We report the first detailed measurement of the shape of the CO luminosity function at high redshift, based on >320 hr of the NSF's Karl G. Jansky Very Large Array (VLA) observations over an area of ∼60 arcmin 2 taken as part of the CO Luminosity Density at High Redshift (COLDz) survey. COLDz "blindly" selects galaxies based on their cold gas content through CO(J=1→0) emission at z∼2-3 and CO(J=2→1) at z∼5-7 down to a CO luminosity limit of log(L CO /K km s −1 pc 2 ) 9.5. We find that the characteristic luminosity and bright end of the CO luminosity function are substantially higher than predicted by semi-analytical models, but consistent with empirical estimates based on the infrared luminosity function at z∼2. We also present the currently most reliable measurement of the cosmic density of cold gas in galaxies at early epochs, i.e., the cold gas history of the universe, as determined over a large cosmic volume of ∼375,000 Mpc 3 . Our measurements are in agreement with an increase of the cold gas density from z∼0 to z∼2-3, followed by a possible decline towards z∼5-7. These findings are consistent with recent surveys based on higher-J CO line measurements, upon which COLDz improves in terms of statistical uncertainties by probing ∼50-100 times larger areas and in the reliability of total gas mass estimates by probing the low-J CO lines accessible to the VLA. Our results thus appear to suggest that the cosmic star-formation rate density follows an increased cold molecular gas content in galaxies towards its peak about 10 billion years ago, and that its decline towards the earliest epochs is likely related to a lower overall amount of cold molecular gas (as traced by CO) bound in galaxies towards the first billion years after the Big Bang.

Dusty star-forming galaxies at high redshift (1 < z < 3) represent the most intense star-forming regions in the universe. Key aspects to these processes are the gas heating and cooling mechanisms, and although it is well known that these galaxies are gas-rich, little is known about the gas excitation conditions. Only a few detailed radiative transfer studies have been carried out owing to a lack of multiple line detections per galaxy. Here we examine these processes in a sample of 24 strongly lensed star-forming galaxies identified by the Planck satellite (LPs) at z ∼ 1.1–3.5. We analyze 162 CO rotational transitions (ranging from J up = 1 to 12) and 37 atomic carbon fine-structure lines ([C i]) in order to characterize the physical conditions of the gas in the sample of LPs. We simultaneously fit the CO and [C i] lines and the dust continuum emission, using two different non-LTE, radiative transfer models. The first model represents a two-component gas density, while the second assumes a turbulence-driven lognormal gas density distribution. These LPs are among the most gas-rich, IR-luminous galaxies ever observed (μ L L IR ( 8 − 1000 μ m ) ∼ 10 13 − 14.6 L ⊙; 〈 μ L M ISM 〉  = (2.7 ± 1.2) × 1012 M ⊙, with μ L ∼ 10–30 the average lens magnification factor). Our results suggest that the turbulent interstellar medium present in the LPs can be well characterized by a high turbulent velocity dispersion ( 〈 ΔV turb 〉  ∼ 100 km s−1) and ratios of gas kinetic temperature to dust temperature 〈 T kin/T d 〉  ∼ 2.5, sustained on scales larger than a few kiloparsecs. We speculate that the average surface density of the molecular gas mass and IR luminosity, Σ M ISM  ∼ 103–4 M ⊙ pc−2 and Σ L IR  ∼ 1011–12 L ⊙ kpc−2, arise from both stellar mechanical feedback and a steady momentum injection from the accretion of intergalactic gas.

We report detections of three z ∼ 2.5 submillimeter-selected galaxies (SMGs; SMM J14011+0252, SMM J14009+0252, SMM J04431+0210) in the lowest rotational transition of the carbon monoxide molecule (CO J = 1 − 0) and one nondetection (SMM J04433+0210). For the three galaxies we detected, we find a line-integrated brightness temperature ratio of the J = 3 − 2 and 1 − 0 lines of 0.68±0.08; the 1−0 line is stronger than predicted by the frequent assumption of equal brightnesses in the two lines and by most single-component models. The observed ratio suggests that mass estimates for SMGs based on J = 3 − 2 observations and J = 1 − 0 column density or mass conversion factors are low by a factor of 1.5. Comparison of the 1 − 0 line intensities with intensities of higher-J transitions indicates that single-component models for the interstellar media in SMGs are incomplete. The small dispersion in the ratio, along with published detections of CO lines with J upper > 3 in most of the sources, indicates that the emission is from multi-component interstellar media with physical structures common to many classes of galaxies. This result tends to rule out the lowest scaling factors between CO luminosity and molecular gas mass, and further increases molecular mass estimates calibrated against observations of galaxies in the local universe. We also describe and demonstrate a statistically sound method for finding weak lines in broadband spectra that will find application in searches for molecular lines from sources at unknown redshifts.