Molecular Hydrogen along Two Lines of Sight through the Large Magellanic Cloud

Publications of the Astronomical Society of the Pacific, 110:60 67, 1998 January 1998. Astronomical Society of the Pacific. All rights reserved. Printed in U.S.A. Molecular Hydrogen along Two Lines of Sight through the Large Magellanic Cloud Kurt S. Gunderson Center for Astrophysics and Space Astronomy, University of Colorado, Boulder, CO 80309; gunderso@casa.colorado.edu Geoffrey C. Clayton Department of Physics and Astronomy, Louisiana State University, Baton Rouge, LA 70803; gclayton@fenway.phys.lsu.edu and James C. Green Center for Astrophysics and Space Astronomy, University of Colorado, Boulder, CO 80309; jgreen@casa.colorado.edu Received 1997 April 8; accepted 1997 October 29 ABSTRACT. We have remeasured the H 2 column densities along lines of sight to the Large Magellanic Cloud (LMC) using spectra taken by the Hopkins Ultraviolet Telescope (HUT). Three lines of sight were analyzed using the pair method and previously determined Fitzpatrick & Massa extinction parameters to allow H 2 absorption models to be fitted. Substantial improvements over the values first reported by Clayton et al. were made for two of the three lines of sight. Comparisons between LMC and Galactic environments suggest that the factors which lead to extinction curve variations (e.g., metallicity or shocks) do not affect the efficiency of H 2 formation on the surfaces of the dust grains: N H 2 /E(B V) values are approximately the same in both galaxies despite large differences in N /E(B V). HI 1. INTRODUCTION Molecular hydrogen plays a crucial role in interstellar environments as a link between dust and other molecules. H 2 is thought to form first on the surfaces of dust grains. Then it catalyzes reactions that form more complicated molecules such as H 2 O,H 3 O, CH 2, or CO (Spitzer 1978). However, the difficulty in observing H 2 has prevented a thorough investigation of these relationships. In fact, standard methods of measuring N H2, the column density along a line of sight in units of cm 2 have relied upon the value X for the inferred ratio X N K 1 (km s 1 ) 1 cm 2 H /I CO (2.4 2.8) # 10 (Buss et al. 2 1994) for cool gas in dark molecular clouds. These methods can predict values that agree with observations of Galactic H 2 (Buss et al. 1994) but should not be trusted for extragalactic observations in which variations in dust grain distributions, metallicities, or star formation activity are likely to influence molecular hydrogen s role. Cohen et al. (1988) report a ratio of X LMC that is 6 times that of the Galaxy, but this measurement is based on the relationship between absolute CO luminosity and line width in LMC and Galactic clouds rather than direct N H2Observations absorption measurements. indicate that environmental differences exist between the Large Magellanic Cloud and the Galaxy which might affect either the dust or the gas. For example, Cardelli, Clayton, & Mathis (1989, hereafter CCM) found a Galactic extinction law that depends on one parameter ( R v ) and is valid for sight lines through a variety of environments over the wavelength range 3.5 0.125 mm. However, several studies have found that LMC sight lines have extinction curves steeper than CCM curves in the FUV. This suggests that a larger population of smaller grains exists in the LMC (e.g., Clayton & Martin 1985; Fitzpatrick 1985) because small grains attenuate FUV radiation more efficiently. Also, 2175 Å bump strengths in the LMC are weaker than the average Galactic bump strengths. A study of H ii regions in the LMC by Dufour, Shields, & Talbot (1982) demonstrates lower abundances compared to solar for all metals that were considered. Decreased metallicity might affect the efficiency of dust formation and result in a different grain size distribution. Small Magellanic Cloud (SMC) observations that show even lower metallicities (Russell & Dopita 1992) and more extreme extinction differences (Fitzpatrick 1986) support this explanation. However, recent work indicates that typical Galactic metallicity is also lower than solar (Snow & Witt 1996) so the differences between the Galaxy and the LMC might not be as large. In fact, results of Russell & Dopita (1992) suggest that carbon abundances in both galaxies are the same. Further evidence for environmental differences arises from color analysis, which indicates that star formation rates in the LMC have been relatively constant compared to the Galaxy (Hyland 1991). The resultant variations in the stellar populations affect the interstellar environments differently through the associated activities of the stars. For example, the 60

MOLECULAR HYDROGEN 61 TABLE 1 The Observed Sample Sk HDE Spectral Type E(B V) LMC N(H i) LMC References 66 19....... ) B1 I 0.25 a 7.0 # 10 21 1, 2 69 83....... 269244 B0 Ia ) ) b 1 69 270...... 269997 B3 Ia 0.19 3.5 # 10 21 1 67 78....... 269371 B3 Ia ) ) b 1 a Fitzpatrick reports an E(B V) value of 0.28 for this star. However, the intrinsic reddening of Sk 66 19 is greater than that of Sk 69 83 by 0.03, so we have corrected E(B V) LMC by this amount. b When measuring LMC H i, Fitzpatrick 1985 assumes a Milky Way foreground component of N(H i) MW 5.5 # 10 cm 2 ; the unreddened stars of this sample revealed no H i absorption beyond this value. References. (1) Fitzpatrick 1985; (2) Clayton & Martin 1985. 30 Dor region in the LMC is an extreme example of a starforming region containing a large population of early-type stars that are likely to produce winds and higher supernovae rates. Grain-grain collisions in supernovae shocks can shift the grain size distribution to smaller grains (Seab & Shull 1983; Jones, Tielens, & Hollenbach 1996), while ion bombardment can amorphitize their surface layers (Tielens et al. 1994). Grains with different surface layers could have different rates of H 2 production. Also, variations in the size distributions of grains can affect the formation and dissociation of molecules. For example, increased shielding of FUV radiation by a population of smaller dust grains in the LMC will inhibit molecular photodissociations (Cardelli & Savage 1988) and also can impede radiative formation of certain molecules (Snow 1993). Despite the demonstrated increase in FUV extinction along LMC lines of sight, LMC molecular abundances (other than H 2 ) relative to CO scale similarly to those of the Galaxy (Booth & de Graauw 1991). Clearly, measurements of H 2 column densities would make interesting contributions to the existing assembly of data on molecules and dust. Such measurements form the basis of this study. Clayton et al. (1996) presented the first analysis of cold H 2 in the LMC in a study combining IUE (International Ultraviolet Explorer), WUPPE (Wisconsin Ultraviolet Photo-Polarimeter Experiment), and HUT (Hopkins Ultraviolet Telescope) data to measure extinction, polarization, and gas columns (including H 2 ) for three lines of sight. However, they analyzed the HUT data before the final calibrations were made and considered only data taken at night to minimize geocoronal airglow. Since inaccurate calibrations and noisier spectra due to signal losses could result in higher H 2 measurement errors, we remeasured the H 2 columns with the newer calibrations and attempted better airglow removal. We also used a better H 2 absorption model as an improvement on the model of Clayton et al. (1996). 2. DATA ANALYSIS 2.1. Observations The Hopkins Ultraviolet Telescope (HUT) observed six LMC stars during the Astro-2 mission of the Space Shuttle Endeavor on 1995 March 13 16 as part of the guest investigator program of G. Clayton. The data set, identified by Sanduleak (Sk) numbers (Sanduleak 1970), includes spectra of two pairs of reddened and unreddened stars, Sk 66 19 (reddened) and Sk 69 83 (unreddened) and Sk 69 270 (reddened) and Sk 67 78 (unreddened) (see Table 1). Details of the observations and the instrument may be found in Clayton et al. (1996) and the references therein. Standard HUT procedures were used to reduce and calibrate the spectra. IRAF procedures first performed initial reductions on the data and generated raw spectra, pitch and yaw errors versus time, and count rates versus time. The temporal data allowed identifications of bad data time spans. Criteria for bad data included: (1) Time spans where pitch and yaw errors exceed 10, which would displace the target star from the aperture and introduce errors in the flux calibration. (2) Time spans where geocoronal airglow contributions were significant. The time when the flux near the O i line at 1304 Å starts to rise as daytime approaches made a good dividing line between night and day. (3) Time spans where the total count rate over the entire bandpass formed a spike. A spike in the count rate is likely to be an event not associated with the target star s emission and a contributor to noise. The summed exposure times used for Sk 66 19, 69 270, 69 83, and 67 78 were 2394, 2492, 670, and 1613 s, respectively. After good data were summed, HUT calibrations were applied. These included a wavelength calibration, a pulse-persistence correction to error bins, a subtraction of a constant dark rate, a correction to remove counts from scattered light, a correction to remove second-order light above 912 Å, a flat-field correction, pitch/yaw error corrections, and the final flux calibration (Kruk et al. 1997). Figures 1 and 2 show the resulting spectra for the four stars in this sample. Next corrections were applied for interstellar H i absorption, airglow contamination, and extinction. Then the target star spectra were divided by the reference star spectra. We took H i column density values from Fitzpatrick (1985) and used templates from the HUT database to remove the effects of the H i. To remove the effects of geocoronal airglow, templates

62 PASP, 110: GUNDERSON, CLAYTON, & GREEN Fig. 1. Spectra of Sk 66 19 (top) and its reference star, Sk 69 83 (bottom). The spectra are uncorrected for airglow, H i absorption, and extinction. The errors in each wavelength bin have been averaged and multiplied by 5 so that error bars could be plotted in the upper left corners of each spectrum. from the HUT database were also used. The primary contaminants included Lyb, Lyg, Lyd, O i 989 Å, and O i 1040 Å. Fitzpatrick & Massa (1986, hereafter FM) extinction curves modeled the extinction using parameter values taken from Clayton et al. (1996) for Sk 69 270. However, due to systematic errors resulting from spectral mismatch between Sk 66 19 and 69 83, Clayton et al. did not report a set of FM parameter values. Once classified as a B0 Ia star (Fitzpatrick 1985), Sk 69 83 has been reclassified as spectral type O7.5 Iaf (Fitzpatrick & Massa 1988). Therefore, to model the extinction for Sk 66 19 we fixed the parameters to which the H 2 absorption region is insensitive to the means of the corresponding Sk 69 270 and Sk 69 239 values that were reported by Clayton et al. (1996). Since the relevant H 2 lines span the range 930 1110 Å, the frozen parameters were those associated with 1 the 2175 Å bump: l 0, g, and C 3. Then C 1, C 2, and C 4 were varied to find the best fit. As a result, the extinction curve for Sk 66 19 should be viewed with caution, but its uncertainty does not affect the accuracy of the H 2 measurement significantly. Finally, the spectra of Sk 66 19 and 69 270 and the spectra of their respective reference stars were divided, resulting in flat spectra to which molecular hydrogen fits could be made. 2.2. Fitting Molecular Hydrogen Absorption Profiles To determine the H 2 column densities, H 2 absorption models were tested against the spectra to find the best x 2 fits. The model data set (S. McCandliss 1996, private communication) included optical depths as a function of wavelength for the 1 1 1 1 Lyman ( B Su r X Sg ) and Werner ( C P u r X Sg ) bands originating from the ground vibrational state ( n 0) to excited vibrational states ( n 0 for Lyman; 0 6 for Werner). There was a file for each of the ground rotational states J 0 5 at column densities of log[n(h 2 )] 21.0, spanning a wavelength region of 900 1159.99 Å. Only models with Doppler parameters of 1 km s 1 were available, but b 1 corresponds to T 80 K for hydrogen molecules, which is a reasonable temperature assumption (Spitzer 1978). The cross sections for 15 13 these transitions lie in the range 10! j cm 2 H 2! 10 so that any column densities above 10 15 cm 2 will result in saturated absorption lines, but only above 10 19 cm 2 will the weakest lines be damped. Therefore, the attenuations for column density values that are typical of the data considered here scale logarithmically with variations in column density. With this consideration J {0, 1, 2, 3, 4, 5} model sets were constructed for log[n(h 2 )] {19.0, 19.3, 19.7,.0,.3,.5,.7}.

MOLECULAR HYDROGEN 63 Fig. 2. Same as Fig. 1, but for Sk 69 270 and its reference star, Sk 67 78 Profiles for columns less than 10 19 cm 2 were not generated because at this value the attenuations of the weaker lines cease to scale logarithmically. However, contributions to N H2 below this level fall within statistical uncertainties in the model fits, as will be discussed in 2.3. Then the optical depth (t l ) profiles t were converted to attenuations ( e l ), redshifted to zlmc 4 0.9 # 10, and degraded to the 3 Å HUT resolution by means of a resolution degrading code (B. Espey 1996, private communication). The fitting routine assumed flat continua of varying strengths attenuated by varying columns of H 2 in the different J states. Logarithmic interpolations of the fully processed absorption profiles were performed to reach column density values between log N {19.0, 19.3, 19.7,.0,.3,.7, 21.0}. To test the accuracy of interpolation after redshifting and degrading the profiles, an interpolated profile for log N.5 was com- pared to the profile that was generated for that value before it was redshifted and degraded. Agreement between the exact and interpolated profiles was to within 1.5%. To allow for the possibility of NLTE population deviations and temperature variations along the lines of sight, no assumptions were made about the J populations except that N(H 2) J i 1 N(H 2) J i. This constraint entered the fitting procedure in two ways. First, the best fit was found by considering only J 0. Then the best fit was found considering J {0, 1}. If the x 2 -value for the best J {0, 1} fit was lower than that of the J 0 fit, then the case of J {0, 1, 2} was considered. J sets including pro- gressively higher values of J were considered until adding higher values only worsened the fit. For example, in the case of Sk 66 19, adding J 3 into the J {0, 1, 2} set always resulted in fits with higher x 2 -values than with only J {0, 1, 2}. Therefore, we assumed that absorption by H 2 in states with J 1 3 also would be negligible. Second, the fitting algorithm tested random combinations of H 2 column density values in the different J states from among the J set being tested by first guessing a N(H 2) J i value and then guessing N(H 2) J i 1 values from between N(H 2) J i and a lower limit. Since the damping approximation breaks down below log (N H 2 ) 19.0, we chose this value as a lower limit and the columns of each species were considered to be either greater than this value or zero. After enough points were tested, the columns associated with the minimum x 2 2 -value, x min, then determined the best fit. Fits for Sk 66 19 and 69 270 are shown in Figure 3. 2.3. Determining Errors The sensitivity of x 2 to column density variations in the fitted absorption models determined the statistical errors (Lampton, Margon, & Bowyer 1976; Cash 1976). According to this pro-

64 PASP, 110: GUNDERSON, CLAYTON, & GREEN Fig. 3. Comparisons of the divided target and reference spectra (solid binned lines) with the best-fitting H 2 absorption models (dotted lines). Prior to division all spectra were corrected for H i absorption, extinction, and airglow contamination. A dashed line at zero relative flux separates the data from airglow line models, which are plotted to help assess the degree of airglow contamination in the spectral regions of the lines. Error bars in the upper left corners of each plot represent average errors in flux value for each wavelength bin of the divided spectra. (Top) Spectrum of Sk 66 19 divided by Sk 69 83. Two arrows indicate the locations of the S iv doublet, which might be responsible for the two poorly fitted peaks nearby. Airglow lines displayed on the bottom are Lyd, Lyg, O i 989 Å (#0.5), Lyb, and O i 1040 Å. (Bottom) Spectrum of Sk 69 270 divided by Sk 67 78. The airglow lines are Lyb and O i 1040 Å. cedure, the number of free parameters determines the width of the x 2 confidence interval for each free parameter. In this study, the total H 2 column density values are of primary interest and not the J populations or the continuum values. Therefore, to determine the total H 2 columns, the population ratio errors and continuum errors are set to, leaving N H2 as the one free parameter whose errors are to be determined. Subsequently, a 2 2 contour defined as the locus of points where x xmin 1.0 demarcates the 1j confidence region. These contours are evident in Figures 4 and 5, which are plots of continuum value against the total H 2 column. The asterisks represent points falling within 1j (68% confidence); the dots represent points within 2.6j (99% confidence). The projection of the contours onto the log N H2 axis then define the statistical errors to within the re- spective confidences. Reporting column density values of individual J states and their 1j confidence intervals requires determining errors for two and three free parameters simultaneously ( J {0, 1} for Sk 69 270 and J {0, 1, 2} for Sk 66 19). This widens the confidence intervals in x 2 -space 2 2 2 2 to x xmin 2.3 for Sk 69 270 and x xmin 3.5 for Sk 66 19. Because of concerns that the high continuum values used in the models ( 1.4 for Sk 66 19 and 1.7 for Sk 69 270) have resulted from errant extinction corrections and might, in turn, have resulted in inaccurate N H2 measurements, we have reexamined the extinction corrections and conclude that the fidelity of the N H2 measurements is unaffected. The continuum values for Sk 66 19 are high because of the way the extinction curve was fitted. This was done before H 2 absorption was considered so that the best model was fitted through the average relative flux value of the data rather than through the peaks in the data. Therefore, after correcting for the extinction with the best-fitting model, the data are centered around a relative flux value of 1.0; the constant continuum values against which H 2 absorption occurs must elevate to accommodate the addition of the absorption templates. The same is true to a lesser extent in the case of Sk 69 270, except that the average relative flux value still lies slightly above 1.0. For both stars the extinction curve parameters were varied such that the extinction curves ran through peaks in the data, resulting in extinction corrected data for which continuum relative flux values are approximately 1.0. Overlaying these data on the spectra in

MOLECULAR HYDROGEN 65 Fig. 4. Continuum values are plotted against the total N H2 -values for a collection of randomly generated absorption models that are fitted to the Sk 66 19 data. The asterisks (*) represent models whose x 2 -values fall within a 68% confidence interval (1j) of the best-fitting model. The dots represent models that fall within a 99% (2.6j) confidence interval of the best-fitting model. Projections of the contours surrounding the asterisks and dots onto the log N H2 axis demarcate the upper and lower error limits for the respective degrees of confidence. Figure 3 showed that the overlaid spectra differ by an additive constant, so that only the continuum values in the fits will have been affected and not the N H2 -values. Sources of systematic error include spectral mismatch between the reddened and unreddened stars, airglow contamination, and the neglect of columns of J populations below 10 19 cm 2. As shown in Figure 3, spectral mismatch of Sk 69 83 and Sk 66 19 seems to have left two peaks near 1062 and 1072 Å that the models cannot fit. These features might be due to mismatched line widths of the S iv doublet at 1063 and 1073 Å; arrows indicating these wavelength positions have been added to this figure. To identify regions where imperfect airglow removal might have interfered with the fits for Sk 66 19 and Sk 69 270, airglow models have been plotted beneath the fitted spectra. Although they have been scaled to fit in the plots, their line ratios, except for the O i 989 Å line, which has been reduced by a factor of 2, have been conserved. The modeled airglow features shown in the plot for Sk 66 19 are Lyd, Lyg, O i 989 Å (#0.5), Lyb, and O i 1040 Å. Only Lyb contamination appears significant in the spectrum of Sk 66 19. The modeled airglow features that appear in the plot for Sk 69 270 are Lyb and O i 1040 Å. The severity of their contamination is difficult to surmise due to the poorer fit to and lower S/N of the Sk 69 270 spectrum. Zero-point wavelength shifts might also be responsible for the poorer quality fit to the Sk 69 270 data but might also be argued against because positions of actual absorption features relative to the corresponding model features are not shifted consistently toward either the blue or the red. Sk Fig. 5. Same as Fig. 3, but for Sk 69 270 Contributions to the total H 2 columns by J states whose columns are less than 10 19 cm 2 might also be significant. However, at 3 Å resolution the absorption missed by neglecting low column, high J contributions could in part have been accounted for by the fitting of higher columns of low J absorption lines, which overlap with the high J lines. Also, since the J populations drop with increasing J it seems unlikely that much more than 10 19 cm 2 is overlooked. To account for these systematic errors we expand the errors from those determined statistically by 10 19 cm 2 and report NH 2 (1.6 0.2) # cm 2 10 for Sk 66 19 and N (0.7 0.4) # 10 cm 2 H 2 for Sk 69 270 (see Table 2). 3. RESULTS Measurements of gas and dust column densities done in this study and in previous studies are compiled in Table 3. The new results are the improved measurements of N H2 along the lines of sight to Sk 66 19 and Sk 69 270. The errors reported in this work have been reduced by factors of 2 (for Sk 69 270) and 4 (for Sk 66 19) from those reported by Clayton et al. (1996). Values for E(B V) and H i, which have been corrected for foreground H i absorption by the Galaxy, were taken from Fitzpatrick (1985). For both LMC sight lines N H I/E(B V) is more than 4 times the Galactic average (Bohlin et al. 1978). Since interstellar gas consists primarily of 5 hydrogen ( N C/N H! 10 in the comparatively metal-rich Galaxy) and only a small fraction (!3% in this study) of the H is N total (10 cm 2 ) TABLE 2 H 2 Column Densities N J 0 (10 19 cm 2 ) N J 1 (10 19 cm 2 ) N J 2 (10 19 cm 2 ) 66 19... 1.6 0.2 7.5 1.2 7.4 1.1!1.1 69 270... 0.7 0.4 5.3 2.3!2.8 )

66 PASP, 110: GUNDERSON, CLAYTON, & GREEN Object TABLE 3 Column Density Ratios N /E(B V) H I (cm 2 mag 1 ) N /E(B V) H2 (cm 2 mag 1 ) N /N H I H2 References Sk 66 19........... 22 2.5 # 10 (6.4 0.8) # 10 44 1, 2 Sk 69 270......... 22 1.8 # 10 (3.7 2.4) # 10 49 1, 2 LMC average........ 22 2.4 # 10 (5.1 2.5) # 10 51 1, 2 Galaxy average...... 21 5 # 10 5 # 10 10 3, 4 References. (1) This study; (2) Fitzpatrick 1985; (3) Bohlin et al. 1978; (4) Dufour et al. 1982. in H 2, this ratio can be interpreted as the overall gas to dust ratio for the respective lines of sight and suggests that much more gas relative to dust exists in the LMC than in the Galaxy. Although the LMC gas to dust values quadruple the average Galactic value, the LMC values of N H 2 /E(B V) are the same as the Galactic value to within 2j. However, the lines of sight through both LMC environments have similar N H I/E(B V) values and similar N H 2 /E(B V) values. Before proceeding with analysis of these results, it must be noted that the usable LMC sample consists of only two data points: one data point for 30 Dor, an extreme star-forming region, (Sk 69 270) and one data point for the diffuse ISM (Sk 66 19). Both have low E(B V) so their respective lines of sight sample diffuse clouds for Sk 66 19 and either diffuse clouds or the outer edges of dense clouds for Sk 69 270. Since each environment is sampled once, a possibility of the H 2 columns being anomalous exists. However, the correlation between E(B V) and N H2 argues against this since it suggests similar mechanisms of regulating H 2 abundances in all environments so that the N H2 -values are normal. Also, the extinction curves of Sk 66 19 and Sk 69 270 that are shown in Fitzpatrick (1985) appear typical of their respective environments. Therefore, we will assume by association that the H 2 columns are also typical and discuss them accordingly. If the dust and H 2 coexist along the lines of sight, the similarity of the N H 2 /E(B V) values carries interesting implications concerning the relationships between the gas and dust. First, the smaller variances of N H 2 /E(B V) between the LMC and the Galaxy compared to the larger variances of N H I/E(B V) between those galaxies is consistent with the belief that H 2 molecules form mostly on the surfaces of grains rather than in direct interactions of H atoms. As discussed in 1, metallicity, the ubiquity of supernova shocks, and stellar winds are examples of environmental properties whose effects on dust grain distributions might be responsible for peculiarities in the LMC extinction curves, but it appears that any resultant changes in the H 2 -forming surface layers of the grains do not interfere with the efficiency of H 2 production. Second, the scaling of NH 2, with E(B V) suggests that dust, as measured by E(B V) regulates the amount of H2 along a line of sight. That there is no such correlation with FUV extinction argues against the likelihood that radiative shielding from FUV photons protects the H 2 molecules significantly. Measuring relative populations of H 2 rotational levels might indicate the degree to which winds, SN shocks, and the UV radiation field are affecting the H 2 (Hollenbach, Chu & McCray 1976). However, since the data for the lines of sight studied in this paper lack the spectral resolution to fit J templates individually, the J population values are too inaccurate to draw significant conclusions. Lastly, the possibilities also exist that a number of effects governing the abundance of H 2 relative to E(B V) might be in competition with each other in a way that maintains a constant N H 2 /E(B V), or that the dust and gas are not even spatially coincident. In the Galaxy, H 2 abundances frequently are inferred from direct observations of spatially coincident CO abundances using the ratio N K 1 (km s 1 ) 1 cm 2 H /I CO (2.4 2.8) # 10. 2 However, the differences in metallicity, dust grain properties, and the UV photon field between the Galaxy and the LMC suggest that the Galactic ratio value might not be accurate elsewhere. Cohen et al. (1988) infer a N H /ICO value for the 2 LMC that is 6 times higher than that of the Galaxy from relationships between N H2, L CO, and Dn. As discussed earlier, the lower LMC metallicity might limit the formation of CO or affect the dust grain properties. In turn, the dust grain properties, which have been observed to be different, might affect the production of H 2. Both CO and H 2 abundances might be affected by the UV photon field. Unless all of these differing environmental properties either counteract each other s effects or have no effects at all on the H 2 /CO ratio, N /I should be H2 CO N H2 different. But how different, if at all? And can the -values presented here be used to make a direct measurement of this ratio? Unfortunately, studies of CO and [C ii], a tracer of photodissociated CO, show that both gasses are spatially coincident in 30 Dor to within a 45 beam width, or down to scales of tens of parsecs (Poglitsch et al. 1995). This indicates a highly fragmented structure of the interstellar medium that allows UV radiation to penetrate deep into the molecular clouds and photodissociate CO, but it also means that observations of CO emission along the line of sight to Sk 69 270 could, due to variations within a single beam, sample a different set of environments than those along which we have measured N H2 ; i.e., a large-beam CO measurement and a single sight line H 2 measurement are incompatible. In the absence of CO observations along the line of sight to Sk 66 19, current H 2 and CO measurements do not permit accurate N /I H2 CO measurements.

MOLECULAR HYDROGEN 67 Regardless, in both the diffuse ISM and 30 Dor regions of the LMC, N H 2 /E(B V) values are roughly constant and comparable to typical values for Galactic lines of sight even though LMC and Galactic values of N H I/E(B V) are very different. The N H 2 /E(B V) correlation between the two LMC lines of sight examined in this paper and the Galaxy might, however, be coincidental, or be in part due to a selection effect of the survey for which Galactic N H2 -values were determined. Copernicus sampled lines of sight toward O and B stars with E(B V) values similar to those presented here (Spitzer & Cochran 1973; Spitzer, Cochran, & Hirshfeld 1974; Spitzer & Jenkins 1975; Bohlin et al. 1978) implying that they, too, pass through diffuse clouds, the diffuse edges of denser clouds, or perhaps circumstellar shells (Hollenbach et al. 1976). However, the fact remains that the LMC and Galactic extinctions, and therefore the dust size distributions and/or compositions, are not similar. Then, if N H 2 /E(B V) is assumed constant, this means that the mechanism(s) responsible for the extinction differences between 30 Dor and the diffuse ISM, as well as for the differences between the LMC and the Galaxy, does not appear to alter the grain populations in a way that affects the efficiency of H 2 production. We would like to thank Stephan McCandliss and Brian Espey for valuable discussions and contributions. This work was supported by NASA grants NAG8-1048 and NSG-5303. REFERENCES Bianchi, L., Clayton, G. C., Bohlin, R. C., Hutchings, J. B., & Massey, P. 1996, ApJ, 471, 3 Bohlin, R. C., Savage, B. D., & Drake, J. F. 1978, ApJ, 224, 132 Booth, R. S., & degraauw, Th. 1991, in IAU Symp. 148, The Magellanic Clouds, ed. R. Haynes & D. Milne (Dordrecht: Kluwer), 415 Buss, R. H., Jr., Allen, M., McCandliss, S., Kruk, J., Liu, J.-C., & Brown, T. 1994, ApJ, 430, 630 Cardelli, J. A., & Clayton, G. C. 1991, AJ, 101, 1021 Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245 (CCM) Cardelli, J. A., & Savage, B. D. 1988, ApJ, 325, 864 Cash, W. 1976, A&A, 52, 307 Clayton, G. 1996, PASP, 108, 225 Clayton, G., Green, J., Wolff, M., Zellner, N., Code, A. D., & Davidsen, A. 1996, ApJ, 460, 313 Clayton, G., & Martin, P. G. 1985, ApJ, 288, 558 Cohen, R. S., Dame, T. M., Garay, G., Montani, J., Rubio, M., & Thaddeus, P. 1988, ApJ, 331, L95 Dufour, R. J., Shields, G. A., & Talbot, R. J., Jr. 1982, ApJ, 252, 461 Fitzpatrick, E. L. 1985, ApJ, 299, 219. 1986, AJ, 92, 1068 Fitzpatrick, E. L., & Massa, D. 1986, ApJ, 307, 286. 1988, ApJ, 328, 734 Hollenbach, D., Chu, S.-I., & McCray, R. 1976, ApJ, 8, 458 Hyland, A. R. 1991, in IAU Symp. 148, The Magellanic Clouds, ed. R. Haynes & D. Milne (Dordrecht: Kluwer), 125 Israel, F. P., et al. 1993, A&A, 276, 25 Jones, A. P., Tielens, A. G. G. M., & Hollenbach, D. J. 1996, ApJ, 469, 740 Kruk, J. W., Kimble, R. A., Buss, R. H., Jr., Davidsen, A. F., Durrance, S. T., Finley, D. S., Holberg, J. B., & Kriss, G. A. 1997, ApJ, 482, 546 Lampton, M., Margon, B., & Bowyer, S. 1976, ApJ, 8, 177 Mochizuki, K., et al. 1994, ApJ, 430, L37 Pak, S., & Jaffe, D. T. 1995, BAAS, 187, 11.01 Poglitsch, A., Krabbe, A., Madden, S. C., Nikola, T., Geis, N., Johansson, L. E. B., Stacey, G. J., & Sternberg, A. 1995, ApJ, 454, 293 Russell, S. C., & Dopita, M. A. 1992, ApJ, 384, 508 Sanduleak, N. 1970, Contr. C. T. I. O., 89 Seab, C. G., & Shull, J. M. 1983, ApJ, 275, 652 Shull, J. M., & York, D. G. 1977, ApJ, 211, 803 Snow, T. P. 1993, ApJ, 402, L73 Snow, T. P., & Witt, A. N. 1996, ApJ, 468, L65 Spitzer, L. 1978, Physical Processes in the Interstellar Medium (New York: Wiley) Spitzer, L., & Cochran, W. D. 1973, ApJ, 186, L23 Spitzer, L., Cochran, W. D., & Hirshfeld, A. 1974, ApJS, 28, 373 Spitzer, L., & Jenkins, E. B. 1975, ARA&A, 13, 133 Tielens, A. G. G. M., McKee, C. F., Seab, C. G., & Hollenbach, D. J. 1994, ApJ, 431, 321