2020MNRAS.492.1465S

We analyse the CO-dark molecular gas content of simulated molecular clouds from the SILCC-Zoom project. The simulations reach a resolution of 0.1 pc and include H2 and CO formation, radiative stellar feedback and magnetic fields. CO-dark gas is found in regions with local visual extinctions A_{V, 3D} ∼ 0.2-1.5, number densities of 10-103 cm-3 and gas temperatures of few 10-100 K. CO-bright gas is found at number densities above 300 cm-3 and temperatures below 50 K. The CO-dark gas fractions range from 40 per cent to 95 per cent and scale inversely with the amount of well-shielded gas (A_{V, 3D} >= 1.5), which is smaller in magnetized molecular clouds. We show that the density, chemical abundances and A_{V, 3D} along a given line-of-sight cannot be properly determined from projected quantities. As an example, pixels with a projected visual extinction of A_{V, 2D} ∼eq 2.5-5 can be both, CO-bright or CO-dark, which can be attributed to the presence or absence of strong density enhancements along the line-of-sight. By producing synthetic CO(1-0) emission maps of the simulations with RADMC-3D, we show that about 15-65 per cent of the H2 is in regions with intensities below the detection limit. Our clouds have X_{CO}-factors around 1.5 x 1020 cm-2 (K km s-1)-1 with a spread of up to a factor ∼ 4, implying a similar uncertainty in the derived total H2 masses and even worse for individual pixels. Based on our results, we suggest a new approach to determine the H2 mass, which relies on the availability of CO(1-0) emission and A_{V, 2D} maps. It reduces the uncertainty of the clouds' overall H2 mass to a factor of <= 1.8 and for individual pixels, i.e. on sub-pc scales, to a factor of <= 3.