Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.stsci.edu/~kgordon/papers/PS_files/lmc_ext.ps.gz Äàòà èçìåíåíèÿ: Thu Apr 24 02:49:35 2003 Äàòà èíäåêñèðîâàíèÿ: Tue May 27 07:53:59 2008 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: molecular cloud

A Reanalysis of the Ultraviolet Extinction from Interstellar Dust
in the Large Magellanic Cloud
K. A. Misselt, Geoffrey C. Clayton & Karl D. Gordon
Department of Physics and Astronomy, Louisiana State University,
Baton Rouge, LA 70803­4001
email: misselt,gclayton,gordon@rouge.phys.lsu.edu
ABSTRACT
We have reanalyzed the Large Magellanic Cloud's (LMC) ultraviolet (UV)
extinction using data from the IUE final archive. Our new analysis takes
advantage of the improved signal--to--noise of the IUE NEWSIPS reduction, the
exclusion of stars with very low reddening, the careful selection of well matched
comparison stars, and an analysis of the effects of Galactic foreground dust.
Differences between the average extinction curves of the 30 Dor region and the
rest of the LMC are reduced compared to previous studies. We find that there
is a group of stars with very weak 2175 š A bumps that lie in or near the region
occupied by the supergiant shell, LMC 2, on the southeast side of 30 Dor. The
average extinction curves inside and outside LMC 2 show a very significant
difference in 2175 š A bump strength, but their far--UV extinctions are similar.
While it is unclear whether or not the extinction outside the LMC 2 region can
be fit with the relation of Cardelli, Clayton & Mathis (CCM), sightlines near
LMC 2 cannot be fit with CCM due to their weak 2175 š A bumps. While the
extinction properties seen in the LMC lie within the range of properties seen in
the Galaxy, the correlations of UV extinction properties with environment seen
in the Galaxy do not appear to hold in the LMC.
Subject headings: dust, extinction --- Magellanic Clouds --- galaxies: individual
(LMC) --- galaxies: ISM --- ultraviolet: ISM
1. Introduction
As our nearest galactic neighbors, the Magellanic Clouds offer a unique opportunity to
study the effects of different galactic environments on dust properties. Their importance
has increased with the recent discovery that the dust in starburst galaxies appears to be

-- 2 --
similar to that in the star forming bar of the Small Magellanic Cloud (SMC) (Calzetti et.
al. 1994; Gordon, Calzetti & Witt 1997, Gordon & Clayton 1998 [GC]). Understanding
dust extinction properties in nearby galaxies is a useful tool in interpreting and modeling
observations in a wide range of extragalactic systems.
Previous studies of the LMC extinction have all arrived at similar conclusions, e.g. the
average LMC extinction curve is characterized by a weaker 2175 š A bump and a stronger
far--UV rise than the average Galactic extinction curve. Two early studies (Nandy et. al.
1981; Koornneef & Code 1981) found little spatial variation in the LMC extinction and
computed an average LMC extinction curve. However, both samples were dominated by
stars near the 30 Doradus star forming region and it was thus unclear whether their average
curves applied to the LMC as a whole. A study by Clayton & Martin (1985) expanded the
sample to include a larger number of non--30 Dor stars and reported tentative evidence for
differences between the extinction curves observed in the 30 Dor region and the rest of the
LMC. Fitzpatrick (1985, hereafter F85) expanded the number of available reddened stars
to 19 including 7 outside of the 30 Dor region, allowing a more detailed analysis of regional
variations. F85 found a significant difference between the UV extinction characteristic of
the 30 Dor region and that outside the 30 Dor region. The average 30 Dor UV extinction
curve was found to have a lower bump strength and stronger far--UV rise (¸2 units at
– \Gamma1 = 7¯m \Gamma1 ) than the non--30 Dor stars. Fitzpatrick (1986, hereafter F86) expanded the
sample by 8 lightly reddened stars located outside the 30 Dor region and confirmed the
results of F85. Clayton et al. (1996) measured the extinction of two LMC stars, one in 30
Dor and one outside 30 Dor down to ¸ 1000 š A. Both extensions seem to be consistent with
extrapolations of the IUE extinction curves to shorter wavelengths.
As part of a program to quantify the range of extinction behavior in the Local Group,
we have reanalyzed the extinction in the Magellanic Clouds. In particular, no analysis has
been done since the discovery that the UV extinction along most Galactic sightlines could be
described by one parameter, the ratio of total--to--selective extinction, R V = A V =E(B \Gamma V )
(Cardelli, Clayton, & Mathis 1989, hereafter CCM). It is of great interest whether such a
relation exists for the Magellanic clouds. In this paper, we discuss the results for the LMC.
An analysis of the SMC extinction appears in GC.

-- 3 --
2. The Data and the Computation of Extinction Curves
2.1. The Sample
Our initial sample of reddened stars consisted of that defined by F85. In an effort to
expand the sample we searched the updated electronic catalog of Rousseau et. al. (1978),
available via the SIMBAD database, which consists of ¸ 1800 LMC stars. Two initial cuts
of the catalog were made: (1) stars with spectral types later than B4 were discarded and
(2) we required B \Gamma V – 0. The first criterion limits the effects of spectral type mismatches
in the resulting extinction curves, which can be quite large for spectral types later than
about B3 (e.g. F85). The second criterion removes unreddened or lightly reddened stars
from consideration. We note that all the F85 stars were included in the resulting sample
of ¸ 250 stars while none of the F86 stars were included as they all had B \Gamma V ! 0. We
then eliminated emission line stars and composite--spectrum objects from our list. The
remaining stars were checked against the IUE database and all of those for which both long
and short wavelength low­dispersion spectra existed (54) were examined in more detail.
Only five stars from this sample were found to be both significantly reddened and have
high S/N IUE spectra. These stars were added to our sample and their IUE spectra are
listed in Table 1. We selected 67 unreddened comparison stars from the sample of LMC
supergiants in Fitzpatrick (1988) for use in constructing extinction curves using the pair
method (Massa, Savage & Fitzpatrick 1983). Approximate UV spectral types for all of
our reddened stars and their respective comparison stars (see below for a discussion of the
selection of extinction pairs) were estimated from a visual comparison of the IUE spectra
to the grid of LMC stars with UV spectral types given in Neubig & Bruhweiler (1998). The
estimated UV spectral types are reported in Table 2.
Table 1. ``New'' Reddened LMC Stars
SWP LWP/LWR
SK Images Images
\Gamma66 88 39129,45383,45384 LWP18165,23730
\Gamma68 23 39155 LWP18198
\Gamma69 206 36552,39832 LWP15751
\Gamma69 210 23270 LWR17442
\Gamma69 279 08924 LWR07672

-- 4 --
Table 2. Reddened/Unreddened Pairs: Properties
Photometry a Spectral Type b
SK E(B\GammaV) Gal
c V B\GammaV U\GammaV J\GammaV H\GammaV K\GammaV Optical UV Key d
\Gamma66 19 0.09: 12.79 0.12 \Gamma0.66 \Gamma0.35 \Gamma0.45 -- B4 I B0 Ia 1
\Gamma66 169 0.03 11.56 \Gamma0.13 \Gamma1.13 -- -- -- O9.7 Ia O9 Ia
\Gamma66 88 0.06 12.70 0.20 \Gamma0.45 -- -- -- B2: B3 Ia 2
\Gamma66 106 0.07 11.72 \Gamma0.08 \Gamma0.99 -- -- -- B2 Ia B3 Ia
\Gamma67 2 0.06 11.26 0.08 \Gamma0.69 \Gamma0.18 \Gamma0.21 \Gamma0.28 B1.5 Ia B2 Ia 3
\Gamma66 35 0.07: 11.55 \Gamma0.07 \Gamma0.95 -- -- -- B1 Ia B2 Ia
\Gamma68 23 0.06 12.81 0.22 \Gamma0.39 -- -- -- OB B4 Ia 4
\Gamma67 36 0.07 12.01 \Gamma0.08 \Gamma0.89 -- -- -- B2.5 Ia B3 Ia
\Gamma68 26 0.04 11.67 0.13 \Gamma0.62 -- -- -- B8: I B3 Ia 5
\Gamma66 35 0.07: 11.55 \Gamma0.07 \Gamma0.95 -- -- -- B1 Ia B2 Ia
\Gamma69 108 0.08 12.10 0.27 \Gamma0.22 \Gamma0.57 \Gamma0.67 \Gamma0.75 B3 I B3 Ia 6
\Gamma67 78 0.05 11.26 \Gamma0.04 \Gamma0.77 -- -- -- B3 Ia B3 Ia
\Gamma70 116 0.05 12.05 0.11 \Gamma0.61 \Gamma0.37 \Gamma0.43 \Gamma0.57 B2 Ia B3 Ia 7
\Gamma67 256 0.07 11.90 \Gamma0.08 \Gamma0.97 -- -- -- B1 Ia B3 Ia
\Gamma68 129 0.07 12.77 0.03 \Gamma0.81 -- -- -- B0.5 O9 Ia 8
\Gamma68 41 0.05 12.0 \Gamma0.14 \Gamma1.10 -- -- -- B0.5 Ia B0 Ia
\Gamma68 140 0.04 12.72 0.06 \Gamma0.77 \Gamma0.26 \Gamma0.31 \Gamma0.37 B0: B0 Ia 9
\Gamma68 41 0.05 12.0 \Gamma0.14 \Gamma1.10 -- -- -- B0.5 Ia B0 Ia
\Gamma68 155 0.02 12.72 0.03 \Gamma0.79 -- -- -- B0.5 O8 Ia 10
\Gamma67 168 0.03 12.08 \Gamma0.17 \Gamma1.17 -- -- -- O8 Iaf O8 Ia
\Gamma69 206 0.08 12.84 0.14 \Gamma0.62 -- -- -- B2: O9 Ia 11
\Gamma67 5 0.06 11.34 \Gamma0.12 \Gamma1.07 -- -- -- O9.7 Ib B0 Ia
\Gamma69 210 0.07 12.59 0.36 \Gamma0.23 -- -- -- B1.5: B1 Ia 12
\Gamma66 118 0.08 11.81 \Gamma0.05 \Gamma0.91 -- -- -- B2 Ia B3 Ia
\Gamma69 213 0.08 11.97 0.10 \Gamma0.65 \Gamma0.26 \Gamma0.29 \Gamma0.33 B1 B1 Ia 13
\Gamma70 120 0.06 11.59 \Gamma0.06 \Gamma0.94 0.21 0.25 0.14 B1 Ia B1.5 Ia
\Gamma69 228 0.06 12.12 0.07 \Gamma0.69 \Gamma0.10 \Gamma0.14 \Gamma0.14 OB B2 Ia 14

-- 5 --
An implicit assumption of the pair method is that the Galactic foreground reddening is
the same for both the program and comparison stars and, hence, cancels out of the resulting
LMC extinction curve. As pointed out by several authors (e.g. Schwering & Israel 1991;
Oestreicher, Gochermann, & Schmidt--Kaler 1995), the Galactic foreground towards the
LMC is quite variable ranging from E(B \Gamma V ) Gal = 0:00 to 0.17. Schwering & Israel (1991)
constructed a foreground reddening map towards the LMC using HI data and a relationship
between E(B \Gamma V ) and the HI column density. They examined the F85 and F86 stars at
a spatial resolution of 48 0 (the resolution of the HI data) and found systematically higher
Galactic foreground reddening associated with the comparison stars than the reddened
stars. Accounting for this systematic affect reduced the difference between the 30 Dor and
non--30 Dor extinction curves. Oestreicher et. al. (1995) used reddenings to ¸ 1400 LMC
foreground stars to construct a Galactic foreground reddening map with a resolution of
¸10 0 . We have quantified the differences in Galactic foreground reddening for our sample
using the higher resolution map of Oestreicher et. al. (1995). For all but one of our pairs
in the 30 Dor sample, the difference in the Galactic foreground reddening between the
reddened and comparison stars, j\DeltaE(B\GammaV) Gal j Ÿ 0:02 while for the non--30 Dor sample,
j\DeltaE(B\GammaV) Gal j Ÿ 0:03 for all but one pair as well. There is no systematic difference in the
foreground reddening between program and comparison stars in either sub--sample with the
average \DeltaE(B\GammaV) Gal being near 0 for both samples. The values for the Galactic foreground
component of the reddening for each star used in the analysis is given in Table 2. For the
two pairs with large foreground differences (SK \Gamma66 19/SK \Gamma66 169, SK \Gamma69 228/SK \Gamma65
15) we have estimated the maximum effect on the extinction curve to be less than the
photometric uncertainties. Therefore, we have not corrected the individual curves for the
differences in the Galactic foreground.
2.2. The Extinction Curves
Extinction curves were constructed using the standard pair method (e.g. Massa, Savage
& Fitzpatrick 1983). Short and long wavelength IUE spectra were extracted using the
IUE NEWSIPS reduction, co--added, binned to the instrumental resolution of ¸5 š A and
merged at the maximum wavelength in the short wavelength spectrum. Uncertainties
in the extinction curve contain terms that depend both on the broadband photometric
uncertainties as well as uncertainties in the IUE fluxes. The flux uncertainties are now
calculated directly in the NEWSIPS reduction. For details of our error analysis, the reader
is referred to GC.
Previous studies suffered from systematic temperature and luminosity mismatches

-- 6 --
Table 2---Continued
Photometry a Spectral Type b
SK E(B\GammaV) Gal
c V B\GammaV U\GammaV J\GammaV H\GammaV K\GammaV Optical UV Key d
\Gamma65 15 0.12 12.14 \Gamma0.10 \Gamma1.02 -- -- -- B1 Ia B1 Ia
\Gamma69 256 0.07 12.61 0.03 \Gamma0.80 0.03 0.04 \Gamma0.02 B0.5 B1 Ia 15
\Gamma68 41 0.05 12.0 \Gamma0.14 \Gamma1.00 -- -- -- B0.5 Ia B0 Ia
\Gamma69 265 0.06 11.88 0.12 \Gamma0.51 -- -- -- B3 I B3 Ia 16
\Gamma68 40 0.05 11.71 \Gamma0.07 \Gamma0.86 -- -- -- B2.5 Ia B3 Ia
\Gamma69 270 0.05 11.27 0.14 \Gamma0.52 \Gamma0.32 \Gamma0.40 \Gamma0.46 B3 Ia B2 Ia 17
\Gamma67 228 0.03 11.49 \Gamma0.05 \Gamma0.87 -- -- -- B2 Ia B2 Ia
\Gamma69 279 0.02 12.79 0.05 \Gamma0.79 \Gamma0.19 \Gamma0.28 \Gamma0.34 OB0 O9 Ia 18
\Gamma65 63 0.03 12.56 \Gamma0.16 \Gamma1.18 -- -- -- O9.7 I: O9 Ia
\Gamma69 280 0.05 12.66 0.09 \Gamma0.65 \Gamma0.22 \Gamma0.22 \Gamma0.33 B1 B1.5 Ia 19
\Gamma67 100 0.05 11.95 \Gamma0.09 \Gamma0.95 -- -- -- B1 Ia B1 Ia
a Optical photometry from Rousseau et. al. (1978), F85 and Fitzpatrick (1988). IR photometry from
Morgan & Nandy (1982) and Clayton & Martin (1985).
b Optical spectral types from Rousseau et. al. (1978), F85 and Fitzpatrick (1988). UV spectral types
estimated by comparison with LMC UV spectral types of Neubig & Bruhweiler (1998).
c Galactic foreground reddening from Oestreicher et. al. (1995); Colon designates uncertain value.
d Key to position in Figure 1.

-- 7 --
between the unreddened/reddened star pairs. These mismatches were evident in the
imperfect line cancellations seen in the extinction curves, especially the Fe III blend near
5.1 ¯m \Gamma1 . This study minimizes mismatches by using a larger sample of comparison
stars than was available to previous studies. Comparison stars for each reddened star
were selected to satisfy the three Fitzpatrick criteria (F85); in addition, we required
\Delta(B \Gamma V ) – 0:15 between the reddened and comparison stars to minimize the uncertainties
in the extinction curve. The first criterion requires that \Delta(U \Gamma B)=\Delta(B \Gamma V ) be appropriate
to dust reddening. The average value of \Delta(U \Gamma B)=\Delta(B \Gamma V ) for the LMC is 0:83 \Sigma 0:1
(F85). Stars with 0:63 Ÿ \Delta(U \Gamma B)=\Delta(B \Gamma V ) Ÿ 1:03 were selected. The second
criterion requires that the difference in intrinsic V magnitudes between the comparison and
reddened stars be ``small'' (j \DeltaV j! 0:8). The V magnitudes of our program stars were
dereddened assuming R V = 3:1. As all LMC stars are at roughly the same distance, this
criterion amounts to assuring comparable absolute magnitudes between the comparison and
reddened stars thus minimizing luminosity mismatches. The third criterion requires that
the comparison and reddened star UV spectra be well--matched. This minimizes residual
features in the extinction curve not due to extinction. This procedure resulted in 3--10
potential comparison stars for each reddened star. Each potential comparison star was used
to compute an extinction curve. The reddened/comparison star pair which resulted in a
curve with the smallest line residuals was adopted. Five stars from the F85 sample had
\Delta(B \Gamma V ) ! 0:15 and were discarded, leaving a total of 19 reddened stars in our study.
These included three 30 Dor stars (SK \Gamma68 126, \Gamma69 199 and \Gamma69 282) and two non--30
Dor stars (SK \Gamma68 107 and \Gamma71 52). In addition, five stars have been added, three to the
30 Dor sample (SK \Gamma69 206, \Gamma69 210 and \Gamma69 279) and two to the non--30 Dor sample
(SK \Gamma66 88 and \Gamma68 23). We have indicated the positions of all of our stars on an Hff map
of the LMC in Figure 1. A key to the numbering of the stellar positions in Figure 1 is given
in Table 2.
The final extinction curves computed for each pair are shown in Figure 2 and the star
pairs are listed in Table 2. The extinction curves have been fit using the Fitzpatrick &
Massa (1990, hereafter FM) parameterization. The FM fit is a six parameter fit including a
linear background, a Drude profile representing the 2175 š A bump, and a far--UV curvature
term. We emphasize that this parameterization is empirical and the individual functions
describing the extinction curve probably have limited physical significance (Mathis &
Cardelli 1992). The FM fits to individual extinction curves are plotted in Figure 2 and
the best fit parameters for each curve are given in Table 3; the functional form of the
parameterization is given as a footnote to Table 3. In determining the uncertainties on
the individual fit parameters we have considered the effects of two sources of uncertainty,
photometric and spectral mismatch. The photometric uncertainties include those in the

-- 8 --
Fig. 1.--- Positions of reddened stars plotted on an Hff image. A key to the numbering is
provided in Table 2.
broad band optical photometry as well as those in the IUE fluxes (for a detailed discussion
of these uncertainties, see GC). We estimate their effect on the FM parameters by shifting
the extinction curves upward by 1oe and downward by 1oe point--by--point. FM fits were
made to both of the shifted extinction curves and the error in each individual parameter
is taken as one--half the absolute value of the difference in the fit parameters between the
two curves. The photometric uncertainties contribute most significantly to errors in the FM
parameters C 1 ; C 2 ; C 3 , and C 4 ; they have little effect on the bump parameters x 0 and fl. The
effects of mismatch errors on the FM parameters were taken from Cardelli et. al. (1992).
By varying the spectral type of the comparison star and fitting the resulting extinction
curve, they were able to estimate the uncertainties introduced in the FM fit parameters
(Table 6 of Cardelli et. al. 1992). We adopt the quadrature sum of the these two sources of
uncertainty as our estimate of the uncertainties in the individual FM fit parameters (Table

-- 9 --
3). For weak features (ie. the weak bump lines of sight in our sample), the uncertainties
introduced by spectral mismatches may be underestimated by the adopted uncertainties.
We determined R V values for all of the reddened stars in our sample which had R, I,
J, H, and/or K observations. Eleven reddened stars had measurements in at least three
of these bands (Morgan & Nandy 1982; Clayton & Martin 1985). Intrinsic colors were
taken from Johnson (1966) and Winkler (1997) assuming the reddened stars' UV spectral
types. The R V values were determined by assuming all extinction laws take the form of
Rieke & Lebofsky (1985) (CCM). The uncertainties were calculated from the range of R V
values which were 67% probable using the reduced ü 2 statistic (Taylor 1982). The R V
values and uncertainties are give in Table 4. We do not include \Gamma69 256 in Table 4 or any
of the subsequent analysis using measured R V values as it value of R V is very uncertain
(1.55\Sigma1.18).
3. Discussion
3.1. Average Curves
3.1.1. 30 Dor/Non--30 Dor
A very important result from previous work on the LMC was the apparent difference
between UV extinction properties in the 30 Dor region and other sightlines in the LMC
(Clayton & Martin 1985; F85, F86). Reddened stars were assigned to the non--30 Dor
(d proj – 1 kpc, 7 objects) and 30 Dor (d proj ! 1 kpc, 12 objects) samples based on
their projected distance from R 136 as in previous studies. We have calculated average
extinction curves for our new 30 Dor and non--30 Dor samples, weighting the individual
curves by their uncertainties. The FM parameters of the average curves were calculated as
the sample mean and the uncertainties for the average FM parameters are the standard
deviation of the mean for the respective samples, eg. oe i =
p
N . Formal FM fits to the
average curves yielded identical parameters within the uncertainties. In Figure 3, the new
average extinction curves for 30 Dor and non--30 Dor are shown with the results of F86
plotted for comparison. The extinction curves of F85 and F86 are virtually the same but
their uncertainty estimates are quite different. At 7.0 ¯m \Gamma1 , the difference between the
Fitzpatrick 30 Dor and non--30 Dor curves is 1.86 \Sigma 0.41 (F85). In F86, the uncertainties
are estimated to be about twice as large making the difference about 2oe. Our results are
similar to F86 but the 30 Dor curve is slightly lower and the non--30 Dor curve slightly
higher in our averages. We find the difference between the average curves at 7.0 ¯m \Gamma1 to
be 0.89 \Sigma 0.53. So the significance of differences in far--UV extinction between the 30 Dor

-- 10 --
Table 3. FM Fit Parameters
FM Fit Parameters a
SK \Delta(B\GammaV) x 0 fl C 1 C 2 C 3 C 4 C 3 =fl 2
LMC--Average Sample
\Gamma66 19 0.25 4.653\Sigma0.010 0.97\Sigma0.07 +0.09\Sigma0.44 0.75\Sigma0.11 2.34\Sigma0.42 0.91\Sigma0.12 2.49\Sigma0.57
\Gamma66 88 0.28 4.579\Sigma0.019 1.03\Sigma0.06 \Gamma0.88\Sigma0.38 1.00\Sigma0.13 2.77\Sigma0.46 0.48\Sigma0.10 2.61\Sigma0.53
\Gamma67 2 0.15 4.625\Sigma0.010 1.08\Sigma0.07 \Gamma3.59\Sigma0.40 1.67\Sigma0.26 3.71\Sigma0.46 0.91\Sigma0.20 3.18\Sigma0.57
\Gamma68 23 0.30 4.513\Sigma0.037 1.05\Sigma0.06 +0.11\Sigma0.42 0.65\Sigma0.10 4.28\Sigma0.84 0.71\Sigma0.14 3.88\Sigma0.88
\Gamma68 26 0.20 4.671\Sigma0.012 1.10\Sigma0.06 \Gamma0.64\Sigma0.43 0.90\Sigma0.13 3.76\Sigma0.44 0.43\Sigma0.11 3.11\Sigma0.50
\Gamma68 129 0.17 4.587\Sigma0.011 0.73\Sigma0.06 \Gamma1.48\Sigma0.39 1.26\Sigma0.19 1.50\Sigma0.42 0.72\Sigma0.16 2.81\Sigma0.91
\Gamma69 108 0.31 4.574\Sigma0.011 1.04\Sigma0.06 \Gamma1.25\Sigma0.39 0.98\Sigma0.11 4.31\Sigma0.44 0.54\Sigma0.10 3.98\Sigma0.61
\Gamma69 206 0.26 4.519\Sigma0.034 0.65\Sigma0.05 \Gamma1.40\Sigma0.38 1.23\Sigma0.14 1.08\Sigma0.43 0.38\Sigma0.11 2.56\Sigma1.09
\Gamma69 210 0.41 4.669\Sigma0.011 0.67\Sigma0.06 \Gamma1.15\Sigma0.37 1.12\Sigma0.11 1.42\Sigma0.41 0.52\Sigma0.11 3.16\Sigma1.07
\Gamma69 213 0.16 4.570\Sigma0.017 0.77\Sigma0.05 \Gamma2.62\Sigma0.37 1.56\Sigma0.24 2.08\Sigma0.46 0.83\Sigma0.21 3.51\Sigma0.90
Average b 0.25 4.596\Sigma0.017 0.91\Sigma0.05 \Gamma1.28\Sigma0.34 1.11\Sigma0.10 2.73\Sigma0.37 0.64\Sigma0.06 3.13 pm0.16
LMC 2 Sample
\Gamma68 140 0.20 4.559\Sigma0.022 1.07\Sigma0.09 \Gamma1.02\Sigma0.40 1.13\Sigma0.17 1.62\Sigma0.41 0.77\Sigma0.14 1.41\Sigma0.43
\Gamma68 155 0.20 4.663\Sigma0.011 0.91\Sigma0.07 \Gamma4.38\Sigma0.42 1.82\Sigma0.23 2.06\Sigma0.41 0.30\Sigma0.11 2.49\Sigma0.62
\Gamma69 228 0.17 4.658\Sigma0.016 1.26\Sigma0.13 \Gamma2.33\Sigma0.38 1.20\Sigma0.18 2.30\Sigma0.42 0.17\Sigma0.09 1.45\Sigma0.40
\Gamma69 256 0.17 4.622\Sigma0.038 1.21\Sigma0.05 \Gamma2.50\Sigma0.37 1.30\Sigma0.19 2.15\Sigma0.51 0.30\Sigma0.11 1.47\Sigma0.37
\Gamma69 265 0.19 4.627\Sigma0.018 0.92\Sigma0.10 \Gamma2.47\Sigma0.39 1.37\Sigma0.18 0.88\Sigma0.41 0.18\Sigma0.11 1.04\Sigma0.54
\Gamma69 270 0.19 4.651\Sigma0.011 1.12\Sigma0.09 \Gamma2.26\Sigma0.37 1.53\Sigma0.21 2.66\Sigma0.45 0.74\Sigma0.15 2.12\Sigma0.49
\Gamma69 279 0.21 4.603\Sigma0.016 0.84\Sigma0.06 \Gamma2.73\Sigma0.37 1.36\Sigma0.16 1.33\Sigma0.42 0.17\Sigma0.10 1.88\Sigma0.65
\Gamma69 280 0.18 4.618\Sigma0.016 0.74\Sigma0.06 \Gamma0.51\Sigma0.49 0.96\Sigma0.14 1.15\Sigma0.41 0.64\Sigma0.15 2.10\Sigma0.82
\Gamma70 116 0.19 4.637\Sigma0.024 1.42\Sigma0.10 \Gamma1.22\Sigma0.45 1.09\Sigma0.15 3.13\Sigma0.41 0.54\Sigma0.13 1.55\Sigma0.30
Average b 0.19 4.626\Sigma0.010 1.05\Sigma0.07 \Gamma2.16\Sigma0.36 1.31\Sigma0.08 1.92\Sigma0.23 0.42\Sigma0.08 1.72 pm0.14
Milky Way average c
-- 4.596\Sigma0.002 0.96\Sigma0.01 0.12\Sigma0.11 0.63\Sigma0.04 3.26\Sigma0.11 0.41\Sigma0.02 3.49\Sigma0.07
a Analytic fit to extinction curve following FM:
\Delta(–\GammaV )
\Delta(B\GammaV ) = C 1 + C 2 x + C 3 D(x) + C 4 F (x);
where x = – \Gamma1 , and
D(x) = x 2
(x 2 \Gammax 2
0 ) 2 +x 2 fl 2
:
F (x) = 0:5329(x \Gamma 5:9) 2 + 0:05644(x \Gamma 5:9) 3 (x ? 5:9)
and F (x) = 0 otherwise.
b Uncertainties in the averages quoted as the standard deviation of the sample mean for the respective samples, eg.
oe i =
p
N.
c From the Galactic data of FM. Errors are the standard deviation of the sample mean.

-- 11 --

-- 12 --

-- 13 --
Fig. 2.--- Individual LMC extinction curves. Optical data are included and, when available,
IR. We have plotted the FM fits and CCM curves offset by 4 and 9 units, respectively along
with the re--binned extinction curve. Where measured values of R V are available, CCM
curves for the measured R V (solid line) and R V \Sigma oe R V
are plotted (dotted line). When no
measured value of R V was available, the ``best fit'' CCM curve is plotted. If no single value
of R V provided an adequate fit, a CCM curve with R V = 3:1 is plotted.

-- 14 --
Fig. 3.--- Comparison of average 30 Dor (dashed line this study, upper dotted line F86) and
non--30 Dor (solid line this study, lower dotted line F86) extinction curves. 1oe error bars for
the new 30 Dor and non--30 Dor average curves have been plotted at various wavelengths.
and non--30 Dor samples is less, being only slightly greater than 1.5oe. The difference in
bump strength between our 30 Dor and non--30 Dor samples is slightly more significant.
Our average non--30 Dor bump strength (A bump = C 3 =fl 2 = 2.97\Sigma0.30) is slightly larger
than that of F86 (A bump = 2:58). This is not unexpected as we have included two new
lines of sight with strong bumps in our non--30 Dor average. In addition, the improvements
realized by using IUE spectra reduced with NEWSIPS are most apparent near the bump.
Our average 30 Dor bump strength (A bump = 2:12 \Sigma 0:20) is only slightly larger than that
found by F86 (A bump = 1:86). The difference in bump strength between our 30 Dor and
non--30 Dor samples is \DeltaA bump = 0:85 \Sigma 0:36, slightly greater than 2oe.

-- 15 --
3.1.2. LMC 2/LMC--general
However, the conclusion drawn by F86 that there are significant intrinsic variations
between extinction curves within each of the 30 Dor and non--30 Dor samples is strengthened
by the additional lines of sight included in this study. In the non--30 Dor sample, for
instance, SK \Gamma68 23 has a strong bump and SK \Gamma70 116 has almost no bump. Similar
differences are seen in the 30 Dor sample. To try and isolate a sample of sightlines with
weak bumps, we have plotted bump strength versus \Delta(B\GammaV) in Figure 4. We discovered
that there is a group of stars with similar reddenings (0.17 Ÿ \Delta(B \Gamma V ) Ÿ 0.21) and bump
strengths that also lie close together in the LMC. These stars lie in or near the region
occupied by the supergiant shell LMC 2 on the southeast side of 30 Dor (Meaburn 1980;
see Figure 1). This structure, which is 475 pc in radius, was formed by the combined stellar
winds and supernovae explosions from the stellar association within (Caulet & Newell
1996). There are nine stars in the LMC 2 group, eight of which are from the 30 Dor sample
and one (SK \Gamma70 116) from the non--30 Dor sample. Four 30 Dor stars (SK \Gamma68 129, \Gamma69
206, \Gamma69 210 and \Gamma69 213) are removed from our new LMC 2 sample. These four stars
lie in or near a prominent dust lane separating the 30 Dor star formation region from the
LH 89 and NGC 2042 stellar associations; SK \Gamma69 206 is on the south--eastern edge of the
dust lane near the stellar association LH 90 while SK \Gamma69 210 is in the middle of the dust
lane, coincident with CO clouds 7 & 8 of Johansson et. al. (1998). While located in the
traditional 30 Dor region, these sightlines have strong bumps, typical of the non--30 Dor
dust.
An average extinction curve has been calculated for the LMC 2 stars and also for the
remaining ten stars which we will call LMC--general. FM parameters and their respective
uncertainties were calculated as above and are reported in Table 3. The parameters for the
average Galactic curve as derived from FM are also shown for comparison. The average
curves for LMC 2 and LMC--general samples are plotted in Figure 5. These two curves
show a very significant difference in bump strength (\DeltaA bump = 1.41 \Sigma 0.21) but the far--UV
curves lie within one sigma of each other. It is worth noting that the average Galactic
bump strength is very similar to that of the LMC--general sample. In Figure 6 we have
over--plotted individual curves within each sample. The dispersion about the mean bump
strength is significantly less for both the LMC--general sample compared to the non--30 Dor
sample (0.50 and 0.78, respectively; Figure 6a) and for the LMC 2 sample compared to the
30 Dor sample (0.43 and 0.72 respectively; Figure 6b).

-- 16 --
Fig. 4.--- Plot of the bump strength normalized to \Delta(B\GammaV) vs. \Delta (B\GammaV). Symbols represent
the samples discussed in the text.

-- 17 --
Fig. 5.--- Comparison of 30 Dor/non--30 Dor average curves from this study (dashed line
and solid line, respectively) with the LMC average and LMC 2 average curves discussed in
the text (dotted line and dot­dash line, respectively). 1oe error bars for the new 30 Dor and
non--30 Dor average curves have been plotted at various wavelengths.

-- 18 --
Table 4. Measured R V Values.
SK R V oe RV
\Gamma66 19 2.46 0.25
\Gamma67 2 2.31 0.44
\Gamma69 108 2.61 0.15
\Gamma70 116 3.31 0.20
\Gamma68 140 2.76 0.35
\Gamma69 213 2.16 0.30
\Gamma69 228 2.23 0.74
\Gamma69 270 2.71 0.11
\Gamma69 279 2.43 0.31
\Gamma69 280 2.56 0.39
Fig. 6.--- (a) Comparison of the individual curves within the non--30 Dor sample (lower
curves) and LMC--general sample (upper curves). The LMC--general curves have been offset
6 units for clarity. (b) Same as (a) for the 30 Dor sample (lower curves) and LMC 2 sample
(upper curves). The LMC 2 curves have been offset by 6 units for clarity.

-- 19 --
3.2. Variations Within the Samples
The form of the UV extinction, as parameterized by FM, along a given line of sight is
potentially a powerful diagnostic of the dust grains responsible for the extinction but the
physical interpretation of variations and correlations among the FM parameters is unclear.
However, to the degree that they represent underlying physical processes it is useful to
examine them within our two LMC samples. The coefficients of the linear component of
the UV extinction (C 1 + C 2 x) are not independent in the Galaxy (Fitzpatrick & Massa
1988) but are in fact themselves linearly related. Fitzpatrick & Massa (1988) interpret
the relationship as either a single grain population modified by evolutionary processes or
a varying mixture of several grain populations with different UV extinction slopes or a
combination of both factors. While C 1 and C 2 have similar values between the two LMC
samples, both LMC samples exhibit systematically smaller values of C 1 and systematically
larger values of C 2 relative to the Galaxy. In the SMC, the values of C 1 and C 2 are even
more extreme than in the LMC (GC). However, the values of C 1 and C 2 for all these
galaxies follow the same linear relationship (see Figure 7a). Hence, whatever underlying
physical processes or dust components are responsible for the variations in the linear part of
the UV extinction must operate similarly in the Galaxy, the LMC and the SMC. Fitzpatrick
& Massa (1988) suggested a possible correlation between the FM parameters C 4 , which
measures the far UV curvature, and fl, the bump width. In Figure 7b we plot C 4 against
fl for the Galaxy, the LMC, and the SMC. Only one SMC sightline (AzV 456) is included
since the remaining three sightlines have no bump and fl is undefined (GC). There is no
correlation between these parameters in the LMC extinction data. The physical significance
of C 4 is unclear; the far UV extinction is a combination of the linear term and the C 4
polynomial term and the separation is mathematical rather than physical (CCM). However,
such a correlation may arise if C 4 and fl are due to different grain populations provided that
the different populations respond to environmental factors in a similar way (Fitzpatrick &
Massa 1988). This is consistent with the conclusion of CCM that the processes producing
changes in extinction must be efficient over a range of particle sizes and compositions. In
this case, the absence of correlation in the LMC would suggest that environmental processes
are affecting the different grain populations differently.
The FM parameters C 1 , C 2 , C 3 , and C 4 depend on R V and so interpreting relations
among them in the absence of R V information is difficult. CCM found that the general
shapes of the UV extinction curves in the Galaxy, expressed as A – =A V , are well represented
by a one parameter family of curves characterized by the value of R V . It is of interest to
determine whether the UV extinction in the LMC follows the relation of CCM and whether
the deviations from CCM in the LMC can be related to deviations seen in the Galaxy We
will discuss the FM bump parameters (x 0 and fl) separately in x3.2.2.

-- 20 --
Fig. 7.--- (a) Plot of C 1 vs. C 2 for the LMC--general sample, LMC 2 sample, the SMC,and
the Galaxy. The dashed line is the least--squares fit the Galactic data given by Fitzpatrick
& Massa (1988). (b) Plot of C 4 vs. fl. Symbols are as in Figure 7a. In both figures, Galactic
data from FM, SMC data from GC.
3.2.1. CCM and the LMC
There is an average Galactic extinction relation, A – =A V , over the wavelength range
0.125 ¯m to 3.5 ¯m, which is applicable to a wide range of interstellar dust environments,
including lines of sight through diffuse dust, dark cloud dust, as well as that associated with
star formation (CCM; Cardelli & Clayton 1991; Mathis & Cardelli 1992). The existence
of this relation, valid over a large wavelength interval, suggests that the environmental
processes which modify the grains are efficient and affect all grains. The CCM relation
depends on only one parameter R V , which is a crude measure of the size distribution of
interstellar dust. Only eleven LMC sightlines in our sample have measured values of R V .
Seven of these are in the LMC 2 sample. The CCM curves for these eleven stars are plotted
in Figure 2. The LMC 2 curves cannot be fit by a CCM curve with any value of R V because
of their weak bumps. The average LMC--general curve is very similar to a Galactic CCM
extinction curve with R V = 2:4. However, only four stars in this sample have measured R V
values so it is not clear how well their extinction curves follow CCM. SK \Gamma67 2 and \Gamma69 213
have stronger FUV extinctions than their respective CCM curves while SK \Gamma66 19 appears
too weak in the bump. Only SK \Gamma69 108 clearly follows the CCM relationship. In Figures 8
and 9 we plot bump strength and A 1300 =A V versus R \Gamma1
V for ten stars with measured R V 's; SK
\Gamma69 256 is excluded due to its very uncertain R V value. Figure 8 shows that bump strength
is consistent with CCM for the LMC--general sample while the LMC 2 sample has bumps
which fall below the typical CCM values. Very little can be said about the relationship

-- 21 --
with R V in the far UV as seen in Figure 9. The uncertainties are quite large and though
both the LMC 2 and LMC--general sample appear to be consistent with CCM in the UV,
they are also consistent with no R V dependence. More accurate values of R V along more
sightlines must be obtained before it can be determined whether a CCM--like relationship
may hold in the LMC. According to CCM, C 3 and therefore A bump are proportional to R V
(Mathis & Cardelli 1992). However, since C 3 (LMC--general)/C 3 (LMC 2) = 2.25, that would
imply that the average value of R V should be more than twice as large in the LMC--general
sample if a CCM--like relationship exists. There is no indication from the available data
that this is true. In fact, the sightlines in both samples appear to have low values of R V
relative to the Galaxy (Table 4). This may indicate that dust grains in the LMC may be
systematically smaller than in the Galaxy.
Although the general shape of the UV extinction in the Galaxy is well represented by
the R V parameterization of CCM, significant deviations are seen, both in the far UV and
the bump (Cardelli & Clayton 1991; Mathis & Cardelli 1992; Fitzpatrick 1998). There
are well known Galactic sightlines which deviate from CCM in much the same way that
the LMC 2 sample does. Three deviant Galactic stars are plotted in Figures 8 and 9
for comparison. HD 29647, 62542, and 210121 all show weak bumps and strong far--UV
extinction for their measured values of R V (3.62, 3.24 & 2.1, respectively; Messinger et. al
1997; Whittet et. al. 1993; Larson et. al. 1996). The bumps seen for HD 29647 and 62542
are not just weak but they are very broad and shifted to the blue (Cardelli & Savage 1988).
The unusual extinction curve characteristics along these lines of sight have been attributed
to their dust environments which are quite diverse. The dust toward HD 62542 has been
swept up by bubbles blown by two nearby O stars and has been subject to shocks while
the HD 29647 sightline passes through a very dense, quiescent environment (Cardelli &
Savage 1988). HD 210121 lies behind a single cloud in the halo. There is no present activity
near this cloud although it was ejected into the halo at some time in the past (Welty &
Fowler 1992; Larson et. al. 1996). These deviations from CCM in the Galaxy indicate
that something other than the size distribution of dust grains as measured by R V must be
important in determining extinction properties along a given line of sight. Evidently, the
same is true in the LMC. Even though all the LMC sightlines have similar, low values of
R V , they exhibit a variety of extinction curves.
3.2.2. The Bump
While the physical significance of the linear and far UV functions in the FM
parameterization is unclear, the Drude profile fitting function for the 2175 š A absorption

-- 22 --
Fig. 8.--- Bump strength normalized to A V plotted vs. R \Gamma1
V for LMC stars with measured
values of R V . The solid line represents the mean CCM relationship and the dotted lines the
approximate dispersion around the mean for the CCM sample of Galactic stars. Symbols
represent the samples discussed in the text. For comparison, several Galactic stars with
``unusual'' extinction curves are plotted. Bump strengths for the Galactic stars were taken
from Cardelli & Savage (1988) (HD 29647) and Welty & Fowler (1992) (HD 62542 & HD
210121). R V values are from Messinger et. al. (1997) (HD 29647), Whittet et. al. (1993)
(HD 62542) and Larson et. al. (1996) (HD 210121).

-- 23 --
Fig. 9.--- Plot of the extinction ratio A 1300 =A V vs. R \Gamma1
V where A 1300 is the extinction at
– = 1300 š A plotted as in Figure 8.

-- 24 --
bump which is part of the FM parameterization does have some physical significance as
the expression of the absorption cross section of a damped harmonic oscillator (CCM; FM;
Mathis & Cardelli 1992). Further, neither x 0 or fl depend on R V and so variations in these
parameters are directly tied to variations in the grains responsible for the bump feature.
There is no evidence for a systematic shift in the central position of the bump in either
LMC sample. The weakness of the bump in the LMC 2 sample means that x 0 and fl are not
strongly constrained in that sample. In the LMC--general sample, there are no systematic
redward shifts of the bump but three sightlines are significantly shifted to the blue. The
range of variation in the LMC is consistent with that seen in the Galactic sample. Several
Galactic lines of sight, eg. HD 62542 and HD 29647 have bumps that are significantly
shifted to shorter wavelengths (x 0 = 4:74 and 4.70, respectively; Cardelli & Savage 1992).
Several possibilities have been suggested to account for this including mantling of the grains
and hydrogenation (Cardelli & Savage 1992; Mathis 1994).
As in the Galaxy, there is real variation in the width of the bump between various lines
of sight in the LMC (Table 3). Five lines of sight in our sample have bump widths which
nominally fall below the narrowest Galactic bump (fl = 0:8, FM). Of these five sightlines
two (SK \Gamma69 213 and SK \Gamma69 280) are affected by spectral mismatches in the bump region.
The true bump widths for these two sightlines are not well constrained by the FM fitting
procedure. The remaining three narrow bump sightlines (SK \Gamma68 129, SK \Gamma69 206, SK
\Gamma69 210), all in the LMC--general sample, interestingly all fall in or near the dust lane on
the northwest edge of 30 Dor. The bumps are well defined and the narrowness of the bump
is real. An expanded view of the SK \Gamma69 210 profile can be seen Figure 10. There is a
strong relationship between environment and fl in the Galaxy. The narrowest bumps are
associated with bright nebulosity while wide bumps are associated with dark, dense clouds
(Cardelli & Clayton 1991, Mathis 1994). Therefore, it has been suggested that mantles form
on the bump grains in dark clouds resulting in broad bumps. In bright nebulae, there are
no mantles and narrower bumps result from the bare grains. In this scenario, mantles are
able to form in dense clouds shielded from the interstellar radiation field while the mantles
on grains near H II regions are removed by the stronger radiation field. However, the three
small fl lines of sight in the LMC appear to be associated with a dense environment even
though they are near the 30 Dor star forming region. Several stars in the LMC--general
sample (eg. SK \Gamma66 19 and SK \Gamma66 88) are associated with bright H II regions and yet
have normal Galactic bump widths. Accepting the explanation for the narrow bumps based
on the Galactic data, we would expect to find narrow bumps in the LMC 2 sample. In
contrast with this expectation, the data presented in Table 3 indicates that the LMC 2
bump widths are comfortably within the average Galactic range. It doesn't appear that the
trend in fl with environment seen the Galaxy holds in the LMC. There are no exceptionally

-- 25 --
wide bumps in our sample save SK \Gamma70 116 with fl = 1:4; however, the bump is extremely
weak and fl is not strongly constrained.
The weak bumps in the LMC 2 region are not unique. As discussed above, several
Galactic lines of sight also have extinction curves with very weak bumps (HD 29647, HD
62542, HD 210121). However, the LMC 2 environment seems to have little in common with
these Galactic lines of sight which in turn seem to have little in common with each other.
Though the swept up, shocked environment near HD 62542 may be similar to the LMC 2
environment (but on a vastly reduced scale), the other two Galactic sightlines sample
relatively quiescent environments. HD 210121 lies behind a single diffuse, translucent cloud
about 150 pc from the Galactic plane. The interstellar radiation field is weaker than in the
general interstellar medium and shocks do not appear to be important (Welty & Fowler
1992). Larson et. al. (1996) suggest that the apparent preponderance of small grains along
the HD 210121 line of sight is due to lack of grain growth through coagulation as a result
of lack of time spent in a dense environment. It appears that very diverse environmental
conditions result in rather similar bump profiles. It is not known whether the bump grains
are being modified in a similar fashion in different environments or substantially different
modifications of the bump grains can result in a similar UV extinction in the bump.
4. Conclusions
Evidently the relationship between the UV extinction, dust grain properties, and
environment is a complicated one. Similar variations in the form of the UV extinction can
arise in a variety of environments. The environmental dependences seen in the Galaxy
do not seem to hold in the LMC. Since large variations in UV extinction are seen within
both the LMC and the Galaxy, global parameters such as metallicity cannot be directly
responsible for the observed variations from galaxy to galaxy as has been suggested (e.g.,
Clayton & Martin 1985). However, one effect of decreased metallicity in the LMC is that
the typical molecular cloud is bigger but more diffuse than those in the Galaxy (Pak
et. al. 1998). Hence, dust grains in the LMC may not spend as much time in dense,
shielded environments as grains in the Galaxy. The lack of time in dense environments may
contribute to the apparent small size of the LMC grains as indicated by the low values
of R V measured in this study. In addition, the weak and narrow bump lines of sight in
the LMC all lie near the 30 Dor star forming region which has no analog in the Galaxy.
The dust along these sightlines has probably been affected by the proximity to the harsh
environment of the copious star formation associated with 30 Dor. However, it must be
pointed out that the most extreme UV extinction curves, having virtually no bumps and a

-- 26 --
Fig. 10.--- Drude profile of SK \Gamma69 210 (filled circles) binned to ¸30 š A resolution compared
to the average Galactic Drude profile.

-- 27 --
very steep far UV are found in the SMC. The SMC dust lies near regions of star formation
but they are very modest compared to 30 Dor. These SMC sightlines have optical depths
similar to those in LMC 2 (GC). Due to very low metallicity of of the SMC, its molecular
clouds are very diffuse (Pak et. al. 1998). One might expect that values of R V in the SMC
be even smaller than in the LMC; the current observations, however, show no evidence for
this (GC).
Even with the improved and expanded samples of extinction in the LMC and
SMC, the link between particular environments and dust characteristics is still unclear.
The combination of the Galactic and Magellanic cloud data show that the extinction
curve/environment links are not as simple as previously proposed. But the different times
spent by grains in dense molecular environments may be a significant factor as suggested for
the Galactic star HD 210121 (Larson et. al. 1996). The processing history of dust grains (ie.
coagulation and mantling in dense clouds environments and exposure to strong shocks and
radiation field outside of clouds) is probably quite different in these three galaxies owing to
the different molecular cloud environments and the varying intensity of star formation. The
interplay between at least these two factors likely plays an important role in determining
the form the UV extinction. The fact that starburst galaxies appear to have SMC--type
dust regardless of metallicity (Calzetti et. al. 1994; Gordon et. al. 1997) implies that the
star formation history of a galaxy plays an important role in determining the extinction
properties. However, the complicated relationship between extinction properties in the UV
and environment implied by the Galactic and Magellanic Cloud data suggests that great
care must be taken in assuming the form of the UV extinction in external galaxies.
This research has made use of the SIMBAD database. IUE spectra were down loaded
from the IUE final archive at ESA. This work has been partially supported through NASA
ATP grant NAG5 3531 to GCC. We thank M. Oestreicher for providing the source code and
data files used for generating the foreground reddening map and M. Bessel for supplying
the Hff image.
REFERENCES
Calzetti, D., Kinney, A. L. & Storchi--Bergmann, T. 1994, ApJ, 429, 582.
Cardelli, J. A. & Savage, B. D. 1988, ApJ, 325, 864.
Cardelli, J. A., Clayton, G. C. & Mathis, J. S. 1989, ApJ, 345, 245 (CCM).
Cardelli, J. A. & Clayton, G. C. 1991, AJ, 101, 1021.

-- 28 --
Caulet, A. & Newell, R. 1996, ApJ, 465, 205.
Clayton, G. C. & Martin, P. G. 1985, ApJ, 288, 558.
Clayton, G. C. et. al. 1996, ApJ, 460, 313.
Fitzpatrick, E. L. 1985, ApJ, 299, 219 (F85).
Fitzpatrick, E. L. 1986, AJ, 92, 1068 (F86).
Fitzpatrick, E. L. 1988, ApJ, 335, 703.
Fitzpatrick, E. L. 1998, in `UV Astrophysics Beyond the IUE Final Archive', eds, W.
Wamsteker & R. Gonzalez Riestra. (Noordwijk: ESA Publications Division), p.461.
Fitzpatrick, E. L. & Massa, D. 1988, ApJ328, 734.
Fitzpatrick, E. L. & Massa, D. 1990, ApJS, 72, 163 (FM).
Gordon, K. D., Calzetti, D. & Witt, A. N. 1997, ApJ, 487, 652.
Gordon, K. D. & Clayton, G. C. 1998, ApJ, 500, 816 (GC).
Johansson, L. E. B. et. al. 1998, A&A, 331, 857.
Johnson, H. L. 1966 ARA&A, 4, 193.
Koornneef, J. & Code, A. D. 1981, ApJ, 247, 860.
Larson, K. A., Whittet, D. C. B. & Hough, J. H. 1996, ApJ, 472, 755.
Massa, D., Savage, B. D. & Fitzpatrick, E. L. 1983, ApJ, 266, 662.
Mathis, J. S. 1994, ApJ, 422, 176.
Mathis, J. S. & Cardelli, J. A. 1992, ApJ, 398, 610.
Meaburn, J. 1980, MNRAS, 192, 365.
Messinger, D. W., Whittet, D. C. B. & Roberge, W. G. 1997, ApJ, 487, 314.
Morgan, D. H. & Nandy, K. 1982, MNRAS, 199, 979.
Nandy, K., Morgan, D. H., Willis, A. J., Wilson, R. & Gondhalekar, P. M.1981, MNRAS, 1
96, 955.
Neubig, M. M. S. & Bruhweiler, F. C. 1998, AJ, submitted.

-- 29 --
Pak, S., Jaffe, D. T., van Dishoeck, E. F., Johansson, L. E. B. & Booth, R. S. 1998, ApJ,
498, 735.
Oestreicher, M. O., Gochermann, J. & Schmidt--Kaler, T. 1995, A&AS, 112, 495.
Rieke, G. H. & Lebofsky, M. J. 1985, ApJ, 288, 618.
Rousseau J.M., Martin N., Prevot L., Rebeirot E., Robin A. & Brunet J.P. 1978, A&AS,
31, 243.
Schwering, P. B. W. & Israel, F. P. 1991, A&A, 246, 231.
Welty, D. E. & Fowler, J. R. 1992, ApJ, 393, 193.
Whittet, D. C. B., Martin, P. G., Fitzpatrick, E. L. & Massa, D. 1993, ApJ, 408, 573.
This preprint was prepared with the AAS L A T E X macros v4.0.