HI 21-centimetre emission from an ensemble of galaxies at an average redshift of one

The baryonic processes in galaxy evolution include gas infall onto galaxies to form neutral atomic hydrogen (HI), the conversion of HI to the molecular state (H$_2$), and, finally, the conversion of H$_2$ to stars. Understanding galaxy evolution thus requires understanding the evolution of both the stars, and the neutral atomic and molecular gas, the primary fuel for star-formation, in galaxies. For the stars, the cosmic star-formation rate density is known to peak in the redshift range $z \approx 1-3$, and to decline by an order of magnitude over the next $\approx 10$ billion years; the causes of this decline are not known. For the gas, the weakness of the hyperfine HI 21cm transition, the main tracer of the HI content of galaxies, has meant that it has not hitherto been possible to measure the atomic gas mass of galaxies at redshifts higher than $\approx 0.4$; this is a critical lacuna in our understanding of galaxy evolution. Here, we report a measurement of the average HI mass of star-forming galaxies at a redshift $z \approx 1$, by stacking their individual HI 21 cm emission signals. We obtain an average HI mass similar to the average stellar mass of the sample. We also estimate the average star-formation rate of the same galaxies from the 1.4 GHz radio continuum, and find that the HI mass can fuel the observed star-formation rates for only $\approx 1-2$ billion years in the absence of fresh gas infall. This suggests that gas accretion onto galaxies at $z < 1$ may have been insufficient to sustain high star-formation rates in star-forming galaxies. This is likely to be the cause of the decline in the cosmic star-formation rate density at redshifts below 1.Comment: 43 pages, 8 figures. Published in Nature: https://www.nature.com/articles/s41586-020-2794-7. In the original version of the paper, the upper x-axis "Lookback time (Gyr)" of Fig. 3 had incorrectly-placed tick marks. This is now fixed in the updated version. We thank Sambit Roychowdhury for pointing out the erro