We examine several different simplified approaches for modelling the chemistry of CO in 3D numerical simulations of turbulent molecular clouds. We compare the different models both by looking at the behaviour of integrated quantities such as the mean CO fraction or the cloud-averaged CO-to-H2 conversion factor, and also by studying the detailed distribution of CO as a function of gas density and visual extinction. In addition, we examine the extent to which the density and temperature distributions depend on our choice of chemical model.
We find that all of the models predict the same density probability density function (PDF) and also agree very well on the form of the temperature PDF for temperatures T > 30 K, although at lower temperatures, some differences become apparent. All of the models also predict the same CO-to-H2 conversion factor, to within a factor of a few. However, when we look more closely at the details of the CO distribution, we find larger differences. The more complex models tend to produce less CO and more atomic carbon than the simpler models, suggesting that the C/CO ratio may be a useful observational tool for determining which model best fits the observational data. Nevertheless, the fact that these chemical differences do not appear to have a strong effect on the density or temperature distributions of the gas suggests that the dynamical behaviour of the molecular clouds on large scales is not particularly sensitive to how accurately the small-scale chemistry is modelled.