1: \documentclass[12pt]{article} \usepackage{aaspp4,rotate} \parskip 0pt
2: %\documentclass{article} \usepackage{emulateapj,rotate,apjfonts}
3:
4: \begin{document}
5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6: \newif\ifFigures \Figurestrue
7: \def\Figure#1{\ifFigures{#1}\fi}
8: \def\eq#1{\begin{equation} #1 \end{equation}}
9: \def\E#1{\hbox{$10^{#1}$}}
10: \def\Ca {[25--12]}
11: \def\Cb {[60--25]}
12: \def\Cc {[100--60]}
13: \def\ccd{color--color diagram}
14: \def\ccds{color--color diagrams}
15: \def\about {\hbox{$\sim$}}
16: \def\la {\hbox{$\lesssim$}}
17: \def\ga {\hbox{$\gtrsim$}}
18: \def\x {\hbox{$\times$}}
19: \def\mic {\hbox{$\mu$m}}
20: \def\Lo {\hbox{$L_\odot$}}
21: \def\tV {\hbox{$\tau_V$}}
22: \def\half {\onehalf}
23: \def\threehalf {\slantfrac{3}{2}}
24: \def\Ref{\reference{}}
25: \def\HII{H{\small II}}
26: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
27:
28: \def\Msg{{\em ApJ Letters}, accepted for publication March 15, 2000}
29: %\submitted{\Msg} %
30: \rightline{\Msg}
31:
32: \title{INFRARED CLASSIFICATION OF GALACTIC OBJECTS}
33:
34: \author{\v{Z}eljko Ivezi\'c\altaffilmark{1} and
35: Moshe Elitzur\altaffilmark{2}}
36:
37: \altaffiltext{1} {Princeton University, Department of Astrophysical Sciences,
38: Princeton, NJ 08544--1001; ivezic@astro.Princeton.edu}
39: \altaffiltext{2} {Department of Physics and Astronomy, University of Kentucky,
40: Lexington, KY 40506--0055; moshe@pa.uky.edu}
41:
42: \begin{abstract}
43:
44: Unbiased analysis shows that IRAS data reliably differentiate between the
45: early and late stages of stellar evolution because objects at these stages
46: clearly segregate in infrared color-color diagrams. Structure in these
47: diagrams is primarily controlled by the density distribution of circumstellar
48: dust. The density profile around older objects is the steepest, declining as
49: $r^{-2}$, while young objects have profiles that vary as $r^{-3/2}$ and
50: flatter. The different density profiles reflect the different dynamics that
51: govern the different environments. Our analysis also shows that high mass star
52: formation is strongly concentrated within \about\ 5 kpc around the Galactic
53: center, in support of other studies.
54:
55: \end{abstract}
56:
57: \keywords{catalogs --- stars: AGB and post-AGB --- stars: formation
58: --- infrared: general --- Galaxy: stellar content --- Galaxy: structure}
59:
60: \section{INTRODUCTION}
61:
62: The IRAS all-sky survey provides a unique opportunity to classify the infrared
63: properties of astronomical objects from a homogeneous data set obtained with a
64: single facility. Cross-correlations of various catalogues with the IRAS Point
65: Source Catalogue (PSC) showed that certain Galactic objects tend to cluster in
66: well defined regions of IRAS \ccds. Notable examples include \HII\ regions
67: (e.g. Hughes \& MacLeod 1989; Wood \& Churchwell 1989 [WC]) and AGB stars (van
68: der Veen \& Habing 1988; VH). The reason for this clustering was not
69: understood, nor was it clear whether such biased analysis based on
70: pre-selection implies reliable selection criteria. From detailed modeling of
71: dusty winds we were able to validate the VH selection criterion proposed for
72: AGB stars and to explain its origin (Ivezi\'c \& Elitzur 1995; IE95). Here we
73: extend this approach to all Galactic PSC sources in an unbiased analysis of
74: IRAS fluxes without prior selections.
75:
76: \section{IRAS DATA AND ITS CLASSES}
77:
78: There are 6338 PSC sources with flux quality of at least 2 in all four IRAS
79: channels. Removing the 651 sources identified as extra-galactic (Beichman et
80: al 1985) produces our basic data set of 5687 Galactic IR objects. For AGB stars
81: we have shown (IE95) that a high IRAS quality does not guarantee the flux is
82: intrinsic to the point source itself; high-quality 60 and 100 \mic\ PSC fluxes
83: can have a cirrus origin instead. Figure 1 shows that the problem afflicts all
84: sources, not just AGB stars. Its x-axis is cirr3/$F_{60}$, a measure of the
85: ratio of cirrus background noise to the 60 \mic\ signal. Intrinsic fluxes
86: should have nothing to do with background emission, yet the \Cc\ color is
87: strongly correlated with cirr3/$F_{60}$ when this noise indicator exceeds a
88: certain threshold. The PSC fluxes of sources above this threshold reflect
89: cirrus, not intrinsic emission. Following IE95, we remove all sources with
90: cirr3 $> 2\,F_{60}$ to eliminate cirrus contamination. This leaves 1493
91: objects that can be considered a reliable representation of Galactic infrared
92: point sources.
93:
94: We submitted these data to the program AutoClass for an unbiased search of
95: possible structure. AutoClass employs Bayesian probability analysis to
96: automatically abstract a given data base into classes (Goebel et al 1989).
97: AutoClass detected that the PSC sources belong to four distinct classes that
98: occupy separate regions in the 4-dimensional space spanned by IRAS fluxes.
99: Figure 2a shows the 2-dimensional projection of this space onto the \Cc--\Ca\
100: color plane. The IRAS colors of black body emission, which are independent of
101: the black body temperature as long as it exceeds \about\ 700~K, are marked by
102: the large dot. There is hardly any data at that point. Instead, the data are
103: spread far away from it, indicating that IRAS fluxes are dominated by
104: surrounding dust. But rather than random scatter, the data show clear
105: structure, with the four AutoClass classes occupying well defined color
106: regions. As a further check on the reality of these classes we constructed
107: their Galactic distributions, shown in Figure 2b.
108:
109: Classes A and B clearly separate in the \ccd. Classes C and D, on the other
110: hand, are mixed together and are distinguished by their flux levels --- class C
111: fluxes are typically 10--100 times higher in all 4 bands. In principle, a
112: single family could produce such behavior if split into nearby and distant
113: objects, but this is not the case here. Extragalactic and heliocentric
114: selection effects are ruled out by the Galactic distributions. Both classes
115: are comprised of Galactic disk objects, but the median flux of class C is 16
116: times higher. If the two were drawn from the same population, class C sources
117: would be on the average 4 times closer and their Galactic latitude
118: distribution 4 times wider. Instead, the latitude histograms are essentially
119: the same, both are centered on $b = 0\arcdeg $ with full-width at half-maximum
120: $1.5\arcdeg \pm 0.3\arcdeg$ for class C and $1.5\arcdeg \pm 0.1\arcdeg$ for
121: class D. Class D fluxes vary by more than \E3, consistent with a population
122: distributed throughout the entire Galactic plane.\footnote{Class A fluxes have
123: a similar dynamic range.} In contrast, class C fluxes vary by a factor of only
124: \about\ 20 and exceed \E3 Jy at both 60 and 100 \mic\ for all sources. This is
125: consistent only with a population confined to, at most, \about\ 5 kpc around
126: the Galactic center with a minimal luminosity of \E3 \Lo. The Galactic
127: longitude distributions corroborate these conclusions. Although both class C
128: and D sources channel most of their luminosity to the far-IR, their spatial
129: and luminosity distributions are very different.
130:
131: \section{THEORY}
132:
133: In IE95 we identified the cause of the particular IRAS colors of AGB stars: The
134: dusty wind problem possesses general scaling properties. For a given dust
135: composition, the solutions are predominantly determined by a single input
136: parameter --- the overall optical depth. All other input is largely
137: irrelevant. As a result, the solutions occupy well defined regions in \ccds.
138: Indeed, the IRAS measured colors of AGB stars fall in the regions outlined by
139: wind solutions with visual optical depth \tV\ \la\ 100.
140:
141: %\Figure{\bigskip \centerline{\epsfclipon \epsfxsize=\hsize \epsfbox{fig1.ps}}}
142: %
143: %\figcaption[fig1.ps]{\Cc\ color vs.\ cirr3/$F_{60}$, where $F_{60}$ is the 60
144: %\mic\ IRAS flux and cirr3 is the IRAS surface brightness at 100 \mic\ around
145: %the point source. The flux quality for all sources is at least 2 in all four
146: %IRAS bands, and 3 for those marked with dark dots. In spite of the high flux
147: %quality, the strong correlation with the cirrus emission at large values of
148: %cirr3/$F_{60}$ is evident.}
149:
150: In Ivezi\'c \& Elitzur (1997; IE97) we showed that scaling is a general
151: property of radiatively heated dust under all circumstances, not just in AGB
152: winds. The most general dust radiative transfer problem contains only two
153: input quantities whose magnitudes matter --- \tV\ and the dust temperature
154: $T_1$ at some point. All other input is defined by dimensionless, normalized
155: profiles that describe (1) the spectral shape of the external radiation, (2)
156: the spectral shape of the dust absorption and scattering coefficients, and (3)
157: the dust spatial distribution. Physical dimensions such as, e.g., luminosity
158: and linear sizes are irrelevant. Scaling applies to arbitrary geometries, and
159: in IE97 we conducted extensive numerical studies of its consequences for
160: spherical shells. With a black body spectral shape for the heating radiation,
161: the temperature $T$ hardly affects IRAS colors when $T >$ 2000~K. The reason
162: is that $T$ is much higher than the Planck equivalent of the shortest IRAS
163: wavelength (12 \mic), which is only 240 K. Although this result was obtained
164: in spherical solutions, its validity is general. Since neither its luminosity
165: nor spectral shape matter, the heating source is quite irrelevant for the IRAS
166: colors. Different objects segregate in IRAS \ccds\ not because they have
167: different central sources, but rather because their dust shells are different.
168:
169: In our extensive study of spherical shells we examined also the effect of
170: $T_1$, which was selected as the dust temperature on the shell inner boundary.
171: As with the heating radiation, and for the same reasons, $T_1$ hardly affects
172: IRAS colors when varied over the plausible range of dust sublimation
173: temperatures 700--2000 K (IE97). This leaves the dust's composition, optical
174: depth and spatial distribution as the only potentially significant reasons for
175: the variation of IRAS colors among different families of Galactic objects.
176: However, dust properties generally do not show large variations among
177: different objects and thus cannot be expected to induce substantial variations
178: in infrared emission. The same applies to \tV, whose range of values is similar
179: in most families. We conclude that {\em objects segregate in IRAS \ccds\
180: primarily because their spatial dust distributions are different.}
181:
182: %\Figure{\bigskip \centerline{\epsfclipon \epsfxsize=\hsize \epsfbox{fig3.ps}}}
183: %
184: %\figcaption[fig3.ps]{\Cb--\Ca\ \ccd. Data are for the same sources presented
185: %in figure 2 using the same symbols. The full lines are model tracks for
186: %spherical dust shells with $T_1$ = 1000~K on their inner boundary and $r^{-p}$
187: %density profiles with $p$ = 0, \half, 1, \threehalf\ and 2, as marked.
188: %Position along each track increases with dust optical depth \tV\ away from the
189: %common origin \tV\ = 0 at the black-body colors, with \tV\ = 0.1 marked by
190: %solid triangle and \tV\ = 10 by solid square. The endpoint for each track is
191: %\tV\ = 100. Dashed-line tracks show the effect on the $p$ = 2 track of lowering
192: %$T_1$ to 300 K (lowermost track), 200 K, 150 K and 100 K (uppermost track).
193: %These tracks end at \tV\ = 10.}
194:
195: We corroborate these conclusions in detailed modeling with the code DUSTY
196: (Ivezi\'c, Nenkova \& Elitzur 1999). In these calculations, a point source
197: emitting as a black body with temperature 5000~K is surrounded by dust whose
198: absorption and scattering coefficients are those of standard interstellar mix.
199: The dust is distributed in a spherical shell whose temperature on the inner
200: boundary is $T_1$ = 1000~K. The shell density profile is taken as a power law
201: $r^{-p}$, with $p$ a free parameter. Scaling ensures that the solution set for
202: each $p$ is a one-parameter family characterized by \tV, leading to the
203: distinct tracks shown in the \Cb--\Ca\ \ccd\ in figure 3. These tracks
204: properly delineate the distribution of IRAS sources and we have verified that
205: the same applies also to the \Cc\ colors shown in fig.\ 2. Tracks for
206: different $p$ differ in the color-color regions that they cross and in the
207: distance induced by \tV\ variation; the steeper the density profile, the
208: smaller is the distance along the track for the same change in \tV. Both
209: properties arise from the differences in relative amounts of material placed
210: at different temperatures, and both are reflected in the data. Class A sources
211: are well explained by the $p$ = 2 track, indeed they cluster close to the
212: black-body point. In contrast, classes C and D are explained by tracks of
213: much flatter density distributions and are located rather far from the
214: black-body colors; it takes only \tV\ \about\ 0.1 to move an object along
215: these tracks all the way from the black body point to the region populated by
216: IRAS sources.
217:
218: Class B cannot be reasonably explained by the same models, even though its
219: colors can be produced by extending the $p$ = 2 track beyond \tV\ = 100. In
220: addition to the implausible optical depths this requires, class B is unlikely
221: to be the high-\tV\ end of class A given their different Galactic
222: distributions. An alternative to \tV-variation is to modify colors by removal
223: of hot dust. The displayed tracks have $T_1$ = 1000 K, as appropriate for
224: shells whose inner boundaries are controlled by dust sublimation, but in
225: detached shells $T_1$ is both lower and arbitrary. The dashed-line tracks in
226: figure 3 show the effect of lowering $T_1$ on the $p$ = 2 track. The
227: difference between $T_1$ = 1000 K and 300 K is marginal because both are
228: higher than the Planck equivalent of 12 \mic. However, further reduction in
229: $T_1$ alters the track significantly, adequately explaining class B colors.
230:
231: \section{DISCUSSION}
232:
233: Our modeling confirms that the primary reason for different IRAS classes is
234: different dust density distributions. Only 5\% of the sources require an
235: additional variation of the dust temperature on the shell inner boundary.
236: Spherical shells were employed here as the simplest method to model extended,
237: three dimensional dust distributions. Power laws were used just for
238: illustration purposes and although their actual values should not be taken
239: literally, they provide a good indication of the overall behavior of the
240: density distributions in the four IRAS classes. The presence of disks,
241: expected in various sources, should not significantly affect our conclusions.
242: The standard disk spectrum $F_\nu \propto \nu^{1/3}$ produces a single point
243: in \ccds\ and thus cannot effect the observed scatter. This spectrum is
244: modified if the disk is embedded in an extended shell, but then the disk is
245: expected to dominate only at sub-mm and mm wavelengths, longer than those
246: observed by IRAS (Miroshnichenko et al 1999). Flared disk emission can be
247: shown equivalent to that from a flat disk embedded in an appropriate spherical
248: shell. Therefore, such a configuration, too, cannot modify our conclusions.
249:
250: We have queried the SIMBAD database about our sample sources. SIMBAD
251: identifications are occasionally ambiguous (``maser"), sometimes less than
252: informative (``part of cloud"), and reliability is not always certain.
253: Nevertheless, they provide useful clues when there are major trends in the
254: data. Class A returned the most decisive results --- 88\% of its members have
255: possible optical identifications, of which roughly 90\% are commensurate with
256: AGB stars. Since class A obeys the VH criterion, this corroborates our earlier
257: finding (IE95) that this criterion is both sufficient and necessary for AGB
258: selection. In IE95 we present a thorough analysis of the IRAS fluxes of AGB
259: stars, including color tracks geared specifically for these objects (single
260: chemistry grains, appropriate stellar temperature, etc.). While that analysis
261: was necessary to verify the VH criterion, the $p$ = 2 track presented here
262: captures the essence of the more detailed study, demonstrating that the
263: density distribution is the leading factor in controlling IRAS colors. These
264: circumstellar shells have the steepest density distribution, setting them
265: apart in the \ccds. In addition, the color tracks are primarily controlled by a
266: single parameter, \tV, hence the compactness of this class color region.
267:
268: Class B had the same identification rate but its composition not as
269: homogeneous. Planetary nebulae comprise 40\% of positive and possible
270: identifications. At 13\%, the only other significant group is ``emission-line
271: stars", a classification consistent with planetary and reflection nebulae. The
272: remaining identifications span a variety of objects, indicating that detached
273: shells may occasionally form under different circumstances. Significantly, the
274: optical depths required for class B are lower than for the others, the
275: dashed-line tracks in fig.\ 3 terminate at \tV\ = 10.
276:
277: SIMBAD identification rates for the two other classes are much lower --- only
278: 32\% for class C and 21\% for D. Among the 38 class C identifications, \HII\
279: regions comprise the only significant group with 15 (40\%). In class D, too,
280: \HII\ regions comprise the single largest group with 23\% of the
281: identifications, followed by young stellar objects at 11\% and planetary
282: nebulae at 10\%. These two classes are clearly dominated by star formation and
283: early stages of stellar evolution, in agreement with their Galactic
284: distributions and with previous attempts to associate star-forming regions
285: with IRAS colors. In the most extensive study of this kind, 83\% of 1302 IRAS
286: selected sources were found to be embedded in molecular clouds and thus trace
287: star formation, and the selection criterion (\Ca\ $\ge$ 0, \Cb\ $\ge$ 0.4)
288: essentially identifies classes C and D (Wouterloot \& Brand 1989; Wouterloot
289: et al 1990).
290:
291: The IRAS colors of classes C and D imply dust density distributions flatter
292: than for AGB stars. These colors are spread over large regions, reflecting
293: variation in density profiles in addition to optical depth. The spread in \Cc\
294: colors is smaller than in \Cb\ because all shells are optically thin at both
295: 100 and 60 \mic\ while their 25 \mic\ optical depth can become significant
296: (IE97). Class C colors occupy a sub-region of class D and are produced by the
297: optically-thick end (\tV\ \ga\ 1) of flat density distributions. Objects whose
298: colors fall in that region can belong to either class C or D. However, among
299: all sources with class D colors, those with high fluxes ($>$ 1000 Jy at both
300: 60 and 100 \mic) concentrate in a compact color region, hence the separate
301: class C. Since all class C sources have $L >$ \E3 \Lo, they are high-mass
302: objects and their concentration in the inner \about\ 5 kpc of the Galaxy shows
303: that the high-mass star formation rate decreases with distance from the
304: Galactic center. This result is in agreement with studies of the initial mass
305: function inside and outside the solar circle (Garmany et al 1982; Wouterloot et
306: al 1995; Casassus et al 1999).
307:
308: Our results explain the findings of all earlier studies that were based on
309: object pre-selection, and reveal the limitations of that approach. By example,
310: consider the WC study. After identifying IRAS counterparts of known
311: ultracompact \HII\ regions, WC proposed the corresponding colors as a necessary
312: and sufficient selection for all ultracompact \HII\ regions and proceeded to
313: estimate the birthrate of O stars. Codella et al (1994) then found that most
314: \HII\ regions in a more extended catalog indeed obeyed the WC color criterion.
315: However, that included also diffuse, not just compact \HII\ regions, therefore
316: WC overestimated the O star birthrate. This clearly demonstrates the
317: shortcomings of any classification based on a pre-selected population. The
318: unbiased analysis presented here shows that IRAS colors reflect primarily the
319: dust density profiles of circumstellar shells and provide a unique indication
320: of the underlying object only in the case of AGB stars. IRAS data in itself is
321: sufficient for differentiating young and old stellar objects; apart from a
322: limited number of detached shells, IRAS sources belong to two distinct groups
323: as is evident from both the \ccds\ and the Galactic distributions: (1) class A
324: sources are at the late stages of stellar evolution and (2) class C and D
325: sources are objects at the early evolutionary stages. This differentiation
326: occurs because the density distributions of dust around young stellar objects
327: have flatter profiles, reflecting the different dynamics that govern the
328: different environments.
329:
330: \acknowledgments
331:
332: Support by NASA and NSF is gratefully acknowledged.
333:
334: \begin{references}
335:
336: \Ref Beichman, C.A., Neugebauer, G., Habing, H.J., Clegg, P.E.
337: \& Chester, T.J., 1985, {\em IRAS Catalogs and Atlases}
338: (US GPO, Washington, DC)
339: \Ref Casassus, S., Bronfman, L., May, J. \& Nyman, L-\AA, 1999, A\&A submitted
340: (astro-ph/9912340)
341: \Ref Codella, C., Felli, M. \& Natale, V., 1994, A\&A, 284, 233
342: \Ref Goebel, J., et al 1989, A\&A, 222, L5
343: \Ref Garmany, C.D., Conti, P.S. \& Chiosi, C., 1982, ApJ, 263, 777
344: \Ref Hughes, V. A. \& MacLeod, G. C. 1989, AJ, 97, 786
345: \Ref Ivezi\'c, \v Z., Elitzur M., 1995, ApJ, 445, 415 (IE95)
346: \Ref Ivezi\'c, \v Z., Elitzur M., 1997, MNRAS, 287, 799 (IE97)
347: \Ref Ivezi\'c, \v Z., Nenkova, M., Elitzur, M., 1999, User Manual for DUSTY,
348: University of Kentucky Internal Report, accessible at
349: http://www.pa.uky.edu/$\sim$moshe/dusty
350: \Ref Miroshnichenko, A., Ivezi\'c, \v Z., Vinkovi\'c, D. \& Elitzur M., 1999,
351: ApJ, 520, L115
352: \Ref van der Veen, W.E.C.J., \& Habing, H.J. 1988, A\&A, 194, 125 (VH)
353: \Ref Wood, D. O. S. \& Churchwell, E. 1989, ApJ, 340, 265 (WC)
354: \Ref Wouterloot, J.G.A. \& Brand, J., 1989, A\&AS, 80, 149
355: \Ref Wouterloot, J.G.A., Brand, J., Burton, W.B. \& Kwee, K.K., 1990,
356: A\&A, 230, 21
357: \Ref Wouterloot, J.G.A., Fiegle, K., Brand, J. \& Winnewisser, G., 1995,
358: A\&A, 301, 236
359:
360: \end{references}
361:
362: %\Figure{\epsfclipon \centerline{\epsfxsize=0.6\hsize
363: %\rotate[r]{\epsfbox{fig2.ps}}}}
364: %
365: %\figcaption[fig2.ps]{(a) \Cc--\Ca\ \ccd\ for all the Galactic IRAS sources free
366: %from cirrus contamination. The four classes identified by AutoClass are marked
367: %with different symbols: Class A (282 sources) --- crosses, B (80 sources) ---
368: %circles, C (119 sources) --- squares, D (1013 sources) --- dots. The large dot
369: %marked BB denotes the colors of black body emission with temperature $T > 700$
370: %K. (b) Galactic distributions of the four classes.}
371: %
372: %\end{document}
373:
374:
375: \newpage
376: %%%%%%%%%%%%%%%%%%% FIGURES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
377:
378: \figcaption[fig1.ps]{\Cc\ color vs.\ cirr3/$F_{60}$, where $F_{60}$ is the 60
379: \mic\ IRAS flux and cirr3 is the IRAS surface brightness at 100 \mic\ around
380: the point source. The flux quality for all sources is at least 2 in all four
381: IRAS bands, and 3 for those marked with dark dots. In spite of the high flux
382: quality, the strong correlation with the cirrus emission at large values of
383: cirr3/$F_{60}$ is evident.}
384: \bigskip
385:
386: \figcaption[fig2.ps]{(a) \Cc--\Ca\ \ccd\ for all the Galactic IRAS sources free
387: from cirrus contamination. The four classes identified by AutoClass are marked
388: with different symbols: Class A (282 sources) --- crosses, B (80 sources) ---
389: circles, C (119 sources) --- squares, D (1013 sources) --- dots. The large dot
390: marked BB denotes the colors of black body emission with temperature $T > 700$
391: K. (b) Galactic distributions of the four classes.}
392: \bigskip
393:
394: \figcaption[fig3.ps]{\Cb--\Ca\ \ccd. Data are for the same sources presented
395: in figure 2 using the same symbols. The full lines are model tracks for
396: spherical dust shells with $T_1$ = 1000~K on their inner boundary and $r^{-p}$
397: density profiles with $p$ = 0, \half, 1, \threehalf\ and 2, as marked.
398: Position along each track increases with dust optical depth \tV\ away from the
399: common origin \tV\ = 0 at the black-body colors, with \tV\ = 0.1 marked by
400: solid triangle and \tV\ = 10 by solid square. The endpoint for each track is
401: \tV\ = 100. Dashed-line tracks show the effect on the $p$ = 2 track of lowering
402: $T_1$ to 300 K (lowermost track), 200 K, 150 K and 100 K (uppermost track).
403: These tracks end at \tV\ = 10.}
404: \bigskip
405:
406: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
407:
408: \Figure{\bigskip \centerline{\epsfclipon \epsfxsize=\hsize \epsfbox{IE1.ps}}}
409:
410: \Figure{\bigskip \epsfclipon \epsfxsize=0.9\hsize \epsfbox{IE2.ps}}
411:
412: \Figure{\bigskip \centerline{\epsfclipon \epsfxsize=\hsize \epsfbox{IE3.ps}}}
413:
414: \end{document}
415: