We compare the CO J =(1-0) and HI emission in the Large Magellanic Cloud (LMC) in three dimensions, i.e. including a velocity axis in addition to the two spatial axes, with the aim of elucidating the physical connection between giant molecular clouds (GMCs) and their surrounding HI gas. The CO J =1-0 dataset is from the second NANTEN CO survey and the HI dataset is from the merged Australia Telescope Compact Array (ATCA) and Parkes Telescope surveys. The major findings of our analysis are: 1) GMCs are associated with an envelope of HI emission, 2) in GMCs [average CO intensity] is proportional to [average HI intensity]^[1.1+-0.1] and 3) the HI intensity tends to increase with the star formation activity within GMCs, from Type I to Type III. An analysis of the HI envelopes associated with GMCs shows that their average linewidth is 14 km s-1 and the mean density in the envelope is 10 cm-3. We argue that the HI envelopes are gravitationally bound by GMCs. These findings are consistent with a continual increase in the mass of GMCs via HI accretion at an accretion rate of 0.05 Msun/yr over a time scale of 10 Myr. The growth of GMCs is terminated via dissipative ionization and/or stellar-wind disruption in the final stage of GMC evolution.

We analyze the conditions for detection of CO(1-0) emission in the Large Magellanic Cloud (LMC), using the recently completed second NANTEN CO survey. In particular, we investigate correlations between CO integrated intensity and HI integrated intensity, peak brightness temperature, and line width at a resolution of 2.6' (~40 pc). We find that significant HI column density and peak brightness temperature are necessary but not sufficient conditions for CO detection, with many regions of strong HI emission not associated with molecular clouds. The large scatter in CO intensities for a given HI intensity persists even when averaging on scales of >200 pc, indicating that the scatter is not solely due to local conversion of HI into H_2 near GMCs. We focus on two possibilities to account for this scatter: either there exist spatial variations in the I(CO) to N(H_2) conversion factor, or a significant fraction of the atomic gas is not involved in molecular cloud formation. A weak tendency for CO emission to be suppressed for large HI linewidths supports the second hypothesis, insofar as large linewidths may be indicative of warm HI, and calls into question the likelihood of forming molecular clouds from colliding HI flows. We also find that the ratio of molecular to atomic gas shows no significant correlation (or anti-correlation) with the stellar surface density, though a correlation with midplane hydrostatic pressure P_h is found when the data are binned in P_h. The latter correlation largely reflects the increasing likelihood of CO detection at high HI column density.