The properties of the molecular gas in distant galaxies on and above the main sequence and how they impact our view what a "normal" object is

ALMA has given a decisive impulse to the study of the molecular gas and dust in distant star-forming galaxies. In particular, it opened the door for studies of main-sequence galaxies representative of the bulk of the star-forming population, overcoming the biases of samples limited to the brightest starbursting objects. In this framework, we have been conducting a multi-cycle and multi-wavelength ALMA survey of several tens of infrared-selected galaxies on and above the main-sequence at z~1.1-1.7. We targeted several carbon monoxide (CO) transitions (J=2,4,5,7) and their underlying dust continuum emission to investigate the properties of cold gas and dust of "normal" and "extreme" star-forming galaxies and study the physics behind their different growing modes.

The CO(5-4) and (7-6) L' luminosities of main-sequence and starburst galaxies linearly correlate with the total infrared luminosity from star formation L_IR over several orders of magnitude and across redshifts, making them good tracers of the obscured star formation SFR. The CO(2-1) luminosity sub-linearly correlates with L_IR as expected for a classical total molecular gas mass tracer and the Schmidt-Kennicutt relation with a smooth variation from main-sequence to starburst galaxies and a redshift evolution. The former have longer depletion timescales or lower star formation efficiencies than starbursts, in agreement with past work.

Using the CO(5-4)/CO(2-1) L' luminosity ratio as a proxy for the gas excitation, we find that the distance from the main sequence is not the best predictor of the conditions of the molecular gas in galaxies: a population of compact objects with starburst-like line ratios is evident on the main sequence, pairing with shorter depletion timescales and higher star formation efficiencies (Puglisi, Daddi, Liu+2019; Puglisi, Daddi, Valentino+2021). This is also evident from the average full spectral line energy distribution (SLED): we do find a trend with the distance from the main sequence, but the properties of main-sequence and starburst galaxies are not significantly different up to J=5, 7 at this stage despite the statistics. A large intrinsic variety of high/low-J line ratios is present even in a homogeneously infrared-selected population of main-sequence galaxies.

The surface density of SFR (~SFR/size²) offers an empirical way out to predict the properties of the cold interstellar medium by naturally accounting for compact objects on and above the main sequence (Valentino, Daddi, Puglisi+2020c). It is also theoretically well supported (Narayanan & Krumholz 2014): compact gas geometries trigger similarly dense SFR configurations where the highly efficient formation of new stars and the copious production of UV, X-ray photons, and cosmic rays plus the dust-gas interaction generate the observed trends. This also suggests a better definition of different growing modes based on the surface density of SFR, surpassing the classical classification of "main-sequence" and "starburst" galaxies relying on SFRs and stellar masses.  

In parallel to the study of CO emission, we have been carrying on systematic observations of both transitions of the neutral atomic carbon [CI] in a subsample of main-sequence and starburst galaxies from our ALMA survey (Valentino, Magdis, Daddi 2018; Valentino+2020b). This species has been proposed as a tracer of the molecular gas mass potentially superior to CO: it has a simple quantum structure and it is optically thin in most conditions, it does not suffer from a large excitation bias as high-J CO transitions, and it benefits (rather than suffers from) the destruction of CO molecules due to cosmic rays, predicted to be abundant is strongly star-forming galaxies as main-sequence and starburst objects at high redshift (Papadopoulos+2004). 

We find an excellent correlation between the [CI](1-0) emission and both low-J CO transitions and the optically thin dust continuum emission, both standard tracers of the total molecular gas mass. We also do not retrieve any strong variation across galaxy star-forming populations and redshifts (main-sequence galaxies at z~1.2, starbursts, submillimeter galaxies at z~2.5-4, local Luminous Infrared Galaxies). This suggests that [CI] is an effective tracer of the global amount of cold gas reservoirs in galaxies at high redshift.

By comparing CO and dust-based gas mass estimates with the [CI] emission, we can derive the [CI]/[H₂] abundance - the conversion factor to derive the gas mass from an optically thin tracer. However, this purely empirical calibration inevitably falls back onto the well-known uncertainties on the alpha-CO and gas-to-dust conversion factors. By applying standard values, we retrieve lower [CI]/[H2] abundances in main-sequence than in starburst galaxies.

Besides providing an alternative way to estimate the molecular gas mass, both the [CI](1-0) and [CI](2-1) transitions greatly expand the information on the ISM that we can collect with reasonable telecope time, allowing for a direct comparison with ISM heating or chemical models or even a first peak into a cosmological perspective offered by semi-analytical models and simulations. As an example, the [CI](2-1)/[CI](1-0) ratio is a proxy for the gas kinetic temperature (and excitation temperature under local thermal equilibrium). Interestingly, this ratio is rather insensitive to different conditions and stays roughly constant over several orders of magnitude of IR luminosity, despite its large scatter.

An analysis of the ISM conditions including [CI] is now feasible not only for bright and biased starburst populations, but also for more normal main-sequence galaxies. We showed the potential of including [CI] in the study of the ISM by applying classical 1D photon-dominated region models (PDR, Kaufman+1999), which give an order-of-magnitude idea of the systematically higher densities and stronger radiation fields in distant starbursts than main sequence galaxies, but oversimplifying the true physics regulating the gas emission. We also show the successes and improvements to be implemented in the Santa Cruz semi-analytical model (Popping+2019), correctly predicting the emission of CO, but not that of [CI], in galaxies across redshifts.   

We studied the effect of AGN onto the dust and molecular gas properties in our sample of IR-selected main-sequence and starburst galaxies at z~1.1-1.7. The simple selection function allows for an assessment of AGN effects onto a general population of star-forming objects typically underrepresented in surveys specifically targeting bright QSO and sub-millimetre galaxies.

By modeling the long-wavelength SED with the code STARDUST (Kokorev+2021, submitted), we derive basic dust properties: masses, temperatures, and infrared luminosities, where we can separate the contribution from dusty torii surrounding AGN and heating by star formation. We find AGN to boost the mid-IR emission, while having no effect on the longest wavelengths in the Rayleigh-Jeans tail proportional to the dust mass. This is clear both from full SED modeling and a simpler observed color S_24um/S_1.2mm, where the strength of the AGN is gauged by its contribution to L_IR (f_AGN=L_{IR, AGN} / L_IR) or its X-ray emission. The dust (and thus gas) fraction seems not to be affected by the presence of AGN in our sample of galaxies representative of the bulk of the star-forming population.

We split our sample according to the AGN contribution to L_IR (f_AGN) and the X-ray emission and stack the CO SLEDs. We do not find any difference between AGN hosts and star-forming galaxies up to J=7 on average. However, there is a broad variety of line ratios and shapes to which AGN might contribute. We identify a small subsample of AGN hosts with large f_AGN that are overluminous in a high-J transition as CO(5-4) compared to what expected from their L_{IR, SFR.} However, a simple linear model L'CO(5-4) = a x L_{IR, SFR} + b x L_{IR, AGN} is sufficient to explain the excess CO(5-4) emission as due to the AGN via, e.g., X-ray or mechanical heating.