Vigorous star formation with low efficiency in massive disk galaxies at \H@@footnotemark \@makefnmark Based on observations carried out with the IRAM Plateau de Bure Interferometer. IRAM is supported by INSU/CNRS (France), MPG (Germany) and IGN (Spain).\@finalstrut
We present the first detection of molecular gas cooling CO emission lines from ordinary massive galaxies at . Two sources were observed with the IRAM Plateau de Bure Interferometer, selected to lie in the mass-star formation rate correlation at their redshift, thus being representative of massive high- galaxies. Both sources were detected with high confidence, yielding K km s pc. For one of the sources we find evidence for velocity shear, implying CO sizes of kpc. With an infrared luminosity of , these disk-like galaxies are borderline ULIRGs but with star formation efficiency similar to that of local spirals, and an order of magnitude lower than that in submm galaxies. This suggests a CO to total gas conversion factor similar to local spirals, gas consumption timescales approaching 1 Gyr or longer and molecular gas masses reaching , comparable to or larger than the estimated stellar masses. These results support a major role of in situ gas consumption over cosmological timescales and with relatively low star formation efficiency, analogous to that of local spiral disks, for the formation of today’s most massive galaxies and their central black holes. Given the high space density of similar galaxies, Mpc, this implies a widespread presence of gas rich galaxies in the early Universe, many of which might be within reach of detailed investigations of current and planned facilities.
Galaxies in the distant Universe were forming stars with much higher rates than today (Madau et al 1996; Elbaz et al 2002; Le Floc’h et al. 2005). For galaxies of a fixed stellar mass, e.g. of the order of , the typical star formation rate (SFR) has been steadily increasing with respect to local values, up to times higher at (Daddi et al. 2007a; D07a hereafter). This increase can be accounted for either by postulating a higher gas content or a higher star formation efficiency, or both. Galaxy mergers and interactions, expected to be much more common in the distant Universe than today, can increase the density of gas in galaxies, and thus help transform gas into stars more rapidly and with a higher star formation efficiency. The relative importance for massive galaxy formation and assembly of the hierarchical merging of small sub galactic units, versus the in situ formation of stars through gas consumption is largely unknown observationally.
The molecular gas content of galaxies is best probed through the detection of redshifted cooling emission lines of CO. Combining observations for local and distant galaxies, all the way from spirals to (ultra) luminous infrared galaxies (LIRGs, , and ULIRGs, ), a nonlinear correlation is observed (e.g., Solomon & van den Bout 2005) between CO luminosity (proportional to molecular gas content) and FIR luminosity (proportional to the SFR), with more luminous sources having higher FIR to CO luminosities, i.e. higher efficiencies than spiral disks in converting gas into stars, likely due to their higher gas densities (Tacconi et al. 2006; Downes & Solomon 1998). Mapping the physics of molecular gas in representative galaxies at high redshifts would be a major step forward into understanding the processes regulating galaxy formation, and is a key target of future astrophysical facilities such as ALMA.
However, if the local correlations hold at high redshifts, current techniques and sensitivities allow us to detect CO lines, and thus study the molecular gas content, only for extreme objects such as bright ( mJy) submm-selected galaxies (Genzel et al. 2004; Neri et al. 2003; Greve et al. 2005). (SMGs), quasars (Walter et al. 2003; Maiolino et al. 2005), or in the case of strong gravitational lensing. Indirect evidence has been mounting very recently that this might not be the general case. Vigorous starbursts at high redshift appear to be the ordinary star formation mode in massive galaxies. The existence of tight correlations between SFRs and galaxy masses (Elbaz et al. 2007; Noeske et al. 2007; D07a) and the predominance of ULIRGs inside mass limited samples of galaxies (Daddi et al 2005; D07a; Caputi et al 2006) suggests that vigorous starbursts were typically long lasting among ordinary massive galaxies in the distant Universe, which therefore might be hosting large amounts of gas.
2 CO observations of BK galaxies
Motivated by this evidence, we observed two representative massive galaxies at high redshift with the IRAM Plateau de Bure Interferometer, searching for the redshifted emission lines from the CO [2-1] transition at rest frame 230.539 GHz, with the aim of constraining their molecular gas content. The two galaxies, -4171 (J123626.53+620835.3, VLA radio frame) and -21000 (J123710.60+622234.6), at redshift and , respectively (measured from Keck DEIMOS spectroscopy), were selected in the Great Observatories Origins Deep Survey Northern field (GOODS-N) using the technique (Daddi et al 2004b), which provides complete -limited samples of normal, massive galaxies at . SFRs of about 100-150 yr were derived by averaging estimates based on the VLA radio detections of Jy and Jy (using the radio- correlation, Condon et al. 1992) and UV estimates (correcting for dust reddening using the Calzetti et al. 2000 law), which agree within 0.2 dex or better. For the 24m estimate (based on the Chary and Elbaz 2001 luminosity dependent templates) also agrees within this factor, while for the 24m estimate of SFR is 2-2.5 larger than the radio and UV ones, mild evidence for possible mid-IR excess usually linked to obscured AGN activity (Daddi et al. 2007b; D07b hereafter). Stellar masses of 0.5–0.8 were derived from optical and UV rest frame photometry, following the method of Daddi et al. (2004b). SFRs and stellar masses are given here for the Chabrier (2003) stellar initial mass function. The two target galaxies lie in the mass-SFR correlation established at their redshifts (D07a), and are thus ordinary massive high- sources with ongoing star formation.
The sources were observed using newly refurbished receivers for 5.7 and 7.5 hours during April 24-25th 2007 for -4171 and -21000, respectively, during stable and high quality conditions. A compact configuration-D with 6 antennas was used. Additional 7.5 hours of poorer quality weather data in configuration-D and with 5 antennas only were obtained for -4171 on August 1st 2007, reducing the noise by only 25%. Standard data reduction was performed using the GILDAS software. Both sources observed were detected in CO with high confidence (Fig. LABEL:Pic, top panels), with a significance of 6.8 and 8.9. Fluxes are Jy km s and Jy km s, for -4171 and -21000, respectively. Fig. LABEL:Spe shows the CO spectra of the two galaxies. The CO lines wavelengths are in agreement with the Keck redshifts, within 50 km s or better. The CO positions agree with the VLA positions within . -4171 is best fitted by a double horn profile, indicative of a rotating disk and consistent with the UV morphology of a nearly edge on disk. -21000 shows a narrower velocity profile with only a hint of double-horns. Fig. LABEL:Pic (bottom panels) shows position-velocity diagram, derived using a pseudo-slit oriented along the major axis defined in the HST+ACS imaging. -4171 is unresolved, while -21000 shows evidence for shear, overall consistent with rotation in a disk. As a cross-check, we separately fitted velocity channels bluer and redder than the systemic velocities. For -4171 the positions of the blue and red horns are separated by . For -21000 we find separation ( kpc; ), definying a position angle of deg, in very good agreement with the UV position angle, confirming that the CO emission is extended.
3 Gas rich massive galaxies at
These results unveil a new and different galaxy formation mode for the most luminous sources like LIRGs and ULIRGs. In Fig. LABEL:fig:XCO we compare CO to FIR luminosities of our galaxies to local and distant sources. We adapted published luminosities for local (Solomon et al. 1997; Yao et al 2003) and distant sources (Greve et al. 2005) to WMAP cosmological parameters (Spergel et al. 2007). We assume here roughly constant brightness temperature (see Braine & Combes 1992) and thus conversion ratios among the different CO tracers to be of order unity, i.e. CO[1-0]CO[2-1]CO[3-2] (Weiß et al 2005). Compared to local ULIRGs, these ordinary massive galaxies have similar bolometric luminosities but larger CO luminosities by a factor of 3–5. On the other hand, SMGs have estimated and therefore (Kennicutt et al. 1998) yr, one order of magnitude higher than our two targets, but the observed CO luminosities are comparable. The two galaxies are outliers of the versus correlation defined by local spirals, LIRGs/ULIRGs and distant SMGs (Fig. LABEL:fig:XCO). Instead, they lie close to a linear to relation normalized to local spirals. The observed ratio in the two massive galaxies is an order of magnitude lower than that of SMGs at similar redshifts, implying much lower star formation efficiency. This is consistent with the fact that SMGs are outliers of the mass-SFR correlation (D07a), probably due to the higher efficiency in forming stars, for a similar stellar mass and CO luminosity. The to ratios of the two galaxies are within the semi-interquartile range of values from local spirals (Fig. LABEL:fig:XCO, right panel). No object with similar to ratio and large was previously known in the distant Universe (Fig. LABEL:fig:XCO). This implies that these distant BzK galaxies behave as scaled up versions of local spirals, with proportionally higher SFR and molecular gas content for a fixed galaxy stellar mass.
We use a conversion rate (Kennicutt et al. 1998) [ yr], adapted to a Chabrier (2003) stellar initial mass function and in the assumption, appropriate for these galaxies, that is about 80% of the bolometric luminosity. In general, we can write , where is the total molecular gas mass (including He) and the conversion factor (Downes & Solomon 1998; Solomon & van den Bout 2005) is (K km s pc) for the Milky Way and as low as (K km s pc) for ULIRGs and likely for SMGs. The high detected CO luminosities imply that giant molecular gas reservoirs are present, of order of , for a ULIRG-like , comparable to SMGs and larger than in local ULIRGs. The typical gas consumption timescales (Fig. LABEL:fig:XCO) are about 2–4 yr for SMGs, a factor of larger for local ULIRGs and reach 2–3 yr for the massive galaxies at , in the case of . The SFR could last for even longer than that, as further gas accretion will be substantial during such long timescales. Hydrodynamical simulations from Keres et al. (2005) predict gas accretion rates of about 10-70 yr for similarly massive galaxies at (i.e., for dark matter halo masses of ; Hayashi et al 2007). This is the same order of magnitude of the estimated SFR in massive galaxies. These are likely lower limits to the SFR duration and gas masses as we are entirely neglecting atomic gas. As the star formation efficiencies are consistent with those of local spirals, in analogy to the latter it might be that or even that the Milky Way conversion factor applies to these massive sources.
This is further supported by the evidence that, unlike local ULIRGs (Downes & Solomon 1998) and distant SMGs (Tacconi et al 2006), these two galaxies, like massive star forming galaxies (Daddi et al 2004a; Bouche et al 2007) in general, are not compact merger-induced starbursts. From detailed Sérsic fitting of their UV rest frame light profiles from Hubble Space Telescope imaging data obtained with Advanced Camera for Surveys (Giavalisco et al. 2004), we derive half-light radii of 4.4 and 5.5 kpc (proper size, or about 0.5, accurate to better than 5%). Both galaxies have disk-like (exponential) morphology with Sérsic index and 0.6 (accurate to better than 20%). BzK-4171 appears indeed as a nearly edge on disk, consistent with the Sérsic fitting. BzK-21000 has a clumpy structure, still consistent with a disk but also with the possible presence of minor merging events. The UV rest frame sizes are likely reasonably accurate measurements of the starbursts sizes, because for these sources reliable SFRs measurements can be obtained from the UV (Daddi et al 2005; D07a; Dannerbauer et al 2006; Reddy et al 2005), i.e. their UV emission is not entirely opaque, while local ULIRGs and distant SMGs are optically thick in the UV (Champan et al 2005; Goldader et al 2002). This is confirmed by the evidence for extended CO emission, at least for BzK-21000. The inferred sizes are a factor of 2-3 larger than distant SMGs (Tacconi et al 2006) and 10 times larger than local ULIRGs (Downes & Solomon 1998). This converts into 1–2 order of magnitudes lower gas densities and justifies the relatively modest star formation efficiencies. Independent estimates of the molecular gas masses and thus SFR timescales, can also be obtained. First, if we apply the Schmidt-Kennicutt law (using its latest derivation from Bouche et al 2007, that includes also distant SMGs), using the UV size for the typical gas and SFR size, we estimate for each galaxy. Second, from the CO velocity widths and (UV and/or CO) size measurements of the galaxies we estimate dynamical masses of about 1–1.5, within the half light radii, including gas, stars and dark matter. Given the estimated total stellar masses of –0.8, roughly half of which presumably lying within the half-light radii, it looks plausible that of gas can be present. We conclude that it would take up to 1 Gyr to consume all of the gas in these typical, massive high-redshift galaxies, assuming the star-formation rate remained roughly constant. Clearly, the star formation episodes could last longer since additional gas is accreted over such long timescales.
In order to evaluate the implications of these results for massive galaxy formation, recall that the two observed galaxies are representative of massive galaxies at high redshifts. They were selected for CO observations simply due to the availability of known spectroscopic redshifts. The two galaxies lie on the SFR to stellar mass correlation inferred at , and their colors and morphologies are fully typical of massive galaxies. Star forming galaxies in that redshift range with similar stellar masses are common in the distant Universe, with space densities of order of Mpc, quite a bit larger than the Mpc observed for SMGs by Chapman et al. (2003). Clearly, these results point to the existence of a much larger population of gas rich galaxies in the distant Universe than previously thought.
We notice that the far-IR luminosity estimates for our objects are based on the UV, radio and mid-IR luminosities, and are solid given that self-consistent measurements of SFR activity have been shown to be obtainable for these kind of objects at high redshift (D07a). It is very unlikely that we are strongly underestimating their star formation activity (which would lead us to overestimate the duration timescales). Instead, the for SMGs from Greve et al. (2005) is based only on the m flux and is likely more uncertain. Using radio and Spitzer, Pope et al. (2006) suggest lower for SMGs. This could imply that SMGs are closer in their SFR modes to massive galaxies, but would change little about our conclusion because it is the latter which represent the dominating star formation mode at high redshift.
The large molecular gas fraction in these galaxies, possibly up to 50% or more, would likely give rise to large scale instabilities, with the formation of clumps (e.g., Immeli et al. 2004; Bournaud, Elmegreen & Elmegreen 2007). This could explain the clumpy structure of these two galaxies, and in general of a large fraction of the high- massive galaxies (e.g., Daddi et al. 2004a), and the high velocity dispersions observed in many of the high- massive disk galaxies (e.g., Genzel et al. 2006; Forster-Schreiber et al. 2006).
The long duration timescale for star formation activity implies, as a corollary, long duration activity for the galactic nuclei. In fact, it has been recently unveiled by D07b that up to 20-50% of massive galaxies in likely contain strongly obscured but relatively luminous AGNs. Some of the large gas reservoir is likely slowly funneled, in some way, into the central black hole with similar timescales to star formation, at a rate roughly consistent with what is required to assemble the correlation between stellar and black hole masses as we observe today (D07b). Our results underline the major role of gas consumption over long timescales and with low efficiencies, as opposed to rapid and strong merger driven bursts, as a major growth mode for both stellar mass and black holes in the distant Universe.
We thank Roberto Neri and Arancha Castro-Carrizo for assistence with taking and reducing IRAM Plateau de Bure Interferometer data. We acknowledge very useful discussions, at various stages of this work, with Kartik Sheth, Frederic Bournaud, Pierre-Alain Duc and Fabian Walter. Comments by an anonymous referee improved the paper. We acknowledge the use of GILDAS software (http://www.iram.fr/IRAMFR/GILDAS). Support for this work, part of the Spitzer Space Telescope Legacy Science Program, was provided by NASA, Contract Number 1224666 issued by the JPL, Caltech, under NASA contract 1407.
- ( ) Bouche N., et al., 2007, ApJ in press (arXiv:0706.2656)
- ( ) Bournaud F., Elmegreen B.G & Elmegreen D.M., 2007, ApJ 670, 237
- ( ) Braine J., Combes F., 1992, A&A, 264, 433
- ( ) Calzetti D., Armus L., Bohlin R. C., et al., 2000, ApJ, 533, 682
- ( ) Caputi K.I., et al., 2006, ApJ, 637, 727
- ( ) Chabrier G., 2003, ApJ, 586, L133
- ( ) Chary R., Elbaz D., 2001, ApJ, 556, 562
- ( ) Chapman S. C., Blain A. W., Smail I., Ivison R. J., 2005, ApJ, 622, 772
- ( ) Chapman S. C., Blain A. W., Ivison R. J., Smail I. R., 2003, Natur, 422, 695
- ( ) Condon J. J., 1992, ARA&A, 30, 575
- ( ) Daddi E., et al., 2004a, ApJ, 600, L127
- ( ) Daddi E., Cimatti A., Renzini A., et al., 2004b, ApJ, 617, 746
- ( ) Daddi E., Dickinson M., Chary R., et al, 2005b, ApJ, 631, L13
- ( ) Daddi E., et al., 2007a, ApJ 670, 156 (D07a)
- ( ) Daddi E., et al., 2007b, ApJ 670, 173 (D07b)
- ( ) Dannerbauer H., et al., 2006, ApJ, 637, L5
- ( ) Downes D., Solomon P. M., 1998, ApJ, 507, 615
- ( ) Elbaz D., Cesarsky C. J., Chanial P., et al., 2002, A&A 384, 848
- ( ) Elbaz D., et al., 2007, A&A, 468, 33
- ( ) Förster Schreiber N. M., et al., 2006, ApJ, 645, 1062
- ( ) Giavalisco M., Ferguson H.C., Koekemoer A., et al., 2004, ApJ. 600, L93
- ( ) Genzel R., et al., 2003, astro-ph/0403183
- ( ) Genzel R., et al., 2006, Natur, 442, 786
- ( ) Greve T. R., et al., 2005, MNRAS, 359, 1165
- ( ) Goldader J. D., Meurer G., Heckman T. M., et al., 2002, ApJ, 568, 651
- ( ) Hayashi M., Shimasaku K., Motohara K., et al., 2007, ApJ, 660, 72
- ( ) Immeli A., Samland M., Westera P., Gerhard O., 2004, ApJ, 611, 20
- ( ) Kennicutt R. C., Jr., 1998, ARA&A, 36, 189
- ( ) Kereš D., Katz N., Weinberg D. H., Davé R., 2005, MNRAS, 363, 2
- ( ) Le Floc’h E., et al., 2005, ApJ, 632, 169
- ( ) Madau P., Ferguson H. C., Dickinson M. E., et al., 1996, MNRAS, 283, 1388
- ( ) Maiolino R., et al., 2007, A&A, 472, L33
- ( ) Neri R., et al., 2003, ApJ, 597, L113
- ( ) Noeske K. G., et al., 2007, ApJ, 660, L43
- ( ) Pope A., et al., 2006, MNRAS, 370, 1185
- ( ) Reddy N.A., et al., 2005, ApJ, 633, 748
- ( ) Solomon P. M., Downes D., Radford S., & Barrett J., 1997, ApJ, 478, 144
- ( ) Solomon P. M., Vanden Bout P. A., 2005, ARA&A, 43, 677
- ( ) Spergel D.N., et al., 2007, ApJS, 170, 377
- ( ) Tacconi L. J., et al., ApJ, 640, 228
- ( ) Yao L., Seaquist E.R., Kuno N., & Dunne L., 2003, ApJ, 588, 771
- ( ) Walter F., et al., 2003, Natur, 424, 406
- ( ) Weiß A., Walter F., Scoville N. Z., 2005, A&A, 438, 533