Is atomic carbon a good tracer of molecular gas in metal-poor galaxies?

Carbon monoxide (CO) is widely used as a tracer of molecular hydrogen (H2) in metal-rich galaxies, but is known to become ineffective in low metallicity dwarf galaxies. Atomic carbon has been suggested as a superior tracer of H2 in these metal-poor systems, but its suitability remains unproven. To help us to assess how well atomic carbon traces H2 at low metallicity, we have performed a series of numerical simulations of turbulent molecular clouds that cover a wide range of different metallicities. Our simulations demonstrate that in star-forming clouds, the conversion factor between [CI] emission and H2 mass, $X_{\mathrm{C}\mathrm{I}}$, scales approximately as $X_{\mathrm{C}\mathrm{I}}\propto Z^{-1}$. We recover a similar scaling for the CO-to-H2 conversion factor, $X_{\mathrm{C}\mathrm{O}}$, but find that at this point in the evolution of the clouds, $X_{\mathrm{C}\mathrm{O}}$ is consistently smaller than $X_{\mathrm{C}\mathrm{I}}$, by a factor of a few or more. We have also examined how $X_{\mathrm{C}\mathrm{I}}$ and $X_{\mathrm{C}\mathrm{O}}$ evolve with time. We find that $X_{\mathrm{C}\mathrm{I}}$ does not vary strongly with time, demonstrating that atomic carbon remains a good tracer of H2 in metal-poor systems even at times significantly before the onset of star formation. On the other hand, $X_{\mathrm{C}\mathrm{O}}$ varies very strongly with time in metal-poor clouds, showing that CO does not trace H2 well in starless clouds at low metallicity.