We present the results of a comparative study of HI narrow self–absorption ( HINSA ) , OH , ^ { 13 } CO , and C ^ { 18 } O in five dark clouds .
We find that the HINSA generally follows the distribution of the emission of the carbon monoxide isotopologues , and has a characteristic size close to that of ^ { 13 } CO .
This confirms earlier work ( 65 ) which determined that the HINSA is produced by cold HI which is well mixed with molecular gas in well–shielded regions .
The OH and ^ { 13 } CO column densities are essentially uncorrelated for the sources other than L1544 .
Our observations indicate that the central number densities of HI are between 2 and 6 cm ^ { -3 } , and that the ratio of the hydrogen density to total proton density for these sources is 5 to 27 \times 10 ^ { -4 } .
Using cloud temperatures and the density of atomic hydrogen , we set an upper limit to the cosmic ray ionization rate of 10 ^ { -16 } s ^ { -1 } .
We present an idealized model for HI to H _ { 2 } conversion in well–shielded regions , which includes cosmic ray destruction of H _ { 2 } and formation of this species on grain surfaces .
We include the effect of a distribution of grain sizes , and find that for a MRN distribution , the rate of H _ { 2 } formation is increased by a factor of 3.4 .

Comparison of observed and modeled fractional HI abundances indicates ages for these clouds , defined as the time since initiation of HI \rightarrow H _ { 2 } conversion , to be 10 ^ { 6.5 } to 10 ^ { 7 } yr. Several effects may make this time a lower limit , but the low values of n _ { HI } we have determined make it certain that the time scale for evolution from a possibly less dense atomic phase to almost entirely molecular phase , must be a minimum of several million years .
This clearly sets a lower limit to the overall time scale for the process of star formation and the lifetime of molecular clouds .