1: \documentclass[12pt,preprint]{aastex}
2: %%\documentclass{emulateapj}
3: %% manuscript produces a one-column, double-spaced document:
4:
5: %%\documentclass[manuscript]{aastex}
6:
7: %% preprint2 produces a double-column, single-spaced document:
8:
9: %%\documentclass[preprint2]{aastex}
10:
11: \newcommand{\afe}{[\alpha/\mathrm{Fe}]}
12: \newcommand{\Ms}{M_{\odot}}
13:
14: \slugcomment{To appear in ApJ Letters}
15:
16: \shorttitle{HB Morphology and Mass Loss}
17: \shortauthors{A. Dotter}
18:
19: \begin{document}
20:
21: \title{Horizontal Branch Morphology and Mass Loss in Globular Clusters}
22:
23: \author{Aaron Dotter}
24: \affil{Department of Physics and Astronomy, University of Victoria,\\
25: P. O. Box 3055, Victoria, British Columbia, V8W 3P6, Canada}
26: \email{dotter@uvic.ca}
27:
28:
29: \begin{abstract}
30: The connection between mass loss on the red giant branch (RGB) and horizontal
31: branch (HB) morphology in globular clusters (GCs) has long been acknowledged
32: but the mechanisms governing mass loss remains poorly understood from a theoretical
33: perspective. The present study uses synthetic HB models to demonstrate for the first
34: time that $\alpha$-enhancement and a simple relation between mass loss and metallicity
35: can explain the entire range of HB morphology (characterized by the HB type index)
36: observed in old, coeval GCs. The mass loss-metallicity relation accounts naturally
37: for the fact that the most metal poor GCs ([Fe/H] $< -2$) have redder HBs than is
38: typical of GCs with $-2 <$ [Fe/H] $< -1.5$ without invoking younger ages. These results
39: may prove useful in studying the contribution of HB stars to integrated light via
40: stellar population synthesis.
41: \end{abstract}
42:
43: \keywords{globular clusters: general --- stars: horizontal branch}
44:
45:
46: \section{Introduction}
47: \citet{z93} recommended that Galactic halo GCs be divided into two groups:
48: one for old, roughly coeval GCs that follow the mean trend in [Fe/H] vs. HB
49: type in the inner halo (the Old Halo; OH) and one for GCs that deviate from
50: this trend by a significant amount (the Younger Halo; YH). Due to the lack
51: of very metal poor GCs in the inner halo, Zinn used more distant GCs to
52: extend the line below [Fe/H] $< -2$, see his Figure 1.
53: % Figure 1 of \citet{z93} shows the locations of all the well-measured GCs
54: % at that time as well as the trend line he used to define the OH population.
55:
56: Zinn's relation was empirically motivated and differs from later studies---such
57: as those by \citet[hereafter LDZ]{ldz2} and \citet{re01} that are based
58: on synthetic HB isochrones---in that it turns back to the red at the lowest
59: metallicities ([Fe/H] $<$ -2). Trends derived from synthetic HB calculations
60: with constant mass loss (LDZ) or mass loss that increases with [Fe/H] and age
61: \citep{re01} indicate that HB type (as defined by LDZ 1994) rises monotonically
62: to +1 as [Fe/H] decreases and remains there for all lower values of [Fe/H].
63: \citet{l91} showed that allowing mass loss to vary with [Fe/H], in a manner
64: consistent with the observed trend in GCs, resulted in better agreement between
65: his models and the pulsational properties of RR Lyrae stars in $\omega$ Cen than
66: was achieved by assuming constant mass loss. Aside from Lee's application to
67: $\omega$ Cen, the concept of searching for a mass loss relation that matches the
68: observed trend in old, coeval GCs has not been explored.
69:
70: The purpose of this Letter is to argue that it is possible to account for
71: the fact that the most metal poor GCs have HB types less than +1 without
72: requiring them to be significantly younger than more metal rich GCs if
73: mass loss on the RGB varies as a function of [Fe/H] in a manner qualitatively
74: similar to that proposed by \citet{l91}.
75:
76: \section{Synthetic HB Model Construction}
77: \citet{ro73} introduced an effective means of modeling the distribution
78: of HB stars in GCs by assuming the stars that arrive on the HB are drawn at
79: random from a Gaussian mass distribution that is truncated on both ends by
80: physical limitations: namely the initial mass and the He core mass of the
81: precursor red giant stars. The method further assumes that stars arrive on
82: the HB at a constant rate and spend roughly the same amount of time there
83: regardless of mass. Under these assumptions, one constructs a synthetic
84: HB by drawing mass and age at random and then interpolating in a grid of
85: He-burning evolutionary tracks to obtain the relevant quantities.
86:
87: Many investigators, such as \citet{ro73}, LDZ (1990, 1994), \citet{ca94},
88: \citet{re01}, and \citet{ca04}, have used the synthetic HB method to show how
89: various model parameters influence HB morphology and the pulsational properties
90: of RR Lyrae stars in GCs.
91: LDZ (1994) generated isochrones in the [Fe/H] vs. HB type diagram for different
92: ages and He abundances while assuming a fixed amount of mass loss and dispersion.
93: \citet{re01} incorporated the Reimers mass loss formula into their synthetic
94: HB analysis (such that the average mass loss increased with [Fe/H] and age) and
95: found that, while the sensitivity of HB morphology increased compared to the
96: findings of LDZ (1994), the general trend was unchanged. These isochrones were
97: used to argue that GCs with anomalously red HBs (compared to the oldest synthetic
98: HB isochrone) are younger than the OH GCs: one of strongest arguments in favor of
99: age as the second parameter governing HB morphology in GCs.
100:
101: This study employs the He core masses, RGB tip masses, and He-burning
102: tracks from \citet{d07} to create synthetic HB models (each consisting of 500
103: stars) spanning the range of [Fe/H] and $\afe$ observed in the Galactic halo.
104: Simulations of 1000-2000 synthetic HBs are generated by allowing [Fe/H] to
105: vary between $-2.5$ and $-0.5$ while $\afe$, age, mass loss ($\Delta$M), and
106: dispersion ($\sigma_M$) are held constant or varied in a specified manner.
107: % The results are analyzed in the [Fe/H] vs. HB type diagram.
108:
109: Figure \ref{param} indicates how some of the governing parameters influence
110: the [Fe/H] vs. HB type trend. Panel A compares trend lines derived from a
111: simulation with constant mass loss ($\Delta$M=0.15 $\Ms$), dispersion
112: ($\sigma_M$=0.02 $\Ms$), and $\afe$=+0.4 for ages of 9, 11, and 13 Gyr; panel B
113: shows the same mass loss and dispersion assumed in A but at a fixed age (13 Gyr)
114: with $\afe$=0, +0.2, +0.4, and +0.6; panel C illustrates the effect of varying
115: the mass loss ($\Delta$M=0, 0.1, 0.2, 0.3 $\Ms$) when the dispersion (0.02 $\Ms$),
116: age (13 Gyr), and $\afe$ (+0.4) are held constant; finally, panel D compares
117: synthetic HBs derived from the He-burning evolutionary tracks of \citet[solid
118: lines]{d07} and \citet[dashed lines]{ca04} at 13 Gyr with $\Delta$M=0, 0.1, and
119: 0.2 $\Ms$. The \citet{ca04} models were calculated at Y=0.23 with a scaled-solar
120: mixture for all [Fe/H] while the \citet[see their Table 4]{d07} models range from
121: Y=0.245 (at [Fe/H]=$-2.5$, $\afe$=0) to Y=0.264 (at [Fe/H]=$-0.5$, $\afe$=0) so it
122: is not surprising that the the \citet{d07} models yield consistently bluer HB types
123: in most regions of the diagram.
124: % Although the trend lines derived from the two sets of models
125: % differ, the general result that HB type shifts from blue to red
126: % at higher [Fe/H] as $\Delta$M increases is present in both cases.
127: Figure \ref{param} shows that there is considerable degeneracy in the [Fe/H] vs. HB
128: type diagram among age, $\afe$, and $\Delta$M. Though not shown,
129: stochastic effects are also significant because even though each synthetic HB
130: consists of 500 stars (a number observed only in rich GCs) there is still a range
131: of possible outcomes ($\Delta$HB type$\sim$0.1) for the same input parameters.
132: Only a few tens of HB stars have been observed in some GCs and this, combined
133: with stochastic effects, can make the parameters governing HB morphology
134: difficult to ascertain from model comparisons.
135:
136:
137:
138: Based on the relationships that were found by varying individual parameters
139: in the synthetic HB models, additional simulations were generated
140: in order to attempt to understand the observed distribution of old,
141: roughly coeval GCs in the Galactic halo. In these simulations, [Fe/H]
142: was allowed to vary over $-2.5 <$ [Fe/H] $< -0.5$ while age was held fixed at
143: 13 Gyr. The level of $\afe$ was drawn at random from a flat distribution in
144: such a way that the average value of $\afe$ decreased as [Fe/H] increased.
145: Figure \ref{alpha} shows how the fraction of synthetic HB models with a given
146: value of $\afe$ is assumed to vary with [Fe/H].
147: %For example, of all the
148: %randomly generated synthetic HBs with $-2.5 <$ [Fe/H] $< -2$, approximately
149: %30\% had $\afe$ = +0.6, 40\% had $\afe$ = +0.4, and 30\% had $\afe$ = +0.2.
150: This method of distributing $\afe$ is based on observations of the
151: $\alpha$-elements as a function of [Fe/H] in halo GCs and field stars
152: \citep[see their Figure 3]{pvi05}.
153: %\footnote{Note that \citet{pvi05} define $\afe$
154: %as the average of [Ca/Fe] and [Ti/Fe] but in this paper all of the
155: %$\alpha$-elements considered by those authors (Mg, Si, Ca, and Ti) are
156: %accounted for in $\afe$ so their Figure 3 is more relevant than their Figure
157: %4.}]{pvi05}.
158: Mass loss ($\Delta$M; measured in $\Ms$) was assumed to vary with [Fe/H] as
159: shown in Figure \ref{delsig}. The minimum value was set as a function of [Fe/H]
160: as in equation \ref{dM}.
161:
162: \begin{equation}
163: \Delta M_{min}(x)=c\frac{exp(\frac{x-a}{b})}{1+exp(\frac{x-a}{b})}\label{dM}
164: \end{equation}
165:
166: Where $x$=[Fe/H] and $a$, $b$, and $c$ are free parameters.
167: The variation of $\Delta$M with [Fe/H] could also have been described by a
168: broken linear relation; the use of equation \ref{dM} was chosen because it provides
169: a smooth variation with [Fe/H]. With the minimum value of $\Delta$M set by equation
170: \ref{dM}, the actual value for a single synthetic HB realization $i$ was set by
171: equation \ref{dMR}.
172:
173: \begin{equation}
174: \Delta M_i = \Delta M_{min}(x_i) + |R_G(0,0.03)|\label{dMR}
175: \end{equation}
176:
177: Where $R_G(y,z)$ is a random number drawn from a Gaussian mass distribution having a
178: mean value $y$ and dispersion $z$. Thus the value of $\Delta$M was more likely to
179: lie near the minimum value but, given a large enough simulation, values at the
180: $\sim$3 $\sigma$ level can be found. Figure \ref{delsig} shows the distribution of
181: $\Delta$M as a function of [Fe/H] using equations \ref{dM} and \ref{dMR} for a
182: simulation of 2000 synthetic HB models. The minimum value as defined by equation
183: \ref{dM} is shown as the solid line while the dashed lines represent the minimum
184: value plus 1, 2, and 3$\sigma$.
185:
186:
187:
188: The value of the dispersion in the mass distribution ($\sigma_M$; also measured in
189: $\Ms$) was set in a similar fashion to that for $\Delta$M (equation \ref{dMR}) except
190: that the minimum value was held constant at 0.02 and the dispersion in $R_G$ was 0.01.
191:
192: \section{Results}
193:
194: Figure \ref{deltam} shows the [Fe/H] vs. HB type diagrams that are obtained
195: when the variable $\Delta$M$_{min}$ given by equation \ref{dM} with $a=-2,
196: b=0.05$, and $c$=0.14 (solid circles) and a constant $\Delta$M$_{min}$=0.14
197: (open squares) are assumed. Both simulations consist of 2000 synthetic HB models at
198: 13 Gyr and both used equation \ref{dMR} to determine $\Delta$M and $\sigma_M$ as
199: outlined in the preceding section. That synthetic HB models become redder as $\afe$
200: increases at a given [Fe/H] may be a necessary part of the explanation for why at
201: least some of the most metal poor GCs have redder HBs than can be found at [Fe/H]
202: $\ga -2$ but $\afe$ alone is not sufficient to explain this behavior. Assuming a
203: constant mass loss and the $\afe$ distribution shown in Figure \ref{alpha} produces
204: the same sort of [Fe/H] vs. HB type behavior demonstrated by LDZ (1994) and \citet{re01}.
205: Allowing $\Delta$M to vary with [Fe/H] (see equations \ref{dM} and \ref{dMR}) as
206: described above produces a trend that matches the original OH trend of \citet{z93}.
207:
208:
209:
210: The GC catalog compiled by \citet{mvdb} places GCs into sub-populations
211: along the lines of \citet{z93} but with additional categories that reflect
212: discoveries made in the intervening years. \citet{mvdb} used the HB isochrones
213: of \citet{re01} to assign a sub-population to each GC. Many of the lowest
214: metallicity GCs are considered part of the YH despite evidence that their
215: ages are similar to other metal poor OH GCs, such as M\,92 and NGC\,2419, with
216: bluer HBs \citep{vdb00,sw02}. Given the lack of a substantial
217: age difference between the lowest metallicity GCs with different HB types, the
218: metal poor GCs NGC\,5053, 5466, 6426, and 7078 (M\,15) have been included in
219: the comparison presented here.
220: %\footnote{NGC\,5053 inhabits the same region of
221: %the [Fe/H] vs. HB type diagram as NGC\,5466, 6426, and 7078 but its age
222: %and HB type are somewhat uncertain. \citet{sw02} found NGC\,5053 to be younger
223: %than the other three GCs but the isochrone comparisons shown by \citet{sa07}
224: %suggest that NGC\,5053 and NGC\,5466 have very similar ages. The HB of NGC\,5053
225: %is very sparsely populated and thus susceptible to stochastic effects.}.
226: Figure \ref{gcshb} compares a synthetic HB simulation with $\alpha$-enhancement
227: and variable mass loss with the OH GCs (filled circles) and four metal poor YH
228: GCs (open circles) from \citet{mvdb}. In order to avoid complications, as much as
229: currently possible, that arise from the HB morphologies of GCs with multiple
230: stellar populations, NGC\,1851 \citep{mi08} and NGC\,2808 \citep{pi07} are not
231: shown in Figure \ref{gcshb}.
232:
233:
234:
235: The variable mass loss suggested in the previous section reproduces the distribution
236: of old GCs in the halo. Indeed, if the He-burning tracks of \citet{ca04} are used
237: instead of the \citet{d07} tracks, approximating $\alpha$-enhancement
238: by [M/H] $\simeq$ [Fe/H] + log(0.638\,x\,10$^{\afe}$ + 0.362) \citep{scs93}, variable
239: mass loss with the same values of $a$, $b$, and $c$ listed at the beginning of this
240: section produces a nearly identical result to that shown in Figure \ref{gcshb}.
241: Based on these results pertaining to mass loss and HB morphology as well as the
242: evidence that the four old, metal poor GCs with relatively red HBs are not especially
243: young \citep{vdb00,sw02}, it is recommended that those GCs be considered part of the
244: OH as originally intended by \citet{z93}.
245:
246:
247:
248: \section{Discussion and Conclusion}
249:
250: \citet{ca05} provides a thorough discussion of mass loss and HB morphology in GCs.
251: Catelan considers a handful of empirically motivated formulae that describe the mass
252: loss rate as a function of stellar parameters such as luminosity, radius, and surface
253: gravity. Each predicts that mass loss increases smoothly with
254: [Fe/H] \citep[Figure 4]{ca05}. Some of the formulae are consistent with the observed
255: difference in HB morphology between second parameter GC pairs, but not all, and no
256: single formula is successful for each of the GC pairs considered \citep[Figures
257: 10-13]{ca05}. There is also a discrepancy between the formulae and observational
258: efforts to measure mass lost by red giants in GCs by \citet{or02} who
259: found that mass loss is not strongly correlated with metallicity and only occurs
260: near the tip of the RGB. On the other hand, \citet{or07} found that mass loss is
261: evident at the level of HB and increases with luminosity up to the RGB tip
262: in 47\,Tuc. The proposed mass loss-[Fe/H] relation should be straightforward
263: to test observationally with mass loss estimates for a large enough GC sample.
264:
265: These examples illustrate that mass loss remains a complex and poorly understood
266: process. It is by no means guaranteed that the information provided by standard
267: stellar evolution models is sufficient to calculate mass loss as a function of
268: age and metallicity. The observed range of HB morphologies in OH GCs (which are
269: assumed to be roughly coeval) of similar metallicity suggests that more than
270: one factor must be involved. It seems quite plausible that mass loss may be
271: influenced at the stellar level (by, metallicity and rotation, for example) and
272: at the cluster level (by cluster mass or concentration, for example). Uncertainties
273: in composition, coming from the lack of accurate abundance determinations in some
274: lesser known GCs, or the possible presence of multiple populations with distinct
275: compositions in a single GC, complicate the analysis and interpretation.
276:
277: The situation is further complicated by the discovery of bi- and multi-modal HBs
278: in some GCs, see \citet{fe98} for examples and a review of the subject. It has
279: been suggested by, e.g. \citet{da08} and references therein, that these (and perhaps
280: all) GCs are comprised of chemically distinct populations that create the complex
281: distribution of HB stars. Complications arise even in GCs with uni-modal HBs.
282: The analysis of M\,3 by \citet{va08} indicates that while the synthetic HB method
283: developed by \citet{ro73}, particularly the assumed Gaussian mass distribution, is
284: `a reasonable first approximation, [it] fails to account for the detailed shape of
285: M\,3's HB mass distribution.' \citet{va08} also warn that there is a mismatch
286: between observed and theoretical lifetime-luminosity relations on the HB.
287:
288: In the absence of a firm theoretical foundation for mass loss in stars, it is
289: still possible to further our understanding of this subject and to apply that
290: knowledge. One application of the results presented here is to the modeling of
291: integrated light via stellar population synthesis. Investigators such as
292: \citet{le00} and \citet{le02} have explored the dependence of integrated light
293: properties on synthetic HB morphology and compared them with observations of
294: extragalactic GCs. The results presented here may provide useful input for future
295: studies along these lines.
296:
297: To summarize, synthetic HB models that include the effects of $\alpha$-enhancement
298: and metallicity-dependent mass loss (via equations \ref{dM} and \ref{dMR}) can
299: account for the entire distribution of HB morphologies observed in old, coeval GCs
300: in the Galactic halo. Essentially the same result is obtained from two distinct sets
301: of He-burning tracks, suggesting this is not a spurious or model-dependent result.
302: As a corollary, it is recommended that NGC\,5053, 5466, 6426, and 7078, recently
303: associated with the YH, be considered members of the OH since their ages are
304: consistent with other OH GCs and their HB morphologies can be explained without
305: invoking substantially younger ages.
306:
307:
308: \acknowledgments
309:
310: Thanks to A. Sarajedini for the motivation to pursue this work and to D.
311: VandenBerg for helpful discussions and comments on the manuscript.
312: Thanks also to the anonymous referee.
313: This work was supported by a CITA National Fellowship to AD.
314: Support from an NSERC grant to D. VandenBerg is also acknowledged.
315:
316: \begin{thebibliography}{}
317: \bibitem[Cassisi et al.(2004)]{ca04}Cassisi, S., Castellani, M., Caputo, F., \& Castellani, V. 2004, \aap, 426, 641
318: \bibitem[Catelan(2005)]{ca05}Catelan, M. 2005 arxiv:astro-ph/0507464
319: \bibitem[Catelan \& de Freitas Pacheco(1994)]{ca94}Catelan, M. \& de Freitas Pacheco, J. A. 1994, \aap, 289, 394
320: \bibitem[D'Antona \& Caloi(2008)]{da08}D'Antona, F. \& Caloi, V. 2008, \mnras, in press
321: \bibitem[Dotter et al.(2007)]{d07}Dotter, A. et al. 2007, \aj, 134, 376
322: \bibitem[Ferraro et al.(1998)]{fe98}Ferraro, F.~R., Paltrinieri, B., Pecci, F.~F., Rood, R.~T., \& Dorman, B. 1998, \apj, 500, 311
323: \bibitem[Lee(1991)]{l91}Lee, Y.-W. 1991, \apjl, 373, 43
324: \bibitem[Lee, Demarque, \& Zinn(1990)]{ldz1}Lee, Y.-W., Demarque, P., \& Zinn, R. 1990, \apj, 350, 155
325: \bibitem[Lee, Demarque, \& Zinn(1994)]{ldz2}Lee, Y.-W., Demarque, P., \& Zinn, R. 1994, \apj, 423, 248
326: \bibitem[Lee, Lee, \& Gibson(2002)]{le02}Lee, H.-C., Lee, Y.-W., \& Gibson, B. K. 2002, \aj, 124, 2664
327: \bibitem[Lee, Yoon, \& Lee(2000)]{le00}Lee, H.-C., Yoon, S.-J., Lee, Y-.W. 2000, \aj, 120, 998
328: \bibitem[Mackey \& van den Bergh(2005)]{mvdb}Mackey, A. D. \& van den Bergh, S. 2005, \mnras, 360, 631
329: \bibitem[Origlia et al.(2002)]{or02}Origlia, L., Ferraro, F. R., Fusi Pecci, F., \& Rood, R. T. 2002, \apj, 571, 458
330: \bibitem[Origlia et al.(2007)]{or07}Origlia, L. Rood, R. T., Fabbri, S., Ferraro, F. R., Fusi Pecci, F., \& Rich, R. M. 2007, \apjl, 667, 850
331: \bibitem[Pritzl, Venn, \& Irwin(2005)]{pvi05}Pritzl, B. J., Venn, K. A., \& Irwin, M. 2005, \aj, 130, 2140
332: \bibitem[Milone et al.(2008)]{mi08}Milone, A. et al. 2008, \apj, 673, 241
333: \bibitem[Piotto et al.(2007)]{pi07}Piotto, G. et al. 2007, \apjl, 661, 53
334: \bibitem[Rey et al.(2001)]{re01}Rey, S.-C., Yoon, S.-J., Lee, Y.-W., Chaboyer, B., \& Sarajedini, A. 2001, \aj, 122, 3219
335: \bibitem[Rood(1973)]{ro73}Rood, R. T. 1973, \apj, 184, 815
336: \bibitem[Salaris, Chieffi, \& Straniero(1993)]{scs93}Salaris, M., Chieffi, A., \& Straniero, O. 1993, \apj, 414, 580
337: \bibitem[Salaris \& Weiss(2002)]{sw02}Salaris, M. \& Weiss, A. 2002, \aap, 388, 492
338: \bibitem[Sarajedini et al.(2007)]{sa07}Sarajedini, A. et al. 2007, \aj, 133, 1658
339: \bibitem[Valcarce \& Catelan(2008)]{va08}Valcarce, A. A. R. \& Catelan, M. 2008, \aap, 487, 185
340: \bibitem[VandenBerg(2000)]{vdb00}VandenBerg, D. A. 2000, \apjs, 129, 315
341: \bibitem[Zinn(1993)]{z93}Zinn, R. 1993, ASPC, 48, 38
342: \end{thebibliography}
343:
344: \clearpage
345:
346: \begin{figure}
347: \plotone{f1}
348: \caption{The effect of varying: (A) age at constant $\Delta$M=0.15, $\sigma_M$=0.02,
349: and $\afe$=+0.4; (B) $\afe$ at constant $\Delta$M=0.15, $\sigma_M$=0.02, and 13
350: Gyr. (C) $\Delta$M at constant $\sigma_M$=0.02, age of 13 Gyr, and $\afe$=+0.4; and
351: (D) $\Delta$M at constant $\sigma_M$=0.02 using the He-burning tracks of
352: \citet[dashed line; Y=0.23, $\afe$=0, 13 Gyr]{ca04} and \citet[solid line;
353: Y=0.245-0.264, $\afe$=0, 13 Gyr]{d07}.\label{param}}
354: \end{figure}
355:
356: \begin{figure}
357: \plotone{f2}
358: \caption{The approximate fraction of all synthetic HBs with a given value of
359: $\afe$ as a function of [Fe/H] in the synthetic HB simulations. Since the
360: number of realizations in each simulation was finite, the actual fractions
361: only approach the values shown here.
362: \label{alpha}}
363: \end{figure}
364:
365: \begin{figure}
366: \plotone{f3}
367: \caption{Distribution of $\Delta$M (in $\Ms$) as a function of [Fe/H]. The points are
368: from one simulation of 2000 synthetic HB models. The simulation used $a=-2$, $b=0.05$,
369: and $c=0.14$ in equation \ref{dM}. The solid line is the minimum trend
370: (equation \ref{dM}) and the dashed lines are 1, 2, and 3$\sigma$ above the minimum.
371: \label{delsig}}
372: \end{figure}
373:
374: \begin{figure}
375: \plotone{f4}
376: \caption{Comparison of synthetic HB simulations for which $\Delta$M$_{min}$ was
377: either fixed (open squares) or varied according to equation \ref{dM}
378: (solid circles). In both cases, $\Delta$M was set by equation \ref{dMR}.
379: HB type varies monotonically with [Fe/H] in the fixed $\Delta$M$_{min}$ case but
380: turns back to the red in the variable $\Delta$M$_{min}$ case below [Fe/H] $\la -2$.
381: \label{deltam}}
382: \end{figure}
383:
384: \begin{figure}
385: \plotone{f5}
386: \caption{The synthetic HB simulation with variable mass loss (small filled cirlces)
387: compared to the OH GC sample (large filled circles) and four YH GC (large open circles)
388: of \citet{mvdb} with omissions as described in the text. The $\Delta$M-[Fe/H] relation
389: given by equation \ref{dM} is capable of simultaneously reproducing the OH GCs and the
390: four metal poor YH GCs without assuming any variation in age.\label{gcshb}}
391: \end{figure}
392:
393: \end{document}
394: