of star formation

929: activity. 

930: Figure \ref{fig;mstar_ssfr} shows SSFR as a function of stellar

931: mass. A negative relation is seen between stellar mass and SSFR, which

932: is consistent with previous results for $z \sim 2$ galaxies

933: \citep{red06a,erb06b}. 

934: This negative relation implies that a larger growth of stellar mass

935: occurs in less massive galaxies.  

936: In comparison with \citet{erb06b}, it should be noted that we apply

937: a different IMF and dust correction as described in the previous

938: section. In Figure \ref{fig;mstar_ssfr}, the results of \citet{erb06b}

939: are converted into those with the same manner as ours.

940: 

941: We must consider the effect of flux-limited bias on this correlation, as

942: discussed in the previous section.  

943: However, it seems that there is a genuine negative correlation even if

944: the range of validity is limited to stellar masses larger than

945: $1\times10^{10}\Msun$.  

946: The steep relation between stellar mass and SSFR results from the

947: shallow correlation between stellar mass and SFR, which shows that the

948: increase in SFR is not as rapid as that in stellar mass. 

949: If we miss galaxies with low SFRs of $\lesssim87\Msun {\rm yr^{-1}}$ as

950: discussed in section \ref{sec;RtoMstar}, that is, the correlation

951: between stellar mass and SFR is steeper, the relation with SSFR would be

952: shallower than that in Figure \ref{fig;mstar_ssfr}.   

953: 

954: \begin{figure}[tb]

955:  \begin{center}

956:   \plotone{f10.eps}

957:   \caption{The relation between the specific ${\rm SFR}_{\Ha}$ and the

958:   stellar mass. 

959:   Filled circles are our results, open triangles show those of

960:   \citet{red06a} for sBzKs in their UV-selected sample, and crosses are

961:   those of \citet{erb06b} for UV-selected sample.

962:   All stellar masses and SSFRs are converted into those derived with

963:   Salpeter IMF and the same dust correction as we apply. 

964:   The dotted line shows the track of galaxies with a constant SFR

965:   of 87$\Msun {\rm yr^{-1}}$. 

966:            }

967:   \label{fig;mstar_ssfr}

968:  \end{center}

969: \end{figure}

970: 

971: 

972: \subsection{Equivalent Width}

973: \label{sec;ew}

974: The rest-frame equivalent widths of $\Ha$ are derived from 

975: Gaussian fitting. 

976: Both $\Ha$ and continuum fluxes have been corrected for dust

977: extinction.

978: As discussed in \citet{erb06b}, one can infer the star formation history

979: of galaxies from the equivalent width. 

980: While the flux of $\Ha$ reflects the current star forming activity,

981: the continuum flux emitted from stars reflects the past star

982: formation. That is to say, the equivalent width represents how 

983: actively a galaxy is forming stars compared with the past.

984: Thus, the equivalent width is a similar indicator to SSFR.

985: Figure \ref{fig;ew_sfr} shows the equivalent width as a function of the

986: SSFR, indicating a mild correlation. 

987: This supports the result of the previous section. 

988: Less massive galaxies are more actively forming stars, which results in

989: a larger growth of stellar mass.

990: 

991: 

992: \begin{figure}[tb]

993:  \begin{center}

994:   \plotone{f11.eps}

995:   \caption{The relation between specific star formation rate and

996:   $\Ha$ equivalent width in rest frame. Rest-frame equivalent widths of

997:   $\Ha$ are derived from line and continuum fluxes after correction for

998:   dust extinction. 

999:            }

1000:   \label{fig;ew_sfr}

1001:  \end{center}

1002: \end{figure}

1003: 

1004: 

1005: \subsection{Metallicity}

1006: \label{sec;metallicity}

1007: Average metallicity is one of the most important properties of a galaxy. 

1008: Since gas and stellar metallicities are related to the past star

1009: formation, they are crucial for understanding the star formation history

1010: and galaxy evolution. 

1011: It is well-known that metallicity is correlated with stellar mass in the

1012: local universe \citep[e.g.,][]{tre04}. 

1013: \citet{erb06a} have reported a similar correlation to the local one

1014: for UV-selected galaxies at $z \sim 2$.  However, their mass-metallicity

1015: ($M$-$Z$) relation at $z \sim 2$ is shifted to lower metallicity by 

1016: $\sim 0.3$ dex. 

1017: This $M$-$Z$ relation at $z \sim 2$ obtained by \citet{erb06a} is widely

1018: used as representative of $z \sim 2$ star forming galaxies

1019: in comparison with both observational and theoretical studies

1020: \citep{liu08,mai08,bro07,der07,fin08}. However, because the UV selection

1021: misses star forming galaxies with a large dust extinction, we cannot

1022: be confident that the $M$-$Z$ relation of \citet{erb06a} 

1023: applies to all star-forming galaxies at $z \sim 2$. 

1024: Here, we determine a $M$-$Z$ relation for sBzKs.

1025: Our sBzK sample, which approximately corresponds to a stellar 

1026: mass-limited sample, should be more suitable for deriving the $M$-$Z$

1027: relation at $z \sim 2$. 

1028: 

1029: It is well known that electron temperature reflects

1030: gas metallicity \citep{kew02,kob04,erb06a}. 

1031: Because auroral lines, which are used to derive electron temperature,

1032: are weak, and observed in only galaxies with low metallicities, 

1033: it is very difficult even in local galaxies to estimate the metallicity

1034: of galaxies from them.

1035: We can only use the ratio of strong emission lines which are emitted

1036: from different ionization levels to derive metallicity for high-$z$

1037: galaxies. At present, there are various metallicity diagnostics, such as

1038: the $R_{23}$ and N2 methods, to estimate gas-phase oxygen abundance,

1039: 12+log(O/H). One problem with these diagnostics is a disagreement

1040: between the resulting oxygen abundances inferred from the different

1041: diagnostics. Even if the same diagnostic is used, some different

1042: calibrations give a large difference in the resulting metallicities

1043: \citep{kew02,kob04}. The same diagnostics and calibration should be used

1044: to properly compare with other results. 

1045: 

1046: We derive 12+log(O/H) using the N2 index method proposed by \citet{sto94},

1047: where N2 is defined as the flux ratio of [\ion{N}{2}]($\lambda6584$)/$\Ha$;

1048: \begin{equation}

1049: {\rm N2} \equiv \log\frac{[{\rm NII}](\lambda6584)}{\Ha}.

1050: \end{equation}

1051: The gas-phase oxygen abundances are derived with the following relation

1052: given in \citet{pet04};

1053: \begin{equation}

1054: \begin{array}{l}

1055: 12+\log({\rm O/H}) = 8.90 + 0.57\times{\rm N2}.

1056: \end{array}

1057: \end{equation}

1058: The reasons we use this relation are that it requires only two

1059: lines of $\Ha$ and [\ion{N}{2}], and the result can be compared

1060: directly with that of \citet{erb06a}, 

1061: where the same relation was used to derive the gas abundance.    

1062: Also, it should be noted that the lines of $\Ha$ and [\ion{N}{2}] are close

1063: enough that the difference in dust extinction is negligible. 

1064: The $1\sigma$ dispersion in the abundance derived from the relation is

1065: 0.18 \citep{pet04}.

1066: 

1067: Figure \ref{fig;mstar_metal} shows the resulting relation between

1068: stellar mass and metallicity. 

1069: The broken line shows the abundance for galaxies with the flux ratio of

1070: [\ion{N}{2}]/$\Ha$=0.63, where N2 saturates \citep{kew02}, suggesting

1071: that the linear relation between the oxygen abundance and N2 index is

1072: acceptable up to the abundance of $\sim8.79$.  

1073: As discussed in section \ref{sec;agn}, the three sBzKs with a large

1074: abundance, ID21193, ID22424, and ID31340, may harbor AGNs, because they

1075: have flux ratios of [\ion{N}{2}]/$\Ha$ larger than 0.63. However, their SEDs

1076: suggest that their continuum is dominated by stellar emission, and that

1077: the AGN contribution is small. 

1078: Even if the three AGN candidates are excluded from our sample, our 

1079: conclusions in the following discussion are unchanged.    

1080: 

1081: We derive upper limits of metallicity for sBzKs without a secure

1082: detection of [\ion{N}{2}], assuming that the peak flux is 3 times sky noise,

1083: and that the line width is the same as that of $\Ha$. 

1084: These galaxies are shown by crosses in the figure.

1085: Also, we estimate mean metallicity of the five sBzKs without

1086: [\ion{N}{2}] detection. We derive it from the stacked spectrum using the

1087: equations (3) and (4), finding that the mean abundance is 8.55.

1088: This is shown by the double circle in the figure, whose stellar mass is the

1089: average of those of the five sBzKs. 

1090: 

1091: 

1092: \begin{figure*}[htb]

1093:  \begin{center}

1094:   \plotone{f12.eps}

1095:   \caption{Stellar mass vs. metallicity relation. Circles and crosses

1096:   show our results. Filled circles show those of sBzKs with the detection

1097:   of both $\Ha$ and [\ion{N}{2}]. 

1098:   Crosses show upper limits of sBzKs without the detection of [\ion{N}{2}],

1099:   which are derived using the $3\sigma$ upper limit on [\ion{N}{2}] flux. 

1100:   The double circle shows the mean abundance of the five sBzKs without a

1101:   detection of [\ion{N}{2}] against the mean stellar mass.

1102:   Open squares show UV-selected $z \sim 2$ galaxies, and

1103:   open triangles are those of the local $\sim 53,000$ SDSS galaxies 

1104:   \citep{erb06a}. All abundances are estimated using the same N2 index

1105:   method. The dotted line is the solar oxygen abundance of 

1106:   12+log(O/H)=8.66 \citep{asp04}.

1107:   The broken line shows the abundance for galaxies with the flux ratio

1108:   of [\ion{N}{2}]/$\Ha$=0.63.  

1109:   Stellar masses of the \citet{erb06a} are converted into those derived 

1110:   with Salpeter IMF. 

1111:   } 

1112:   \label{fig;mstar_metal}

1113:  \end{center}

1114: \end{figure*}

1115: 

1116: 

1117: Figure \ref{fig;mstar_metal} shows that more massive sBzKs tend to be

1118: more metal-rich, a trend qualitatively consistent with that of

1119: \citet{erb06a}. 

1120: However, the metallicities of sBzKs with [\ion{N}{2}] detection are

1121: higher by $\sim 0.2$ dex than those of UV-selected galaxies with the

1122: same stellar mass at the similar redshift. Considering the

1123: uncertainties, a significant shift exists at the 2.7$\sigma$ level

1124: between the two $M$-$Z$ relations.  

1125: If three AGN candidates are excluded from the sample, the confidence

1126: level in the shift slightly decreases, to the 1.8$\sigma$ level.

1127: 

1128: We again emphasize that the N2 flux ratios are converted into the oxygen

1129: abundances using the same relation as \citet{erb06a}.

1130: We make sure that the N2 indices are larger than those of

1131: \citet{erb06a} for galaxies with the given stellar masses, suggesting

1132: that the cause of the difference in $M$-$Z$ relation is not a systematic

1133: difference in our abundance estimate methodology. 

1134: 

1135: \citet{pet04} have found a relation between gas oxygen abundance and

1136: O3N2 index, which is defined as 

1137: ${\rm O3N2}$ $\equiv$

1138: $\log$[([\ion{O}{3}]($\lambda5007$)/$\Hb$) /

1139: ([\ion{N}{2}]($\lambda6584$)/$\Ha$)].  

1140: We check the abundance with the O3N2 diagnostics for three sBzKs with

1141: the necessary lines observed. The abundances from the O3N2 ratios are

1142: lower than those from the N2 ratios by $\sim 0.13$ dex. This is

1143: consistent with \citet{erb06a}, who found that the abundances from O3N2

1144: are 0.17 dex lower on average than those from N2.  

1145: In addition, it is not likely that the systematic difference in

1146: estimates of stellar masses accounts for the difference in the $M$-$Z$

1147: relation.   

1148: The stellar masses of \citet{erb06a} are converted into those derived

1149: with Salpeter IMF. 

1150: As discussed in section \ref{sec;sedfit}, the systematic difference of

1151: stellar masses should be less than a factor of 2.

1152: This indicates that there is no way that it is the cause of the

1153: difference in $M$-$Z$ relation. 

1154: 

1155: We consider the possibility that our $M$-$Z$ relation

1156: suffers from a selection bias that only sBzKs with high metallicities

1157: are plotted, since metallicities are properly measured only for sBzKs

1158: with both $\Ha$ and [\ion{N}{2}] lines detected.   

1159: The stacking analysis for sBzKs without [\ion{N}{2}] detection

1160: suggests that there are also sBzKs with lower metallicities

1161: (Figure \ref{fig;mstar_metal}), supporting the idea that

1162: selection bias cannot be completely ruled out.

1163: This may imply that the mean metallicities of sBzKs with given

1164: stellar masses are slightly lower than those plotted in Figure 

1165: \ref{fig;mstar_metal}. 

1166: However, the mean metallicity of only sBzKs without [\ion{N}{2}]

1167: detection is similar to that of UV-selected galaxies, implying that

1168: sBzKs would still be more metal rich on average than UV-selected

1169: galaxies. This fact suggests that it is likely that the UV selection

1170: misses star forming galaxies at $z\sim2$ with comparatively higher

1171: metallicities, and that metallicities of star forming galaxies at 

1172: $z \sim 2$ have a large dispersion.     

1173: 

1174: 

1175: Therefore, we conclude that the oxygen abundance of sBzKs is greater

1176: than that of UV-selected $z \sim 2$ galaxies. 

1177: As described in section \ref{sec;sfr}, $K$-selected star forming

1178: galaxies at $z\sim2$ have slightly higher SFRs but similar to

1179: UV-selected galaxies, if the same correction for dust extinction is

1180: applied, while their metallicities are significantly different.   

1181: This fact may imply that the two star forming galaxy populations at 

1182: $z \sim 2$ pass through different star forming histories before this

1183: epoch. 

1184: 

1185: 

1186: 

1187: This difference in the $M$-$Z$ relation between sBzKs and UV-selected

1188: galaxies may be related to the difference in color of the two

1189: populations. 

1190: Figure \ref{fig;k_rk} shows $R-K$ color as a function of $K$ for our

1191: sBzK sample and the UV-selected galaxy sample of \citet{erb06a}.

1192: This figure shows that the sBzK spectroscopic sample galaxies are

1193: redder than the UV-selected galaxies.

1194: Indeed, the sBzK ID31569, which is as blue as UV-selected galaxies, is

1195: located on the $M$-$Z$ relation of \citet{erb06a}. The difference in

1196: $R-K$ color may imply that galaxies with different colors have different

1197: ages and/or dust contents, since the 4000\AA \ break falls

1198: between the $R$ and $K$-bands. This fact indicates that star forming

1199: galaxies at $z \sim 2$ with redder colors have larger metal abundances,

1200: and that the $M$-$Z$ relation of \citet{erb06a} may not represent the

1201: entire population of $z \sim 2$ galaxies. 

1202: 

1203: 

1204: Another possibility is that the difference in redshifts of galaxies

1205: between the two samples accounts for the difference in the $M$-$Z$

1206: relation.  

1207: While the median redshift of galaxies in our sample is 1.7, the mean

1208: redshift is $\left<z\right>=2.2$ in the \citet{erb06a} sample. Because the

1209: galaxies in our sample are on average at a lower redshift, the fact that 

1210: our sBzKs have higher metallicities than UV-selected galaxies of

1211: \citet{erb06a} may plausibly be attributed to cosmic evolution.  

1212: However, this interpretation would require a surprisingly rapid

1213: evolution of the $M$-$Z$ relation over 800 Myr.

1214: 

1215: We plot results of star forming galaxies at $z \sim 0$ from

1216: SDSS data in Figure \ref{fig;mstar_metal} \citep{tre04,erb06a}. It

1217: should be noted that we cannot directly discuss redshift evolution of

1218: the $M$-$Z$ relation, since these SDSS galaxies are not necessarily

1219: descendants of sBzKs.    

1220: Our $M$-$Z$ relation seems to be similar to that of SDSS galaxies

1221: \citep{tre04}, but we may not necessarily claim that the metallicities

1222: of sBzKs are the same as those of the local SDSS galaxies.  

1223: There is the possibility that the true metallicities of the SDSS

1224: galaxies are higher than those derived by the N2 method, if the

1225: saturation of the N2 index at high metallicity is taken into 

1226: account. Thus, the oxygen abundances of sBzKs can be as high as those of

1227: the SDSS galaxies.  

1228: Then, our result that the metallicities of sBzKs are close to solar

1229: abundance is consistent with the assumption of the SEDs with solar

1230: abundance in SED fitting.

1231: 

1232: It is also found that there are $M$-$Z$ correlations at other

1233: redshifts. 

1234: \citet{sav05} have reported the relation for NIR-selected

1235: galaxies at $z\sim0.7$, finding that the metallicities of galaxies with

1236: a given stellar mass have a dispersion of $\sim0.2$ dex around the mean 

1237: abundances. The correlation with large scatter is similar to what we

1238: found for galaxies at $z\sim2$.

1239: \citet{mai08} have found the relation shifted to much lower metallicity

1240: for 9 Lyman Break Galaxies at $z\sim3.5$, suggesting that the

1241: metallicity of galaxies decreases with increasing redshift. 

1242: However, since metallicity measurements for high-$z$ galaxies are very

1243: limited, especially at $z>2$, the $M$-$Z$ relations of high-$z$ galaxies

1244: obtained so far may have significant statistical and systematic

1245: uncertainties. Therefore, in future, it is essential to derive more

1246: reliable $M$-$Z$ relations at high redshifts from a large, unbiased

1247: sample. 

1248: 

1249: \begin{figure}[tb]

1250:  \begin{center}

1251:   \plotone{f13.eps}

1252:   \caption{The color magnitude diagram of $R-K$ vs. $K$ for our sBzK

1253:   spectroscopic sample (filled circles), the whole sBzK sample in the SDF

1254:   (dots) and the NIRSPEC sample of \citet{erb06c} (open triangles).  

1255:            }

1256:   \label{fig;k_rk}

1257:  \end{center}

1258: \end{figure}

1259: 

1260: 

1261: 

1262: 

1263: \section{Summary}

1264: \label{sec;conclusion}

1265: We have carried out NIR spectroscopy of 44 BzK galaxies with 

1266: $\KAB \lesssim 23$ in the Subaru Deep Field with MOIRCS on the Subaru

1267: Telescope.  The $zJ$ and

1268: $HK$ grism spectra cover wavelengths from 0.9 to 2.3 $\micron$ at a

1269: resolution of $R$=500. 

1270: The 44 spectroscopic targets are selected from a large sample of BzKs in

1271: 180 arcmin$^2$ of the SDF. 

1272: Among the 44 targets, 40 are classified as sBzKs and 4 as pBzKs. Half

1273: of the 44 targets are fainter than $\KAB=22$. We detect $\Ha$ and some other

1274: emission lines in 15 sBzKs, nine of which have $\KAB > 22$.  

1275: 

1276: Spectroscopic redshifts are derived from $\Ha$, and fluxes

1277: of emission lines are measured.

1278: Stellar mass and dust extinction are also obtained from SED-fitting to

1279: multiwavelength imaging data of $BVRi'z'K$.    

1280: We have then obtained the following results.

1281: 

1282: The spectroscopic redshift distribution of the sBzKs does

1283: not peak at $z \sim 2$, and objects at $z < 1.4$ are also

1284: selected. These lower-$z$ objects are near the selection boundary 

1285: in the $BzK$ diagram. On the other hand, no $z > 2$ objects are

1286: detected. This result suggests that a fraction of star forming galaxies

1287: at low redshifts can meet the sBzK criterion. However, the lack of

1288: objects at $z > 2$ may be due to a flux-limited bias. 

1289: 

1290: The $\Ha$ luminosities are converted into SFRs. 

1291: It is found that sBzKs with higher stellar masses have larger SFRs.

1292: The correlation between SFR and stellar mass may be attributed to a

1293: larger amount of gas for more massive galaxies.

1294: Dividing SFR by stellar mass, specific SFRs are also derived.

1295: A negative relation is seen between stellar mass and SSFR, which is

1296: consistent with previous results for $z \sim 2$ galaxies. 

1297: This negative relation implies that a proportionately larger growth of stellar

1298: mass occurs in less massive galaxies. 

1299: 

1300: Gas-phase oxygen abundances, 12+log(O/H), are derived from the ratio of

1301: [\ion{N}{2}]($\lambda6584$) to $\Ha$ using the N2 index method. 

1302: We have found a correlation between stellar mass and oxygen abundances,

1303: in the sense that more massive sBzKs tend to be more metal-rich, which

1304: is qualitatively consistent with the relation for UV-selected $z \sim 2$

1305: galaxies. However, the metallicities of sBzKs are on average higher by

1306: $\sim 0.2$ dex than that of UV-selected galaxies at similar redshifts,

1307: and are close to those of local galaxies. 

1308: Compared with the uncertainties, this difference in metallicity between

1309: sBzKs and UV-selected galaxies is significant, and may be due to the

1310: color difference between the two populations. 

1311: Our sBzKs have redder $R-K$ colors than the UV-selected

1312: galaxies. Further, an sBzK with a blue color similar to the UV-selected

1313: galaxies has a lower abundance, comparable to those of the UV-selected

1314: galaxies. These facts suggest that there is a color-dependence in the

1315: oxygen abundance in galaxies at $z \sim 2$, which may imply that

1316: galaxies with different ages and/or dust amounts have different

1317: metallicities.   

1318: 

1319: 

1320: 

1321: 

1322: 

1323: \acknowledgments

1324: We would like to thank the Subaru Telescope staff for their invaluable

1325: help in our observation with Subaru/MOIRCS. 

1326: We also thank an anonymous referee for useful comments, which have

1327: greatly improved the paper.

1328: M. Hayashi acknowledges support from the Japan Society for the Promotion

1329: of Science (JSPS) through JSPS Research Fellowship for Young Scientists.   

1330: 

1331: 

1332: \begin{thebibliography}{}

1333: \bibitem[Alonso-Herrero et al.(2007)]{alo07} Alonso-Herrero, A., et

1334: 				 al. 2008, arXiv:0712.3121

1335: 

1336: \bibitem[Asplund et al.(2004)]{asp04} Asplund, M., Grevesse, N., Sauval, A. J.,

1337: 				 Allende Prieto, C., Kiselman, D. 2004,

1338: 				 \aap, 417, 751

1339: 

1340: \bibitem[Baldwin et al.(1981)]{bpt81} Baldwin, J. A., Phillips, M. M.,

1341: 				     \& Terlevich, R. 1981, \pasp, 93, 5 

1342: 

1343: \bibitem[Blanc et al.(2008)]{bla08} Blanc, G. A., et al. 2008, arXiv:0803.0763

1344: 

1345: \bibitem[Bolzonella et al.(2000)]{bol00} Bolzonella, M, Miralles, J. M.,

1346: 				     \& Pell$\rm \acute{o}$, R. 2000, \aap,

1347: 				     363, 476

1348: 

1349: \bibitem[Brooks et al.(2007)]{bro07} Brooks, A. M., et al. 2007, \apj,

1350: 				 655, L17 

1351: 

1352: \bibitem[Bruzual \& Charlot (2003)]{bru03} Bruzual, A. G., \& Charlot,

1353: 				   S. 2003, \mnras, 344, 100

1354: 

1355: 

1356: \bibitem[Calzetti et al.(2000)]{cal00} Calzetti, D., et al. 2000, \apj,

1357: 				   533, 682

1358: 

1359: \bibitem[Chabrier(2003)]{cha03} Chabrier, G. 2003, \pasp, 115, 763

1360: 

1361: \bibitem[Colbert et al (2005)]{colbert05} Colbert, J. W., Teplitz,

1362: 				     H. I., Yan, L., Malkan, M. A.,

1363: 				     McCarthy, P. J. 2005, \apj, 621,

1364: 				     587  

1365: 

1366: \bibitem[Coleman et al.(1980)]{cww80} Coleman, G. D., Wu, C.-C., \&

1367:                                  Weedman, D. W. 1980, \apjs, 43, 393 

1368: 

1369: 

1370: \bibitem[Daddi et al.(2004)]{dad04} Daddi, E., et al. 2004, \apj, 617, 746

1371: 

1372: 

1373: \bibitem[Daddi et al.(2007)]{dad07} Daddi, E., et al. 2007, \apj, 670,

1374: 				 156

1375: 

1376: \bibitem[Daddi et al.(2008)]{dad08} Daddi, E., Dannerbauer, H., Elbaz,

1377: 				 D., Dickinson, M., Morrison, G., Stern,

1378: 				 D., \& Ravindranath, S. 2008, \apj, 673, L21

1379: 

1380: \bibitem[Dannerbauer et al.(2006)]{dan06} Dannerbauer, H., et al. 2006,

1381: 				 \apj, 637, L5 

1382: 

1383: \bibitem[De Dossi et al.(2007)]{der07} De Rossi, M. E., Tissera, P. B.

1384: 				 \& Scannapieco, C. 2007, \mnras, 374, 323

1385: 

1386: \bibitem[Dickinson et al.(2003)]{dic03} Dickinson, M., Papovich, C.,

1387: 			    Ferguson, H. C., \& Budav$\rm \acute{a}$ri,

1388: 			    T. 2003, \apj, 587, 25 

1389: 

1390: 

1391: \bibitem[Erb et al.(2006a)]{erb06a} Erb, D. K., Shapley, A. E.,

1392: 				 Pettini., M., Steidel, C. C, Reddy,

1393: 				 N. A., \& Adelberger, K. L. 2006a,

1394: 				 \apj, 644, 813 

1395: 

1396: \bibitem[Erb et al.(2006b)]{erb06b} Erb, D. K., Steidel, C. C, Shapley,

1397: 				 A. E., Pettini., M., Reddy, N. A., \&

1398: 				 Adelberger, K. L. 2006b, \apj, 647, 128

1399: 

1400: \bibitem[Erb et al.(2006c)]{erb06c} Erb, D. K., Steidel, C. C, Shapley,

1401: 				 A. E., Pettini., M., Reddy, N. A., \&

1402: 				 Adelberger, K. L. 2006c, \apj, 646, 107

1403: 

1404: \bibitem[Gallazzi et al.(2005)]{gal05} Gallazzi, A., Chrlot, S.,

1405: 				 Brinchmann, J., White, S. D. M.,

1406: 				 \& Tremonti, C. A. 2005, \mnras, 362, 41

1407: 

1408: 

1409: \bibitem[Grazian et al.(2007)]{gra07} Grazian, A., et al. 2007, \aap, 465, 393

1410: 

1411: 

1412: \bibitem[Finlator \& Dav$\rm \acute{e}$(2008)]{fin08} Finlator, K. \&

1413: 				 Dav$\rm \acute{e}$, R. 2008, \mnras, 385, 2181

1414: 

1415: \bibitem[Fontana et al.(2003)]{fon03} Fontana, A., et al. 2003, \apj, 594, L9

1416: 

1417: 

1418: \bibitem[Halliday et al.(2008)]{hal08} Halliday, C., et al. 2008,

1419: 				      arXiv:0801.1193 

1420: 

1421: \bibitem[Hayashi et al.(2007)]{hay07} Hayashi, M., Shimasaku, K.,

1422: 			    Motohara, K., Yoshida, M., Okamura S. \&

1423: 			    Kashikawa, N. 2007, \apj, 660, 72 

1424: 

1425: \bibitem[Hicks et al.(2002)]{hicks02} Hicks, E. K., Malkan, M. A.,

1426: 				     Teplitz, H. I., McCarthy, P. J.,

1427: 				     and Yan, L. 2002, \apj, 581, 205  

1428: 

1429: \bibitem[Ichikawa et al.(2006)]{ich06} Ichikawa, T., et al. 2006

1430: 				 Proc. SPIE, 6269, 38

1431: 

1432: \bibitem[Kajisawa \& Yamada(2001)]{kaj01} Kajisawa, M., \& Yamada,

1433: 			    T. 2001, \pasj, 53, 833 

1434: 

1435: \bibitem[Kashikawa et al.(2004)]{kas04} Kashikawa, N., et al. 2004,

1436: 				     \pasj, 56, 1011 

1437: 

1438: \bibitem[Kauffmann et al.(2003)]{kau03} Kauffmann, G., et al. 2003,

1439: 				     \mnras, 346, 1055 

1440: 

1441: \bibitem[Kennicutt(1998)]{ken98} Kennicutt, R. C. 1998, \araa, 36, 189

1442: 

1443: \bibitem[Kewley \& Dopita(2002)]{kew02} Kewley, L. J. \& Dopita,

1444: 				 M. A. 2002, \apjs, 142, 35

1445: 

1446: \bibitem[Kobulnicky \& Kewley(2004)]{kob04} Kobulnicky, H. A. \& Kewley,

1447: 				 L. J. 2002, \apj, 617, 240 

1448: 

1449: \bibitem[Kriek et al.(2008)]{kri08} Kriek, M., et al. 2008,  arXiv:0801.1110

1450: 

1451: \bibitem[Kong et al.(2006)]{kon06} Kong, X., et al. 2006, \apj, 638, 72

1452: 

1453: \bibitem[Lane et al.(2007)]{lan07} Lane, K. P., et al. 2007, \mnras, 379, L25

1454: 

1455: \bibitem[Liu et al.(2008)]{liu08} Liu, X., Shapley, A. E., Coil, A. L.,

1456: 				 Brinchmann, J. \& Ma, C. P. 2008, \apj,

1457: 				 678, 758

1458: 

1459: \bibitem[Ly et al.(2008)]{ly08} Ly, C. et al. 2008, in preparation

1460: 

1461: \bibitem[Ly et al.(2007)]{ly07} Ly, C. et al., 2007, \apj, 657, 738

1462: 

1463: \bibitem[Maiolino et al.(2008)]{mai08} Maiolino, R., et al. 2008, \aap,

1464: 				     488, 463 

1465: 

1466: \bibitem[Malkan \& Filippenko(1983)]{mal83} Malkan, A. M., \&

1467: 				 Filippenko, A. V. 1983, \apj, 275, 477 

1468: 

1469: \bibitem[McCarthy et al.(1999)]{mccarthy99} McCarthy, P. J. et al.,

1470: 				 1999, \apj, 520, 548 

1471: 

1472: \bibitem[Peng et al.(2002)]{pen02} Peng, C. Y., Ho, L. C., Impey, C. D. \& Pix,

1473: 				 H.-W. 2002 \aj, 124, 266  

1474: 

1475: \bibitem[Pettini \& Pagel(2004)]{pet04} Pettini, M. \& Pagel,

1476: 				 B. E. J. 2004 \mnras, 348, L59  

1477: 

1478: \bibitem[Pickles(1998)]{pic98} Pickles, A. J. 1998, \pasp, 110, 863

1479: 

1480: \bibitem[Press et al.(1995)]{pre} Press, W. H., et al. 1995, Numerical recipes

1481: 				 (Cambridge: Cambridge University Press) 

1482: 

1483: 

1484: \bibitem[Reddy et al.(2006a)]{red06a} Reddy, N. A., et al. 2006a \apj, 644, 792

1485: 

1486: \bibitem[Reddy et al.(2006b)]{red06b} Reddy, N. A., Steidel, C. C., Erb,

1487: 				   D. K., Shapley, A. E., \& Pettini,

1488: 				   M. 2006b \apj, 653, 1004 

1489: 				 

1490: \bibitem[Richards et al.(2006)]{ric06} Richards, G. T., et al. 2006,

1491: 				     \aj, 131, 2766 

1492: 

1493: \bibitem[Salpeter(1955)]{sal55} Salpeter, E. E. 1955, \apj, 121, 161

1494: 

1495: \bibitem[Savaglio et al.(2005)]{sav05} Savaglio, S., et al. 2005, \apj,

1496: 				     635, 260

1497: 

1498: \bibitem[Sawicki(2002)]{saw02} Sawicki, M. 2002, \aj, 124, 3050

1499: 

1500: \bibitem[Spinoglio et al.(1995)]{spin95} Spinoglio, L., Malkan, M. A.,

1501: 				 Rush, B., Carrasco, L. \&

1502: 				 Recillas-Cruz, E. 1995, \apj, 453, 616 

1503: 

1504: \bibitem[Storchi-Bergmann et al.(1994)]{sto94} Storchi-Bergmann, T.,

1505: 				   Calzetti, D., \& Kinney, A. L. 1994,

1506: 				   \apj, 429, 572 

1507: 

1508: \bibitem[Takagi et al.(2003)]{takagi03} Takagi, T., Arimoto, N. \&

1509: 				 Hanami, H. 2003, \mnras, 340, 813

1510: 

1511: \bibitem[Tremonti et al.(2004)]{tre04} Tremonti, C. A., et al. 2004,

1512: 				   \apj, 613, 898 

1513: 

1514: \bibitem[Veilleux \& Osterbrock(1987)]{vei87} Veilleux S. \& Osterbrock

1515: 				     D. E. 1987, \apjs, 63, 295

1516: \end{thebibliography}

1517: 

1518: \clearpage

1519: \begin{deluxetable}{cccccc}

1520: %\rotate

1521: \tabletypesize{\small}

1522: \tablewidth{0pt}

1523: %\tablenum{1}

1524: \tablecolumns{6}

1525: \tablecaption{Photometric properties of 15 sBzKs with emission lines detected. \label{table;photo}}

1526: \tablehead{

1527:  \colhead{} &  Magnitude &

1528:  \multicolumn{2}{c}{Color} & \multicolumn{2}{c}{SED Fitting}            \\  

1529:  \colhead{ID}& \colhead{$K$} & \colhead{$B-z$} & \colhead{$z-K$} &

1530:  \colhead{stellar mass\tablenotemark{a}}& \colhead{$E(B-V)$}\\}

1531: 

1532: \startdata

1533: 21193&  21.3& 1.69& 1.89& 3.8 & 0.51\\ 

1534: 21360&  21.6& 1.61& 1.53& 1.5 & 0.43\\

1535: 21667&  22.1& 1.24& 2.15& 5.5 & 0.38\\

1536: 21973&  22.4& 1.09& 1.53& 1.9 & 0.31\\

1537: 22424&  22.2& 1.72& 2.65& 15  & 0.45\\

1538: 22602&  22.2& 1.23& 1.98& 5.1 & 0.35\\

1539: 25528&  22.3& 1.21& 1.49& 1.2 & 0.41\\

1540: 26411&  21.7& 1.47& 2.38& 5.4 & 0.44\\

1541: 26821&  21.8& 1.10& 1.24& 2.1 & 0.38\\

1542: 26840&  23.0& 0.73& 0.89& 0.61& 0.29\\

1543: 27677&  22.4& 1.37& 2.57& 3.6 & 0.45\\

1544: 27815&  22.9& 1.06& 1.45& 1.4 & 0.36\\

1545: 28761&  21.9& 1.04& 1.02& 1.5 & 0.35\\

1546: 31340&  21.6& 1.46& 2.53& 11  & 0.41\\

1547: 31569&  22.0& 0.78& 0.74& 0.61& 0.23\\

1548: \enddata

1549: 

1550: \tablenotetext{a}{The unit is $10^{10} \Msun$.}

1551: 

1552: \end{deluxetable}

1553: 

1554: \clearpage

1555: \begin{deluxetable}{cccccccc}

1556: %\rotate

1557: \tabletypesize{\small}

1558: \tablewidth{0pt}

1559: %\tablenum{1}

1560: \tablecolumns{8}

1561: \tablecaption{Spectroscopic properties of 15 sBzKs with emission lines detected.\tablenotemark{a} \label{table;spec}}

1562: \tablehead{

1563:  \colhead{ID} & \colhead{Detected Lines\tablenotemark{b}}&

1564:  \colhead{redshift}& \colhead{$F_{\Ha}$\tablenotemark{c}} &

1565:  \colhead{$L_{\Ha}$\tablenotemark{d}} &

1566:  \colhead{SFR$_{\Ha}$\tablenotemark{e}} &

1567:  \colhead{EW$_{\Ha}$\tablenotemark{f}} &

1568:  \colhead{12+log(O/H)\tablenotemark{g}}\\}

1569: 

1570: \startdata

1571: 21193& $\Ha$,[\ion{N}{2}]& 1.479& 

1572:      10.9$\pm$1.77& 1.5$\pm$0.24& 4.2$\pm$0.68& 1.8$\pm$0.56&

1573: 8.78$^{+0.08}_{-0.13}$\\ 

1574: 

1575: 21360& $\Ha$      & 1.174& 

1576:      4.16$\pm$2.25& 0.32$\pm$0.18& 0.49$\pm$0.27& 0.86$\pm$0.53&

1577: $<$8.84\\

1578: 

1579: 21667& $\Ha$      & 1.750& 

1580:      11.1$\pm$2.26& 2.3$\pm$0.47& 2.5$\pm$0.51& 4.5$\pm$1.8&

1581: $<$8.68\\

1582: 

1583: 21973& $\Ha$,[\ion{O}{3}]& 1.504& 

1584:      18.3$\pm$3.58& 2.6$\pm$0.51& 1.8$\pm$0.36& 8.5$\pm$8.4&

1585: $<$8.75\\

1586: 

1587: 22424& $\Ha$,[\ion{N}{2}] & 2.018& 

1588:      7.25$\pm$1.24& 2.1$\pm$0.37& 3.9$\pm$0.66& 7.0$\pm$6.9&

1589: 8.80$^{+0.07}_{-0.09}$\\

1590: 

1591: 22602& $\Ha$       & 1.872& 

1592:      7.72$\pm$2.35& 1.9$\pm$0.58& 1.7$\pm$0.52& 1.1$\pm$0.46&

1593: $<$8.88\\

1594: 

1595: 25528& $\Ha$,[\ion{N}{2}] & 1.506& 

1596:      9.99$\pm$2.00& 1.4$\pm$0.29& 2.0$\pm$0.40& 3.3$\pm$1.4&

1597: 8.94$^{+0.08}_{-0.12}$\\

1598: 

1599: 26411& $\Ha$,[\ion{N}{2}] & 1.498& 

1600:      14.0$\pm$2.67& 2.0$\pm$0.38& 3.3$\pm$0.63& 2.7$\pm$1.2&

1601: 8.70$^{+0.12}_{-0.23}$\\

1602: 

1603: 26821& $\Ha$,$\Hb$,[\ion{O}{3}],[\ion{N}{2}]& 2.044& 

1604:      22.2$\pm$1.43& 6.8$\pm$0.44& 7.3$\pm$0.47& 5.9$\pm$1.2&

1605: 8.66$^{+0.06}_{-0.07}$\\

1606: 

1607: 26840& $\Ha$,$\Hb$,[\ion{O}{2}],[\ion{O}{3}],[\ion{N}{2}]& 2.012& 

1608:      8.98$\pm$1.37& 2.6$\pm$0.40& 1.5$\pm$0.23& 6.6$\pm$4.6&

1609: 8.52$^{+0.09}_{-0.14}$\\

1610: 

1611: 27677& $\Ha$       & 1.710& 

1612:      4.53$\pm$1.39& 0.89$\pm$0.27& 1.6$\pm$0.50& 2.8$\pm$1.6&

1613: $<$8.88\\

1614: 

1615: 27815& $\Ha$,[\ion{N}{2}] & 1.796& 

1616:      7.75$\pm$1.01& 1.7$\pm$0.22& 1.7$\pm$0.22& 5.5$\pm$4.0&

1617: 8.60$^{+0.10}_{-0.18}$\\

1618: 

1619: 28761& $\Ha$,[\ion{N}{2}] & 1.818& 

1620:      10.3$\pm$4.90& 2.3$\pm$1.1& 2.1$\pm$1.0& 1.8$\pm$0.91&

1621: 8.68$^{+0.19}_{-8.68}$\\

1622: 

1623: 31340& $\Ha$,[\ion{N}{2}] & 1.719& 

1624:      7.10$\pm$1.39& 1.4$\pm$0.28& 2.0$\pm$0.39& 2.5$\pm$0.73&

1625: 8.81$^{+0.08}_{-0.13}$\\

1626: 

1627: 31569& $\Ha$,$\Hb$,[\ion{O}{3}],[\ion{N}{2}]& 1.250& 

1628:      33.0$\pm$2.05& 3.0$\pm$0.19& 1.1$\pm$0.07& 9.9$\pm$5.5&

1629: 8.28$^{+0.12}_{-0.25}$\\

1630: \enddata

1631: 

1632: \tablenotetext{a}{

1633:  For the two sBzKs classified as an AGN and excluded

1634:  from the discussions in \S 5, the SFR, equivalent width, and oxygen

1635:  abundance derived by the same manners as the other sBzKs are also

1636:  listed.}   

1637: \tablenotetext{b}{[\ion{N}{2}] means [\ion{N}{2}]($\lambda6584$).}

1638: \tablenotetext{c}{Observed flux. The unit is $10^{-17}$ ergs s$^{-1}$ cm$^{-2}$.}

1639: \tablenotetext{d}{Observed luminosity. The unit is $10^{42}$ ergs s$^{-1}$.}

1640: \tablenotetext{e}{The unit is $10^2 \Msun {\rm yr^{-1}}$. Dust-corrected.}

1641: \tablenotetext{f}{The unit is $10^2$ \AA. Rest-frame and dust-corrected.}

1642: \tablenotetext{g}{12+log(O/H) is derived from N2 index method.}

1643: \end{deluxetable}

1644: 

1645: \end{document}

1646: 

1647: