RESUMO
At redshifts beyond [Formula: see text], the 21 cm line from neutral hydrogen is expected to be essentially the only viable probe of the three-dimensional matter distribution. The lunar far-side is an extremely appealing site for future radio arrays that target this signal, as it is protected from terrestrial radio frequency interference, and has no ionosphere to attenuate and absorb radio emission at low frequencies (tens of MHz and below). We forecast the sensitivity of low-frequency lunar radio arrays to the bispectrum of the 21 cm brightness temperature field, which can in turn be used to probe primordial non-Gaussianity generated by particular early universe models. We account for the loss of particular regions of Fourier space due to instrumental limitations and systematic effects, and predict the sensitivity of different representative array designs to local-type non-Gaussianity in the bispectrum, parametrized by [Formula: see text]. Under the most optimistic assumption of sample variance-limited observations, we find that [Formula: see text] could be achieved for several broad redshift bins at [Formula: see text] if foregrounds can be removed effectively. These values degrade to between [Formula: see text] and 0.7 for [Formula: see text] to [Formula: see text], respectively, when a large foreground wedge region is excluded. This article is part of a discussion meeting issue 'Astronomy from the Moon: the next decades (part 2)'.