The apparent lack of cold molecular gas in blue compact dwarf (BCD) galaxies is at variance with their intense star-formation episode. The CO molecule, often used a tracer of H2 through a conversion function, is selectively photodissociated in dust-poor environments and, as a result, a potentially large fraction of H2 is expected to reside in the so-called CO-dark gas, where it could be traced instead by infrared cooling lines [CI], [CII], and [OI]. Although the fraction of CO-dark gas to total molecular gas is in theory expected to be relatively large in metal-poor galaxies, many uncertainties remain due to the difficulty in identifying the main heating mechanism associated to the cooling lines observed in such galaxies.

Investigations of the Herschel Dwarf Galaxy Survey (DGS; Madden et al.2013) show that the heating mechanism in the neutral gas of BCDs cannot be dominated by the photoelectric effect on dust grains below some threshold metallicity due to the low abundance of dust and polycyclic aromatic hydrocarbons, implying that other heating mechanisms need to be invoked, along with a new interpretation of the corresponding infrared line diagnostics. In the study presented here and in Lebouteiller et al. (2017), we use optical and infrared lines to constrain the physical conditions in the HII region + HI region of the BCD I Zw 18 (18 Mpc; ≍2% solar metallicity) within a consistent photoionization and photodissociation model. We show that the HI region is entirely heated by a single ultraluminous X-ray source with important consequences on the applicability of [CII] to trace the star-formation rate and to trace the CO-dark gas. We derive stringent upper limits on the size of H2 clumps that may be detected in the future with JWST and IRAM/NOEMA. We also show that the nature of the X-ray source can be inferred from the corresponding signatures in the ISM. Finally we speculate that star formation may be quenched in extremely metal-poor dwarf galaxies due to X-ray photoionization.