ON THE EVOLUTION OF THE MOLECULAR GAS FRACTION OF STAR-FORMING GALAXIES

Molecular hydrogen is arguably the most important component of the interstellar medium (ISM) in high-redshift galaxies, since it is the phase necessary for, and immediately preceding, star formation. However, observing cool (≲50 K) molecular hydrogen is difficult; the molecule has no permanent electric dipole and so one must rely on tracer molecular emission (the most common being rovibrational emission from the ¹²CO isotopomer, hereafter "CO").

Unfortunately, uncertainty of the precise calibration of CO luminosity to total gas mass in both low- and high-z galaxies has made interpretation of results challenging; even in the local universe there is seen to be metallicity and luminosity dependence on the conversion (Boissier et al. 2003; Blitz et al. 2007; Komugi et al. 2010). Furthermore, although the bulk of the molecular gas in star-forming disks is contained in virialized clouds, it is clear that the most luminous galaxies—mostly mergers—have molecular gas distributions and star formation modes dramatically different to quiescent disks; they require an alternative molecular mass calibration related to the physical conditions of their interstellar media (Solomon et al. 1997). The importance of understanding how the thermodynamic state of the gas affects the calibration also remains controversial (high-z observations typically target J_upper>1, requiring some correction to J = (1 → 0)) based on the excitation of the gas; although see Ivison et al. (2010) and Harris et al. (2010).

The majority of molecular gas studies at high-z have, necessarily, focused on the most active systems (generally submillimeter-selected galaxies and quasars), and so comparatively little is known about the evolution of the molecular gas properties of the more common, but less active population. Nevertheless, recent studies have started to make progress in expanding the parameter space of CO observations at high-z, and several studies have detected CO emission from "normal" disks at z = 1–2 (Daddi et al. 2008; 2010; Tacconi et al. 2010). The striking result is that the efficiency of star formation in these distant galaxies is claimed to be similar to that seen in local "quiescent" disks (Dannerbauer et al. 2009; Daddi et al. 2010), with enhanced star formation rates (SFRs) driven simply by larger gas reservoirs, drained by star formation (and feedback) on Gyr timescales. This hints that a secular mode of star formation might be ubiquitous in L^⋆ disks independent of cosmic epoch (but see the higher-z results of Coppin et al. 2007 and Riechers et al. 2010).

The aim of this Letter is to further expand the parameter space of molecular gas studies in normal galaxies, building on our observations of luminous infrared galaxies (LIRGs) at z = 0.4 (Geach et al. 2009a, hereafter G09). We present new IRAM Plateau de Bure interferometric (PdBI) observations of a sample of 24 μm-selected galaxies in the outskirts of the rich cluster Cl 0024+16 at z = 0.395. These galaxies bridge the gap between quiescent, local spirals and luminous, high-redshift starbursts and disks. Throughout we adopt a Ω_m = 0.3, Ω_Λ = 0.7, and H₀ = 70 km s⁻¹ Mpc⁻¹ cosmology.

Our sample consists of five Spitzer MIPS 24 μm detected LIRG-class galaxies in the outskirts of the rich cluster Cl 0024+16 (z = 0.395; see Geach et al. 2006). We obtained mid-infrared spectroscopy with the Spitzer Infrared Spectrograph to confirm the presence of aromatic features, a clear indication of star formation (rather than infrared emission dominated by an active galactic nucleus; Geach et al. 2009b). Combined with the two galaxies observed with IRAM PdBI in G09, this represents an effort to obtain CO detections or limits for all the galaxies where we have detected significant emission from the polycyclic aromatic hydrocarbon 7.7 μm band. The line luminosities imply SFRs of ∼30–60 M_☉ yr⁻¹, thus allowing us to link molecular gas with moderate levels of dusty star formation in the galaxies.

In this Letter, we present the results in the context of the broader star-forming population at z = 0.4, rather than focusing on potential environmental effects in the cluster. We believe this is a valid approach, since these galaxies are expected to be randomly accreted onto the cluster from the surrounding field, and the galaxies are seen at a stage where strong environmental effects associated with clusters (specifically ram-pressure stripping and harassment) are yet to have an effect (Treu et al. 2003; Moran et al. 2007). They lie at clustocentric radii of 1.8–4.6 Mpc (∼1–3 × the virial radius of Cl 0024+16). Intrinsically, they are likely to be representative of disk galaxies of equivalent mass sampled from a random field and can be compared to galaxies with similar properties at low and high redshift, i.e., normal star-forming disks.

The new IRAM observations were conducted over 2009 June–October in configuration "D," using five antennae, and we adopted the same strategy as in G09. Sensitivities ranged between 0.51 and 1.17 mJy (median average for 10 MHz wide channels and two polarizations), and the on-source exposure times ranged between 6.4 and 12.8 hr. We targeted the CO (J = 1 → 0) 115.27 GHz rotational transition at ν_obs ≃ 82.63 GHz in five sources. The 3 mm receiver was tuned to the frequency of the redshifted CO (J = 1 → 0) line at the systemic redshift of each galaxy derived from optical spectroscopy (Czoske et al. 2001; Moran et al. 2007). As in G09, the correlator was set up with 2.5 MHz spacing (2 × 64 channels, 320 MHz bandwidth). The phase and flux calibrators were the sources 3C454.3, 0119+115, and 0007+171. The observing conditions were good or excellent in terms of atmospheric phase stability; however, any anomalous and high phase-noise visibilities were flagged in the calibration stage. Data were calibrated, mapped, and analyzed using gildas (Guilloteau & Lucas 2000).

3.1. Plateau de Bure CO Detections

We detect CO (J = 1 → 0) emission in three galaxies (>4σ detections within 2'' of the phase tracking center, significances determined from the integrated line flux). Velocity-integrated maps and millimeter spectra are shown in Figure 1. Total fluxes are evaluated from single or double Gaussian fits to the spectra depending on whether the profiles are very broad and have hints that they are double peaked. The line luminosities are in the range L'_CO = (0.26–0.64) × 10¹⁰ K km s⁻¹ pc² (Table 1). Uncertainties on the luminosities are estimated by re-evaluating the Gaussian fits after artificially adding noise to each channel based on the observed rms fluctuations in each data cube. Upper 3σ limits on L'_CO for the two non-detections are based on the rms noise in the observations of each source (note that the two non-detections are also the two galaxies with the lowest SFRs in the sample). The luminosities and line widths are listed in Table 1; we also list the properties of the two galaxies from G09 which are included in the following analysis.

Zoom In Zoom Out Reset image size

Table 1. Properties of the Galaxy Sample

Target	αJ2000	δJ2000	$z_{\rm [O \,\mathsc {ii}]}$	LFIR	SFR	VCO(1-0)FWHM	L'CO(1-0)	M(H2 + He)a	M⋆b	fgas
	(h m s)	(° ' '')		(1011L☉)	(M☉ yr−1)	(km s−1)	(1010 K km s−1 pc2)	(1010 M☉)	(1011 M☉)
MIPS J002652.5	00 26 52.5	+17 13 59.9	0.3799	3.5 ± 0.5	62 ± 19	140 ± 10	0.64 ± 0.05	2.9 ± 0.2	0.95	0.23 ± 0.02
MIPS J002703.6	00 27 03.6	+17 11 27.9	0.3956	2.3 ± 0.3	42 ± 12	250 ± 30	0.43 ± 0.06	2.0 ± 0.3	0.87	0.19 ± 0.03
MIPS J002715.0	00 27 15.0	+17 12 45.6	0.3813	1.9 ± 0.3	35 ± 11	340 ± 40	0.26 ± 0.03	1.2 ± 0.3	1.10	0.10 ± 0.03
MIPS J002609.1	00 26 09.1	+17 15 11.5	0.3940	1.5 ± 0.3	28 ± 9	⋅⋅⋅	<0.39	<1.8	0.41	<0.30
MIPS J002606.1	00 26 06.1	+17 04 16.4	0.3904	1.9 ± 0.3	34 ± 10	⋅⋅⋅	<0.25	<1.1	0.49	<0.19
MIPS J002621.7c	00 26 21.7	+17 19 26.4	0.3803	3.1 ± 0.2	56 ± 16	140 ± 10	0.68 ± 0.06	3.1 ± 0.3	1.12	0.22 ± 0.02
MIPS J002721.0c	00 27 21.1	+16 59 49.9	0.3964	3.2 ± 0.2	59 ± 16	160 ± 30	1.14 ± 0.11	5.2 ± 0.5	0.98	0.34 ± 0.04

Notes. ^aAssuming a CO luminosity to total gas mass conversion factor of α = 4.6 M_☉(K km s^-1 pc²)⁻¹ that includes helium. ^bUncertainties ∼0.02 dex based on refitting spectral templates 1000 times with random perturbation within 1σ photometric uncertainties. ^cGeach et al. (2009a).

Download table as:  ASCIITypeset image

The spectra of MIPS J002652.5 and MIPS J002715.0, like MIPS J002721.0 (G09), exhibit broad CO (J = 1 → 0) emission and can be well fit by double Gaussian profiles. None of the new sample show obvious signs of major mergers or strong tidal interaction in the deep optical imaging (Figure 1), and so we conclude that we are most likely observing CO (J = 1 → 0) emission tracing molecular gas distributed over rotationally supported disks.

3.2. Physical Properties of the Galaxies

Figure 2 shows the observed f_gas in M_⋆ ⩾ 10¹⁰ M_☉ galaxies at z = 0 (Leroy et al. 2008), z = 0.4 (this work), and z = 1.5–2 (Daddi et al. 2010; Tacconi et al. 2010) showing the clear decline in 〈f_gas〉 in the 10 Gyr since z ∼ 2. As a guide, we cast the evolution of f_gas relative to the cosmological abundance of baryons in stars and molecular gas at z = 0: $\Omega _{\rm H_2}/(\Omega _{\rm H_2}+\Omega _\star) = 0.078^{+0.160}_{-0.039}$ from the "baryon census" of Fukugita et al. (1998). At z = 0, the observed f_gas in star-forming disks is close to that inferred from the total relative abundance of molecular gas and stars in the universe, but at z ∼ 2 it was a factor ∼5–10 larger than the local value.⁶ The decline in molecular gas fraction in the stellar mass limited sample is broadly characterized by f_gas ∝ (1 + z)^γ, with γ ∼ 1.5–2.5, shallower than the rate of decline of the average specific SFR in galaxies over the same epoch, which falls off more like (1 + z)⁴ (e.g., Karim et al. 2010). This is expected, since molecular reservoirs can be replenished with fresh material via cooling; thus, it is the relative evolution of the cooling rate and SFR that shape the evolution of f_gas.

Zoom In Zoom Out Reset image size

Recent semi-analytic prescriptions for galaxy formation make predictions for the evolution of the molecular ISM. In Figure 2, we show the average H₂ + He fraction in M_⋆ ⩾ 10¹⁰ M_☉ galaxies selected in halos of various mass from the galform model (galaxies populated within the Millennium Simulation ΛCDM framework). This latest model (Lagos et al. 2010) implements an empirically based star formation law (Blitz & Rosolowsky 2006) to estimate the molecular gas mass. The model predicts that (at z < 3), f_gas is on average lower, and its evolution stronger in more massive halos. In the mass regime pertinent to our observed sample, 10¹² M_☉, the model tracks are well-matched to the observations over 0 < z < 2. The results are broadly consistent with other models. For example, in smoothed particle hydrodynamic simulations, Kereš et al. (2005) predict that the average accretion rate of gas onto galaxies within halos of M_vir = 10¹² M_☉ decreases from $\dot{M}_{\rm cold}\sim 20\ M_\odot \ {\rm yr}^{-1}$ at z = 2 to $\dot{M}_{\rm cold}\sim 3\ M_\odot \ {\rm yr}^{-1}$ at z = 0.4. The observed evolution is also consistent with the predicted cosmological evolution of the accretion rate at fixed halo mass in other models (see Dutton et al. 2010).

The evolution of f_gas is expected to be strongly dependent on galaxy mass, since the evolution of the global average SFR appears to be halo mass dependent, as does the expected evolution of the gas accretion rate. Cooling and star formation are the key drivers of the molecular gas fraction; but it is also sensitive to the merger rate (which can deliver molecular material, trigger star formation, and reconfigure the baryonic content), feedback (heating or ejection of cold gas and H₂ dissociation), environmental effects (e.g., ram-pressure stripping), and growth in size of galaxy disks. The latter has an important effect since it (on average) reduces the disk hydrostatic pressure (∝r²) and therefore the efficiency of H₂ formation (reflected in the H₂/H i ratio; Elmegreen 1993; Wong & Blitz 2002), as can the metallicity and strength of the interstellar UV radiation field.

Selection effects will affect Figure 2 severely, and we should discuss how these might affect our interpretation. Sensitivity limits do not allow us to detect galaxies with very low gas fractions at high-z, and in low-z surveys the very gas rich systems appear to be rare. Here we have attempted to compare similar galaxies across a range of epochs, but the current sample sizes are small, and inhomogeneously selected; any conclusions we draw from Figure 2 should be treated as indicative for future surveys that can more precisely control selection and have the efficiency to survey larger numbers.

A further caveat to note is that both Daddi et al. (2010) and Tacconi et al. (2010) apply corrections to their observed CO luminosities on account of the difference in the Rayleigh–Jeans brightness temperature of the higher J transitions targeted. Tacconi et al. (2010) apply R₃₁ = 0.5 to their CO (J = 3 → 2) luminosities (defined R₃₁ = L'_CO(3→2)/L'_CO(1→0)), and Daddi et al. (2010) apply R₂₁ = 0.86 to their CO (J = 2 → 1) luminosities (see also Dannerbauer et al. 2009). Leroy et al. (2008) also apply R₂₁ = 0.8 to a sub-sample of the z ∼ 0 galaxies that were mapped in CO (J = 2 → 1). Furthermore, Daddi et al. (2010) assume an α = 3.6 ± 0.8, inferred from dynamical mass arguments. Although this is within ∼1σ of the Galactic value, assuming α = 4.6 increases the gas fraction by ∼12%. These systematic uncertainties in CO–H₂ conversions should be taken as a corollary when interpreting Figure 2, however, a large change in the gas mass (or stellar mass) would be required to remove the trend in f_gas. It is unlikely that we have overestimated the gas mass in the z = 0.4 LIRGs, since we have applied α = 4.6. In the high-z sources, the gas masses would have to be overestimated by a factor ∼5, or the stellar masses underestimated by a factor ∼4 to remove the trend.

We have presented new constraints on the molecular gas mass of z = 0.4 LIRGs, showing strong evolution of f_gas since z ∼ 2. The observed trend encodes critical information on the evolution of the growth of galaxies, including gas accretion, feedback, and stellar mass assembly. Future surveys that are able to split samples based on environment, mass, and other parameters will disentangle the relative effects that shape the evolution of the gas fraction. These observations will soon be supplemented and surpassed by progress made with the Atacama Large Millimeter Array and the Expanded Very Large Array, which will provide unprecedented sensitivity and resolution with which to probe the evolution of gas in galaxies.

This work was based on observations made with the IRAM PdBI, supported by INSU/CNRS (France), MPG (Germany), and IGN (Spain). We thank the anonymous referee for a constructive report. It is a pleasure to thank Roberto Neri for his support in this project. J.E.G. acknowledges the National Sciences and Engineering Research Council of Canada, support from the endowment of the Lorne Trottier Chair in Astrophysics and Cosmology (McGill) and the UK Science and Technology Facilities Council (STFC). I.R.S. and A.C.E. also acknowledge STFC.