1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3: % FM EXTINCTION SERIES #V %
4: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6:
7: %\documentclass{aastex}
8: \documentclass[10pt,preprint]{aastex}
9:
10: \slugcomment{To appear in the Astrophysical Journal}
11: \shortauthors{Fitzpatrick \& Massa}
12: \shorttitle{Extinction Survey}
13:
14: %------------------------------------------------------------------------%
15: % END OF PREAMBLE
16: %
17: %------------------------------------------------------------------------%
18:
19: \begin{document}
20:
21: \newcommand{\atlas}{{ATLAS9}}
22: \newcommand{\tlusty}{{TLUSTY}}
23: \newcommand{\synspec}{{SYNSPEC}}
24: \newcommand{\iras}{{\it IRAS}}
25: \newcommand{\ans}{{\it ANS}}
26: \newcommand{\hst}{{\it HST}}
27: \newcommand{\iue}{{\it IUE}}
28: \newcommand{\oao}{{\it OAO-2}}
29: \newcommand{\td}{{\it TD-1}}
30: \newcommand{\tmass}{{2MASS}}
31: \newcommand{\mast}{{MAST}}
32: \newcommand{\irsa}{{IRSA}}
33: \newcommand{\hip}{{\it Hipparcos}}
34: \newcommand{\simbad}{{\it SIMBAD}}
35:
36: \newcommand{\teff}{\mbox{$T_{\rm eff}$}}
37: \newcommand{\logg}{{$\log g$}}
38: \newcommand{\vturb}{$v_{turb}$}
39: \newcommand{\abund}{[m/H]}
40: \newcommand{\vsini}{$v \sin i$}
41:
42: \newcommand{\kms}{km\,s$^{-1}$}
43: \newcommand{\msun}{${\rm M}_\sun$}
44: \newcommand{\ebv}{\mbox{$E(B\!-\!V)$}}
45: \newcommand{\invmic}{\mbox{$\mu{\rm m}^{-1}$}}
46:
47: \title{An Analysis of the Shapes of Interstellar Extinction Curves. V.
48: The IR-Through-UV Curve Morphology}
49:
50: \author{E.L.~Fitzpatrick\altaffilmark{1}, D.~Massa\altaffilmark{2}}
51: \altaffiltext{1}{Department of Astronomy \& Astrophysics, Villanova
52: University, 800 Lancaster Avenue, Villanova, PA 19085, USA;
53: fitz@astronomy.villanova.edu}
54: \altaffiltext{2}{SGT, Inc., NASA/GSFC, Mailstop 665.0, Greenbelt,
55: MD 20771; massa@derckmassa.net}
56:
57: \begin{abstract}
58: We study the IR-through-UV wavelength dependence of 328 Galactic
59: interstellar extinction curves affecting normal, near-main sequence B
60: and late O stars. We derive the curves using a new technique which
61: employs stellar atmosphere models in lieu of unreddened ``standard''
62: stars. Under ideal conditions, this technique is capable of virtually
63: eliminating spectral mismatch errors in the curves. In general, it
64: lends itself to a quantitative assessment of the errors and enables a
65: rigorous testing of the significance of relationships between various
66: curve parameters, regardless of whether their uncertainties are
67: correlated. Analysis of the curves gives the following results:
68: \begin{enumerate}
69: \item In accord with our previous findings, the central position of the
70: 2175 \AA\/ extinction bump is mildly variable, its width is highly
71: variable, and the two variations are unrelated. \item Strong
72: correlations are found among some extinction properties within the UV
73: region, and within the IR region.
74: \item With the exception of a few curves with extreme (i.e., large) values of $R(V)$,
75: {\it the UV and IR portions of Galactic extinction curves are not
76: correlated with each other.}
77: \item The large sightline-to-sightline variation seen in our sample
78: implies that any average Galactic extinction curve will always reflect the
79: biases of its parent sample.
80: \item The use of an average curve to deredden a spectral energy
81: distribution (SED) will result in significant errors, and a realistic
82: error budget for the dereddened SED must include the observed variance of
83: Galactic curves.
84: \end{enumerate}
85: While the observed large sightline-to-sightline variations, and the
86: lack of correlation among the various features of the curves, make it
87: difficult to meaningfully characterize average extinction properties,
88: they demonstrate that extinction curves respond sensitively to local
89: conditions. Thus, each curve contains potentially unique information
90: about the grains along its sightline.
91: \end{abstract}
92:
93: \keywords{ISM:dust,extinction --- methods:data analysis}
94:
95: %%%%%%%%%%%%%%%%%%%%%%%%%
96: % INTRODUCTION
97: %%%%%%%%%%%%%%%%%%%%%%%%%
98:
99: \section{INTRODUCTION\label{secINTRO}}
100: In the previous paper in this series (Fitzpatrick \& Massa 2005a,
101: hereafter Paper IV), we introduced a technique,
102: ``extinction-without-standards,'' to determine the shapes of
103: UV-through-IR interstellar extinction curves by modeling the observed
104: spectral energy distributions (SEDs) of reddened early-type stars. The
105: method involves a $\chi^2$-minimization procedure to determine
106: simultaneously the basic properties of a reddened star (namely, \teff,
107: \logg, \abund, and \vturb) {\it and} the shape of its extinction curve,
108: utilizing grids of stellar atmosphere models to represent intrinsic
109: SEDs and an analytical form of the extinction curve, whose shape is
110: determined by a set of adjustable parameters.
111:
112: In general, the benefits of extinction-without-standards are increased
113: accuracy and precision (in most applications) over results generated
114: using the standard Pair Method of extinction curve determination and,
115: very importantly, a reliable estimate of the uncertainties in the
116: resultant extinction curves. Specifically, the advantages of the new
117: method include: (1) the elimination of the requirement for observations
118: of unreddened spectral standard stars, (2) the near-elimination of
119: ``spectral mismatch'' as a source of extinction curve error, (3) the
120: ability to produce accurate curves for much more lightly-reddened
121: sightlines than heretofore possible, (4) the ability to derive accurate
122: ultraviolet curves for later spectral types (i.e., to late-B classes)
123: than previously possible, and (5) the ability to provide quantified
124: estimates of the degree of correlation between various morphological
125: features of the curves.
126:
127: The chief limitation of the method is that the intrinsic SEDs of the
128: reddened stars must be well-represented by model atmosphere
129: calculations. This currently eliminates from consideration such
130: objects as high luminosity stars, Be stars, Wolf-Rayet stars, and any
131: spectrally peculiar star. This restriction also affects the Pair
132: Method, since there are only a small number of unreddened members of
133: these classes which can serve as ``standard stars,'' and it is
134: difficult to be certain that the intrinsic SEDs of the standard stars
135: really represent those of the reddened stars. In addition, the extinction-without-standards technique requires well-calibrated (and absolutely-calibrated) SED observations.
136:
137: In this paper, we apply our extinction-without-standards technique to a
138: sample of 328 Galactic stars for which multi-wavelength SEDs are
139: available. For these stars, we derive normalized UV-through-IR
140: extinction curves, sets of parameters which describe the shapes of the
141: curves, and sets of parameters which characterize the stars
142: themselves. The scope of our discussion focuses on two objectives: (1)
143: the presentation of the broad-ranging results and a description of the
144: methodology employed, and (2) a thorough examination of general
145: extinction curve morphology. Among the issues addressed is the
146: correlation between the IR and UV properties of extinction (see, e.g.,
147: Cardelli, Clayton, \& Mathis 1989). This paper is not intended as a
148: review of Galactic extinction and we explicitly restrict our attention
149: only to issues touched upon by our new analysis. The results, however,
150: do constitute a broad view of Galactic extinction and lend themselves
151: to numerous other investigations, including the general properties of
152: extinction curves, regional trends in extinction properties, the
153: correlation between extinction and other interstellar properties,
154: determination of intrinsic SEDs for objects in clusters containing
155: survey stars, the study of small scale spatial variations in dust grain
156: populations from stars in cluster extinction curves, etc. We plan to
157: pursue some of these in future papers. Preliminary results from an
158: early version of this study were reported by Fitzpatrick (2004;
159: hereafter F04), which also provided a review of then-recent progress in
160: interstellar extinction studies.
161:
162: The sample of stars chosen for this survey and the data used are
163: described in \S \ref{secDATA}. This is followed in \S \ref{secMETHOD}
164: by a brief description of the extinction-without-curves technique,
165: including a discussion of the error analysis --- which is critical to
166: the analyses in the latter parts of the paper. The results of the
167: survey, essentially an atlas of Galactic extinction curves, is
168: presented in \S \ref{secATLAS}, with the full sets of tables and the
169: full set of figures from this section available in the electronic
170: edition of the Journal. In \S \ref{secSTARS} we briefly discuss the
171: stellar properties derived from our analysis and then in \S
172: \ref{secEXTINCTION} present a detailed description of Galactic
173: extinction curve morphology, from the IR to the UV spectral regions.
174: Finally, in \S \ref{secDISCUSS} we provide a brief summary of the chief
175: conclusions of this study.
176:
177: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
178: % THE SURVEY DATA
179: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
180:
181: \section{THE SURVEY STARS AND THEIR DATA\label{secDATA}}
182:
183: In principle, the extinction-without-standards technique can be applied
184: to, or expanded to include, any type of SED data. In practice,
185: however, we have developed the technique with specific datasets in
186: mind, namely, low-resolution UV spectrophotometry from the {\it
187: International Ultraviolet Explorer} (\iue) satellite and ground-based
188: optical and near-IR photometry. In this particular study, we utilize
189: \iue\/ spectrophotometry, {\it UBV} photometry, and {\it JHK}
190: photometry from the {Two-Micron All Sky Survey} (\tmass). Although
191: some of the survey stars have additional data available, e.g.,
192: Str\"{o}mgren $ uvby\beta$ photometry, we elected --- for the sake of
193: uniformity --- to include only the {\it UBV} data in the analysis. As
194: a result, the stellar parameters we derive in this paper will be less
195: accurate than those determined in Paper IV. Nevertheless, the errors
196: in these parameters are still well-determined.
197:
198: The most restrictive of the datasets is that of the \iue, and so we
199: began our search for survey stars with the \iue\/ database. Using the
200: search engine provided by the {Multimission Archive at STScI} (\mast),
201: we examined all low-resolution spectra for stars with \iue\/ Object
202: Class numbers of 12 (main sequence O), 20 (B0-B2 V-IV), 21 (B3-B5
203: V-IV), 22 (B6-B9.5 V-IV), 23 (B0-B2 III-I), 24 (B3-B5 III-I), and 25
204: (B6-B9.5 III-I). Because our goal is to obtain a uniform dataset of
205: reddened stars whose SEDs can be modeled accurately, we eliminated the
206: following types of objects from the available field of candidates: (1)
207: stars without good-quality spectra from both the short-wavelength (SWP)
208: and long-wavelength (LWR or LWP) \iue\/ spectral regions, (2)
209: clearly-unreddened stars, (3) known Be stars, (4) luminosity class I
210: stars, (5) O stars more luminous than class V, (6) O stars earlier than
211: spectral type O5, and (7) stars with peculiar-looking UV spectra (as
212: based on our own assessment).
213:
214: The list of potential candidates from the \iue\/ database was then
215: examined for the availability of {\it UBV} data, using the General
216: Catalog of Photometric Data (GCPD) maintained at the University of
217: Geneva (see Mermilliod, Mermilliod, \& Hauck 1997)\footnotemark
218: \footnotetext{The GCPD catalog was accessed via the website at
219: http://obswww.unige.ch/gcpd/gcpd.html.}. In general, stars without
220: {\it V} and \bv\/ measurements were eliminated from consideration.
221: However, broadband Geneva photometry was available for five stars
222: without {\it UBV} data (BD+56 526, HD62542, HD108927, HD110336, and
223: HD143054), and so we included these stars and utilized the Geneva $U-B$
224: and $V-B$ indices.
225:
226: The final trimming of the survey sample was not performed until the
227: SEDs had been modeled. At this point we imposed the requirement that
228: all survey stars must have values of $\ebv \geq 0.20$ mag. This limit
229: is somewhat arbitrary and, as shown by Paper IV, useful extinction
230: curves can be derived via the extinction-without-standards technique
231: for \ebv\/ values considerably lower than 0.20 mag. However, the
232: uncertainties do rise at low \ebv\/ and for this survey we wanted a
233: sample of stars for which the uncertainties in the final parameters
234: were uniformly small. We plan in the future to examine the
235: extinction properties along lightly reddened sightlines.
236:
237: Near-IR {\it JHK} photometry (and their associated uncertainties) were
238: retrieved for the survey stars from the \tmass\/ database at the
239: NASA/IPAC Infrared Science Archive (\irsa)\footnotemark
240: \footnotetext{The \tmass\/ data were accessed via the \irsa\/ website
241: at http://irsa.ipac.caltech.edu/applications/Gator.}. Only those
242: \tmass\/ magnitudes for which the uncertainties are less than $\pm0.1$
243: mag are used here. For seven stars whose \tmass\/ measurements were
244: either non-existent or of low quality (HD23180, HD23512, HD37022,
245: HD144470, HD147165, HD147933, and HD149757), Johnson {\it JHK}
246: magnitudes were available and were retrieved via the GCPD catalog.
247: Note that the availability of {\it JHK} data was not a requirement for
248: inclusion in this survey although, as will be quantified below, most of
249: our sample have such data.
250:
251: The last data collection activity was to retrieve ancillary information
252: for all stars, such as coordinates, alternate names, and spectral
253: types, using the \simbad\/ database.
254:
255: The only data processing required for this program involved the \iue\/
256: spectrophotometry. The \iue\/ data contained in the \mast\/ archive
257: were processed using the NEWSIPS software (Nichols \& Linsky 1996). As
258: discussed in detail by Massa and Fitzpatrick (2000; hereafter MF00),
259: these data contain significant thermal and temporal dependencies and
260: suffer from an incorrect absolute calibration. We corrected the data
261: for their systematic errors and placed them onto the \hst/FOS flux
262: scale of Bohlin (1996) using the corrections and algorithms described
263: and provided by MF00. This step is absolutely essential for our
264: program since our ``comparison stars'' for deriving extinction curves
265: are stellar atmosphere models and systematic errors in the absolute
266: calibration of the data do not cancel out as they would in the case of
267: the Pair Method. (Note that the thermally- and temporally-dependent
268: errors in the NEWSIPS data would not generally cancel out in the Pair
269: Method --- see MF00.) When multiple spectra were available in one of
270: \iue{\it 's} wavelength ranges (SWP or LWR and LWP), they were combined
271: using the NEWSIPS error arrays as weights. Small aperture data were
272: scaled to the large aperture data and both trailed and point source
273: data were included. Short and long wavelength data were joined at
274: 1978~\AA\ to form a complete spectrum covering the wavelength range
275: $1150 \leq \lambda \leq 3000$~\AA. Data longward of 3000~\AA\ were
276: ignored because they are typically of low quality and subject to
277: residual systematic effects.
278:
279: After all the limits and restrictions were imposed, we arrived at our
280: final sample of 328 stars for which UV spectrophotometry covering
281: 1150-3000~\AA\/ and {\it UBV} photometry are available. Of these, 298
282: stars have at least some near-IR photometry, while 287 have a complete
283: set of {\it J-}, {\it H-}, and {\it K-} band measurements (with 280 of
284: these from the \tmass\/ program). Table \ref{tabSTARS} lists all the
285: survey stars, along with some general descriptive information. (The
286: complete version of Table \ref{tabSTARS} appears in the electronic
287: version of the paper.) The stars in Table 1 are ordered by Right
288: Ascension. For the star names, we adopted the most common form among
289: all the possibilities listed in the \simbad\/ database (i.e., ``HDnnn''
290: was the most preferred, followed by ``BDnnn'', etc.) There are 185
291: survey stars which are members of open clusters or associations. The
292: identity of the cluster or association is either contained in the star
293: name itself (e.g., NGC 457 Pesch 34) or is given in parentheses after
294: the star's name. The spectral types in Table \ref{tabSTARS} were
295: selected from those given in the \simbad\/ database, and the source of
296: the adopted types is shown in the ``Reference'' column of the table.
297: When multiple types were available for a particular star, we selected
298: one based on our own preferred ranking of the sources. For the B
299: stars, the quality of the spectral types varies widely, and the types
300: themselves are given only as a general reference --- {\it they do not
301: play any role in our analysis}. A scan of the types seems to indicate
302: that in some instances we have violated our selection criteria, e.g.,
303: several ``Ib''s and ``e''s can be found. However, in our estimation,
304: these are unreliable types and do not reflect the spectral information
305: we have examined. For example, a number of B stars in clusters have
306: been erroneously classified as emission-line stars based on
307: contamination of their spectra by nebular emission lines. The O stars
308: in the sample are, on the other hand, uniformly well-classified and the
309: types are used in the analysis (see \S \ref{secATLAS}).
310:
311: An overview of the survey sample can be gained from Figures
312: \ref{figCOORDS} and \ref{figSTARSTATS}. The former shows the location
313: of the stars on the sky (in Galactic coordinates) and the latter
314: summarizes the breadth of the stellar and interstellar properties of
315: the sample. Figure \ref{figCOORDS} shows that our sample is clearly
316: biased to sightlines passing through the galactic plane, as would be
317: expected given by the lower limit of $\ebv = 0.20$ mag which we
318: imposed. The locations of open clusters or associations for which
319: five of more members are in the survey sample are indicated on the
320: figure by the large circles. Note that the size of the circles is {\it
321: not} intended to represent the physical extent of the
322: clusters or associations. The data shown in histogram form in Figure
323: \ref{figSTARSTATS} are final results from the analysis, but are useful
324: here to characterize the sample. Most of our stars are mid-to-early B
325: stars ($\teff = 15000-30000$ K) with a median reddening of
326: $\ebv = 0.45$ mag. The median value of $R(V)$ ($\equiv
327: A_V/\ebv$) for the sample is 3.05, essentially identical to the Galactic
328: mean value for the diffuse ISM, and our sample is dominated by
329: sightlines though the diffuse ISM. However --- due to the relatively
330: small sample size and the biases present in the \iue\/ satellite's
331: choice of targets over the years --- our survey does not necessarily
332: constitute a representative sample of the various types of regions
333: present in the ISM. Care must be taken in interpreting average
334: properties derived from our results.
335:
336: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
337: % THE ANALYSIS
338: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
339:
340: \section{THE ANALYSIS\label{secMETHOD}}
341:
342: It was shown by Paper IV that the energy distributions of reddened B-type
343: stars could be modeled successfully using theoretical predictions of
344: the intrinsic SEDs of the stars and a parametrized form of the
345: UV-through-IR extinction curve to account for the distortions
346: introduced by interstellar extinction. A byproduct of the fit is a
347: determination of the wavelength-dependence of the extinction affecting
348: the star. This is the essence of the extinction-without-standards
349: technique. Although it was discussed in detail by Paper IV, for
350: completeness, we use this section to outline the basics of the
351: technique. In addition, some details of the process have changed since
352: Paper IV, as a result of experience gained from the application of the
353: process to several hundred reddened stars. These changes are
354: highlighted here.
355:
356: \subsection{Modeling the SEDs}
357:
358: The observed SED $f_{\lambda}$ of a reddened star can be represented as
359: \begin{equation}
360: \label{eqnFLUX}
361: f_{\lambda} = F_{\lambda}\, \theta_R^2\, 10^{-0.4 E(B-V) [k(\lambda-V) + R(V)]} \;\;,
362: \end{equation}
363: where $F_{\lambda}$ is the intrinsic stellar surface flux, $\theta_R
364: \equiv R/d$ is the stellar angular radius (where $R$ is the physical
365: radius and $d$ is the distance), \ebv\/ is the familiar measure of the
366: amount of interstellar reddening, $k(\lambda-V) \equiv
367: E(\lambda-V)/E(B-V)$ is the normalized extinction curve, and $R(V)
368: \equiv A_\lambda/E(B-V)$ is the ratio of reddening to extinction at
369: $V$. By adopting stellar atmosphere models to represent $F_\lambda$
370: and a using parametrized form of $k(\lambda-V)$, we can treat Equation
371: (\ref{eqnFLUX}) as a non-linear least squares problem and solve for the
372: set of optimal parameters which generate the best fit to the observed
373: flux. As in Paper IV, we perform the least squares minimization using the
374: Interactive Data Language (IDL) procedure MPFIT developed by Craig
375: Markwardt \footnotemark \footnotetext{Markwardt IDL Library available
376: at http://astrog.physics.wisc.edu/$\sim$craigm/idl/idl.html.}. The
377: observed SEDs which are fitted in the process consist of the \iue\/ UV
378: spectrophotometric fluxes, optical {\it UBV} magnitudes, and near-IR
379: {\it JHK} magnitudes discussed \S \ref{secDATA}.
380:
381: As related in Paper IV, we developed this analysis utilizing R.L. Kurucz's
382: (1991) line-blanketed, hydrostatic, LTE, plane-parallel \atlas\/ models,
383: computed in units of erg~cm$^{-2}$~sec$^{-1}$~\AA$^{-1}$ and the
384: synthetic photometry derived from the models by Fitzpatrick \& Massa
385: (2005b). The models are functions of four parameters: \teff\/,
386: \logg\/, [m/H], and $v_t$. All of these parameters can be determined
387: in the fitting process although, because of data quality, it is
388: sometimes necessary to constrain one or more to a reasonable value and
389: solve for the others. We smooth and bin the \iue\/ fluxes to match the
390: sampling of the \atlas\/ models (10 \AA\/ bins over most of the \iue\/
391: range; see Fitzpatrick \& Massa 1999).
392:
393: Because we include some O stars in the current sample, we have expanded
394: our technique to incorporate the \tlusty\/ OSTAR2002 grid of
395: line-blanketed, hydrostatic, NLTE, plane-parallel models by Lanz \&
396: Hubeny (2003). That grid includes 12 \teff\/ values in the range 27500
397: -- 55000 K, 10 chemical compositions from twice-solar to metal-free,
398: and surface gravities ranging from \logg\/ = 4.75 down to the modified
399: Eddington limit. All models were computed with \vturb\/ = 10 \kms. We
400: only consider the solar abundance models in this analysis and, thus,
401: the \tlusty\/ models are considered as functions of two parameters,
402: \teff\/ and \logg. Synthetic {\it UBV} photometry for these models was
403: produced as described above for the \atlas\/ models. To keep the
404: O-star and B-star fitting procedures as similar as possible, the
405: \tlusty\/ models were binned to the same wavelength scale as the
406: \atlas\/ models. While analyzing the O stars we found that their
407: low-dispersion \iue\/ spectra (excluding the strong wind lines) are
408: very insensitive to temperature and our analysis yielded very uncertain
409: results. As a result, we modified the procedure for these stars and
410: adopted values of \teff\/ based on their spectral types, rather than
411: solving for \teff. Table \ref{tabOTEMPS} lists the temperature scale
412: used; it is a compromise between the results of Martins, Schaerer, \&
413: Hillier (2005) and our own analysis of optical line spectra of O-stars
414: (such as published by Walborn \& Fitzpatrick 1990) using the \tlusty\/
415: models, which are appropriate for O stars without massive winds. The
416: results of this investigation will be reported elsewhere. We assume a
417: generous uncertainty in these \teff\/ values (see below) and the
418: extinction curve results are actually very insensitive to the adopted
419: temperatures.
420:
421: The value of the surface gravity $\log g$ is often poorly-determined
422: when using only broadband photometry, because it lacks a specific
423: gravity-sensitive index to help constrain $\log g$. For the
424: cluster stars, however, we can apply ancillary information ---
425: specifically, the cluster distances as listed in Table \ref{tabSTARS}
426: --- to provide strong constraints on $\log g$. We adopt the same
427: procedure as used by Fitzpatrick \& Massa (2005b), in which the Padova
428: grid of stellar structure models allow the Newtonian gravity of a star
429: to be inferred through its unique relation with the star's surface
430: temperature and radius. (When the distance is specified, the physical
431: radius of the star becomes a fit parameter via its influence on the
432: angular radius $\theta_R$. Fitzpatrick and Massa (2005b) used distances
433: determined by \hip\/ parallaxes.) In our iterative fitting procedure,
434: the current values of $T_{eff}$ and $\theta_R$, coupled with the Padova
435: models, determine the current value of $\log g$. Generous 1-$\sigma$
436: uncertainties in the distances are included in the error analysis (see
437: below). For field stars, no such constraints on $\log g$ are possible
438: and we solve for $\log g$ as a free parameter. We do, however, apply a
439: reality check to the results and, if the final value seems physically
440: unlikely (i.e., $\log g >$ $\sim$4.3 or $\log g <$ $\sim$3.0), we
441: replace it with the mean sample value of $\log g = 3.9$.
442: Uncertainties in this assumed value are incorporated in the error
443: analysis.
444:
445: A parametrized representation of the extinction curve, covering the
446: whole UV-through-IR spectral range, is the heart of the current
447: analysis. As in Paper IV, we construct this curve in two parts, joined
448: together at 2700 \AA. An example of this formulation, for one
449: particular set of parameters, is illustrated in Figure
450: \ref{figEXTCRV}. In the UV, ($\lambda \leq 2700$ \AA; the shaded
451: region in the figure), we use a modified form of the UV parametrization
452: scheme of Fitzpatrick \& Massa (1990; hereafter Paper III), as shown by the
453: thick solid curve. At longer wavelengths, we use a cubic spline
454: interpolation through a set of UV, optical, and IR ``anchor points''
455: (the $U_n$, $O_n$, and $I_n$ in the figure), as shown by the thick
456: dashed curve.
457:
458: The modified Paper III extinction curve is defined by
459: \begin{eqnarray}
460: \label{eqnFMFUNC}
461: k(\lambda-V) =
462: \left\{
463: \begin{array}{ll}
464: c_1 + c_2 x + c_3 D(x,x_0,\gamma) & x \leq c_5 \\
465: c_1 + c_2 x + c_3 D(x,x_0,\gamma) + c_4 (x-c_5)^2 & x > c_5 \;\; ,
466: \end{array}
467: \right.
468: \end{eqnarray}
469: where $x \equiv \lambda^{-1}$, in units of inverse microns
470: (\invmic). There are seven free parameters in the formula which
471: correspond to three features in the curve: (1) a linear component
472: underlying the entire UV wavelength range, defined by $c_1$ and $c_2$; (2)
473: a Lorentzian-like 2175 \AA\/ bump, defined by $c_3$, $x_0$, and $\gamma$
474: and expressed as
475: \begin{equation}
476: \label{eqnDRUDE}
477: D(x,x_0,\gamma) = \frac{x^2}{(x^2-x_0^2)^2 +x^2\gamma^2} \;\; ;
478: \end{equation}
479: and (3) a far-UV curvature component (i.e., the departure in the far-UV
480: from the extrapolated bump-plus-linear components), defined by $c_4$
481: and $c_5$. All seven free parameters can be determined by the least
482: squares minimization algorithm.
483:
484: The modification made here to the Paper III formula is in the far-UV
485: curvature term. In Paper III, the value of $c_5$ was fixed at 5.9 \invmic\/
486: and $c_4$ was the scale factor applied to a pre-defined cubic
487: polynomial. In working with the current dataset, we found that the
488: formulation in Equation (\ref{eqnFMFUNC}) significantly improved the fits to
489: many stars, particularly those with weak far-UV curvature, and degraded
490: the fits in almost no cases. In the modified form, the curvature is
491: functionally simpler --- containing a single quadratic term ---
492: although we have added another free parameter to the extinction
493: curve representation. Because the primary goal of the Paper III formula was
494: (and still is) to provide an analytical expression which reproduces as
495: closely as possible observed extinction curves, the cost of an
496: additional free parameter was deemed worthwhile.
497:
498: Using the UV fit parameters above, additional quantities can be defined
499: which help describe the UV curve properties. Particularly useful ones
500: include 1) $\Delta1250 \equiv c_4(8.0-c_5)^2$, which is the value of
501: the FUV curvature term at 1250 \AA\/ and provides a measure of the
502: strength of the FUV curvature; 2) $A_{bump} \equiv \pi \;
503: c_3/(2\gamma)$, which is the area of the 2175 \AA\/ bump; and 3)
504: $E_{bump} \equiv c_3/\gamma^2$, which is the maximum height of the
505: 2175 \AA\/ bump above the background linear extinction.
506:
507: The cubic spline interpolation which produces the optical-through-IR
508: region of our parametrized extinction curve is produced using the IDL
509: procedure SPLINE. The nine anchor points shown in Figure
510: \ref{figEXTCRV} are specified by five free parameters. The UV points,
511: $U_1$ and $U_2$, are simply the values at 2600 \AA\/ and 2700 \AA\/
512: resulting from the modified Paper III formula and require no new free
513: parameters. They assure that the two separate pieces of the extinction
514: curve will join smoothly, although not formally continuously. The
515: three optical points, $O_1$, $O_2$, and $O_3$ (at 3300 \AA, 4000 \AA,
516: and 5530 \AA, respectively) are each treated as free parameters and are
517: adjusted in the fitting procedure to assure the normalization of the
518: final extinction curve. The IR points, $I_1$, $I_2$, $I_3$, $I_4$, and
519: $I_5$ (at 0.0, 0.25, 0.50, 0.75, and 1.0 \invmic, respectively) are
520: functions of two free parameters, $k_{IR}$ and $R(V)$ as follows:
521: \begin{equation}
522: I_n \equiv k(\lambda - V) = k_{IR} \lambda_n^{-1.84} - R(V) \;\; .
523: \label{eqnIR}
524: \end{equation}
525: This assures that the IR portion of our curve follows the the power-law
526: form usually attributed to IR extinction, with a value for its exponent
527: from Martin \& Whittet (1990). As noted in Paper IV, the value of the
528: power law exponent could potentially be determined from an analysis
529: like ours. However, we have found that an IR dataset consisting only
530: of {\it JHK} magnitudes, as in our survey, is insufficient to specify
531: three IR parameters.
532:
533: Note that we adopt the cubic spline formulation for the optical/IR
534: extinction curve simply because we do not have an acceptable analytical
535: expression for the curve shape over this range. The spline approach is
536: very flexible in that the number of anchor points can be modified
537: depending on the datasets available. In the current application, we
538: use only three optical anchor points because we have only three optical
539: data points ($V$, $B-V$, and $U-B$). This approach will bear its best
540: fruit when it can be applied to spectrophotometric data in the near-IR
541: through near-UV region. Then, a large number of anchor points can be
542: used to precisely measure the heretofore poorly-determined shape of
543: extinction in this region, without bias towards a particular analytical
544: expression.
545:
546: In summary, the analysis performed here --- modeling the SEDs of 328
547: reddened stars via Equation (\ref{eqnFLUX}) --- involves determining the
548: best-fit values for as many as 18 free parameters per star, via a
549: non-linear least squares analysis. These include up to 4 parameters to
550: define the theoretical stellar atmosphere model (\teff, \logg, \abund,
551: \vturb), up to 12 parameters to describe the extinction curve shape
552: ($O_1$, $O_2$, $O_3$, $R(V)$, $k_{IR}$, $c_1$ through $c_5$, $x_0$,
553: $\gamma$), the angular radius $\theta_R$, and \ebv. We weight the UV,
554: optical, and IR datasets equally in the fitting procedure.
555:
556: \subsection{Error Analysis
557: \label{secERRORS}}
558:
559: One of the main benefits of the extinction-without-standards technique
560: is the error analysis, which provides a well-quantified estimate of the
561: uncertainties in the best-fit model parameters and allows possible
562: correlations between parameter errors to be explored. This latter
563: benefit is important for assessing the reality of apparent correlations
564: between parameters. The uncertainties in the best-fit parameters
565: derived for our survey stars were determined by running 100 Monte Carlo
566: simulations for each star. In each simulation, the fitting procedure
567: was applied to an input SED consisting of the final best-fit model
568: convolved with a random realization of the observational errors
569: expected to affect the actual data. The adopted 1-$\sigma$
570: uncertainties for each parameter, which will be presented in \S
571: \ref{secATLAS}, were taken as the standard deviations of the values
572: produced by the 100 simulations.
573:
574: Our observational error model for the \iue\/ data consists of random
575: photometric uncertainties and camera zero-point errors as described by
576: Fitzpatrick \& Massa (2005b). The assumed observational errors in the
577: Johnson $B-V$ and $U-B$ indices were as given in Table 7 of that
578: paper. The $V$ magnitudes were assumed to have a 1-$\sigma$
579: uncertainty of 0.015 mag. The uncertainties in the \tmass\/ $JHK$ data
580: were as obtained from the \tmass\/ archive. Johnson $JHK$ magnitudes
581: were assumed to have uncertainties of $\pm$0.03 mag. Random
582: realizations of each of these observational errors, which were added to
583: the best-fit model SED for each Monte Carlo simulation, were determined
584: using the IDL procedure RANDOMN, which produces a normally distributed
585: random variable.
586:
587: In cases where assumptions were made about the values of specific fit
588: parameters, we incorporated uncertainties in the assumptions in the
589: error analysis. In particular: (1) for the O-type stars, the adopted
590: spectral type-dependent temperatures were taken to have 1-$\sigma$
591: uncertainties of $\sim$1000 K; (2) for cluster stars, the adopted
592: distances (used to constrain the surface gravities) were assumed to
593: have 1-$\sigma$ uncertainties of $\pm$20\%; and 3) for the field stars
594: whose values of $\log g$ were taken to be the sample mean of 3.9, this
595: mean was assumed to have a 1-$\sigma$ uncertainty of $\pm$0.2. The
596: values of $\teff$, distance, and $\log g$ used in the the Monte Carlo
597: calculations for the relevant stars were varied randomly in the
598: simulations (using RANDOMN) in accord with these uncertainties.
599:
600: This study certainly does not constitute the first attempt to quantify
601: the uncertainties in interstellar extinction curves. Most pair method
602: studies (see, for example, Cardelli, Sembach, \& Mathis 1992) have
603: incorporated some form of error analysis, often based on the
604: methodologies presented by Massa, Savage, \& Fitzpatrick (1983) and
605: Massa \& Fitzpatrick (1986). However, as long as we restrict our
606: sample to stars which are well represented by the model atmospheres we
607: employ, the advantages of the current technique are great. Because the
608: stellar parameters (temperature, surface gravity, and abundance) are
609: given by continuous mathematical variables (instead of a non-uniformly
610: sampled, discrete sets of standard stars), we are able to perform a
611: well-defined Monte Carlo analysis. The results of this analysis
612: explicitly quantify the uncertainties in all of the input data and
613: assumptions and, thus, the final error bars affecting the derived
614: curves. Moreover, since many realizations of the individual curves are
615: produced, the full shape of the ``error ellipses'' (which describe
616: correlations between the errors) are determined for each specific set
617: of input parameters. Additional discussion of the
618: extinction-without-standards error analysis can be found in Paper IV of
619: this series, along with a demonstration of the quantitative accuracy of
620: the results.
621:
622: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
623: % THE ATLAS
624: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
625:
626: \section{AN ATLAS OF GALACTIC EXTINCTION CURVES\label{secATLAS}}
627:
628: The results of the extinction-without-standards analysis of 328
629: Galactic stars are presented in Table \ref{tabPARMS}, Table
630: \ref{tabEXTINCT}, and Figure \ref{figATLAS}. The 18 free parameters
631: determined by the fitting procedure are divided between Tables
632: \ref{tabPARMS} and \ref{tabEXTINCT}, with the latter containing the 12
633: parameters which define the shape of the normalized interstellar
634: extinction curves $k(\lambda-V) \equiv E(\lambda-V)/E(B-V)$. For both
635: tables, only the first 10 entries are shown here. The full versions of
636: the tables can be viewed in the electronic edition of the Journal. The
637: uncertainties listed in the tables are the 1-$\sigma$ errors derived
638: from the Monte Carlo analysis described in \S \ref{secERRORS}.
639:
640: Figure \ref{figATLAS} shows the normalized extinction curves for the
641: survey sample. The figure consists of 33 panels, each (except the
642: last) containing 10 extinction curves arbitrarily shifted vertically
643: for clarity. Only the first panel is shown here. The full figure is
644: given in the electronic version of this paper. The solid curves in
645: the figure show the parametrized UV-through-IR curves whose shapes were
646: determined by the fitting procedure described in \S \ref{secMETHOD}
647: (the parameters describing the curves are in Table \ref{tabEXTINCT}).
648: An estimate of the shape of the average Galactic extinction curve
649: (corresponding to $R(V)=3.1$; from Fitzpatrick 1999) is shown for
650: reference by the dash-dot curves. The 1-$\sigma$ uncertainties of the
651: survey extinction curves are indicated in Figure~\ref{figATLAS} by the
652: grey shaded regions, which are based on the Monte Carlo error
653: simulations. Their thicknesses indicate the standard deviations of the
654: ensemble of simulations at each wavelength. The actual normalized
655: ratios between the observed stellar SEDs and the atmosphere models are
656: shown by the symbols. Large filled circles indicate {\it JHK} data in
657: the IR region ($\lambda^{-1} < 1 \; \invmic$) and {\it UBV} data in the
658: optical region ($1.5 < \lambda^{-1} < 3.0 \; \invmic$). In the UV
659: region (($\lambda^{-1} > 3.3 \; \invmic$), the small symbols with
660: 1-$\sigma$ error bars show the ratios between the \iue\/ data and the
661: models.
662:
663: A close examination of the curves in Figure \ref{figATLAS} shows that
664: the parametrized curves are extremely good representations of the
665: observed extinction ratios and thus serve as useful proxies for the
666: actual curves themselves. This is particularly apparent in the UV,
667: where the spectrophotometric data show the flexibility of the
668: parametrization scheme. For those wishing to use these curves in
669: extinction studies, we have prepared a tar file containing the
670: parametrized curves for all 328 stars, sampled at 0.087 \invmic\/
671: intervals, and their accompanying fit parameters. Directions for
672: retrieving the file are given in the Appendix.
673:
674: In using or interpreting these curves it is important to recognize that
675: their shapes in the regions between the IR and optical and between the
676: UV and optical are interpolations only and not strongly constrained by
677: data. Additional observations, particularly fully-calibrated
678: spectrophotometric data, would be very useful to constrain the shape of
679: the extinction curves in these regions --- and in the optical where
680: only broadband {\it UBV} measurements are currently employed. It is
681: certainly counter-intuitive that the spectral regions where the
682: detailed shapes of the extinction curves are most poorly-determined are
683: ground-accessible, while the UV data are so well-measured.
684:
685: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
686: % STAR DISCUSSION
687: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
688:
689: \section{Properties of the Sample Stars\label{secSTARS}}
690:
691: Although the main goal of this paper is to explore Galactic extinction,
692: it is nevertheless reasonable to consider briefly the stellar
693: properties revealed by our analysis since they directly affect the
694: extinction results. As discussed above, our reliance on broadband
695: photometry for the current work results in stellar parameters that are
696: not as accurate as those presented in Paper IV, due to the lack of a
697: good surface gravity discriminator. Nevertheless, more than 50\% of
698: the sample stars reside in open clusters and associations and, for
699: these stars, the accuracy will be increased, due the use of ancillary
700: information.
701:
702: Of the four stellar properties determined in the analysis (\teff,
703: \logg, \abund, \vturb), the most significant to our extinction program
704: is \teff\/ because it has the most impact on the shapes of the model
705: atmosphere SEDs used to derive the extinction curves. Figure
706: \ref{figSPTEFF} shows a plot of the derived \teff\/ values against
707: spectral type for the spectral class B0 and later stars (filled and
708: open circles). For comparison, we show several spectral type vs.
709: \teff\/ calibrations from the literature (solid and dashed lines) and
710: data from our photometric calibration study (Fitzpatrick \& Massa
711: 2005b, open squares) in which we modeled the SEDs of 45 unreddened
712: B-type stars. While the survey star data are generally consistent with
713: the comparison data in the figure, considerable scatter is present at
714: some spectral classes --- most particularly at types B2 and B3 --- and
715: there is a general departure between our results and the \teff\/
716: calibrations in the neighborhood of types B1 and B2, in the sense that
717: our results indicate hotter temperatures than the calibrations.
718:
719: We examined the \iue\/ spectra of a number of the survey stars ---
720: those whose temperatures are most discrepant with their claimed
721: spectral types --- and compared them with spectral classification
722: standards to see if somehow our fitting procedure was arriving at
723: grossly incorrect temperatures. An example of such a comparison is
724: shown in Figure \ref{figCOMPARE}. We plot a portion of the \iue\/
725: spectra of the survey stars HD228969 and BD+45 973, the two hottest
726: stars in the ``B2'' and ``B3'' spectral bins, respectively, along with
727: several spectral standards with expected temperatures in the
728: neighborhood of 30000 K. The close match between the survey stars and
729: the hot standards is evident and it is clear that the cool spectral
730: types found in the literature for these stars are unreliable. This is
731: the general conclusion from all such comparisons we have performed.
732: The temperatures found from the fitting procedure are consistent with
733: those expected from a close examination of the UV spectral features of
734: the stars. We conclude that the outliers in Figure \ref{figSPTEFF}
735: result from poor optical spectral types. This is not surprising, since
736: the available types are from a large number of sources and based on a
737: wide variety of observational material of very non-uniform quality.
738:
739: The general discrepancy between our results and the calibrations in the
740: B1-B2 region is a different matter. We derive considerably hotter
741: effective temperatures for the B1 (25000 -- 26000 K) and B2 (23000 --
742: 24000 K) stars than expected from previous calibrations. However,
743: inspection of the UV features in the \atlas\/ models make it difficult
744: to believe that typical B1 and B2 stars are as cool as the spectral
745: type -- \teff\/ calibrations suggest. We must bear in mind, however,
746: that our sample is strongly biased toward cluster stars, which may be
747: considerably younger and more compact than the ``field'' B stars used
748: in the calibrations. Furthermore, the current B star calibrations are
749: all over 20 years old, and it is quite possible that they are in need
750: of revision.
751:
752: Another way to look at the \teff\/ values is shown in Figure
753: \ref{figCBRACKET} where we plot \teff\/ as a function of the
754: Str\"{o}mgren reddening-free index $[c] \equiv c_1 - 0.20(b-y)$, which
755: is a measure of the strength of the Balmer jump. Str\"{o}mgren
756: photometry is not used in this program but is available for 162 of our
757: stars. The symbols in Figure \ref{figCBRACKET} are the same as in
758: Figure \ref{figSPTEFF}, with the addition of the open circles which
759: denote O stars. The figure demonstrates the essentially exact overlap
760: between the current results and those for the unreddened, mid-to-late B
761: stars from Fitzpatrick \& Massa (2005b) as well as the smooth
762: transition from the early B stars into the O stars. There is no
763: indication of any systematic effects present in the results for the
764: early B stars. On the contrary, the spectral type vs. \teff\/
765: relations, transformed into the \teff\/ vs. $[c]$ plane as described in
766: the figure legend, show a number of abrupt and physically unrealistic
767: changes in slope suggestive of inadequacies in the calibrations.
768:
769: We conclude that our derived effective temperatures are reasonable. In
770: most cases where a temperature strongly disagrees with a published MK
771: type, it agrees quite well with the UV type determined by Valencic,
772: Clayton, \& Gordon (2004), indicating that the MK type is of poor
773: quality or else influenced by something else (e.g., the presence of a
774: cooler companion which is invisible in the UV). We also suspect that
775: the spectral type -- \teff\/ calibration for the B1 and B2 stars may
776: need to be revised. Finally, we are gratified by the overall
777: consistency between the current results and those of Paper~IV and
778: Fitzpatrick \& Massa (2005b), where the stellar parameters were more
779: strongly constrained, and with the smooth relation between \teff\/ and
780: [c], which suggests that no strong systematic effects are present.
781:
782: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
783: % EXTINCTION DISCUSSION
784: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
785:
786: \section{Properties of Galactic Interstellar Extinction\label{secEXTINCTION}}
787:
788: \subsection{General Properties\label{secEXT_PROPERTIES}}
789:
790: Although the individual extinction curves for all of the survey stars
791: are displayed in the 33 panels of Figure \ref {figATLAS}, it is
792: nonetheless difficult --- from that figure --- to visualize the range
793: of extinction properties present in the sample. To provide such a
794: view, we plot the analytical fits for the full set of 328 survey curves
795: in Figure \ref{figALLCURVES}. (These curves can be reproduced from the
796: parameters listed in Table \ref{tabEXTINCT}. The curves have been
797: plotted using small dots in those spectral ranges where they are
798: interpolated or extrapolated. The solid portions correspond to the
799: regions constrained by near-IR {\it JHK} photometry, optical {\it UBV}
800: photometry, and UV \iue\/ spectrophotometry. The top panel, in which
801: the curves are plotted in their native form, $E(\lambda-V)/E(B-V)$,
802: shows the wide range of variation observed in Galactic extinction
803: curves, although a clear core of much more restricted variation is
804: evident in the distribution. The convergence of the curves in the
805: range $1.8 < \lambda^{-1} < 2.5~\invmic$ is a product of the
806: normalization and obscures our view of variations in the optical
807: region. The bottom panel of Figure \ref{figALLCURVES} presents the
808: same curves, but normalized to the total extinction at 1~\invmic. The
809: exact overlap of the curves at $\lambda^{-1} < 1~\invmic$ arises
810: because we have adopted a power law form with a fixed exponent of
811: $-1.84$ for all curves in this spectral region (see Eq. [\ref{eqnIR}]).
812: It is unlikely that the exponent is actually so constant --- Larson \&
813: Whittet (2005) have found evidence for a more negative value in a
814: sample of high-latitude clouds --- however, as noted earlier in \S
815: \ref{secMETHOD}, our near-IR dataset consisting of only {\it JHK}
816: measurements is insufficient for independently evaluating the
817: exponent. This IR normalization reveals that the extinction in the
818: optical spectral region can range from very-nearly grey to very
819: strongly wavelength-dependent.
820:
821: One of the goals of many Galactic extinction studies is to derive an
822: estimate of the typical or average wavelength-dependence of
823: interstellar extinction. Such a mean curve is often used as the
824: standard of normalcy against which particular sightlines are judged, or
825: for comparison with results for external galaxies, or for
826: ``dereddening'' the SEDs of objects for which there is no specific
827: extinction knowledge. While constructing an average curve is a
828: straightforward process, the degree to which the result represents
829: ``normal'' or even ``typical'' extinction is problematic. We will take
830: up the issue of an average Galactic curve later in \S
831: \ref{secDISCUSS}. Here we present an average curve for our sample, to
832: help characterize the general extinction properties of our survey
833: sightlines.
834:
835: The bottom panel in Figure \ref{figSTARSTATS} shows a clear peak in the
836: distribution of $R(V)$ values of our sample. A Gaussian fit to the
837: region of this peak, shown as the smooth curve in the figure, has a
838: centroid at $R(V) = 2.99$ and a width given by $\sigma = 0.27$. The
839: peak is within the range of values considered as average for the
840: diffuse ISM (see, e.g., Savage \& Mathis 1979). Thus, as noted in \S
841: \ref{secDATA}, our sample is dominated by sightlines whose $R(V)$
842: values are consistent with the diffuse phase of the ISM (although
843: composite sightlines are undoubtedly present). We have constructed an
844: average curve to represent the properties of these diffuse sightlines
845: by taking the simple mean extinction value at each wavelength using all
846: the curves with $2.4 < R(V) < 3.6$ (i.e., the ~2-$\sigma$ range of the
847: Gaussian fit in Figure \ref{figSTARSTATS}; 243 sightlines in total).
848: This mean extinction curve is shown in Figure \ref{figAVGCURVE} by the
849: thick solid curve. The dark grey shaded region shows the variance of
850: the 243-curve sample. The set of 12 parameters describing this curve
851: are listed in Table \ref{tabAVGCURVE}. Since so many of our sightlines
852: are included in the mean, removing the restriction on $R(V)$ has little
853: effect. If we had included all 298 sightlines for which $R(V)$ has
854: been derived, then the mean curve would differ from that shown only by
855: being several tenths lower in the UV region. The variance of the full
856: sample is larger, and this is shown by the lightly shaded region. For
857: comparison, several other estimates of average Galactic curves are
858: shown in Figure \ref{figAVGCURVE}. The curves from Cardelli et al.
859: (1989), Fitzpatrick (1999), Seaton (1979), and Valencic et al. (2004) are all intended to represent the diffuse ISM mean. The
860: results from Savage et al. (1985) are mean values for the 800
861: sightlines in that study with $E(B-V) \ge 0.20$ mag (matching our
862: survey cutoff). No restriction on $R(V)$ was imposed. The error bars for
863: the Savage et al. (1985) data are sample variances; they are generally
864: similar to the variances of our full sample (lightly shaded region).
865: The much larger value for the 1500 \AA\/ point is likely due to
866: spectral mismatch in the potentially strong C IV stellar wind lines.
867:
868: The differences among the various mean curves in Figure
869: \ref{figAVGCURVE} are instructive. The great intrinsic variety of
870: Galactic extinction curves as seen, for example, in Figure
871: \ref{figALLCURVES} shows that any mean curve is subject to the biases
872: in the sample from which it was produced. It is probably impossible to
873: construct a sample of sightlines whose properties could be claimed to
874: provide a fair representation of all the types of conditions found in
875: the ISM. Thus, there is likely no unique or best estimate of mean
876: Galactic extinction. Any of the mean curves in Figure
877: \ref{figAVGCURVE} would serve as reasonable representations of Galactic
878: extinction. In any situation where an average curve is adopted,
879: however, it is important to recognize the intrinsic variance of the
880: underlying sample and incorporate the uncertainties of the average
881: curve in any error analysis. Because it is derived from such a large
882: sample and is largely free of contributions from spectral mismatch, the
883: sample variance for our diffuse curves (shown in Figure
884: \ref{figAVGCURVE} by dark shaded region) would provide a reasonable
885: estimate of the uncertainty in any version of a mean Galactic diffuse
886: ISM curve. We have included our diffuse ISM mean curve from Figure
887: \ref{figAVGCURVE} --- and its accompanying uncertainty --- in the tar
888: file discussed in \S \ref{secATLAS} (see the Appendix).
889:
890: We can also use our large sample to investigate the ``smoothness'' of
891: UV extinction. In Figure \ref{figIUECURVE}, we plot the simple mean of
892: the actual extinction ratios for our survey sample in the spectral
893: region covered by the \iue\/ data (upper curve, open circles).
894: Overplotted is a parameterized fit to these data, using the extinction
895: formulation given in \S \ref{secMETHOD} (solid curve). This figure
896: illustrates two points; namely, 1) the lack of small-scale structure in
897: UV extinction and 2) the degree to which the modified Paper III UV
898: parametrization scheme reproduces the shape of UV extinction. A
899: detailed discussion of the former point, and an indication of the kinds
900: of features that might be expected in the UV, can be found in Clayton
901: et al. (2003). In short, polycyclic aromatic hydrocarbon (PAH)
902: molecules, which have been suggested as the source of mid-IR emission
903: features, might produce noticeable absorption in the UV, and possibly
904: even contribute to the 2175 \AA\/ bump absorption. Earlier studies
905: have always failed to find structure in UV extinction curves and
906: Clayton et al. were able to place very stringent 3-$\sigma$ upper
907: limits of $\sim0.02A_V$ on any possible 20 \AA-wide features in the
908: extinction curves towards two heavily reddened B stars. The data at
909: the bottom of Figure \ref{figIUECURVE} show the differences between the
910: mean extinction curve and the best-fit model. A small number of points
911: do rise above (or below) the general noise level of the residuals, but
912: these points --- which are labeled in the figure --- are due to
913: interstellar gas absorption lines (which are sometimes strong in the
914: observations and always non-existent in the model atmosphere SEDs),
915: mismatch between the C IV $\lambda$1550 stellar wind lines in the O
916: stars and the static model SEDs, or known inadequacies in the \atlas\/
917: opacity distribution functions (labeled ``b'' in the figure; see Paper
918: IV). Excluding these points, the standard deviation of the mean curve
919: around the best-fit is 0.06\ebv\/ mag, corresponding to $\sim0.02A_V$
920: mag, for the 10 \AA-wide spectral bins. This is not quite so
921: restrictive as the result of Clayton et al., but once again affirms the
922: smoothness of UV extinction.
923:
924: \subsection{Spatial Trends \label{secEXT_SPATIAL}}
925:
926: Studying spatial trends in extinction can serve two purposes. First,
927: identifying strong regional variations can allow a better estimate of
928: the shape of the extinction curve affecting a particular sightline
929: permitting, for example, a more accurate determination of the intrinsic
930: SED shape of an exotic object. This has been the motivation for a
931: number of studies, such as by Massa \& Savage (1981) and Torres (1987)
932: who used extinction curves derived for B stars in open clusters (NGC
933: 2244 and NGC 6530, respectively) to determine the SEDs of cluster O
934: stars. The second purpose is to gain insight into the nature of dust
935: grains and the processes which modify them by observing how extinction
936: curves respond to various environmental properties, such as density or
937: radiation field.
938:
939: In Figure \ref{figLONGITUDE} we show a sweeping view of the regional
940: trends in our data by plotting several extinction curves properties
941: against Galactic longitude for each of our survey sightlines.
942: Sightlines towards stars in the clusters or associations highlighted in
943: Figure \ref{figCOORDS} are indicated by the larger symbols in Figure
944: \ref{figLONGITUDE} (the key is in figure caption) and the rest of the
945: sample by the small filled circles. The dashed lines in the figure
946: panels show the parameter values that correspond to the diffuse mean
947: curve in Figure \ref{figAVGCURVE}.
948:
949: A number of regions stand out in Figure \ref{figLONGITUDE}. For
950: example, NGC 1977 near $l = 210\degr$ (open triangles; includes the Orion
951: Trapezium region) is well-known for its high $R(V)$ sightlines. Figure
952: \ref{figLONGITUDE} shows that this region is also notable for its flat
953: UV extinction curves (i.e., small $c_2$) and weak bumps (i.e., small
954: $A_{bump}$). Interestingly, the far-UV curvature for this region
955: appears typical. An early discussion of UV extinction curves towards
956: NGC 1977 can be found in Panek (1983). The relationship between $R(V)$
957: and other curve properties will be the subject of \S
958: \ref{secEXT_RDEPEND} below.
959:
960: The Carina direction (large filled circles), particularly towards Tr 14
961: and Tr 16, also shows elevated $R(V)$ values. This is another
962: weak-bumped direction which also shows lower than average FUV
963: curvature. Optical and IR extinction studies have been performed for
964: Carina sightlines (e.g., Tapia et al. 1988 and references within), but
965: we are not aware of correspondingly detailed UV extinction studies.
966:
967: Large $R(V)$ values are also seen along a number of sightlines in the
968: general direction of the Galactic center. This includes sightlines to
969: the cluster NGC 6530 (open squares) and towards the $\rho$ Oph dark
970: cloud (small filled circles near $l \simeq 253 \degr$. These
971: sightlines also feature low UV extinction and, in the case of NGC 6530,
972: weaker than average 2175 \AA\/ bumps. Large $R(V)$ are well known in
973: the Ophiuchus region (Chini \& Krugel 1983) and the UV extinction has
974: been examined by Wu, Gilra, \& van Duinen (1980). UV extinction
975: towards NGC 6530 was studied by Torres (1987).
976:
977: Finally we note a broad region from $l \simeq 50 \degr$ to $l \simeq 150
978: \degr$ where the $R(V)$ values are systematically slightly below the mean
979: value. No trends are obvious in the UV parameters. These sightlines
980: are in the direction of the Perseus spiral arm (e.g., Georgelin \&
981: Georgelin 1976) and sample dust in the interarm region and, possibly,
982: the Perseus Arm itself --- for those stars more distant than $\sim$2
983: kpc, such as in h \& $\chi$ Per (x's in Figure \ref{figLONGITUDE}).
984: Extinction towards individual regions in this zone have been studied
985: (e.g., Tr 37 by Clayton \& Fitzpatrick 1987, Cep OB3 by Massa \& Savage
986: 1984, h \& $\chi$ Per by Morgan, McLachlan, \& Nandy 1982, and Cr 457
987: by Rosenzweig \& Morrison 1986), but we are not aware of any
988: comprehensive investigation of the general region.
989:
990: Morgan et al. (1982) found a dependence of UV extinction on Galactic
991: latitude, $b$, for sightlines to stars in h \& $\chi$ Per, in the sense
992: that the extinction at 1550 \AA\/ increases with increasingly negative
993: values of $b$. Our data suggest a similar effect for the UV linear
994: slope $c_2$ (which is closely related to the extinction level at 1550
995: \AA), but not for any other extinction parameter, including $R(V)$. We
996: have examined our dataset for general trends with Galactic latitude, or
997: with distance above and below the plane, and have found none. This is
998: not surprising, however, since our sample is dominated by low-latitude
999: sightlines and we have little leverage for a latitude-dependence
1000: study. Our lower cutoff of 0.20 mag in \ebv\/ eliminated most
1001: high-latitude sightlines from consideration. Previous studies have
1002: shown that it is difficult to uncover latitude dependences in
1003: extinction (e.g., Kiszkurno-Koziej \& Lequeux 1987). Local trends
1004: might be uncovered by examining small zones in Galactic longitude (such
1005: as in the study by Morgan et al.), although such detailed
1006: investigations are beyond the scope of this paper. Likewise, studies
1007: of extinction variations over small spatial scales, such as among
1008: sightlines to cluster members are beyond our scope, but might well
1009: provide important information linking grain populations with other ISM
1010: diagnostics.
1011:
1012: The quantity $\ebv/d$, which can be computed from the data in Tables
1013: \ref{tabSTARS} and \ref{tabPARMS}, provides a crude measure of one
1014: important physical property of the ISM, namely, dust density. While
1015: the shortcomings of this measure as a direct proxy for density are
1016: clear --- for example, a high density dust cloud along a long,
1017: otherwise vacant sightline will yield a misleadingly low value of
1018: $\ebv/d$ --- it is nevertheless useful as a first-look and as a guide for
1019: future studies. Figure \ref{figDENSITY} shows plots of four extinction
1020: curve parameters against $\ebv/d$. The symbols are the same as for
1021: Figure \ref{figLONGITUDE}. The three UV parameters all show evidence
1022: for a weak trend with density, in the sense of flatter slopes, broader
1023: bumps, and increasing FUV curvature with increasing density. Hints of
1024: these three effects were seen in the first two papers in this series
1025: (Fitzpatrick \& Massa 1986, 1988; hereafter Papers I and II). We see
1026: no evidence for a trend with bump strength $A_{bump}$ (not shown in
1027: Figure \ref{figDENSITY}) and the dependence of $R(V)$ on $\ebv/d$ is
1028: complex and difficult to characterize, although the highest density
1029: sightlines all have larger-than-average values of $R(V)$. While not
1030: conclusive, the results in Figure \ref{figDENSITY} certainly suggest
1031: that comparisons of our survey data with detailed measures of ISM
1032: physical conditions could yield interesting results. We note that
1033: Rachford et al. (2002) have found positive trends of bump width and
1034: far-UV curvature with the fraction of hydrogen atoms in the form
1035: H$_2$. This is consistent with the results in Figure \ref{figDENSITY}
1036: in the sense that one would expect higher H$_2$ fractions in denser,
1037: and therefore better-shielded, regions. Other studies of the
1038: density-dependence of extinction have been performed by Massa (1987)
1039: and Clayton, Gordon, \& Wolff (2000).
1040:
1041:
1042: \subsection{Relationships Among the Fit Parameters\label{secEXT_RELATIONS}}
1043:
1044: With twelve parameters to describe the UV-through-IR extinction curves
1045: of each of the 328 survey stars, there are many possible correlations
1046: and relationships to investigate. We have looked at all of these
1047: possibilities and, as an interested reader can verify from the data in
1048: Table \ref{tabEXTINCT}, virtually all the parameters are remarkably
1049: UNcorrelated with each other! In this section, we consider only the
1050: two most striking relationships between extinction parameters: $c_1$
1051: vs. $c_2$ (the UV linear intercept and slope) and $R(V)$ vs. $k_{IR}$
1052: (the ratio of selective-to-total extinction and the IR scale
1053: factor). We also examine the most important non-relationship between
1054: parameters: $x_0$ vs. $\gamma$ (the centroid and width of the 2175
1055: \AA\/ bump). Below, in \S \ref{secEXT_RDEPEND}, we will consider the
1056: possible relationship between IR curve features and UV features.
1057: \subsubsection{$c_1$ vs. $c_2$}
1058:
1059: In Papers I and II we
1060: showed that pair method extinction curves in the region of the 2175 \AA\/
1061: extinction bump could be modeled very precisely using a Lorentzian-like
1062: ``Drude profile'' (see Eq. [\ref{eqnDRUDE}]) combined with a linear
1063: background extinction (defined by the parameters $c_1$ and $c_2$ in
1064: Eq. [\ref{eqnFMFUNC}]). Further, it was shown that the linear
1065: parameters for the 45 sightlines in the study appeared to be very well
1066: correlated, and could likely be replaced with a single parameter
1067: without loss of accuracy. In the pilot study for this paper, F04
1068: showed that this correlation was maintained in a larger sample of 96
1069: curves derived using the extinction-without-standards technique.
1070:
1071: Figure \ref{figC1C2} shows a plot of the linear parameters $c_1$ vs.
1072: $c_2$ for our survey sample of 328 extinction-without-standards
1073: curves. The dotted error bars show the orientation of the 1-$\sigma$
1074: error ellipses of the measurements. These were determined by the
1075: distribution of results from the Monte Carlo simulations. The obvious
1076: correlation between the errors in $c_1$ and $c_2$ is not unanticipated
1077: (see Figure 4 in Paper II), but the ability to
1078: explicitly determine such errors is a major advantage of the
1079: extinction-without-standards approach and is critical for evaluating
1080: the significance of apparent correlations.
1081:
1082: The solid line in Figure \ref{figC1C2} corresponds to the linear relation
1083: \begin{equation}
1084: \label{eqnC1C2}
1085: c_1 = 2.09 - 2.84 \; c_2 \;\; ,
1086: \end{equation}
1087: which is a weighted fit that minimizes the scatter in the direction
1088: perpendicular to the fit. (Because there is uncertainty in both $c_1$
1089: and $c_2$, a normal least squares fit is not appropriate.) The fact
1090: that this relationship is nearly parallel to the long axis of the
1091: correlated error bars explains why the relationship between $c_1$ and
1092: $c_2$ remains so clear, even in the presence of observational error.
1093:
1094: To determine whether the observed scatter about the mean $c_1$ vs $c_2$
1095: relationship is caused only by observational error or is at least
1096: partially the result of ``cosmic'' scatter, we must examine the
1097: residuals to the best-fit relationship. The distribution of residuals
1098: perpendicular to the best-fit line is complex, consisting of a Gaussian
1099: core of values with $\sigma \simeq 0.07$ and a more extended
1100: distribution of outliers reaching out to values of about $\pm0.3$. About
1101: 86\% of the points fall within the 2-$\sigma$ range
1102: of the Gaussian core. The RMS value of the expected observational
1103: errors perpendicular to the best-fit relation is 0.057. The Gaussian
1104: core of the observed residuals is thus only slightly broader than that
1105: expected from observational scatter alone. We conclude that $c_1$ and
1106: $c_2$ are indeed intrinsically well-correlated quantities, with a
1107: cosmic scatter comparable to our measurements errors. However a
1108: significant fraction of the sample ($\sim$10\%) show evidence for a
1109: wider deviation from the mean relationship. In most instances,
1110: Equation (\ref{eqnFMFUNC}) could be simplified without loss of accuracy by
1111: replacing the two linear parameters $c_1$ and $c_2$ with a single
1112: parameter.
1113:
1114: \subsubsection{$R(V)$ vs. $k_{IR}$}
1115:
1116: F04 showed that the two parameters describing the IR portion of the
1117: extinction-without-standards curves, $k_{IR}$ and $R(V)$ are apparently
1118: well-correlated and that the IR curve might be defined by a single
1119: parameter. Figure \ref{figRKIR} shows a plot of $k_{IR}$ vs. $R(V)$
1120: for the survey sample, along with the 1-$\sigma$ error bars. The solid
1121: curve shows the best-fit weighted linear relationship
1122: \begin{equation}
1123: \label{eqnKRV}
1124: k_{IR} = -0.83 + 0.63 \; R(V) \;\; ,
1125: \end{equation}
1126: which minimizes the scatter in the direction perpendicular to the
1127: relation. As in the discussion above, we see that the errors are
1128: strongly correlated and the near coincidence of the long axis of the
1129: error ellipses with the direction of the best-fit relationship would
1130: preserve the appearance of a correlation even in the face of
1131: significant observational error. The residuals in the direction
1132: perpendicular to the best-fit line are distributed in a Gaussian form,
1133: with $\sigma = 0.11$. The RMS value of the observational errors in
1134: this same direction is $\pm0.12$. Thus the observed distribution of
1135: points in Figure \ref{figRKIR} is consistent with perfectly correlated
1136: quantities and the expected observational error.
1137:
1138: The correlation between $R(V)$ and $k_{IR}$ indicates that the shape of
1139: near-IR extinction at $\lambda >$ $\sim$1 $\mu {\rm m}$, over a wide
1140: range of $R(V)$ values, can be characterized by a single parameter.
1141: I.e., the two parameters in Equation (\ref{eqnIR}) are redundant and
1142: could be replaced --- without loss of accuracy --- by a single
1143: parameter, e.g., $R(V$), based on Equation (\ref{eqnKRV}):
1144: \begin{equation}
1145: \label{eqnIREXT}
1146: \frac{E(\lambda-V)}{E(B-V)} = [-0.83+0.63R(V)]\lambda^{-1.84} - R(V) \;\; .
1147: \end{equation}
1148: This is consistent with the results of Martin \& Whittet (1990; see
1149: their Table 2), who utilized IR data out to the $M$ band near 5 $\mu$m.
1150: Our study utilizes only shorter wavelength near-IR bands, but includes
1151: many more sightlines and spans a wider range in $R(V)$ values than
1152: could be studied by Martin and Whittet. Our parameter $k_{IR}$ is
1153: essentially the same as their $e$ parameter, and our results
1154: demonstrate the dependence of $e$ on $R(V)$.
1155:
1156: The significance of the tight correlation between $R(V)$ and
1157: $k_{IR}$ is somewhat difficult to assess. We must keep in mind that,
1158: at one level, this relation simply states that the three IR data points
1159: given by \tmass\/ $JHK$ photometry can be summarized by two parameters
1160: at the level of the errors in the \tmass\/ photometry. On the other
1161: hand, it also bears on the issue of the underlying shape of infrared
1162: extinction and its so-called universality. Our data, which are
1163: dominated by diffuse ISM sightlines, are consistent with the notion of
1164: a universal shape for IR extinction, as given by Equation
1165: (\ref{eqnIREXT}). However, given our restricted IR wavelength coverage
1166: and the typical uncertainties in the \tmass\/ data, our results are
1167: relatively insensitive to departures from universality. For example,
1168: the actual IR power law exponent among our sample stars could vary
1169: significantly around the mean value of $-1.84$ used here, and we would
1170: still find a very strong correlation between $k_{IR}$ and $R(V)$. The
1171: issue of the universality of IR extinction is best left to indepth
1172: studies which utilize more focussed approaches and more appropriate
1173: datasets, such as by Larson \& Whittet (2005) and Nishiyama et al.
1174: (2006), both of which have found a range in values for the IR
1175: exponent.
1176:
1177: Values for $R(V)$ are often estimated from the formula $R(V) = 1.1
1178: E(V-K)/E(B-V)$, which is based on van de Hulst's theoretical extinction
1179: curve No. 15 (e.g., Johnson 1968). In Figure \ref{figEVMINK} we plot
1180: $E(V-K)/E(B-V)$ vs. $R(V)$ for our survey sightlines. The dotted line
1181: shows the van de Hulst relation, which agrees well with the data for
1182: values of $R(V)$ near 3 --- not surprising since it was derived from a
1183: theoretical curve with R = 3.05 --- but systematically deviates at
1184: higher and lower values of $R(V)$. The solid line in the figure shows
1185: the best-fit linear relation which minimizes the residuals
1186: perpendicular to the fit. It is given by
1187: \begin{eqnarray}
1188: \label{eqnEKMINV}
1189: R(V) = -0.26 + 1.19 \; \frac{E(V-K)}{E(B-V)} \;\; .
1190: \end{eqnarray}
1191: This relationship was derived only from the sightlines with \tmass\/
1192: $K$-band measurements (solid circles in Figure \ref{figEVMINK}) but
1193: also agrees with those measurements based on Johnson $K$-band
1194: photometry (open circles). This exact form of this relation depends
1195: slightly on our choice of an IR power law exponent of $-1.84$, and the
1196: small scatter is another indicator that our data are consistent with a
1197: single functional form for IR extinction. The relationship in Equation
1198: (\ref{eqnEKMINV}) can be reproduced by Equation (\ref{eqnIREXT}) with a
1199: wavelength of $\lambda \simeq 2.1 \; \mu$m.
1200:
1201: \subsubsection{$x_0$ vs. $\gamma$\label{secEXT_X0GAMMA}}
1202:
1203: Among the many non-correlations between extinction quantities, one of the
1204: most significant is that between the position of the peak of the 2175
1205: \AA\/ bump (parametrized here by $x_0$) and its FWHM (parametrized by
1206: $\gamma$). The lack of a relationship between these two quantities, as
1207: first reported in Paper I, places strong constraints on the nature of
1208: the dust grains which produce the 2175 \AA\/ feature (see, e.g., Draine
1209: 2003).
1210:
1211: Figure \ref{figGAMMAX0} shows a plot of $\gamma$ vs. $x_0$ for our
1212: survey sightlines, along with their 1-$\sigma$ error bars. As in all
1213: previous studies, the lack of a correlation is clear.
1214:
1215: Figure \ref{figBUMPSTATS} shows the distribution of bump peak positions
1216: (left panel) and widths (right panel) plotted in histogram form (shaded
1217: regions). With the exception of a few outliers, the distribution of
1218: bump peaks can be fitted well with a Gaussian function, as indicated in
1219: Figure \ref{figBUMPSTATS} by the smooth solid curve. The centroid of
1220: the Gaussian is at $x_0 = 4.5903$ \invmic\/ and its width is given by
1221: $\sigma = 0.0191$ \invmic. These correspond to a mean bump position of
1222: 2178.5 \AA\/ with a 1-$\sigma$ range of $\pm9.1$ \AA. The RMS value of
1223: the $x_0$ measurement errors for the full sample is $\pm0.0058$
1224: \invmic\/ (corresponding to $\pm2.8$ \AA).
1225:
1226: While these results suggest small but significant variations in bump
1227: positions --- as reported in Paper I --- the Gaussian-like distribution
1228: of $x_0$ values in Figure \ref{figGAMMAX0} led us to consider that
1229: perhaps our error analysis might underestimate the uncertainties in
1230: $x_0$ and that the width of the Gaussian itself might represent the
1231: true observational error. We examined this issue by considering the
1232: results for sightlines towards stars in open clusters and
1233: associations. We have previously used such sightlines to help estimate
1234: extinction curve measurement uncertainties (Massa \& Fitzpatrick
1235: 1986). If it is assumed that the true extinction curve is identical
1236: for all cluster sightlines (based on the small spatial separation
1237: between the sightlines), then each cluster curve represents an
1238: independent measurement of the same curve and the variations from
1239: sightline-to-sightline give the net measurement errors. Since it is
1240: unlikely that there is no cosmic variability among the sightlines, this
1241: procedure provides upper limits on observational errors. We utilized
1242: data for the 13 clusters with five or more stars in our survey (NGC
1243: 457, Cr 463, NGC 869, NGC 884, NGC 1977, NGC 2244, NGC 3293, Tr 16, NGC
1244: 4755, NGC 6231, NGC 6530, Tr 37, and Cep OB3), yielding a total of 154
1245: sightlines. For each cluster we computed the mean value of $x_0$ and
1246: subtracted it from the individual cluster values. We then examined the
1247: ensemble of residuals for the full cluster sample. The shape of the
1248: residuals distribution is Gaussian-like, although very slightly skewed
1249: towards positive values. A Gaussian fit yields the result shown by the
1250: dotted curve in the $x_0$ panel of Figure \ref{figGAMMAX0} (scaled to
1251: match the height of the main distribution), which has a width given by
1252: $\sigma = 0.011$ \invmic. This is larger than the RMS measurement
1253: error for the cluster sample, i.e., $\pm0.0058$ \invmic\/, but
1254: significantly smaller than the observed width of the full sample. For
1255: a number of the clusters, there are obvious curve-to-curve variations
1256: present, and the width of the residuals distribution must certainly
1257: overestimate the measurement errors. From this more detailed analysis,
1258: our conclusion remains that the small variations of $\pm9.1$ \AA\/ seen
1259: in the full survey sample, are significantly larger than the expected
1260: observational errors and indicate true variations in the position of
1261: the bump peak.
1262:
1263: The distribution of bump widths (right panel of Figure
1264: \ref{figBUMPSTATS}) is decidedly non-Gaussian, with a strong tail in
1265: the direction of large values. The main peak of the distribution,
1266: however, is Gaussian in appearance and a fit to this region (smooth
1267: curve in the figure) yields a centroid of 0.890 \invmic\/ and a width
1268: of $\sigma = 0.050$ \invmic. The RMS value of the observational errors
1269: is $\pm0.031$ \invmic, and so the width of this Gaussian, and the width
1270: of the whole distribution is clearly larger than can be accounted for
1271: by observational errors. Again consistent with earlier results, we
1272: find that the bump widths vary significantly from sightline to
1273: sightline, but with no correlation with the centroid position. The
1274: shape of the $\gamma$ distribution might
1275: suggestion two populations of bumps --- one characterized by the
1276: Gaussian fit and the other characterized by a larger mean centroid
1277: ($\sim$1.1 \invmic) and a wider range of values ($\sigma \simeq 0.1$
1278: \invmic), but this is not the only possible interpretation of the
1279: results. We examined the bump widths for the cluster sample as above.
1280: However, the distribution of cluster residuals is complex, showing the
1281: asymmetry of the full sample, and indicating significant variations
1282: within the clusters, and we were thus not able to confirm the accuracy
1283: of the measurement errors. The range in observed $\gamma$ values is so
1284: large, however, that the evidence for cosmic scatter is unambiguous.
1285:
1286: \subsection{$R(V)$-Dependence\label{secEXT_RDEPEND}}
1287:
1288: Cardelli, Clayton, \& Mathis (1988, 1989) were the first to demonstrate
1289: a link between UV and optical/IR extinction by showing that $R(V)$ is
1290: related to the level of UV extinction. Essentially, sightlines with
1291: large $R(V)$ values tend to have low UV extinction, and vice versa.
1292: Cardelli et al. quantified this relationship in the following way:
1293: \begin{eqnarray}
1294: \label{eqnCCM1}
1295: \frac{A(\lambda)}{A(V)} = a(\lambda) + b(\lambda) \; R(V)^{-1} \;\;\;\; ;
1296: \end{eqnarray}
1297: i.e., the total extinction at wavelength $\lambda$ normalized by the
1298: total extinction at $V$ is a linear function of $R(V)^{-1}$. In the
1299: time since the original work, the perception of this relationship has
1300: evolved to the point where it is often referred to as a ``law'' and
1301: Galactic extinction curves are often stated or assumed to be a
1302: 1-parameter family (with $R(V)^{-1}$ as the parameter). Recently, for
1303: example, Valencic et al. (2004) found that 93\% of a large sample of
1304: Galactic extinction curves obey a modified form of this relation. In
1305: this section, we will show that the relationship in Equation
1306: (\ref{eqnCCM1}) is partially illusory and that Galactic extinction curves
1307: are decidedly not a 1-parameter family in $R(V)^{-1}$.
1308:
1309: The original basis for Equation (\ref{eqnCCM1}) is data such as shown
1310: in Figure \ref{figRVAV}, where we plot $A(\lambda)/A(V)$ vs.
1311: $R(V)^{-1}$ at four different UV wavelengths for our survey sightlines
1312: (see Figure 1 in Cardelli et al. 1989). It is clear why a linear
1313: function would be chosen to quantify the obvious relationships seen in
1314: the figure, and the data give the impression of being reasonably
1315: well-correlated. The solid line in the 2695 \AA\/ panel shows an
1316: example of such a linear relationship. It is a weighted fit which
1317: minimizes the residuals in the direction perpendicular to the fit, and
1318: is given by $A(2695 \; {\rm \AA})/A(V) = 0.58 + 4.73 \; R(V)^{-1}$.
1319: The appearance of Figure \ref{figRVAV} is, however, deceiving. The
1320: normalization used in the y-axis is constructed from the measured
1321: values of $E(\lambda-V)/E(B-V)$ by the transformation \begin{eqnarray}
1322: \label{eqnCCM2} \frac{A(\lambda)}{A(V)} \equiv
1323: \frac{E(\lambda-V)}{E(B-V)} \; R(V)^{-1} + 1\;\;\;\; . \end{eqnarray}
1324: Thus, the four panels in Figure \ref{figRVAV} essentially amount to
1325: plots of $xy$ vs. $x$ and, {\it even if $x$ and $y$ were completely
1326: unrelated}, some degree of apparent correlation would inevitably
1327: appear. In addition, if there actually were an intrinsic relationship
1328: between $x$ and $y$, its significance could be greatly
1329: overinterpreted.
1330:
1331: The true significance of the relationships plotted in Figure
1332: \ref{figRVAV} could be fairly assessed if the measurement errors were
1333: well-determined, including the effects introduced by the transformation
1334: in Equation (\ref{eqnCCM2}). Our Monte Carlo error analysis allows us
1335: to quantify the uncertainties in any combination of fitted or measured
1336: quantities and fully take into account the artificial correlation
1337: induced by Equation (\ref{eqnCCM2}). The principal axes of the
1338: 1-$\sigma$ errors for $A(\lambda)/A(V)$ and $R(V)^{-1}$ are indicated
1339: in Figure \ref{figRVAV} by the dotted error bars and show the strong
1340: correlation produced by the chosen normalization. Ironically, because
1341: of the normalization and the resultant error correlations, the
1342: appearance of correlations in Figure \ref{figRVAV} is actually enhanced
1343: by uncertainties in $R(V)$. When the correlated errors are taken into
1344: account, the linear relation shown in the 2695 \AA\/ panel of Figure
1345: \ref{figRVAV} is found to have a reduced $\chi^2$ value of $\sim$4.2,
1346: indicating that the scatter about the relation is more than four times
1347: greater than accounted for by the known uncertainties in the data.
1348:
1349: Although our error analysis offers us a way to overcome the
1350: complications arising from the normalization chosen by Cardelli et al.,
1351: a more direct approach to evaluating the relationship between UV and IR
1352: extinction is simply to return to the actual quantities determined by
1353: the extinction analysis and look at plots of $E(\lambda-V)/E(B-V)$ vs.
1354: $R(V)$. A small amount of algebra (i.e., combining Equations
1355: [\ref{eqnCCM1}] and [\ref{eqnCCM2}]) will show that, if a linear
1356: relationship as in Equation (\ref{eqnCCM1}) exists, then the following
1357: relation must also hold:
1358: \begin{eqnarray}
1359: \label{eqnCCM3}
1360: \frac{E(\lambda-V)}{E(B-V)} = b(\lambda) + [a(\lambda)-1] \; R(V) \;\;\;\; .
1361: \end{eqnarray}
1362: I.e., any true linear correlation between $A(\lambda)/A(V)$ and
1363: $R(V)^{-1}$ will also be present between $E(\lambda-V)/E(B-V)$ and
1364: $R(V)$, with a simple transformation relating the linear coefficients.
1365: The great advantage to viewing the data in the $E(\lambda-V)/E(B-V)$
1366: vs. $R(V)$ reference frame is that no artificial correlations (in either the parameters or their uncertainties) are
1367: introduced and we only have to deal with the natural correlations which
1368: arise from the dependence of all the extinction parameters on the
1369: properties of the best-fit stellar SEDs. Figure \ref{figRVKLAM}
1370: illustrates this point. It shows plots of $E(\lambda-V)/E(B-V)$ vs.
1371: $R(V)$ (along with their associated 1-$\sigma$ errors) for the same
1372: four wavelengths as in Figure \ref{figRVAV}. The 2695 \AA\/ panel
1373: shows a weighted linear fit, which is given by $E(2695 \; {\rm
1374: \AA}-V)/E(B-V) = 4.80 - 0.44 \; R(V)$. This is nearly exactly what
1375: would be expected from the fit to $A(2695 \; {\rm \AA})/A(V)$ vs.
1376: $R(V)^{-1}$ in Figure \ref{figRVAV} and the transformation in Equation
1377: (\ref{eqnCCM3}). The exact transformation is shown by the nearly
1378: coincident dashed line in Figure \ref{figRVKLAM}, verifying the
1379: argument leading to Equation (\ref{eqnCCM3}). (A careful comparison
1380: between the lines and the data points will show that the two panels do
1381: indeed show the same relationship.)
1382:
1383: From a statistical point of view, the presentations in Figures
1384: \ref{figRVAV} and \ref{figRVKLAM} are identical, as are the linear fits
1385: to the 2965 \AA\/ data. In fact, both fits have nearly identical,
1386: underwhelming, reduced $\chi^2$ values of $\sim$4.2. Fits performed at
1387: wavelengths shortward of 2675 \AA, show increasingly large values of
1388: $\chi^2$ (e.g., $\chi^2 = 6.9$ for a linear fit at 1665 \AA). The
1389: lesson from Figures \ref{figRVAV} and \ref{figRVKLAM} is that the
1390: choice of normalization affects the perception of how well-related
1391: $R(V)$ is to the level of UV extinction. The data in Figure
1392: \ref{figRVAV} look better-correlated than do those in Figure
1393: \ref{figRVKLAM} --- but they are not. The ``cosmic scatter'' is
1394: appreciable and there is no functional relationship between $R(V)$ and
1395: the UV extinction (whether linear or more complex) for which the
1396: scatter approaches the current level of measurement errors. Although
1397: there is a relationship in the sense that large-$R(V)$ curves [i.e., $R(V)
1398: \gtrsim 4.0$] differ systematically from low-$R(V)$ curves, UV extinction
1399: properties cannot be expressed as a 1-parameter family in $R(V)$ at
1400: anywhere near the level of observational accuracy.
1401:
1402: Note that the discussion above is not affected by the likelihood that
1403: some of the sightlines are composites, possibly spanning distinct
1404: regions of very different $R(V)$. If a universal relationship of the
1405: form in Equation (\ref{eqnCCM1}) were to hold, then composite regions
1406: would still lie along the line $a(\lambda) + b(\lambda) \; R(V)^{-1}$
1407: and would not result in ``cosmic scatter.''
1408:
1409: The recent study by Valencic et al. (2004) examined the
1410: $R(V)$-dependence of UV extinction by looking at correlations between
1411: $R(V)^{-1}$ and UV fit parameters based on the Paper III formulation,
1412: as had been done in F99 and F04. As noted earlier, they concluded that
1413: most Galactic curves (i.e., 93\%) are consistent with a
1414: $R(V)$-dependent ``law.'' While the approach was somewhat different,
1415: the results of the Valencic et al. study suffer the same problem as
1416: shown in Figures \ref{figRVAV} and \ref{figRVKLAM} because they
1417: explicitly multiply the UV fit coefficients by $R(V)^{-1}$, thus
1418: forcing a correlation in the errors and enhancing the perception of a
1419: correlation between the quantities. The large percentage of curves
1420: believed to be consistent with a single $R(V)^{-1}$ relation results
1421: from the complications introduced by the choice of curve
1422: normalization. Figure \ref{figRVFM} shows the relationship between
1423: $R(V)$ and the UV fitting parameters in their original form, in which
1424: the correlations among the parameters and their errors are not
1425: artificially enhanced. The figure contains several extra quantities
1426: derived from the fit parameters which help describe features of the UV
1427: curves and which are defined in \S \ref{secMETHOD} (and also in the
1428: figure legend). This figure again shows that, while systematic
1429: differences exist between high-$R(V)$ and low-$R(V)$ curves, there is
1430: no simple (or complex) relation between $R(V)$ and the UV fitting
1431: properties which is consistent with measurement errors.
1432:
1433: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1434: % SUMMARY
1435: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1436:
1437: \section{SUMMARY\label{secDISCUSS}}
1438:
1439: Several of our findings are worth emphasizing.
1440:
1441: \noindent {\underline {\bf Variability of \boldmath{$x_0$} and
1442: \boldmath{$\gamma$}:}}$\;$ We found, in accordance with our previous
1443: analysis, that while the central position of the 2175 \AA\/ bump does not
1444: vary much, it is indeed variable. On the other hand, the bump width
1445: varies considerably. Further, there still appears to be no
1446: relationship between the central position of the bump and and its width
1447: -- verifying the results described in Paper I to a higher degree of
1448: accuracy than previously possible.
1449:
1450: \noindent {\underline {\bf Correlations among the fit parameters:}}$\;$
1451: Generally, the various fit parameters which describe the shape of the
1452: UV-through-IR extinction curves are not related to one another. The
1453: only exceptions are $c_1-c_2$ and $k_{IR}-R(V)$. The first relates the
1454: slope and intercept of the UV portion of the curve and is effectively a
1455: functional relationship for most of the data. The latter relates the
1456: scale factor of the power law used to describe the IR portion of the
1457: curve and the ratio of total-to-selective extinction. At the level of
1458: our measurements errors, these two parameters are functionally related and
1459: consistent with a universal form to IR extinction. However, due to observational uncertainties, the dominance of diffuse ISM sightlines in our study, and the limitations of the IR data we employ, our results
1460: are not ideal for addressing the detailed shape of IR extinction and
1461: significant sightline-to-sightline variations could still exist.
1462:
1463: \noindent {\underline {\bf Correlations with dust density:}}$\;$ As
1464: determined in the past, various properties of the extinction curves are
1465: weakly correlated with the mean line of sight dust density, as measured
1466: by $E(B-V)/d$. Presumably, these correlations reflect the operation of
1467: a physical process, such as grain growth or coagulation.
1468:
1469: \noindent {\underline {\bf Correlations of \boldmath{$R(V)$} with UV
1470: curve properties relationship:}$\;$ Correlations between $R(V)$ and UV
1471: curve properties have received considerable attention over the years
1472: since Cardelli et al. (1988) first pointed out that a large-scale trend
1473: could be found. By expressing the curves in their native form, we
1474: verify that a there is a weak relationship between $R(V)$ and the UV in
1475: the sense that sightlines with extremely large $R(V)$ values tend to
1476: have low normalized UV extinction curves. However, this relationship
1477: is only evident for the largest $R(V)$ sightlines. For the majority of
1478: sightlines, there is no evidence for an $R(V)$-dependence of the
1479: extinction curve shapes. Specifically, inspection of Figure
1480: \ref{figRVKLAM} shows that for diffuse ISM sightlines ($2.4 < R(V) <
1481: 3.6$), which comprise the bulk (82\%) of our sample, no relationship
1482: exists, even though a large range in the extinction is present.
1483: Moreover, Figure \ref{figRVFM} shows that the 2175 \AA\/ bump
1484: properties display a similar behavior, i.e., a trend is evident only
1485: for the largest $R(V)$ values, with the strength of the bump (as
1486: measured by $A_{bump}$ or $E_{bump}$) tending to be slightly weaker
1487: than the average for lower $R(V)$ sightlines. Thus, we conclude that
1488: there is no global 1-parameter family of extinction curves, although
1489: extremely large $R(V)$ curves tend to have distinctive properties.
1490:
1491: \noindent {\underline {\bf Extinction curve variability -- the meaning
1492: and utility of an average curve:}}$\;$ The previous result begs the
1493: meaning of an average extinction curve. As noted in
1494: \S \ref{secEXT_PROPERTIES}, simple mean curves always reflect the biases
1495: of their parent samples. Our mean curve for the diffuse ISM (Figure
1496: \ref{figAVGCURVE}, an average of all curves with $2.4 \leq R(V) \leq
1497: 3.6$, and that from Valencic et al.\ (2004, for R=3.1) are derived from
1498: the largest samples and probably provide the best estimate of mean
1499: Galactic diffuse ISM extinction properties at short wavelengths.
1500: However, one must always be mindful of the dark shaded region in Figure
1501: \ref{figAVGCURVE} which illustrates the RMS variance that can be
1502: expected for an extinction curve along an arbitrary diffuse sightline.
1503: Typical RMS dispersions in $k(\lambda-V)$ are 0.31, 0.68, 0.62, 1.44
1504: at $\lambda = $2695, 2175, 1665, 1245 \AA, respectively. This means
1505: that if the mean curve is used to deredden an object with $E(B-V) =
1506: 0.50$~mag, the uncertainty in the dereddened continuum at these
1507: wavelengths would be 0.15, 0.30, 0.30 and 0.72~mag, respectively, due
1508: to uncertainties in the extinction alone! It is, however, possible to
1509: take advantage of localized uniformity in the extinction to reduce this
1510: error (e.g., Massa \& Savage 1981).
1511:
1512: \noindent {\underline {\bf Physical implications:}}$\;$ Perhaps the
1513: best way to summarize the physical origin of the $R(V)$-dependence for
1514: Galactic extinction is to examine the $R(V)^{-1} - c_2$ plot shown in
1515: Figure~\ref{figRVC2}, which includes a best-fit linear relation (such a
1516: relation was the basis for the $R(V)$-dependent curves produced by F99
1517: and F04 -- see Figure~10 of F04). Figure~\ref{figRVC2} can be
1518: summarized thusly: when extinction curves are steep in the optical
1519: (large $R(V)^{-1}$) they tend to stay steep in the UV (large $c_2$) and
1520: when extinction curves are flat in the optical (small $R(V)^{-1}$) they
1521: tend to stay flat in the UV (small $c_2$). However, as the scatter in
1522: the Figure illustrates, this trend is only apparent for extreme values
1523: of $R(V)^{-1}$ or $c_2$. In general, there is no unique relation
1524: between these parameters over the range spanned by most of the sample.
1525: It is likely that the general connection between UV and optical
1526: extinction slopes simply reflects the fact that the overall grain size
1527: distribution affects all wavelengths. But the presence of such a large
1528: scatter demonstrates that several other factors (e.g., chemical
1529: composition, grain history, coagulation, coating, radiation
1530: environment, etc.) must also be involved. In other words, the large
1531: variance in the relation between UV and optical slopes indicates that
1532: dust grain size distributions do not behave as a 1-parameter family.
1533:
1534: \noindent {\underline {\bf Final remarks:}}$\;$ Having painted a
1535: negative picture of the relationship between $R(V)$ and extinction
1536: properties at other wavelengths, it would be disingenuous of us to
1537: present yet another set of $R(V)$-dependent curves. The results of
1538: Cardelli et al. (1989), F04 (which supersedes those of F99), and
1539: Valencic et al. (2004) are all reasonable, and the differences among
1540: them are instructive of the biases introduced by sample selection and
1541: methodology. It should always be remembered that these curves
1542: represent very general trends in Galactic extinction and do not
1543: constitute a standard of normalcy. Finally, although the prospect of
1544: accurately dereddening an unknown SED using a mean extinction curve is
1545: poor, this same variability demonstrates that extinction curves are
1546: responsive to local conditions, so that each one contains potentially
1547: unique information about the grains along the sightline.
1548:
1549:
1550: %%%%%%%%%%%%%%%%%%%%%%%%%%%
1551: % ACKNOWLEDGEMENTS
1552: %%%%%%%%%%%%%%%%%%%%%%%%%%%
1553:
1554: \begin{acknowledgments}
1555: E.F. acknowledges support from NASA grant NAG5-12137, NAG5-10385, and
1556: NNG04GD46G. D.M. acknowledges support from NASA grant NNG04EC01P. We
1557: are grateful to the referee Geoff Clayton for helpful comments and
1558: suggestions. Some of the data presented in this paper were obtained
1559: from the Multimission Archive at the Space Telescope Science Institute
1560: (MAST). STScI is operated by the Association of Universities for
1561: Research in Astronomy, Inc., under NASA contract NAS5-26555. Support
1562: for MAST for non-HST data is provided by the NASA Office of Space
1563: Science via grant NAG5-7584 and by other grants and contracts. This
1564: publication also makes use of data products from the Two Micron All Sky
1565: Survey, which is a joint project of the University of Massachusetts and
1566: the Infrared Processing and Analysis Center/California Institute of
1567: Technology, funded by the National Aeronautics and Space Administration
1568: and the National Science Foundation.
1569: \end{acknowledgments}
1570:
1571: %%%%%%%%%%%%%%%%%%%%%%%%%%%
1572: % ACKNOWLEDGEMENTS
1573: %%%%%%%%%%%%%%%%%%%%%%%%%%%
1574:
1575: \appendix
1576: \section{Obtaining the Data from this Study}
1577:
1578: The extinction curves and associated information from this study are
1579: available via anonymous ftp at ``ftp.astronomy.villanova.edu''. After
1580: logging in as ``anonymous'', change to the appropriate directory by
1581: typing ``cd fitz$/$FMV\_EXTINCTION''. A ``README'' file, two IDL
1582: procedure files, and two tar files are present. The tar files contain
1583: the extinction curves from this program and the R-dependent curves
1584: from Fitzpatrick 2004 (``FMV\_EXTCURVES.tar'' and ``F04\_RCURVES.tar'',
1585: respectively). The tar files unfold into individual ascii data files
1586: for each extinction curve, containing the 12 extinction parameters
1587: which describe the curve and the curve itself (along with 1-$\sigma$
1588: uncertainties for the curves from this paper). The two IDL procedures
1589: provide tools for reading data from the ascii files and for
1590: constructing extinction curves from the 12 fit parameters
1591: (``READ\_FMV\_FILES.pro'' and ``MAKE\_FMV\_CURVE.pro'', respectively).
1592:
1593:
1594: %%%%%%%%%%%%%%%%%%%%%%%%%%%
1595: % REFERENCES
1596: %%%%%%%%%%%%%%%%%%%%%%%%%%%
1597:
1598: \clearpage
1599: \bibliographystyle{apj}
1600: \begin{thebibliography}
1601:
1602: \bibitem[Balona \& Crampton 1974]{BC74}
1603: Balona, L., \& Crampton, D. 1974, \mnras, 166, 203
1604:
1605: \bibitem[]{} %%% TABLE 1 REFERENCE
1606: Barbier, M., \& Boulon, J. 1960, J. Obs., 43, 69
1607:
1608: \bibitem[Blaauw 1963]{Bl63}
1609: Blaauw, A. 1963, in Basic Astronomical Data, ed. K.A. Strand (University of Chicago Press: Chicago), Chapter 20
1610:
1611: \bibitem[]{} %%% TABLE 1 REFERENCE
1612: Blaauw, A., Hiltner, W.A., \& Johnson, H.L. 1959, \apj, 130, 69
1613:
1614: \bibitem[Bohlin 1996]{Bohlin}
1615: Bohlin, R.~C. 1996, \aj, 111, 1743
1616:
1617: \bibitem[B\"ohm-Vitense 1981]{bv81} %%% FIGURE 5 REFERENCE
1618: B\"ohm-Vitense, E. 1981, \araa, 19, 295
1619:
1620: \bibitem[]{} %%% TABLE 1 REFERENCE
1621: Borgman, J. 1960, \bain, 15, 255
1622:
1623: \bibitem[]{} %%% TABLE 1 REFERENCE
1624: Bouigue, M.R. 1959, Publ. Obs. Haute-Provence, 4, 52
1625:
1626: \bibitem[]{} %%% TABLE 1 REFERENCE
1627: Boulon, J., Duflot, M., \& Fehrenbach, C. 1958, J. Obs., 42, 1
1628:
1629: \bibitem[]{} %%% TABLE 1 REFERENCE
1630: Boulon, J., \& Fehrenbach, C. 1959, Publ. Obs. Haute-Provence, 4, 55
1631:
1632: \bibitem[]{} %%% TABLE 1 REFERENCE
1633: Buscombe, W. 1962, Mount Stromlo Obs. Mimeo., 4, 1
1634:
1635: \bibitem[]{} %%% TABLE 1 REFERENCE
1636: Buscombe, W. 1969, \mnras, 144, 31
1637:
1638: \bibitem[Cardelli et al. 1988]{CCM88}
1639: Cardelli, J.A., Clayton, G.C. \& Mathis, J.S. 1988, \apjl, 329, 33
1640:
1641: \bibitem[Cardelli et al. 1989]{CCM89}
1642: Cardelli, J.A., Clayton, G.C. \& Mathis, J.S. 1989, \apj, 345, 245
1643:
1644: \bibitem[Cardelli et al. 1992]{CSM92}
1645: Cardelli, J.A., Sembach, K.R., \& Mathis, J.S. 1992, \aj, 104, 1916
1646:
1647: \bibitem[Chini and Kr\"{u}gel 1983]{CK83}
1648: Chini, R. \& Kr\"{u}gel, E. 1983, \aap, 117, 289
1649:
1650: \bibitem[]{} %%% TABLE 1 REFERENCE
1651: Claria, J.J. 1974, \aj, 79, 1022
1652:
1653: \bibitem[Clayton \& Fitzpatrick 1987]{CF87}
1654: Clayton, G.C., \& Fitzpatrick, E.L. 1987, \aj, 92, 147
1655:
1656: \bibitem[Clayton et al. 2000]{Clay00}
1657: Clayton, G.C., Gordon, K.D., \& Wolff, M.J. 2000, \apjs, 129, 147
1658:
1659: \bibitem[Clayton et al. 2003]{Clay03}
1660: Clayton, G.C., et al. 2003, \apj, 592, 947
1661:
1662: \bibitem[]{} %%% TABLE 1 REFERENCE
1663: Cowley, A., Cowley, C., Jaschek, M., \& Jaschek, C. 1969, \aj, 74, 375
1664:
1665: \bibitem[Crawford 1978]{Cr78}
1666: Crawford, D.L. 1978, \aj, 83, 48
1667:
1668: \bibitem[Crawford \& Barnes 1970]{CB70}
1669: Crawford, D.L., \& Barnes, J.V. 1970, \aj, 75, 952
1670:
1671: \bibitem[]{} %%% TABLE 1 REFERENCE
1672: Crawford, D., Limber, D.N., Mendoza, E.E., Schulte, D., Steinman, H., \& Swihart, T. 1955, \apj, 121, 24
1673:
1674: \bibitem[]{} %%% TABLE 1 REFERENCE
1675: De Vaucouleurs, A. 1957, \mnras, 117, 449
1676:
1677: \bibitem[]{} %%% TABLE 1 REFERENCE
1678: Denoyelle, J. 1977, \aaps, 27, 343
1679:
1680: \bibitem[]{} %%% TABLE 1 REFERENCE
1681: Divan, L. 1954, Annales d'Astrophysique, 17, 456
1682:
1683: \bibitem[Draine 2003]{Dr03}
1684: Draine, B.T. 2003, \araa, 41, 241
1685:
1686: \bibitem[]{} %%% TABLE 1 REFERENCE
1687: Feast, M.W., Stoy, R.H., Thackeray, A.D., \& Wesselink, A.J. 1961, \mnras, 122, 239
1688:
1689: \bibitem[]{} %%% TABLE 1 REFERENCE
1690: Feast, M.W., Thackeray, A.D., \& Wesselink, A.J. 1955, Mem. R. Astron. Soc, 67, 51
1691:
1692: \bibitem[]{} %%% TABLE 1 REFERENCE
1693: Feast, M.W., Thackeray, A.D., \& Wesselink, A.J. 1957, Mem. R. Astron. Soc, 68, 1
1694:
1695: \bibitem[Fitzpatrick 1999]{Fitz99}
1696: Fitzpatrick, E.L. 1999, \pasp, 111, 63 (F99)
1697:
1698: \bibitem[Fitzpatrick 2004]{Fitz04}
1699: Fitzpatrick, E.L. 2004, in Astrophysics of Dust, eds. A.N. Witt, G.C.
1700: Clayton, \& B.T. Draine, ASP Conference Series, Vol. 309, 33 (F04)
1701:
1702: \bibitem[Fitzpatrick \& Massa 1986]{FM86}
1703: Fitzpatrick, E.L., \& Massa D. 1986, \apj, 307, 286 (Paper I)
1704:
1705: \bibitem[Fitzpatrick \& Massa 1988]{FM88}
1706: Fitzpatrick, E.L., \& Massa D. 1988, \apj, 328, 734 (Paper II)
1707:
1708: \bibitem[Fitzpatrick \& Massa 1990]{Paper III}
1709: Fitzpatrick, E.L., \& Massa D. 1990, \apjs, 72, 163 (Paper III)
1710:
1711: \bibitem[Fitzpatrick \& Massa 1999]{FM99}
1712: Fitzpatrick, E.L., \& Massa D. 1999, \apj, 525, 1011
1713:
1714: \bibitem[Fitzpatrick \& Massa 2005a]{Paper IV}
1715: Fitzpatrick, E.L., \& Massa D. 2005a, \aj, 130, 1127 (Paper IV)
1716:
1717: \bibitem[Fitzpatrick \& Massa 2005b]{FM05b }
1718: Fitzpatrick, E.L., \& Massa D. 2005b, \aj, 129, 1642
1719:
1720: \bibitem[Flower 1977]{Flower77} %%% FIGURE 5 REFERENCE
1721: Flower, P.J. 1977, \aap, 54 31
1722:
1723: \bibitem[]{} %%% TABLE 1 REFERENCE
1724: Garrison, R.F. 1967, \apj, 147, 1003
1725:
1726: \bibitem[]{} %%% TABLE 1 REFERENCE
1727: Garrison, R.F., \& Kormendy, J. 1976, \pasp, 88, 865
1728:
1729: \bibitem[]{} %%% TABLE 1 REFERENCE
1730: Garrison, R.F., \& Schild, R.E. 1979, \aj, 84, 1020
1731:
1732: \bibitem[Georgelin \& Georgelin 1976]{bb76}
1733: Georgelin, Y.M., \& Georgelin, Y.P. 1976, \aaps, 49, 57
1734:
1735: \bibitem[]{} %%% TABLE 1 REFERENCE
1736: Georgelin, Y.M., Georgelin, Y.P., \& Roux, S. 1973, \aap, 25, 337
1737:
1738: \bibitem[]{} %%% TABLE 1 REFERENCE
1739: Guetter, H.H. 1964, Publications of the David Dunlap Observatory, 2, 403
1740:
1741: \bibitem[]{} %%% TABLE 1 REFERENCE
1742: Guetter, H.H. 1968, \pasp, 80, 197
1743:
1744: \bibitem[]{} %%% TABLE 1 REFERENCE
1745: Hack, M. 1953, Annales d'Astrophysique, 16, 417
1746:
1747: \bibitem[]{} %%% TABLE 1 REFERENCE
1748: Hardie R.H., \& Crawford, D.L. 1961, \apj, 133, 843
1749:
1750: \bibitem[]{} %%% TABLE 1 REFERENCE
1751: Herbig, G.H., \& Spalding, J. 1955, \apj, 121, 118
1752:
1753: \bibitem[]{} %%% TABLE 1 REFERENCE
1754: Hill, P.W. 1970, \mnras, 150, 23
1755:
1756: \bibitem[]{} %%% TABLE 1 REFERENCE
1757: Hiltner, W.A. 1956, \apjs, 2, 389
1758:
1759: \bibitem[]{} %%% TABLE 1 REFERENCE
1760: Hiltner, W.A., Garrison R.F., \& Schild R.E. 1969, \apj, 157, 313
1761:
1762: \bibitem[]{} %%% TABLE 1 REFERENCE
1763: Hiltner, W.A., Morgan, W.W., \& Neff, J.S. 1965, \apj, 141, 183
1764:
1765: \bibitem[]{} %%% TABLE 1 REFERENCE
1766: Hoag, A.A., \& Applequist, N.L. 1965, \apjs, 12, 215
1767:
1768: \bibitem[]{} %%% TABLE 1 REFERENCE
1769: Hoag, A.A., \& Smith, E.V.P. 1959, \pasp, 71, 32
1770:
1771: \bibitem[]{} %%% TABLE 1 REFERENCE
1772: Hoffleit, D. 1956, \apj, 124, 61
1773:
1774: \bibitem[]{} %%% TABLE 1 REFERENCE
1775: Houk N., \& Cowley, A.P. 1975, Michigan Catalogue of Two-Dimensional Spectral Types for the HD Stars, C01, 0
1776:
1777: \bibitem[]{} %%% TABLE 1 REFERENCE
1778: Houk, N. 1978, Michigan Catalogue of Two-Dimensional Spectral Types for the HD Stars, C02, 0
1779:
1780: \bibitem[]{} %%% TABLE 1 REFERENCE
1781: Houk, N., \& Smith-Moore, M. 1988, Michigan Catalogue of Two-Dimensional Spectral Types for the HD Stars, C04, 0
1782:
1783: \bibitem[Humphreys \& McElroy 1984]{HM84} %%% FIGURE 5 REFERENCE
1784: Humphreys, R.M. \& McElroy, D.B. 1984, \apj, 284, 565
1785:
1786: \bibitem[]{} %%% TABLE 1 REFERENCE
1787: Johnson, H.L. 1962, \apj, 136, 1135
1788:
1789: \bibitem[Johnson 1968]{joh68}
1790: Johnson, H.L. 1968, in Nebulae and Interstellar Matter, eds. B.M. Middlehurst \& L.H. Aller (University of Chicago Press: Chicago), Chapter 5
1791:
1792: \bibitem[]{} %%% TABLE 1 REFERENCE
1793: Johnson, H.L., \& Morgan, W.W. 1953, \apj, 117, 313
1794:
1795: \bibitem[]{} %%% TABLE 1 REFERENCE
1796: Johnson, H.L., \& Morgan, W.W. 1955, \apj, 122, 429
1797:
1798: \bibitem[Kiszkurno-Koziej \& Lequeux 1987]{KKL87}
1799: Kiszkurno-Koziej, E. \& Lequeux, J. 1987, \aap, 185, 291
1800:
1801: \bibitem[Kurucz 1991]{K91}
1802: Kurucz, R.L. 1991, in Stellar Atmospheres: Beyond Classical Models,
1803: ed. L. Crivellari, I. Hubeny, \& D. G. Hummer (NATO ASI Ser. C.;
1804: Dordrect: Reidel), 441
1805:
1806: \bibitem[Lanz \& Hubeny 2003]{LH03}
1807: Lanz, T., \& Hubeny, I. 2003, \apjs, 146, 417
1808:
1809: \bibitem[Larson \& Whittet 2005]{LW05}
1810: Larson, K.A., \& Whittet, D.C.B. 2005, \apj, 623, 897
1811:
1812: \bibitem[Levato \& Abt 1976]{LA76}
1813: Levato, H., \& Abt, H.A. 1976 \pasp, 88, 712
1814:
1815: \bibitem[]{} %%% TABLE 1 REFERENCE
1816: Levato H., \& Malaroda S. 1980, \pasp, 92, 323
1817:
1818: \bibitem[]{} %%% TABLE 1 REFERENCE
1819: Levato, H., Malaroda, S., Morrell, N., Garcia, B., \& Hernandez, C. 1991, \apjs, 75, 869
1820:
1821: \bibitem[]{} %%% TABLE 1 REFERENCE
1822: Ma\'{i}z-Apell\'{a}niz, J., Walborn, N.R., \& Galu\'{e}, H.A., Wei L. 2004, \apjs, 151, 103
1823:
1824: \bibitem[Malagnini et al. 1984]{mmf84} %%% FIGURE 5 REFERENCE
1825: Malagnini, M.L., Morossi, C., \& Faraggiana, R. 1984, in The MK Process
1826: and Stellar classification, ed. R.F. Garrison (David Dunlap
1827: Observatory:Toronto), 321 375
1828:
1829: \bibitem[Martin \& Whittet 1990]{MW90}
1830: Martin, P.G., \& Whittet, D.C.B. 1990, \apj, 357, 113
1831:
1832: \bibitem[Martins, Schaerer, \& Hillier 2005]{MSH05}
1833: Martins, F., Schaerer, D., \& Hillier, D.J. 2005, \aap, 436,1049
1834:
1835: \bibitem[Massa 1987]{mas87}
1836: Massa, D. 1987, \aj, 94, 1675
1837:
1838: \bibitem[Massa \& Fitzpatrick 1986]{MF86}
1839: Massa, D., \& Fitzpatrick, E.L. 1986, \apjs, 60, 305
1840:
1841: \bibitem[Massa \& Fitzpatrick 2000]{MF00}
1842: Massa, D., \& Fitzpatrick, E.L. 2000, \apjs, 126, 517 (MF00)
1843:
1844: \bibitem[Massa \& Savage 1981]{MS81}
1845: Massa, D., \& Savage, B.D. 1981, \apj, 248, 201
1846:
1847: \bibitem[Massa \& Savage 1984]{MS84}
1848: Massa, D., \& Savage, B.D. 1984, \apj, 279, 310
1849:
1850: \bibitem[Massa et al. 1983]{MSF}
1851: Massa, D., Savage, B.D., \& Fitzpatrick, E.L. 1983, \apj, 266, 662
1852:
1853: \bibitem[]{} %%% TABLE 1 REFERENCE
1854: Massey, P., \& Johnson, J., 1993, \aj, 105, 980
1855:
1856: \bibitem[]{} %%% TABLE 1 REFERENCE
1857: McNamara, B.J. 1976, \aj, 81, 375
1858:
1859: \bibitem[]{} %%% TABLE 1 REFERENCE
1860: Meadows, A.J. 1961, \mnras,123, 81
1861:
1862: \bibitem[]{} %%% TABLE 1 REFERENCE
1863: Mendoza, E.E. 1956, \apj, 123, 54
1864:
1865: \bibitem[Mermilliod et~al., 1997]{MMH}
1866: Mermilliod, J.-C., Mermilliod, M., \& Hauck, B. 1997 \aaps, 124, 349
1867:
1868: \bibitem[]{} %%% TABLE 1 REFERENCE
1869: Metreveli, M.D. 1968, Abastumanskaya Astrofiz. Obs., Byull., 38, 93
1870:
1871: \bibitem[]{} %%% TABLE 1 REFERENCE
1872: Moffat, A.F.J., \& Fitzgerald, M.P. 1974, \aap, 34, 291
1873:
1874: \bibitem[Morgan et al. 1982]{MMN82}
1875: Morgan, D.H., McLachlan, A., \& Nandy, K. 1982, \mnras, 198, 779
1876:
1877: \bibitem[]{} %%% TABLE 1 REFERENCE
1878: Morgan, W.W., Code, A.D., \& Whitford, A.E. 1955, \apjs, 2, 41
1879:
1880: \bibitem[]{} %%% TABLE 1 REFERENCE
1881: Morgan, W.W., Keenan, P.C., \& Kellman, E. 1943, An Atlas of Stellar Spectra, With an Outline of Spectral Classification (University of Chicago Press:Chicago)
1882:
1883: \bibitem[]{} %%% TABLE 1 REFERENCE
1884: Morgan, W.W., Whitford, A.E., \& Code, A.D. 1953, \apj, 118, 318
1885:
1886: \bibitem[]{} %%% TABLE 1 REFERENCE
1887: Morrell, N., Garcia, B., \& Levato, H. 1988, \pasp, 100, 1431
1888:
1889: \bibitem[]{} %%% TABLE 1 REFERENCE
1890: Morris, P.M. 1961, \mnras, 122, 325
1891:
1892: \bibitem[Nichols \& Linsky, 1996]{NL}
1893: Nichols, J.S. \& Linsky, J.L. 1996, \aj, 111, 517
1894:
1895: \bibitem[Nishiyama et al. 2006]{Nishi06}
1896: Nishiyama, S., et al. 2006, \apj, 638, 846
1897:
1898: \bibitem[]{} %%% TABLE 1 REFERENCE
1899: Osawa, K. 1959, \apj, 130, 159
1900:
1901: \bibitem[Panagia 1973]{Pan73} %%% FIGURE 5 REFERENCE
1902: Panagia, N. 1973, \aj, 78, 929
1903:
1904: \bibitem[Panek 1983]{pan83}
1905: Panek, R.J. 1983,\apj, 270, 169
1906:
1907: \bibitem[]{} %%% TABLE 1 REFERENCE
1908: Pecker, Ch. 1953, Annales d'Astrophysique, 16, 321
1909:
1910: \bibitem[]{} %%% TABLE 1 REFERENCE
1911: Penston, M.V., Hunter, J.K., \& O'Neill, T. 1975, \mnras, 171, 219
1912:
1913: \bibitem[Perez et al. 1987]{Per87}
1914: Perez, M.R., The, P.S., \& Westerlund, B.E., 1987, \pasp, 99, 1050
1915:
1916: \bibitem[]{} %%% TABLE 1 REFERENCE
1917: Perry C.L., Hill, G., \& Christodoulou, D.M. 1991, \aaps, 90, 195
1918:
1919: \bibitem[]{} %%% TABLE 1 REFERENCE
1920: Perry, C.L., Franklin, C.B., Landolt, A.U., \& Crawford D.L. 1976, \aj, 81, 632
1921:
1922: \bibitem[Rachford et al 2002]{rach02}
1923: Rachford, B.L., et al. 2002, \apj, 577, 221
1924:
1925: \bibitem[]{} %%% TABLE 1 REFERENCE
1926: Racine, R. 1968, \aj, 73, 233
1927:
1928: \bibitem[]{} %%% TABLE 1 REFERENCE
1929: Racine, R., 1968, \aj, 73, 588
1930:
1931: \bibitem[]{} %%% TABLE 1 REFERENCE
1932: Roman, N.G. 1951, \apj, 114, 492
1933:
1934: \bibitem[]{} %%% TABLE 1 REFERENCE
1935: Roman, N.G. 1955, \apjs, 2, 195
1936:
1937: \bibitem[]{} %%% TABLE 1 REFERENCE
1938: Roman, N.G. 1956, \apj, 123, 246
1939:
1940: \bibitem[Rosenzweig \& Morrison 1986]{RM86}
1941: Rosenzweig, P., \& Morrison, N.D. 1986, \apj, 306, 522
1942:
1943: \bibitem[Savage \& Mathis 1979]{SM79}
1944: Savage, B.D., \& Mathis, J.S. 1979, \araa, 17, 73
1945:
1946: \bibitem[Savage et al. 1985]{SMMW85}
1947: Savage, B.D., Massa, D., Meade, M., Wesselius, P.R. 1985, \apjs,59, 397
1948:
1949: \bibitem[]{} %%% TABLE 1 REFERENCE
1950: Schild R., Neugebauer, G., \& Westphal, J.A. 1971, \aj, 76, 237
1951:
1952: \bibitem[]{} %%% TABLE 1 REFERENCE
1953: Schild R.E., Hiltner, W.A., \& Sanduleak, N., 1969 \apj, 156, 609
1954:
1955: \bibitem[]{} %%% TABLE 1 REFERENCE
1956: Schild, R.E. 1970, \apj, 161, 855
1957:
1958: \bibitem[]{} %%% TABLE 1 REFERENCE
1959: Schild, R.E., 1965 \apj, 142, 979
1960:
1961: \bibitem[]{} %%% TABLE 1 REFERENCE
1962: Schild, R.E., \& Chaffee, F. 1971, \apj, 169, 529
1963:
1964: \bibitem[Schmidt-Kaler 1982]{Schmidt82} %%% FIGURE 5 REFERENCE
1965: Schmidt-Kaler, Th. 1982, in Landolt-Bornstein, Numerical Data and
1966: Functional Relationships in Science and Technology, Vol. 2, eds. K.
1967: Schaifers \& H.H. Voigt (Springer Verlag:Berlin), 451
1968:
1969: \bibitem[Seaton 1979]{Seaton}
1970: Seaton, M.J. 1979, \mnras, 187, 73P
1971:
1972: \bibitem[]{} %%% TABLE 1 REFERENCE
1973: Seidensticker, K.J. 1989, \aaps, 79, 61
1974:
1975: \bibitem[]{} %%% TABLE 1 REFERENCE
1976: Sharpless, S. 1952, \apj, 116, 251
1977:
1978: \bibitem[]{} %%% TABLE 1 REFERENCE
1979: Simonson, S.C. 1968, \apj, 154, 923
1980:
1981: \bibitem[]{} %%% TABLE 1 REFERENCE
1982: Slettebak, A., 1968 \apj, 154, 933
1983:
1984: \bibitem[]{} %%% TABLE 1 REFERENCE
1985: Smith, M.A. 1972, \apj, 175, 765
1986:
1987: \bibitem[]{} %%% TABLE 1 REFERENCE
1988: Stebbins, J., \& Kron, G.E. 1956, \apj, 123, 440
1989:
1990: Tapia, M.; Roth, M.; Marraco, H.; Ruiz, M. T. Royal Astronomical Society, Monthly Notices (ISSN 0035-8711), vol. 232, June 1, 1988, p. 661-681
1991:
1992: \bibitem[Torres 1987]{tor87}
1993: Torres, A.V. 1987, \apj, 322, 949
1994:
1995: \bibitem[Turner 1980]{Tur80}
1996: Turner, D.G. 1980, \apj, 240, 137
1997:
1998: \bibitem[]{} %%% TABLE 1 REFERENCE
1999: Turner, D.G., Grieve, G.R., Herbst, W., \& Harris, W.E. 1980, \aj, 85, 1193
2000:
2001: \bibitem[Valencic et al. 2004]{VCG04}
2002: Valencic, L.A., Clayton, G.C., \& Gordon, K.D. 2004, \apj, 616, 912
2003:
2004: \bibitem[]{} %%% TABLE 1 REFERENCE
2005: Vrba, F.J., \& Rydgren, A.E. 1984, \apj, 283, 123
2006:
2007: \bibitem[Walborn \& Fitzpatrick 1990]{WF90}
2008: Walborn, N.R., \& Fitzpatrick, E.L. 1990, \pasp, 102, 379
2009:
2010: \bibitem[]{} %%% TABLE 1 REFERENCE
2011: Walker, M.F. 1957, \apj, 125, 636
2012:
2013: \bibitem[]{} %%% TABLE 1 REFERENCE
2014: Walker, M.F. 1961, \apj, 133, 438
2015:
2016: \bibitem[]{} %%% TABLE 1 REFERENCE
2017: Walraven, T., \& Walraven, J.H. 1960, \bain, 15, 67
2018:
2019: \bibitem[]{} %%% TABLE 1 REFERENCE
2020: Warren, W.H., \& Hesser J.E. 1977, \apjs, 34, 115
2021:
2022: \bibitem[]{} %%% TABLE 1 REFERENCE
2023: Wenzel, W. 1951, Veroeff. Sternwarte Sonneber, 5, 1
2024:
2025: \bibitem[Wu et al. 1980]{wgd80}
2026: Wu, C.-C., Gilra, D.P., \& van Duinen, R.J. 1980, \apj, 241, 173
2027:
2028: \bibitem[]{} %%% TABLE 1 REFERENCE
2029: Young, A. 1978, \pasp, 90, 144
2030:
2031: \end{thebibliography}
2032: \clearpage
2033:
2034: %%%%%%%%%%%%%%%%%%%%%%%%
2035: % TABLES
2036: %%%%%%%%%%%%%%%%%%%%%%%%
2037:
2038:
2039: %% TABLE 1
2040:
2041: \begin{deluxetable}{lcrcrrc}
2042: \tabletypesize{\scriptsize}
2043: \tablenum{1}
2044: \tablewidth{0pc}
2045: \tablecaption{Basic Data for Survey Stars [The complete version of this
2046: table is in the electronic edition of the Journal. The printed edition
2047: contains only a sample.]
2048: \label{tabSTARS}}
2049: \tablehead{
2050: \colhead{Star\tablenotemark{a}} &
2051: \colhead{Spectral} &
2052: \colhead{V} &
2053: \colhead{Distance\tablenotemark{c}} &
2054: \colhead{$l$} &
2055: \colhead{$b$} &
2056: \colhead{Reference} \\
2057: \colhead{} &
2058: \colhead{Type\tablenotemark{b}} &
2059: \colhead{(mag)} &
2060: \colhead{(pc)} &
2061: \colhead{($\degr$)} &
2062: \colhead{($\degr$)} &
2063: \colhead{} }
2064: \startdata
2065: HD698 & B5 II: SB & 7.10 & 1125 & $117.689$ & $ -4.25$ & 1 \\
2066: HD3191 & B1 IV:nn & 8.58 & 1203 & $121.068$ & $ -1.36$ & 2 \\
2067: BD+57 245 (NGC 457) & \nodata & 9.85 & 2429 & $126.583$ & $ -4.58$ & \nodata \\
2068: BD+57 252 (NGC 457) & B1 IV & 9.51 & 2429 & $126.644$ & $ -4.42$ & 2 \\
2069: NGC 457 Pesch 34 & \nodata & 10.61 & 2429 & $126.646$ & $ -4.38$ & \nodata \\
2070: NGC 457 Pesch 13 & \nodata & 10.78 & 2429 & $126.646$ & $ -4.38$ & \nodata \\
2071: NGC 457 Pesch 9 & B1 V & 9.83 & 2429 & $126.646$ & $ -4.38$ & 3 \\
2072: Cr 463 \#18 & \nodata & 10.35 & \phn702 & $127.091$ & $ 9.20$ & \nodata \\
2073: BD+70 131 (Cr 463) & \nodata & 10.06 & \phn702 & $127.280$ & $ 9.17$ & \nodata \\
2074: Cr 463 \#5 & \nodata & 10.37 & \phn702 & $127.264$ & $ 9.37$ & \nodata \\
2075: \enddata
2076: \tablenotetext{a}{The stars are listed in order of increasing Right Ascension using the most commonly adopted forms of their names. The first preference was ``HDnnn'', followed by ``BDnnn'', etc. There are 185 survey stars which are members of open clusters or associations, or 56\% of the sample. The identity of the cluster or association is either contained in the star name itself (e.g., NGC 457 Pesch 34) or is given in parentheses after the star's name.}
2077: \tablenotetext{b}{Spectral types were selected from those given in the SIMBAD database, and the source of the adopted types is shown in the ``Reference'' column. When multiple types were available for a particular star, we selected one based on our own preferred ranking of the sources.}
2078: \tablenotetext{c}{The NGC 2244 distance is from Perez et al. 1987; the NGC 3293 distance is from Bolona \& Crampton 1974; the Trumpler 14 and 16 distances are from Massey \& Johnson 1993; the Cep OB3 distance is from Crawford \& Barnes 1970. The distances to all other clusters or associations are from the Open Clusters and Galactic Structure database maintained by Wilton S. Dias, Jacques L\'{e}pine, Bruno S. Alessi, and Andr\'{e} Moitinho at http://www.astro.iag.usp.br/~wilton/. For the non-cluster stars, distances were calculated using the $E(B-V)$ values from this study and the absolute magnitudes from Turner 1980 (for mid-B and earlier types) and Blaauw 1963 (for mid-B and later types).}
2079: \tablerefs{(1) Hiltner 1956; (2) Morgan, Code, \& Whitford 1955; (3) Hoag, \& Applequist 1965; (4) Ma\'{i}z-Apell\'{a}niz, Walborn, Galu\'{e}, \& Wei 2005; (5) Johnson, \& Morgan 1955; (6) Slettebak 1968; (7) Schild 1965; (8) Racine 1968; (9) Morgan, Keenan, \& Kellman 1943; (10) Mendoza 1956; (11) Morgan, Whitford, \& Code 1953; (12) Osawa 1959; (13) Cowley, Cowley, Jaschek, \& Jaschek 1969; (14) Metreveli 1968; (15) Guetter 1968; (16) Boulon, \& Fehrenbach 1959; (17) Buscombe 1962; (18) Roman 1955; (19) Georgelin, Georgelin, \& Roux 1973; (20) Bouigue 1959; (21) Wenzel 1951; (22) Smith 1972; (23) Penston, Hunter, \& O'Neill 1975; (24) McNamara 1976; (25) Levato H., \& Abt H.A.; (26) Schild, \& Chaffee 1971; (27) Borgman 1960; (28) Warren, \& Hesser 1977; (29) Sharpless 1952; (30) Racine 1968; (31) Crawford, Limber, Mendoza, Schulte, Steinman, \& Swihart 1955; (32) Johnson, \& Morgan 1953; (33) Meadows 1961; (34) Johnson 1962; (35) Hoag, \& Smith 1959; (36) Barbier, \& Boulon 1960; (37) Young 1978; (38) Claria 1974; (39) Feast, Thackeray, \& Wesselink 1955; (40) Moffat, \& Fitzgerald 1974; (41) Houk 1978; (42) Feast, Stoy, Thackeray,\& Wesselink 1961; (43) Denoyelle 1977; (44) Hoffleit 1956; (45) Turner, Grieve, Herbst, \& Harris 1980; (46) Morrell, Garcia, \& Levato 1988; (47) Massey, \& Johnson 1993; (48) Levato, Malaroda, Morrell, Garcia, \& Hernandez 1991; (49) Morris 1961; (50) Hiltner, Garrison, \& Schild 1969; (51) Houk, \& Cowley 1975; (52) Seidensticker 1989; (53) Feast, Thackeray, \& Wesselink 1957; (54) Perry, Franklin, Landolt, \& Crawford 1976; (55) Schild 1970; (56) De Vaucouleurs 1957; (57) Walraven, \& Walraven 1960; (58) Garrison 1967; (59) Buscombe 1969; (60) Hardie, \& Crawford 1961; (61) Perry, Hill, \& Christodoulou 1991; (62) Schild, Neugebauer, \& Westphal 1971; (63) Schild, Hiltner, \& Sanduleak 1969; (64) Levato, \& Malaroda 1980; (65) Garrison, \& Schild 1979; (66) Roman 1956; (67) Houk, \& Smith-Moore 1988; (68) Hiltner, Morgan, \& Neff 1965; (69) Walker 1957; (70) Walker 1961; (71) Stebbins, \& Kron 1956; (72) Guetter 1964; (73) Hill 1970; (74) Herbig, \& Spalding 1955; (75) Vrba, \& Rydgren 1984; (76) Hack 1953; (77) Roman 1951; (78) Divan 1954; (79) Garrison, \& Kormendy 1976; (80) Simonson 1968; (81) Blaauw, Hiltner, \& Johnson 1959; (82) Pecker 1953; (83) Boulon, Duflot, \& Fehrenbach 1958}
2080: \end{deluxetable}
2081: \clearpage
2082:
2083:
2084: %% TABLE 2
2085:
2086: \begin{deluxetable}{lc}
2087: \tablewidth{0pc}
2088: \tablecaption{Adopted Temperature Scale for Main Sequence O Stars
2089: \label{tabOTEMPS}}
2090: \tablenum{2}
2091: \tablehead{
2092: \colhead{Spectral} &
2093: \colhead{$T_{eff}$} \\
2094: \colhead{Type} &
2095: \colhead{(K)} }
2096: \startdata
2097: O6 & $40000$ \\
2098: O6.5 & $38500$ \\
2099: O7 & $37000$ \\
2100: O7.5 & $36500$ \\
2101: O8 & $36000$ \\
2102: O8.5 & $34750$ \\
2103: O9 & $33500$ \\
2104: O9.5 & $32750$ \\
2105: B0 & $32000$ \\
2106: \enddata
2107: \end{deluxetable}
2108: \clearpage
2109:
2110: %% TABLE 3
2111:
2112: \begin{deluxetable}{llcccrc}
2113: \tabletypesize{\footnotesize}
2114: \tablenum{3}
2115: \tablewidth{0pc}
2116: \tablecaption{Best-Fit Parameters for Survey Stars [The complete version of this
2117: table is in the electronic edition of the Journal. The printed edition
2118: contains only a sample.]
2119: \label{tabPARMS}}
2120: \tablehead{
2121: \colhead{Star} &
2122: \colhead{$T_{eff}$\tablenotemark{a}} &
2123: \colhead{$\log g$\tablenotemark{b}} &
2124: \colhead{[m/H]\tablenotemark{c}} &
2125: \colhead{$v_{turb}$\tablenotemark{d}} &
2126: \colhead{$\theta_R$} &
2127: \colhead{$E(B-V)$} \\
2128: \colhead{} &
2129: \colhead{(K)} &
2130: \colhead{} &
2131: \colhead{} &
2132: \colhead{$\rm (km/s)$} &
2133: \colhead{(mas)} &
2134: \colhead{(mag)} }
2135: \startdata
2136: HD698 & $ 18434\pm 499$ & $ 3.72\pm 0.31$ & $-0.18\pm 0.10$ & $ 3.2\pm 0.5$ & $0.0670\pm0.0014$ & $ 0.37\pm 0.01$ \\
2137: HD3191 & $ 24001\pm 890$ & $ 3.41\pm 0.28$ & $-0.43\pm 0.06$ & $10$ & $0.0328\pm0.0009$ & $ 0.68\pm 0.01$ \\
2138: BD+57 245 (NGC 457) & $ 22885\pm 697$ & $ 3.64\pm 0.20$ & $-0.39\pm 0.08$ & $ 5.5\pm 0.6$ & $0.0160\pm0.0004$ & $ 0.50\pm 0.01$ \\
2139: BD+57 252 (NGC 457) & $ 24924\pm 616$ & $ 3.64\pm 0.16$ & $-0.47\pm 0.06$ & $10$ & $0.0173\pm0.0004$ & $ 0.51\pm 0.01$ \\
2140: NGC 457 Pesch 34 & $ 23594\pm 661$ & $ 3.91\pm 0.17$ & $-0.07\pm 0.06$ & $ 1.2\pm 0.8$ & $0.0110\pm0.0003$ & $ 0.51\pm 0.01$ \\
2141: NGC 457 Pesch 13 & $ 22023\pm 809$ & $ 3.84\pm 0.17$ & $-0.47\pm 0.11$ & $ 2.1\pm 0.8$ & $0.0114\pm0.0003$ & $ 0.51\pm 0.01$ \\
2142: NGC 457 Pesch 9 & $ 25738\pm 577$ & $ 3.77\pm 0.16$ & $-0.75\pm 0.07$ & $ 9.1\pm 0.7$ & $0.0147\pm0.0003$ & $ 0.54\pm 0.01$ \\
2143: Cr 463 \#18 & $ 11891\pm 286$ & $ 3.99\pm 0.16$ & $-0.60\pm 0.10$ & $0$ & $0.0198\pm0.0003$ & $ 0.35\pm 0.01$ \\
2144: BD+70 131 (Cr 463) & $ 11351\pm 204$ & $ 3.93\pm 0.17$ & $-0.73\pm 0.07$ & $ 5.1\pm 0.4$ & $0.0210\pm0.0003$ & $ 0.28\pm 0.01$ \\
2145: Cr 463 \#5 & $ 11859\pm 230$ & $ 4.10\pm 0.19$ & $-0.59\pm 0.09$ & $ 2.0\pm 0.4$ & $0.0170\pm0.0002$ & $ 0.30\pm 0.01$ \\
2146: \enddata
2147: \tablenotetext{a}{For the O stars analyzed using the TLUSTY atmosphere models, the values of $T_{eff}$ were adopted from the spectral type vs. $T_{eff}$ relation given in Table 2. These stars can be identified by their 1-$\sigma$ uncertainties, which are $\pm$1000 K.}
2148:
2149: \tablenotetext{b}{For stars in clusters, the surface gravities are determined as discussed in \S 3.1 and rely on stellar evolution models and cluster distance determinations. Surface gravities for non-cluster stars are not always well-determined, because of a lack of specific spectroscopic indicators. In some cases, the best-fit solutions for these stars indicated physically unlikely results (i.e., $\log g >$ $\sim$4.3 or $\log g <$ $\sim$3.0). For these stars, a value of $\log g$ = 3.9 was assumed (which is the mean $\log g$ of the rest of the sample) and a 1-$\sigma$ uncertainty of $\pm$0.2 was incorporated in the error analysis. These cases can be identified by $\log g$ entries of ``$3.9\pm0.2$.''}
2150:
2151: \tablenotetext{c}{For the O stars in the sample, our fitting procedure utilized solar abundance TLUSTY models. For these stars the values of [m/H] are indicated by entries of ``0'' without uncertainties.}
2152:
2153: \tablenotetext{d}{For the O stars, the adopted TLUSTY models incorporate $v_{turb} = 10$ km\,s$^{-1}$. For these stars the values of $v_{turb}$ are indicated by entries of ``10'' without uncertainties. For the B stars, which were modeled using ATLAS9 models, the values of $v_{turb}$ were determined by the fitting procedure, but were constrained to lie between 0 and 10 km\,s$^{-1}$. Stars whose best-fit SED models required these limiting values are indicated by $v_{turb}$ entries of ``0'' or ``10'', without error bars. The uncertainties for stars with best-fit $v_{turb}$ values close to these limits may be underestimated due to this truncation.}
2154: \end{deluxetable}
2155: \clearpage
2156:
2157: %% TABLE 4
2158:
2159: \topmargin 0.6in
2160: \begin{deluxetable}{lccrcrccccrrccc}
2161: \rotate
2162: \tabletypesize{\scriptsize}
2163: \tablenum{4}
2164: \tablewidth{0pc}
2165: \tablecaption{Best-Fit Extinction Curve Parameters for Survey Stars
2166: [The complete version of this table is in the electronic edition of the
2167: Journal. The printed edition contains only a sample.]
2168: \label{tabEXTINCT}}
2169: \tablehead{
2170: \colhead{} &
2171: \multicolumn{7}{c}{UV Coefficients\tablenotemark{a}} &
2172: \colhead{} &
2173: \multicolumn{3}{c}{Optical Spline Points\tablenotemark{b}} &
2174: \colhead{} &
2175: \multicolumn{2}{c}{IR Coefficients\tablenotemark{c}} \\ \cline{2-8} \cline{10-12} \cline{14-15}
2176: \colhead{Star} &
2177: \colhead{$x_0$} &
2178: \colhead{$\gamma$} &
2179: \colhead{$c_1$} &
2180: \colhead{$c_2$} &
2181: \colhead{$c_3$} &
2182: \colhead{$c_4$} &
2183: \colhead{$c_5$} &
2184: \colhead{} &
2185: \colhead{$O_1$} &
2186: \colhead{$O_2$} &
2187: \colhead{$O_3$} &
2188: \colhead{} &
2189: \colhead{$R(V)$} &
2190: \colhead{$k_{IR}$} }
2191: \startdata
2192: HD698 & $ 4.551\pm 0.006$ & $ 0.96\pm 0.03$ & $ 0.07\pm 0.19$ & $ 0.99\pm 0.05$ & $ 2.95\pm 0.19$ & $ 0.15\pm 0.05$ & $ 6.51\pm 0.29$ & & $ 2.38\pm 0.09$ & $ 1.33$ & $ 0.00$ & & $ 3.94\pm 0.16$ & $ 1.70\pm 0.18$ \\
2193: HD3191 & $ 4.636\pm 0.003$ & $ 0.94\pm 0.02$ & $-0.79\pm 0.20$ & $ 1.00\pm 0.04$ & $ 2.99\pm 0.16$ & $ 0.31\pm 0.02$ & $ 5.76\pm 0.09$ & & $ 2.04\pm 0.06$ & $ 1.33$ & $ 0.01$ & & $ 2.81\pm 0.09$ & $ 0.93\pm 0.14$ \\
2194: BD+57 245 (NGC 457) & $ 4.561\pm 0.005$ & $ 0.89\pm 0.03$ & $-0.47\pm 0.31$ & $ 0.88\pm 0.06$ & $ 3.13\pm 0.19$ & $ 0.24\pm 0.03$ & $ 5.71\pm 0.16$ & & $ 2.05\pm 0.06$ & $ 1.31$ & $ 0.00$ & & $ 2.97\pm 0.13$ & $ 0.96\pm 0.17$ \\
2195: BD+57 252 (NGC 457) & $ 4.577\pm 0.005$ & $ 0.92\pm 0.02$ & $-0.66\pm 0.25$ & $ 0.95\pm 0.05$ & $ 3.39\pm 0.17$ & $ 0.17\pm 0.02$ & $ 5.48\pm 0.23$ & & $ 2.15\pm 0.06$ & $ 1.32$ & $ 0.00$ & & $ 2.97\pm 0.10$ & $ 0.98\pm 0.15$ \\
2196: NGC 457 Pesch 34 & $ 4.579\pm 0.003$ & $ 0.91\pm 0.01$ & $-0.55\pm 0.20$ & $ 0.86\pm 0.04$ & $ 3.32\pm 0.11$ & $ 0.31\pm 0.02$ & $ 6.14\pm 0.09$ & & $ 1.99\pm 0.07$ & $ 1.30$ & $ 0.01$ & & $ 2.94\pm 0.11$ & $ 0.84\pm 0.16$ \\
2197: NGC 457 Pesch 13 & $ 4.587\pm 0.004$ & $ 0.87\pm 0.02$ & $-0.69\pm 0.33$ & $ 0.84\pm 0.07$ & $ 2.91\pm 0.14$ & $ 0.26\pm 0.02$ & $ 5.59\pm 0.17$ & & $ 2.03\pm 0.07$ & $ 1.31$ & $ 0.01$ & & $ 3.11\pm 0.11$ & $ 0.90\pm 0.16$ \\
2198: NGC 457 Pesch 9 & $ 4.578\pm 0.006$ & $ 1.07\pm 0.03$ & $-1.13\pm 0.22$ & $ 1.08\pm 0.04$ & $ 4.31\pm 0.22$ & $ 0.18\pm 0.02$ & $ 5.00\pm 0.18$ & & $ 2.37\pm 0.05$ & $ 1.35$ & $ 0.01$ & & $ 2.76\pm 0.07$ & $ 0.79\pm 0.12$ \\
2199: Cr 463 \#18 & $ 4.606\pm 0.009$ & $ 1.18\pm 0.05$ & $-0.47\pm 0.26$ & $ 1.03\pm 0.07$ & $ 5.23\pm 0.45$ & $ 0.30\pm 0.03$ & $ 5.09\pm 0.20$ & & \nodata & $ 1.37$ & $ 0.01$ & & $ 3.38\pm 0.17$ & $ 1.29$ \\
2200: BD+70 131 (Cr 463) & $ 4.577\pm 0.011$ & $ 1.11\pm 0.05$ & $ 0.17\pm 0.32$ & $ 0.86\pm 0.08$ & $ 5.11\pm 0.49$ & $ 0.35\pm 0.04$ & $ 5.00\pm 0.19$ & & \nodata & $ 1.37$ & $ 0.00$ & & $ 3.40\pm 0.20$ & $ 1.30$ \\
2201: Cr 463 \#5 & $ 4.610\pm 0.008$ & $ 1.08\pm 0.04$ & $ 0.45\pm 0.33$ & $ 0.74\pm 0.08$ & $ 4.77\pm 0.34$ & $ 0.32\pm 0.03$ & $ 5.06\pm 0.18$ & & \nodata & $ 1.37$ & $ 0.00$ & & $ 2.91\pm 0.18$ & $ 0.99$ \\
2202: \enddata
2203: \tablenotetext{a}{For the stars HD237019, HD18352, and HD25443 the long wavelength {\it IUE} spectra are incomplete. For these cases we constrained the UV linear extinction component to follow the relation $c_1 = 2.18 - 2.91 c_2$ from Fitzpatrick 2004. For these stars we list uncertainties for the $c_2$ values but not for the $c_1$ values.}
2204:
2205: \tablenotetext{b}{The uncertainties in the $O_2$ and $O_3$ optical spline points (at wavelengths of 4000 \AA\/ and 5530 \AA, respectively) are typically 0.01 or less and are not listed. For several stars --- those without $U$ band photometry --- we did not solve for the $O_1$ point at 3300 \AA.}
2206:
2207: \tablenotetext{c}{For field stars without IR photometry, we assumed $R(V) = 3.1$ and $k_{IR} = 1.11$, with the latter based on the relation $k_{IR} = 0.63R(V) - 0.84$ from Fitzpatrick 2004. For such stars in clusters, we adopted the mean $R(V)$ of the other cluster members and a value of $k_{IR}$ based on the aforementioned relation. These assumed values are listed in the Table without uncertainties. Several survey stars have apparently noisy {\it JHK} data and yielded very uncertain values of $k_{IR}$. For these, we ultimately derived the extinction curve by solving for the best-fit value of $R(V)$ with $k_{IR}$ constrained to follow the Fitzpatrick 2004 relation. The resultant $R(V)$ values are listed with their uncertainties while the $k_{IR}$ values are listed without uncertainties.}
2208: \end{deluxetable}
2209: \clearpage
2210:
2211: %% TABLE 5
2212:
2213: \topmargin 0.0in
2214: \begin{deluxetable}{cl}
2215: \tablenum{5}
2216: \tablewidth{0pc}
2217: \tablecaption{Extinction Curve Parameters for Mean Curve in Figure 9
2218: \label{tabAVGCURVE}}
2219: \tablehead{
2220: \colhead{Parameter} &
2221: \colhead{Value} }
2222: \startdata
2223: $x_0$ & \phm{$-$}4.592 \invmic \\
2224: $\gamma$ & \phm{$-$}0.922 \invmic \\
2225: $c_1$ & $-0.175$ \\
2226: $c_2$ & \phm{$-$}0.807 \\
2227: $c_3$ & \phm{$-$}2.991 \\
2228: $c_4$ & \phm{$-$}0.319 \\
2229: $c_5$ & \phm{$-$}6.097 \\
2230: $O_1$ & \phm{$-$}2.055 \\
2231: $O_2$ & \phm{$-$}1.322 \\
2232: $O_3$ & \phm{$-$}0.000 \\
2233: $R(V)$ & \phm{$-$}3.001 \\
2234: $k_{IR}$ & \phm{$-$}1.057 \\
2235: \enddata
2236: \end{deluxetable}
2237: \clearpage
2238:
2239:
2240: %%%%%%%%%%%%%%%%%%%%%%%%%%
2241: % FIGURES
2242: %%%%%%%%%%%%%%%%%%%%%%%%%%
2243:
2244: \clearpage
2245:
2246: \begin{figure}[ht]
2247: \epsscale{1.0}
2248: \plotone{f1.eps}
2249: %\plotone{../FIGURES/fig_COORDS.eps}
2250: \caption{Sky distribution of survey stars, shown in a sinusoidal
2251: projection of Galactic coordinates. The Galactic plane is the solid
2252: horizontal line and the Galactic center is in the middle of the
2253: figure. The locations of open clusters or associations containing five
2254: or more survey stars are indicated by the large circles, labeled with
2255: the cluster name. The sizes of the circles do not represent the
2256: physical extent of the clusters or associations.
2257: \label{figCOORDS}}
2258: \end{figure}
2259: \clearpage
2260:
2261: \begin{figure}[ht]
2262: \epsscale{0.50}
2263: \plotone{f2.eps}
2264: %\plotone{../FIGURES/fig_STARSTATS.eps}
2265: \caption{Representative properties of the survey stars. The three
2266: panels show results from the analysis presented in \S \ref{secATLAS},
2267: which help characterize the properties of our sample. The typical
2268: survey member is a mid-B star (top panel), with a median reddening of
2269: $\overline{\ebv} = 0.45$ mag (middle panel), viewed along a
2270: line-of-sight passing through the diffuse ISM (bottom panel). The
2271: bottom panel also shows a Gaussian fit to the values of $R(V) \equiv
2272: A_V/\ebv$ in the neighborhood of the peak in the distribution. The
2273: peak is located at $R(V) = 2.99$, similar to the mean values usually
2274: attributed to the diffuse ISM, and the width of the Gaussian
2275: corresponds to $\sigma = \pm 0.27$. The mean and median values of
2276: $R(V)$ for the whole sample are 3.22 and 3.05, respectively.
2277: \label{figSTARSTATS}}
2278: \end{figure}
2279: \clearpage
2280:
2281:
2282: \begin{figure}[ht]
2283: \epsscale{1.0}
2284: \plotone{f3.eps}
2285: %\plotone{../FIGURES/fig_EXTCRV.eps}
2286: \caption{A parametrized representation of normalized UV-through-IR
2287: extinction (solid and dashed curve). The curve consists of two parts:
2288: 1) $\lambda \leq 2700$ \AA\/ (shaded region), where we adopt a modified
2289: version of the 3-component parametrization scheme of Paper III; and 2)
2290: $\lambda > 2700$ \AA, where we adopt a cubic spline interpolation
2291: through sets of IR ($I_n$), optical ($O_n$), and UV ($U_n$) ``anchor
2292: points.'' The values of the anchor points and the seven parameters
2293: describing the UV portion of the curve are determined by fitting the
2294: observed SED of a reddened star, as described in \S \ref{secMETHOD}.
2295: The particular curve shown in this figure corresponds to that derived
2296: for the star HD147933.
2297: \label{figEXTCRV}}
2298: \end{figure}
2299: \clearpage
2300:
2301:
2302: \begin{figure}[ht]
2303: \epsscale{0.70}
2304: \plotone{f4.eps}
2305: %\plotone{../FIGURES/fig_ATLASCURVES.eps}
2306: \caption{Normalized extinction curves for 328 Galactic stars derived
2307: using the extinction-without-standards approach. The symbols show the
2308: normalized ratios of the model atmosphere fluxes to 1) the \iue\/
2309: spectrophotometry in the UV ($\lambda^{-1} > 3.3 \; \invmic$), 2) Johnson
2310: {\it UBV} photometry in the optical, and 3) Johnson or \tmass\/ {\it
2311: JHK} photometry in the near-IR ($\lambda^{-1} < 1 \; \invmic$).
2312: Individual 1-$\sigma$ error bars are shown for the data points, but are
2313: typically only visible in the region of the 2175 \AA\/ bump and in the
2314: far-UV, where the signal level of the \iue\/ data is lowest. Small
2315: crosses indicate \iue\/ data points excluded from the fit, for the
2316: reasons discussed in Paper IV. \iue\/ data points in the region $1215 \leq
2317: \lambda \leq 1235$ \AA\/ are excluded due to contamination from
2318: scattered solar Ly$\alpha$ photons. The solid curves are the
2319: parametrized fits to the data as determined by the SED-fitting
2320: procedure discussed in \S \ref{secMETHOD}, and the shaded regions show
2321: the 1-$\sigma$ uncertainty in the curves, based on Monte Carlo
2322: simulations. For comparison, the dash-dot curves show an estimate of
2323: the average Galactic extinction curve from F99 (corresponding to $R(V)
2324: = 3.1$). Only the first panel of the figure is shown in the print edition
2325: of the Journal. The entire figure consists of 33 panels and is available
2326: in the electronic edition.
2327: \label{figATLAS}}
2328: \end{figure}
2329: \clearpage
2330:
2331: \begin{figure}[ht]
2332: \epsscale{0.8}
2333: \plotone{f5.eps}
2334: %\plotone{../FIGURES/fig_SPTEFF.eps}
2335: \caption{$\log T_{eff}$ as a function of spectral type for the survey
2336: stars of type B0 or later. Luminosity class V or IV stars are shown by
2337: filled circles; higher luminosity classes by open circles. For
2338: comparison, data for the 44 unreddened stars from the photometric
2339: calibration study of Fitzpatrick \& Massa (2005b) are shown by the open
2340: squares. The various dotted, dashed, and solid lines show a number of
2341: published spectral type vs. $T_{eff}$ calibrations, as indicated in the
2342: figure. Small random horizontal offsets have been added to the data
2343: points to increase their visibility.
2344: \label{figSPTEFF}}
2345: \end{figure}
2346: \clearpage
2347:
2348: \begin{figure}[ht]
2349: \epsscale{0.8}
2350: \plotone{f6.eps}
2351: %\plotone{../FIGURES/fig_COMPARE.eps}
2352: \caption{UV spectra of two survey stars, HD228969 and BD+54 973,
2353: compared with three early-type spectral classification standard stars.
2354: The reddened survey stars are classified as types B2 II: and B3 V,
2355: respectively, although the SED-fitting procedure indicates temperatures
2356: in the neighborhood of 30000 K. Comparison of the UV spectral features
2357: with the classification standards indicates that these two stars are
2358: likely misclassified members of earlier spectral classes. Much of the
2359: scatter in the \teff\/ vs. spectral type diagram in Figure
2360: \ref{figSPTEFF} is probably the result of such classification
2361: uncertainties.
2362: \label{figCOMPARE}}
2363: \end{figure}
2364: \clearpage
2365:
2366: \begin{figure}[ht]
2367: \epsscale{0.8}
2368: \plotone{f7.eps}
2369: %\plotone{../FIGURES/fig_CBRACKET.eps}
2370: \caption{\teff\/ as a function of the Str\"{o}mgren reddening-free
2371: index $[c] \equiv c_1 - 0.20 (b-y)$, which measures the strength of the
2372: Balmer jump, for 162 sample stars. Symbols are the same as for Figure
2373: \ref{figSPTEFF}, with the addition of the open circles which denote
2374: O-type stars. Also shown are the spectral type vs. \teff\/
2375: calibrations from Figure \ref{figSPTEFF}, transformed into the \teff\/
2376: vs. $[c]$ plane using Crawford's (1978) spectral type vs. $c_0$ and
2377: $c_0$ vs. $(b-y)_0$ relations. Note that Str\"{o}mgren photometry was not
2378: used in modeling the SEDs of our survey stars.
2379: \label{figCBRACKET}}
2380: \end{figure}
2381: \clearpage
2382:
2383: \begin{figure}[ht]
2384: \epsscale{0.8}
2385: \plotone{f8.eps}
2386: %\plotone{../FIGURES/fig_ALLCURVES.eps}
2387: \caption{Analytical representations of 328 Galactic UV-through-IR
2388: extinction curves plotted in their native normalization
2389: $E(\lambda-V)/E(B-V)$ ({top panel}) and transformed into an
2390: IR-normalized form $A(\lambda)/A(1\/\mu{\rm m})$ ({bottom panel}). The
2391: curves, in their standard form, can be reproduced from the parameters
2392: given in Table \ref{tabEXTINCT}. The convergence of the curves in the
2393: optical region (top panel) and the IR region (bottom panel) results
2394: from their normalizations. The value of $R(V) \equiv A(V)/E(B-V)$ for
2395: each curve is the negative of its intercept at $\lambda^{-1} = 0 \;
2396: \invmic$ in the $E(\lambda-V)/E(B-V)$ panel.
2397: \label{figALLCURVES}}
2398: \end{figure}
2399: \clearpage
2400:
2401: \begin{figure}[ht]
2402: \epsscale{1.0}
2403: \plotone{f9.eps}
2404: %\plotone{../FIGURES/fig_AVGCURVE.eps}
2405: \caption{An average UV-through-IR extinction curve for our sample
2406: compared with other mean extinction curves. The thick solid curve is
2407: the mean for the 243 stars in our sample with $2.4 < R(V) < 3.6$, i.e.,
2408: for those sightlines with $R(V)$ values indicative of the diffuse ISM.
2409: The dark gray shaded region shows the sample variance about the mean
2410: curve at all wavelengths. The light grey shaded area shows the larger
2411: variance that results when we include all 298 sightlines with measured
2412: $R(V)$ values. The mean curve can be reconstructed from the extinction
2413: parameters listed in Table \ref{tabAVGCURVE}. The dashed and dotted
2414: curves show estimates of mean UV Galactic curves from the sources
2415: indicated in the figure. The large filled circles are means from the
2416: \ans\/ satellite extinction catalog of Savage et al. (1985) for 800
2417: stars with $\ebv \ge 0.20$ mag. These measurements result from filter
2418: photometry centered at wavelengths of 3000, 2500, 2200, 1800, and 1550
2419: \AA. Error bars on the \ans\/ data show the sample variances and
2420: include the effects of spectral mismatch.
2421: \label{figAVGCURVE}}
2422: \end{figure}
2423: \clearpage
2424:
2425: \begin{figure}[ht]
2426: \epsscale{1.0}
2427: \plotone{f10.eps}
2428: %\plotone{../FIGURES/fig_IUECURVE.eps}
2429: \caption{Unweighted mean UV extinction curve for 318 survey stars (open
2430: circles). Ten stars (HD14092, BD+56 501, HD14321, HD18352, HD25443,
2431: CD-42 4819, HD99872, HD326328, HD197702, and HD239710) were excluded
2432: from the mean because they have incomplete \iue\/ spectra. The
2433: individual data points (shown in Figure \ref{figATLAS}) are typically
2434: 10 \AA\/ apart. The solid curve is
2435: a parametrized fit to the mean curve, using the formulation discussed
2436: in \S \ref{secMETHOD}. The $O-C$ residuals are shown as filled points;
2437: they are offset from their zero mean for display. A number of distinct
2438: features are seen in the residuals and labeled in the figure. These arise
2439: from ISM gas-phase absorption lines, mismatch between the
2440: C~{\sc iv}~$\lambda$1550 stellar wind lines in the O stars and the static
2441: model SEDs, and known inadequacies in the \atlas\/ opacity distribution
2442: functions (labeled ``b'' in the figure; see Paper IV). No other credible
2443: features are seen in the residuals. The standard deviation of the
2444: residuals about their mean value of zero (excluding the labeled points) is
2445: $0.06\ebv$ mag, corresponding to $\sim0.02A(V)$ mag. This figure
2446: demonstrates both the intrinsic smoothness of UV extinction and the
2447: ability of our parametrization scheme to reproduce the shape of UV
2448: extinction curve to extremely high accuracy.
2449: \label{figIUECURVE}}
2450: \end{figure}
2451: \clearpage
2452:
2453: \topmargin -0.5in
2454: \begin{figure}[ht]
2455: \epsscale{0.9}
2456: \plotone{f11.eps}
2457: %\plotone{../FIGURES/fig_LONGITUDE.eps}
2458: \caption{Spatial trends in extinction properties. The values of four
2459: parameters which describe the shapes of interstellar extinction curves
2460: are plotted against Galactic longitude for each of our survey
2461: sightlines. Small filled circles show field sightlines. Other symbols
2462: denote sightlines to clusters or associations for which five or more
2463: members are included in our survey. From left to right, the clusters
2464: are: h \& $\chi$ Per (x's); NGC 457 (open circles); Cr 463 (filled
2465: diamonds); Cep OB3 (filled squares); Trumpler 37 (filled triangles);
2466: NGC 6530 (open squares); NGC 6231 (*'s); NGC 4755 (open diamonds);
2467: Carina clusters (large filled circles, includes Trumpler 14 and 16, Cr
2468: 228, and NGC 3293); NGC 1977 (open triangles); and NGC2244 (+'s). The
2469: dashed lines show the parameter values corresponding to the diffuse
2470: mean ISM curve from Figure \ref{figAVGCURVE}.
2471: \label{figLONGITUDE}}
2472: \end{figure}
2473: \clearpage
2474:
2475: \begin{figure}[ht]
2476: \epsscale{0.9}
2477: \plotone{f12.eps}
2478: %\plotone{../FIGURES/fig_DENSITY.eps}
2479: \caption{Density trends in extinction properties. The values of four
2480: parameters which describe the shapes of interstellar extinction curves
2481: are plotted against \ebv$/d$ for our survey sightlines.
2482: Small filled circles are field sightlines. Other symbols (same as in
2483: Figure~\ref{figLONGITUDE}) denote sightlines to clusters or associations
2484: containing five or more survey stars. The dashed lines show the parameter
2485: values corresponding to the diffuse mean ISM curve in
2486: Figure~\ref{figAVGCURVE}.
2487: \ref{figAVGCURVE}.
2488: \label{figDENSITY}}
2489: \end{figure}
2490: \clearpage
2491:
2492: \topmargin 0.0in
2493: \begin{figure}[ht]
2494: \epsscale{1.1}
2495: \plotone{f13.eps}
2496: %\plotone{../FIGURES/fig_C1C2.eps}
2497: \caption{Plot of the slope $c_2$ vs. the intercept $c_1$ for the
2498: linear component of UV extinction. The uncertainties in $c_1$ and
2499: $c_2$ are strongly correlated, as indicated by the 1-$\sigma$ error
2500: bars (dotted lines). The orientation of the error bars was determined
2501: by fitting ellipses to the loci of results from the Monte Carlo error
2502: simulations. The individual sets of error bars are not orthogonal to
2503: each other because the scales of the x- and y-axes are different. The
2504: solid line is a weighted linear fit to the data, designed to minimize
2505: the residuals in the direction perpendicular to the fit. It is given by
2506: $c_1 = 2.09 - 2.84 c_2$.
2507: \label{figC1C2}}
2508: \end{figure}
2509: \clearpage
2510:
2511: \begin{figure}[ht]
2512: \epsscale{1.1}
2513: \plotone{f14.eps}
2514: %\plotone{../FIGURES/fig_RVKIR.eps}
2515: \caption{Plot of $R(V)$ vs. the IR scale factor $k_{IR}$ (see
2516: Eq.~[\ref{eqnIR}]). The uncertainties in $R(V)$ and $k_{IR}$ are
2517: strongly correlated, as indicated by the 1-$\sigma$ error bars (dotted
2518: lines). The orientation of the error bars was determined by fitting
2519: ellipses to the loci of results from the Monte Carlo error
2520: simulations. The solid line is a weighted linear fit to the data,
2521: designed to minimize the residuals in the direction perpendicular to
2522: the fit. It is given by $k_{IR} = -0.83 + 0.63 R(V)$.
2523: \label{figRKIR}}
2524: \end{figure}
2525: \clearpage
2526:
2527: \begin{figure}[ht]
2528: \epsscale{1.0}
2529: \plotone{f15.eps}
2530: %\plotone{../FIGURES/fig_RVEVMINK.eps}
2531: \caption{Plot of $E(V-K)/E(B-V)$ vs. $R(V)$ for sightlines with \tmass\/
2532: $K$-band photometry (filled circles) and Johnson $K$-band photometry
2533: (open circles). The dotted line shows the widely-used relation $R(V) =
2534: 1.1E(V-K)/E(B-V)$. The solid line is a linear fit to the \tmass\/ data,
2535: minimizing the residuals in the direction perpendicular to the fit. It
2536: is given by $R(V) = -0.26 + 1.19E(V-K)/E(B-V)$.
2537: \label{figEVMINK}}
2538: \end{figure}
2539: \clearpage
2540:
2541: \begin{figure}[ht]
2542: \epsscale{1.1}
2543: \plotone{f16.eps}
2544: %\plotone{FIGURES/fig_GAMMAX0.eps}
2545: \caption{Plot of the FWHM of the 2175 \AA\/ bump $\gamma$ vs. its
2546: central position, $x_0$, both in units of \invmic. Dotted lines
2547: show the 1-$\sigma$ error bars, revealing the uncertainties in
2548: $\gamma$ and $x_0$ to be generally uncorrelated. The orientation of
2549: the error bars was determined by fitting ellipses to the loci of
2550: results from the Monte Carlo error simulations. The individual sets of
2551: error bars are not orthogonal to each other because the scales of the
2552: x- and y-axes are different. A number of sightlines which deviate most
2553: from the main distribution have been labeled.
2554: \label{figGAMMAX0}}
2555: \end{figure}
2556: \clearpage
2557:
2558: \begin{figure}[ht]
2559: \epsscale{1.1}
2560: \plotone{f17.eps}
2561: %\plotone{../FIGURES/fig_BUMPSTATS.eps}
2562: \caption{2175 \AA\/ bump statistics. The left panel shows a histogram
2563: of the distribution of the 2175 \AA\/ bump peak positions for our
2564: sample. The smooth curve is a Gaussian fit centered at $x_0 = 4.5903
2565: \; \invmic$ (2178.5 \AA) with a width of $\sigma = 0.0191 \; \invmic$
2566: (9.1 \AA). The RMS value of the measurement errors in $x_0$ is $\pm
2567: 0.0058 \; \invmic$ ($\pm 2.8 \; {\rm \AA}$). The dashed curve shows a
2568: Gaussian fit to the distribution of bump positions for 154 stars
2569: located in the 13 open clusters with more than 5 stars in the survey.
2570: For each cluster, a mean value of $x_0$ was computed and the
2571: distributions constructed relative to the mean (see \S
2572: \ref{secEXT_X0GAMMA}). The width of the cluster Gaussian is given by
2573: $\sigma = 0.011 \; \invmic$ (5.2 \AA). The right panel shows the
2574: distribution of bump FWHM values. The smooth curve is a Gaussian fit
2575: to the main peak in the distribution. Its central position corresponds
2576: to $\gamma = 0.890 \; \invmic$ and its width is $\sigma = 0.050 \;
2577: \invmic$. The RMS measurement error in $\gamma$ is $\pm 0.031 \;
2578: \invmic$.
2579: \label{figBUMPSTATS}}
2580: \end{figure}
2581: \clearpage
2582:
2583: \begin{figure}[ht]
2584: \epsscale{1.0}
2585: \plotone{f18.eps}
2586: %\plotone{../FIGURES/fig_RVAV.eps}
2587: \caption{Plots of the normalized extinction values $A(\lambda)/A(V)$
2588: vs. $R(V)^{-1}$ at four different UV wavelengths. $A(\lambda)/A(V)$ is
2589: derived from the measured values of $E(\lambda-V)/E(B-V)$ by using
2590: Equation (\ref{eqnCCM2}). 1-$\sigma$ error bars are
2591: shown, based on our Monte Carlo simulations. The error bars, which are
2592: not orthogonal to each other because the scales of the x- and y-axes
2593: are different, show the clear correlation in the uncertainties for
2594: $A(\lambda)/A(V)$ and $R(V)^{-1}$. The straight line in the 2695 \AA\/
2595: panel is a weighted linear fit to the data, which minimizes the
2596: scatter perpendicular to the relation. It is given by $A(2695 \;
2597: {\rm \AA})/A(V) = 0.58 + 4.73 \; R(V)^{-1}$.
2598: \label{figRVAV}}
2599: \end{figure}
2600: \clearpage
2601:
2602: \begin{figure}[ht]
2603: \epsscale{1.0}
2604: \plotone{f19.eps}
2605: %\plotone{../FIGURES/fig_RVKLAM.eps}
2606: \caption{Plots of the measured extinction values $E(\lambda-V)/E(B-V)$
2607: vs.\ $R(V)$ at four different UV wavelengths. 1-$\sigma$ error bars
2608: are shown, based on our Monte Carlo simulations as in previous
2609: figures. Any true relationships seen in plots of $A(\lambda)/A(V)$
2610: vs.\ $R(V)^{-1}$ are preserved in plots of $E(\lambda-V)/E(B-V)$
2611: vs.\ $R(V)$, but with the great benefit that the correlations between
2612: the errors are greatly reduced. The solid line in the 2695 \AA\/ panel
2613: is a weighted linear fit to the data, which minimizes the scatter
2614: perpendicular to the relation. This fit is $E(2695 \; {\rm \AA} \;
2615: -V)/E(B-V) = 4.80 - 0.44 \; R(V)$. The nearly coincident dashed line
2616: is the fit in Figure~\ref{figRVAV} transformed by Equation
2617: (\ref{eqnCCM3}). This fit is $E(2695 \; {\rm \AA} \; -V)/E(B-V) = 4.73
2618: - 0.42 \; R(V)$.
2619: \label{figRVKLAM}}
2620: \end{figure}
2621: \clearpage
2622:
2623: \topmargin -0.5in
2624: \begin{figure}[ht]
2625: \epsscale{0.8}
2626: \plotone{f20.eps}
2627: %\plotone{../FIGURES/fig_RVFM.eps}
2628: \caption{Plots of UV fit parameters (and derived quantities) vs.
2629: $R(V)$. 1-$\sigma$ error bars are shown, based on our Monte Carlo
2630: simulations as in previous figures. The quantity $\Delta$1250 is the
2631: difference between the observed value of $E(\lambda-V)/E(B-V)$ at 1250
2632: \AA\/ (8.0 \invmic) and an extrapolation of the linear plus bump
2633: components of UV extinction. It thus measures the strength of the FUV
2634: curvature and is computed by $\Delta1250 = c_4(8.0-c_5)^2$ (see Eq.
2635: [\ref{eqnFMFUNC}]). $A_{bump} \equiv \pi \; c_3/(2\gamma)$ is the area
2636: under the 2175 \AA\/ bump. $E_{bump} \equiv c_3/\gamma^2$ is the
2637: height of the 2175 \AA\/ bump above the linear background extinction.
2638: \label{figRVFM}}
2639: \end{figure}
2640: \clearpage
2641:
2642: \topmargin -0.0in
2643: \begin{figure}[ht]
2644: \epsscale{0.8}
2645: \plotone{f21.eps}
2646: %\plotone{../FIGURES/fig_RVC2.eps}
2647: \caption{Plot of the slope of the UV linear component $c_2$ vs.
2648: $R(V)^{-1}$. 1-$\sigma$ error bars are shown, based on our Monte Carlo
2649: simulations as in previous figures. The solid line is a weighted
2650: linear fit which minimizes the residuals in the direction perpendicular
2651: to the fit. It is given by $R(V)^{-1} = 0.17 + 0.20 \; c_2$.
2652: \label{figRVC2}}
2653: \end{figure}
2654:
2655: \end{document}
2656:
2657: