Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.arcetri.astro.it/science/k20/papers/18200.web.pdf Äàòà èçìåíåíèÿ: Tue Nov 30 19:23:23 2004 Äàòà èíäåêñèðîâàíèÿ: Fri Feb 28 07:57:24 2014 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: http astrokuban.info astrokuban

The Astrophysical Journal, 608:L29­L32, 2004 June 10
2004. The American Astronomical Society. All rights reser ved. Printed in U.S.A.

METAL ENRICHMENT IN NEAR-INFRARED LUMINOUS GALAXIES AT z 2: SIGNATURES OF PROTO­ELLIPTICAL GALAXIES?1 D. F. de Mello,2,3 E. Daddi,4 A. Renzini,4 A. Cimatti,5 S. di Serego Alighieri,5 L. Pozzetti,6 and G. Zamorani6
Received 2004 February 15; accepted 2004 April 23; published 2004 May 10

ABSTRACT ° ° We present the analysis of the co-added rest-frame UV spectrum (1200 A ! z ! 2000 A) of five K-luminous galaxies at z 2 from the K20 sur vey. The composite spectrum is characterized by strong absorption lines over the UV continuum from C, N, O, Al, Si, and Fe in various ionization stages. While some of these lines are interstellar, several among the strongest absorptions are identified with stellar photospheric lines. Most of the photospheric and interstellar features are stronger in the K-luminous composite spectrum than in Lyman break galaxies at z 3. This suggests higher metallicity and possibly also larger interstellar velocity dispersion caused by macroscopic motions. The absorption lines and the slope of the UV continuum are well matched by the spectrum of the nearby luminous infrared galaxy NGC 6090, which is in the process of merging. A metallicity higher than solar is suggested by comparing the pure photospheric lines (Si iii, C iii, and Fe v) with starburst models. The evidence of high metallicity, together with the high masses, high star formation rates, and possibly strong clustering, well qualify these galaxies as progenitors of local massive elliptical galaxies. Subject headings: galaxies: evolution -- galaxies: formation -- galaxies: high-redshift -- galaxies: starburst
1. INTRODUCTION

Despite the fact that as much as 75% of the stellar mass in the universe is in early-type galaxies (e.g., Fukugita et al. 1998), the epoch at which these massive galaxies have assembled and acquired their high metallicity is still unknown. Although the progenitors of these massive galaxies should be seen at high redshifts, connecting them with their high-redshift counterparts remains a difficult task. The currently best-known high-z galaxy population is that of Lyman break galaxies (LBGs; Steidel et al. 2003). As in local starbursts, the strongest features in the rest-frame UV spectrum of LBGs are interstellar and photospheric absorption lines of C, N, O, Si, and Fe, typical of hot massive stars (e.g., Shapley et al. 2003, hereafter S03). They have subsolar metallicity (Pettini et al. 2001), a ver y young stellar population (!0.2 Gyr), and low masses (1010 M,). Moreover, it appears that the spatial clustering of z p 3 LBGs is too low for them to be direct progenitors of z 1 extremely red objects and local massive galaxies (Daddi et al. 2001). While these UV-selected LBGs are certainly an important component of the high-z galaxy population, it is still not clear how representative of the entire population at high-z they are. There might be other objects at high-z that would not be picked up by this selection. In fact, there are evidences that K-selected galaxies at 2 ! z ! 4 are much more strongly clustered than UV-selected ones (Daddi et al. 2003). For instance, the Kselected galaxies that are in the so-called redshift desert at 1.5 ! z spec ! 2.5 (Daddi et al. 2004, hereafter D04) appear to be an important class of objects, with overall physical properties quite distinct from those of LBGs. Comparing the properties
1 Based on obser vations taken with the European Southern Obser vator y, Chile, ESO programs 70.A-0140 and 168.A-0485. 2 Laborator y for Astronomy and Solar Physics, Code 681, Goddard Space Flight Center, Greenbelt, MD 20771. 3 Department of Physics, Catholic University of America, 620 Michigan Avenue, Washington, DC 20064. 4 ESO, Karl-Schwarzschild-Strasse 2, D-85748 Garching, Germany. 5 INAF­Osser vatorio Astronomico di Arcetri, Largo E. Fermi 5, I-50125 Florence, Italy. 6 INAF­Osser vatorio Astronomico di Bologna, via Ranzani 1, I-4027 Bologna, Italy.

of LBGs at z p 3 from S03 with the sample investigated in this Letter taken from D04, it appears that K-band luminous galaxies at z 2 have higher reddening [E( B V ) 0.3­0.4 vs. 0.10­0.17], redder UV slopes ( b 0.54 vs. 0.73­1.09), higher star formation rates (SFRs; 100­500 M, yr 1 vs. 25­ 50 M, yr 1), older ages (700 Myr vs. 100­300 Myr), and, therefore, higher masses (1011 M, vs. 1010 M,). These differences are likely arising as a consequence of the K-selection, as opposed to the UV selection for LBGs. Given their observational properties, it has been argued that these K-luminous galaxies may represent better candidates than LBGs as starforming progenitors of massive early-type galaxies (D04). If these objects are indeed the progenitors of local early-type galaxies, one would expect them to harbor more metal-rich stellar populations and interstellar medium, compared to LBGs. In this Letter we revisit these K-luminous galaxies by analyzing their composite spectrum and comparing it with those of various templates, including LBGs and starburst galaxies, in order to gather new insight on their metal enrichment, on their nature, and on their role in galaxy evolution.
2. THE SPECTRA

The K20 sur vey obtained Ver y Large Telescope (VLT) spectroscopy of a complete sample of about 500 galaxies with K s ! 20 (Cimatti et al. 2002). In this Letter, we focus on the spectral properties of five K20 galaxies in the Chandra Deep Field­South, Great Obser vator y Origins Deep Sur vey­South field at 1.7! z ! 2.3, with the highest signal-to-noise ratio (S/N) among the nine presented in D04. Including the even redder K20 galaxies with no spectroscopic redshifts and 1.7 z phot 2.3, the nine objects represent 20%­30% of z 2 K-selected galaxies. This may introduce a bias toward the UV brightest objects, possibly minimizing the differences with respect to LBGs. The average spectrum nevertheless shows significantly different features when compared with LBGs as shown in the following sections. The individual spectra were reduced to a common resolution ° of 25 A in the obser ved frame, by convolving them with a top-hat function. This allowed to match the lowest resolution of some of the spectra taken with grism 150I at R p 200 with L29


L30

DE MELLO ET AL.

Vol. 608

VLT FORS2. The spectra were normalized to a common mean ° ° flux level within the range 1300 A ! l ! 2000 A, deredshifted and then co-added to produce a representative average spectrum at AzS p 2.1. A flux-normalized version of the average spectrum was obtained by fitting the continuum with a spline func° tion. The resolution of the composite spectrum is low (8 A in the rest frame); however, the strong features seen in their average spectra can give important information on the nature of these objects. In order to evaluate the S/N of the final coadded spectrum, we used relatively line-free regions in the continuum-normalized spectrum. In three such regions at 1270­ ° 1287, 1440­1470, and 1876­1904 A we measure S/ N 30 ° over the 1 A pixel sampling. As all the spectra were boxcar ° smoothed to match the 8 A resolution of the final average, such an estimate may be taken as representative of the S/N over a ° full 8 A resolution element. The true S/N is likely higher considering that those regions will actually contain faint lines, given that the co-added spectrum is completely filled with ab° sorption lines in the 1200­2000 A domain investigated here.
3. COMPARISON WITH LBGs

The LBG class as a whole has been recently analyzed by S03 using the co-added spectra of almost 1000 low S/N spectra of LBGs at z 3, divided in four subsamples according to the EW of Lya. In order to bracket the range of LBG properties, we show in Figure 1 the comparison with the S03 LBGs' templates having the lowest Lya EW (i.e., Lya in absorption, group 1 in S03) and the largest Lya EW (i.e., Lya in emission, group 4 in S03). The former comparison is perhaps the most appropriate, as none of the K-luminous galaxies at z 2 has Lya in emission, and also because the LBGs with no Lya emission have the reddest UV slopes (S03), although not as red as those of the K-selected galaxies. The spectrum of both high-z galaxy classes is dominated by absorption lines. The Lya absorption LBG spectrum and that of the K-luminous galaxies show a rather good match of the Si ii l1526 and C iv l1550 features. These are the two strongest features in the LBGs' composite spectum. As noted by S03, the interstellar rather than the wind component dominates the ° Si iv doublet at 1400 A in the LBGs, while the C iv feature exhibits both the blueshifted broad absorption and the redshifted emission associated with the stellar winds, in addition to a strong, narrower, interstellar absorption component. The lack of resolution in our composite spectrum smooths out the wind signatures; however, at the same resolution, the overall shape of the C iv feature is similar in both spectra, whereas the Si iv interstellar line is stronger in the K-luminous galaxies. Other interstellar lines such as C ii l1335 and Al ii l1671 are also appreciably stronger in the K-luminous galaxies than in LBGs. All absorption features are much weaker in the the other LBGs' composites, having stronger Lya emission (Fig. 1, bottom) than in the K-luminous galaxies. The most striking differences between the LBG composite spectra and the K-luminous composite are the absorption lines ° at 1295, 1380, 1430, and 1485 A, which are remarkably strong in the latter and almost undetected or ver y weak in the former. We have identified these features as the photospheric absorptions of Si iii l1296, C iii l1428, and Si ii l1485 (de Mello et al. 2000). Close to these lines there are also photospheric lines of Fe ii, Fe v, and N iv] that could be blended ° with them because of the low resolution. The 1380 A feature is probably a blend of several Fe photospheric lines in the ° region 1360­1390 A (Leitherer et al. 2001). In addition, the

Fig. 1.--Composite spectra of K-luminous galaxies and LBGs. Top and bottom panels show the LBGs with Lya in absorption and emission, respectively (Shapley et al. 2003). The spectra are continuum normalized, and major absorption lines are identified. Features marked with an asterisk are photospheric. The resolutions of the LBG spectra were degraded to that of our composite.

weak feature seen on the red side of Si ii l1526 is likely due ° to the photospheric lines Si ii and Fe iv at 1533 A, while the red broad edge of Si ii l1304 seems to be due to the presence of the photospheric Si ii l1309. We interpret the strong photospheric lines in our spectrum as a clear indication of a significantly higher metallicity in the stellar component of the Kluminous galaxies at z 2 compared to the LBGs at z 3, which have subsolar metallicity. Besides, to a higher metallicity, the stronger interstellar lines can also be due to a higher interstellar velocity dispersion caused by large-scale inhomogeneities and macroscopic motions in the interstellar medium, as noted in other high-z galaxies such as MS 1512-cB58 (e.g., de Mello et al. 2000).
4. COMPARISON WITH LOCAL STARBURSTS
AND STARBURST MODELS

The detection of several strong stellar lines in the average spectrum of K-luminous galaxies at z 2 unambiguously indicates high metallicity in these galaxies, to an extent that has no counterpart among other known high-redshift galaxies. Therefore, we have checked whether a similar spectrum is exhibited by any of the most well-studied local starbursts galaxies. We found that the local starbursts, NGC 1705 1, NGC 1741, and NGC 4214, have ver y different spectral features than those of our K-selected ones. A good match is obtained only with the spectrum of NGC 6090, both for the absorption lines as well as for the continuum (Fig. 2). Note in particular that NGC 6090 has an almost identical strength of the 1295, 1430, ° and 1485 A absorption systems and a marginally weaker ° 1380 A absorption. These are the undetected or ver y weak lines in LBGs, which were identified as photospheric lines in the previous section. The interstellar lines of NGC 6090 appear ° on average systematically blueshifted by 1­2 A with respect to those of the K-luminous galaxies, unlike the stellar ones,


No. 1, 2004

NEAR-IR LUMINOUS GALAXIES AT z 2

L31

Fig. 2.--Composite spectrum of K-luminous galaxies is compared to NGC 6090 (Gonzalez-Delgado et al. 1998). The spectra are overplotted with only ° a common scaling to the same average flux level in 1400­1600 A, with no alteration of their intrinsic continuum shape. Major absorption lines are identified. The strong emission line in NGC 6090 is Lya.

° Fig. 3.--Zoom of the 1430 A index region. We show a comparison to L99 models: dashed lines are for solar metallicity; dotted lines are for 0.25 Z,. Spectra ° are continuum normalized. L99 spectra are shown with their 1 A resolution. If ° degraded to 8 A resolution also the Si iii l1417 absorption appears weaker than in the K-luminous z 2 galaxies.

suggesting a weaker effect of winds. This difference is not apparent toward LBGs. Although no strong conclusion can be derived from a single object, this match is intriguing because the overall properties of NGC 6090 are similar to the K-selected galaxies (see D04). NGC 6090 is an interacting system in the process of merging with strong star formation. It is a luminous infrared galaxy (log L IR p 11.51; Scoville et al. 2000) with a number of luminous clusters triggered by the galaxy-galaxy interaction. The other local starbursts, on the other hand, are dwarf galaxies having likely their first major burst of star formation. For instance, NGC 1705 1 is a super­star cluster (SSC) in a dwarf galaxy that experienced a strong burst of star formation 10 Myr ago (de Mello et al. 2000); NGC 1741 is an interacting system with young SSCs (a few to 100 Myr) of masses 104­106 M, (Johnson et al. 1999); NGC 4214 is a Magellanic irregular with SSCs of ages 1­3 Myr old (MaizApellaniz et al. 1998). Like NGC 6090, the K-luminous galaxies are massive galaxies with spatially extended starbursts. We have also checked whether the composite spectrum of the K-luminous galaxies can be reproduced by the Starburst99 stellar evolutionar y synthesis models (Leitherer et al. 1999, hereafter L99). We first investigated the contraints given by the strength of the C iv l1550 absorption system. In the models with instantaneous star formation, this feature rapidly disappears after the cessation of star formation (see Fig. 5 in de Mello et al. 2000). Therefore, its detection in the composite spectrum of the z 2 galaxies demonstrates that they contain a population of young stars. This supports the notion that the red slopes of the UV spectra of the K20 star-forming galaxies at z 2 are due to reddening and not to aging of their stellar populations, confirming that these galaxies are active starbursts and thus relevant contributors to the star formation density at z 2, as discussed in D04. At the same time, we do not detect the strong P Cygni shape of C iv l1550, typical of massive short-lived O stars. As a consequence, a single burst that occurred within 3 Myr can be ruled out. Instead, models with

continuous star formation reproduce well the strength of C iv l1550 at practically any older age. Second, we used L99 to obtain constraints on the overall metallicity of the K-luminous galaxies. We used the EW of the absorption features in the ° wavelength range 1415­1435 A as a "metallicity index" that does not strongly depend on age and initial mass function (Leitherer et al. 2001). Figure 3 shows a zoom of this region, which covers the photospheric features Si iii l1417, C iii l1427, and Fe v l1430. As can be seen in Figure 3, the three lines are detected, but because of the low resolution C iii and Fe v are blended. We attempted to deblend the two lines by fitting the feature with two Gaussian profiles and found that the central wavelengths of two Gaussians do coincide with the C iii and Fe v wavelengths. We measured the EW of the index ° to be EW p 2.3 0.4 A. The same absorption system in the ° models has EW p 0.5 and 1.2 A for 0.25 and 1 Z,, respectively. Therefore, the K-luminous galaxies have a metallicity index higher than the L99 0.25 Z, metallicity at the 4­5 j level and higher than the L99 Z, metallicity at the 3 j level. ° The 0.4 A error that we derive for the metallicity index may be slightly underestimated because of mismatches in the continuum normalization. However, the obser ved EW of the 1415­ ° 1435 A feature would be even higher if considering also the wings of the lines that, because of the low resolution, expand ° ° outside the 1415­1435 A region (EW p 2.8 A in the region 1412­1440). Recently, Steidel et al. (2004) also used the ° 1425 A index and L99 to estimate the metallicity of UVselected galaxies at 1.4 ! z ! 2.5 and found that they have metallicity near solar. The fact that our sources have even stronger absorptions in this range suggests that their metallicity is generally higher than that of Steidel et al. UV-selected galaxies at z 2. This is quite plausible given that the K-selected galaxies are likely to be on average more massive than the UV-selected ones. Note, however, that the two classes of objects are not entirely distinct, as 9% of the UV-selected galaxies at z 2 are brighter than K p 20 (Steidel et al. 2004). Finally, we


L32

DE MELLO ET AL.

Vol. 608

° emphasize that besides the 1425 A index, other evidences exist that point toward a high metallicity. As we showed before, there are several photospheric and interstellar lines that are much stronger in the composite spectrum than in the Z, L99 models and than in the LBGs' composite spectrum, suggesting that K-luminous galaxies have high metallicity. Similarly to Savaglio et al. (2004), strong Fe ii and Mn ii lines are detected ° in the near-UV (2200­2700 A) region (not shown here), consistently with high metallicity.
5. DISCUSSION

The most striking feature of the near-UV spectrum of Kband luminous galaxies at z 2 is the presence of strong stellar absorption features, with strengths not previously reported at high redshifts that suggest a high metallicity for this class of high-z galaxies. Stellar photospheric lines are an important diagnostic of the stellar content of galaxies, and their detection also demonstrates a UV continuum without a dominant active galactic nucleus contribution. These stellar features are typical of B-type stars (de Mello et al. 2000), which are dominating the UV flux in the explored spectral region, rather than shortlived ver y massive O stars. This can be well expected given that the strong Balmer breaks detected in the near-IR imply high-mass contents and a continuous SFR for 700 Myr (see D04). The fact that our composite spectrum was generated with a small number of galaxies may affect these results, as a few peculiar objects could be present. Co-addition of a larger number of spectra is desirable for the future in order to smooth out the contribution of individual objects. As these K-selected starbursts are found to have redder UV

slopes and higher reddening than UV-selected galaxies at 1.5 ! z ! 3.5 (the UV selection requires blue and flat slopes), it seems reasonable to find that they have also a higher metal content. It is not clear whether these galaxies are evolutionar y descendants of z 3 LBGs. The higher metallicity and dust content could be a consequence of the continuous star formation that occurred between the two epochs. However, if the D04 tentative estimate of a correlation length of r0 1 7 h 1 Mpc for the K-selected galaxies at z 2 is confirmed, they seem to be more clustered than the LBGs at z 3. This suggests a different evolutionar y path for the two populations. Sawicki & Yee (1998) suggested that LBGs either are the progenitors of present-day sub-L* galaxies or may form luminous galaxies through mergers as they evolve. However, one would expect a stronger clustering in the latter case (Daddi et al. 2001). The signatures of their high metallicity, the confirmation of ver y high SFRs, together with the evidences for high masses and possibly strong clustering reinforce the suggestion that these K-luminous objects at z 2 well qualify as progenitors of local early-type galaxies, caught while still in the act of actively forming stars. Indeed, with their masses being estimated above 1011 M,, they would most likely become earlytype galaxies at z p 0. We thank C. Steidel for providing the composite spectra of LBGs and S. Savaglio, D. Thomas, R. Gonzalez-Delgado, and the anonymous referee for useful suggestions. Support for this work was provided through NASA grant GO09481.1 from STScI, which is operated by the Association of Universities for Research in Astronomy, Inc., under contract NAS5-26555.

REFERENCES Cimatti, A., et al. 2002, A&A, 392, 395 Daddi, E., et al. 2001, A&A, 376, 825 ------. 2003, ApJ, 588, 50 ------. 2004, ApJ, 600, L127 (D04) de Mello, D. F., Leitherer, C., & Heckman, T. M. 2000, ApJ, 530, 251 Fukugita, M., Hogan, C. J., & Peebles, P. J. E. 1998, ApJ, 503, 518 Gonzalez-Delgado, M., et al. 1998, ApJ, 495, 698 Johnson, K. E., et al. 1999, AJ, 117, 1708 Leitherer, C., et al. 1999, ApJS, 123, 3 (L99) Leitherer, C., et al. 2001, ApJ, 550, 724 Maiz-Apellaniz, J., et al. 1998, A&A, 329, 409 Pettini, M., et al. 2001, ApJ, 554, 981 Savaglio, S., et al. 2004, ApJ, 602, 51 Sawicki, M., & Yee, H. K. C. 1998, AJ, 115, 1329 Scoville, et al. 2000, AJ, 119, 991 Shapley, A. E., et al. 2003, ApJ, 588, 65 (S03) Steidel, C. C., et al. 2003, ApJ, 592, 728 ------. 2004, ApJ, 604, 534