1: %\documentclass[manuscript]{aastex}
2: \documentclass[12pt,preprint]{aastex}
3: \usepackage{natbib,graphicx,latexsym}
4: \newcommand{\lsim}{\
5: \raise-2.truept\hbox{\rlap{\hbox{$\sim$}}\raise5.truept\hbox{$<$}\ }}
6: \newcommand{\gsim}{\
7: \raise-2.truept\hbox{\rlap{\hbox{$\sim$}}\raise5.truept\hbox{$>$}\ }}
8: \newcommand{\etal}{et~al.}
9: \newcommand{\vh}{$(V{-}H)$}
10: \newcommand{\vk}{$(V{-}K)$}
11: \newcommand{\ct}{$(C{-}T_1)$}
12: \renewcommand{\ur}{$(U{-}R)$}
13: \renewcommand{\bv}{$(B{-}V)$}
14: \newcommand{\bi}{$(B{-}I)$}
15: \newcommand{\bh}{$(B{-}H)$}
16: \newcommand{\vi}{$(V{-}I)$}
17: \newcommand{\ik}{$(I{-}K)$}
18: \newcommand{\ih}{$(I{-}H)$}
19: \newcommand{\jk}{$(J{-}K)$}
20: \newcommand{\gz}{$(g{-}z)$}
21:
22: \begin{document}
23:
24: \shorttitle{Metallicity-Color Relations in GC Systems}
25: \shortauthors{Cantiello \& Blakeslee}
26:
27: \title{On the Metallicity-Color Relations and Bimodal Color
28: Distributions in Extragalactic Globular Cluster Systems}
29:
30: \author{Michele Cantiello\altaffilmark{1,2}}
31: \and
32: \author{John P. Blakeslee\altaffilmark{1}}
33:
34: \altaffiltext{1}{Department of Physics and Astronomy, Washington State University,
35: Pullman, WA 99164.}
36: \altaffiltext{2}{INAF--Osservatorio Astronomico di
37: Teramo, Via M. Maggini, I-64100 Teramo, Italy}
38:
39: \begin{abstract}
40: %
41: We perform a series of numerical experiments to study how the nonlinear
42: metallicity--color relations predicted by different stellar population
43: models affect the color distributions observed in extragalactic globular
44: cluster systems.
45: %
46: We present simulations in the $UBVRIJHK$ bandpasses
47: based on five different sets of simple stellar population (SSP) models.
48: The presence of photometric scatter in the colors is included as well.
49: %
50: We find that unimodal metallicity distributions frequently ``project''
51: into bimodal color distributions. The likelihood of this effect depends
52: on both the mean and dispersion of the metallicity distribution,
53: as well as of course on the SSP model used for the transformation.
54: %
55: Adopting the Teramo-SPoT SSP models for reference, we find that
56: optical--to--near-IR colors should be favored with respect to other
57: colors to avoid the bias effect in globular cluster color
58: distributions discussed by \citet{yoon06}. In particular, colors such
59: as \vh\ or \vk\ are more robust against nonlinearity of the
60: metallicity--color relation, and an observed bimodal distribution
61: in such colors is more likely to indicate a true underlying bimodal
62: metallicity distribution. Similar conclusions come from the
63: simulations based on different SSP models, although we also identify
64: exceptions to this result.
65: %
66: \end{abstract}
67:
68: \keywords{galaxies: star clusters -- galaxies: elliptical and lenticular, cD --
69: globular clusters: general}
70:
71:
72: \section{Introduction}
73:
74: The globular cluster (GC) system of a galaxy provides a
75: unique view into the formation history of the galaxy.
76: Apart from some rare exceptions, GCs are known to represent a
77: relatively simple class of objects, with homogeneous ages
78: and chemical compositions for the stars composing each GC. Thus, GCs
79: are at the same time reasonably simple objects, and good tracers
80: of the early star formation histories of the host galaxy.
81: %
82: For these reasons, in recent years a great deal of effort has been made
83: to study the observational properties of extragalactic GC systems. As
84: a consequence, many GC features have been discovered, providing
85: valuable constraints on the evolutionary paths of galaxies.
86:
87: One commonly observed property is that the GC populations in galaxies
88: tend to be bimodal in their color distributions. A combination of
89: photometric and spectroscopic observations indicates that GC systems
90: are fairly homogeneous in terms of age, so differences in color mainly
91: reflect metallicity differences. Thus, the bimodal color
92: distributions have usually been interpreted as bimodal metallicity
93: distributions; see the reviews by \citet{west04}, \citet{brodie06},
94: and references therein, for further details.
95: %
96: However, more recently some authors
97: \citep[e.g.][]{richtler06,yoon06} have shown that a bimodal color
98: distribution can be enhanced, or even originated, by the effect of
99: nonlinear metallicity-color (MC hereafter) relations.
100:
101: For instance, \citet{richtler06} has shown that the observed MC
102: relation for the Washington system \ct\ color, coupled with a
103: Gaussian scatter of 0.08 mag around the mean relation, can transform a
104: nearly flat GC metallicity distribution into a bimodal \ct\
105: distribution. \citet[][YYL06 hereafter]{yoon06}, instead, exploited
106: the fact that, when the Horizontal-Branch (HB) morphology is
107: realistically modeled in stellar population simulations, the color
108: indices sensitive to stars in this evolutionary stage follow ``wavy''
109: MC relations. Such a nonlinear feature has in fact been observed for the
110: metallicity versus $(g-z)$ color of GCs in the Milky Way and Virgo
111: ellipticals \citep{peng06}. This feature causes evenly-spaced
112: metallicity bins to be ``projected'' into larger/narrower color bins,
113: depending on the location on the MC relation. YYL06 consequently conclude
114: that it is not necessary to invoke a
115: bimodal metallicity distribution to have a bimodal color
116: distribution. We will refer to this effect
117: as {\it metallicity projection bias}.
118:
119: In this paper we study how nonlinear MC relations can affect the
120: color distributions observed in different passbands. Our aim is ($i$)\,to test how
121: various colors ``suffer'' from the nonlinear effects described above,
122: and, consequently, ($ii$)\,to suggest the optimal color(s) for revealing
123: the presence of real bimodal GC metallicity distributions.
124:
125: We do this by first carrying out multiple sets of simulations based on the
126: Simple Stellar Population (SSP) models developed by the Teramo-SPoT group \citep[][SPoT
127: hereafter\footnote{The SPoT models are available at the Teramo-SPoT website:
128: www.oa-teramo.inaf.it/SPoT}]{raimondo05}. We then perform the same tests using
129: simulations based on four other sets of SSP models and compare the results.
130: Finally, we summarize the most robust conclusions on GC
131: colors and their underlying metallicity distributions from this work.
132:
133: \section{Models simulations}
134: For this study
135: we adopt the SPoT SSP models as our reference models for two reasons.
136: First, these models have proven to match fairly well
137: the observed integrated photometric properties
138: of galaxies, i.e. colors, surface brightness fluctuation magnitudes,
139: etc., in different passbands for a large sample of objects with very
140: different physical properties
141: \citep{cantiello05,raimondo05,cantiello07}.
142: %
143: Second, the SPoT models also provide a good match to the observed
144: color--magnitude diagrams (CMDs) for star clusters with a wide range of
145: ages and chemical compositions \citep{brocato00,raimondo05}. In
146: particular, these models are optimized to simulate the HB spread
147: observed in Galactic GCs.
148:
149: The detailed numerical synthesis of the CMD features is a key point
150: for the aims of the present study since, as shown by YYL06, the wavy
151: feature that can produce a ``projected'' color bimodality is due to
152: a realistic treatment of the HB morphology. It is worth noting here
153: that this feature is {\it not} a peculiarity of the YYL06 models; in
154: fact it was already presented by \citet[see their Figures 2 and
155: 4]{lee02} and, as we will show, it is present also in the SPoT
156: models. Not surprisingly, this feature is not noticeable in those
157: models where the HB morphology is not properly simulated to match the
158: observed Galactic GC properties, as for example in the \citet{bc03}
159: models, which adopt a fixed red HB morphology.
160:
161: In fact, our reference SPoT models attempt to simulate in a realistic
162: way all features of the observed CMD, that is all the stellar
163: evolutionary stages including the fast and bright phases of the Giant
164: Branch stage. These models are computed according to the following
165: prescriptions: \citet{scalo98} Initial Mass Function; solar
166: scaled stellar evolution tracks from \citet{pietrinferni04}; HB
167: morphology reproduced taking into account the effects due to age,
168: metallicity, and the stellar mass spread due to the stochasticity of
169: the mass-loss along the RGB. The RGB mass--loss rate is evaluated
170: according to Reimers' law \citep{reimers75}, with efficiency
171: $\eta_{RGB}=0.4$. Thermal pulses are simulated using the analytic
172: formulations by \citet{wagenhuber98}. Finally, the atmosphere models
173: are from \citet{westera02}. See \citet{raimondo05} for further
174: details.
175: %
176: Throughout this paper, we will consider the $t=13$ Gyr age models for
177: reference, if not stated otherwise.
178:
179: Figure~\ref{cmspot} shows the MC relations from the SPoT models for
180: several different colors. Data for the Galactic GCs are also
181: shown. The optical colors and the [Fe/H] values for
182: the Galactic GCs are taken from the \citet{harris96} updated online
183: catalog\footnote{http://www.physics.mcmaster.ca/$\sim$harris/mwgc.dat},
184: while the near-IR photometric data are from \citet{brocato90} and
185: \citet{cohen07}. As seen in Fig.~\ref{cmspot}, the models provide
186: a good match to the integrated properties of the Galactic GC system.
187: The ``wavy'' behavior of the MC relations for the \vi,
188: \bi, \ur, and \bv\ colors is clearly evident.
189: Furthermore, it is worth emphasizing the general nonlinearity
190: of the MC relations for all the colors shown in the figure.
191:
192: \subsection{GC Simulations: SPoT models}
193:
194: We have developed a procedure to simulate a GC population with an
195: arbitrary metallicity distribution and number of objects. Throughout
196: this paper, however, we will consider the case of Gaussian metallicity
197: distributions, and we simulate GC populations composed of 1000 objects.
198: %
199: Armed with the MC relations of our reference models, the metallicity
200: distribution of the GC system is randomly populated and projected into
201: a color distribution. Finally, we use the KMM code
202: \citep{mclachlan88,ashman94} to test whether the GC color distribution is
203: best fit by a single or double Gaussian function.
204:
205: In Figure \ref{bimodal} (left panel), we show the results of one of
206: these simulations. Specifically, in this case we have simulated a
207: metallicity distribution similar to the one adopted by YYL06, that is
208: a Gaussian with peak at [Fe/H]$\,=-0.65$ dex and dispersion
209: $\sigma_{[Fe/H]}=0.5$ dex. It is clearly recognized from the \vi\
210: panel that the projected color distribution is bimodal. By running
211: the KMM code, we find that, for this specific simulation, all
212: the optical colors\footnote{We consider the \ur\ as it is the
213: nearest color the Washington system \ct\ color, not provided
214: with the SPoT models. The \ct\ index is interesting because it is
215: known to be one for which the GCs distribution is bimodal in all of the
216: limited number of observed galaxies \citep{richtler03}.} and the \jk\ color
217: distributions are significantly bimodal, while the optical to
218: near-infrared colors, including \ih, have unimodal distributions.
219:
220: As a check to these simulations, we have also made some numerical
221: experiments adopting a bimodal metallicity distribution. In
222: particular, Figure \ref{bimodal} (right panel) shows a simulation
223: carried out adopting the bimodal metallicity distribution of the
224: Galactic GC system, obtained using the prescriptions of
225: \citet{cote99}. In the Figure also the observed metallicity and color
226: distributions of Galactic GCs are shown. The histograms of the
227: simulated GC population are shown with solid lines in the panels,
228: while the histograms for the actual observed Galactic GCs are shown
229: with dotted lines. To simulate observational scatter of the data, we
230: have included a 10\% Gaussian scatter in the colors. As can be seen
231: from this comparison, there is generally a good match between the
232: simulated and actual color histograms of the Galactic GC system.
233:
234: Since our goal here is to identify the colors least affected by the
235: projection bias, regardless of the underlying metallicity
236: distribution, we have performed various tests assuming unimodal
237: Gaussian distributions with peaks at [Fe/H]$\,=-1.65, -1.15, -0.65,
238: -0.15$ dex and three values for the dispersion: $\sigma_{[Fe/H]}=0.25,
239: 0.5, 0.75$ dex. For each of the twelve ([Fe/H], $\sigma_{[Fe/H]}$)
240: pairs, we have simulated a GC system with a unimodal metallicity
241: distribution and evaluated the colors of each GC according to the
242: adopted MC relations. Afterwards, by using the KMM code, we estimate
243: the likelihood, $P(bimodal)$, that the color distribution is better
244: represented by two Gaussians than a single Gaussian, for various color
245: choices. Values of $P(bimodal){\,\approx\,}1$ mean that the color
246: distribution is likely bimodal; conversely, color distributions with
247: $P(bimodal){\,\approx\,}0$ are likely unimodal. We have run the
248: simulations both with and without including a 10\% Gaussian scatter in
249: the simulated colors.
250:
251: Table \ref{tab_spot} gives the results of these simulations.
252: For each color index, the table lists the locations of the best-fitting
253: blue and red peaks and the value of $P(bimodal)$
254: for each choice of mean metallicity and dispersion. The results are
255: also shown graphically in the Figure \ref{plotspot}, where solid dots
256: mark the results for simulations without any color errors, and
257: open circles mark results obtained with the random 10\% color scatter.
258: The different rows and columns refer to
259: different mean [Fe/H] and $\sigma$ values, respectively, as labeled.
260:
261: Two considerations emerge from inspection of Figure \ref{plotspot}.
262: First, according to the SPoT models, the projection effect that
263: causes a unimodal metallicity distribution to be observed as a bimodal
264: color distribution is not a unique characteristic of the HB-sensitive
265: colors. It is, instead, present for most of the analyzed colors. For
266: example, in the case of \jk, almost half of the numerical
267: experiments carried out give bimodal color distributions
268: [$P(bimodal)\sim 1.0$]. Thus the nonlinearity of the MC relation
269: is not specific to just one or a few colors,
270: such as the $(g-z)$ and \vi\ colors discussed by YYL06.
271: Although, as shown in Figure \ref{cmspot}, different colors are
272: affected differently by nonlinearity in the MC relation.
273:
274: The second consideration that emerges from these simulations regards how
275: the presence of color scatter (i.e. the photometric uncertainty) can
276: affect the probability of obtaining a bimodal color distribution.
277: The addition of color scatter can of course decrease the probability
278: of bimodality by smoothing
279: out the separation between the peaks. More surprisingly, it can also
280: make bimodality appear more probable by removing sharp features from
281: the color distributions, and thus significantly improving the
282: goodness-of-fit of the double Gaussian model used by KMM.
283:
284: It is interesting to note that the two extreme metallicities,
285: [Fe/H]$\,=-1.65$ and $-$0.15 dex, result in strictly-unimodal, and
286: generally-bimodal, color distributions, respectively \citep[see
287: also][their Fig. 3]{yoon06}. Moreover, simulations with larger
288: $\sigma$ values are in almost all cases more bimodal. Thus, for the
289: combination ([Fe/H]$\,=-$0.15, $\sigma{\,=\,}0.75$), the color
290: distributions are significantly bimodal for all simulated colors, but
291: again this is an extreme case.
292: % and should not be considered very realistic.
293: The color distributions (as indicated by the peaks in Table \ref{tab_spot})
294: obtained from the most extreme simulations are not typical of those
295: normally observed for extragalactic GC systems.
296:
297: In order to refine our study, we now focus on those simulations best
298: matching real GC systems. We have compared the color peaks from Table
299: \ref{tab_spot} with observed color peaks from literature. In
300: particular, we have selected as ``realistic'' the simulations with:
301: \begin{itemize}
302: \item $(V-I)_{0,blue} \sim 0.95$ and $(V-I)_{0,red}\sim 1.15$, based
303: on the \citet{brodie06} compilation
304: for bright (mainly E and S0) galaxies with $M_B \leq -18.5$~mag.
305:
306: \item $(B-I)_{0,blue} \sim 1.94$ and $(B-I)_{0,red} \sim 2.06$,
307: derived from the \citet{harris06} sample of bright galaxies.
308:
309: \item $(I-H)_{0,blue} \sim 1.3$ and $(I-H)_{0,red} \sim 1.7$,
310: from \citet{kundu07}, based on M\,87.
311:
312: \end{itemize}
313:
314: In order to avoid any bias towards bimodal distributions, we have also
315: considered as realistic those unimodal color distributions whose peak
316: is equal to the averaged blue and red peak colors reported above. By
317: matching these criteria with the simulations, we have found that only
318: the subset of simulations with [Fe/H]$\,=-$1.15, $-$0.65
319: and $\sigma_{[Fe/H]}=0.5, 0.75$
320: provide realistic ranges for the color peaks. For example, the peaks
321: for the \vi, \bi, and \ih\ colors for the case of ([Fe/H]$\,=-$1.15,
322: $\sigma_{[Fe/H]}=0.75$) are all in good agreement with the observational
323: values listed above, even though most of these colors are found to
324: have unimodal color distributions for this particular simulation.
325:
326: By inspecting only the panels for the
327: \textit{\sffamily s}elected ``realistic'' simulations in
328: Figure~\ref{plotspot} (the panels labeled with an ``\textsf{S}''),
329: one can see that the colors \bv, \vi, and \bi\ have, on average,
330: an increased probability of being projected to a bimodal distribution,
331: while colors such as \ik, \vh and \vk\ have lower probabilities.
332: Thus, if one wants to minimize the bias from the MC
333: projection effect in real observations, i.e. if the
334: contribution to bimodality due to a nonlinear MC relation
335: {is} to be neglected, then the \ik, \vh, and \vk\
336: colors are to be preferred.
337:
338: Finally, we must emphasize that the above conclusions do not change if
339: we adopt different ages. In fact, although there is some shift in
340: color, the MC {\it profiles} are not strongly affected even when the
341: $t=5$ Gyr models are considered (Figure \ref{cmspot}). Since the
342: projection effect is due to the shape of the MC relation (that is, to
343: the changing derivative of the relation), and not to the absolute
344: color values, this explains why the outcome of the simulations does
345: not change significantly with age. In more detail, for the
346: \citet{raimondo05} models at an age of $t=5$ Gyr, the ``wavy'' MC
347: relation is mostly related to the appearance of the HB at
348: metallicities [Fe/H]$\lsim -0.4$ dex, while no HB is present at higher
349: [Fe/H]. Finally, it is worth noting that the above results do not
350: change substantially if the numerical experiments are carried out
351: using a different number of simulated GCs\footnote{We have found that
352: the locations of the color peaks change on average $\lsim$0.05 mag,
353: and $P(bimodal)$ by less than 25\% if $>$50 up to $\sim$2000 GC are
354: considered in the simulations. Numerical experiments with less than 50
355: sources can significantly deviate from the results given in Table
356: \ref{tab_spot}. Thus, our results should be compared with observations
357: that include color data for more than 50 GCs.}.
358:
359:
360: \subsection{Other SSP models}
361: The results presented in the previous Section are based on a
362: particular choice of the MC relations derived from the SPoT SSP
363: models. In order to verify the robustness of those results, in this
364: section we perform the same analysis discussed above, but with the MC
365: relations derived from four other sets of SSP models. We consider the
366: \citet{maraston05}, \citet{anders03}, \citet{bc03} and \citet{lee02}
367: models (hereafter Mar05, And03, BC03, and Lee02, respectively).
368:
369: We emphasize that, with the lone exception of the Lee02 models, the
370: quoted models are computed with the primary aim of deriving the
371: integrated photometric properties of stellar systems. This means that,
372: in contrast to the SPoT models, they are not constrained to match
373: as well with the specific features of observed CMDs. As a
374: consequence, the detailed shape of the MC relation may not take
375: into account the effect of stars in a particular evolutionary
376: phase, which is a key point for a detailed modeling of the MC
377: relations. Keeping in mind this warning, we perform for these models
378: the same analysis discussed above for the SPoT models.
379: %%
380: For these simulations we again adopt a model age of 13 Gyr, except for the
381: Lee02 models, which do not include this age, so we use their 14~Gyr models.
382: The results of the simulations are presented in Figure \ref{plotabsel}, where we
383: show only the results for the selected ``realistic'' simulations, although this
384: choice does not substantially affect the our conclusions.
385:
386: Inspecting the panels of Figure \ref{plotabsel}, we find some
387: differences with the results based on the SPoT models. For example,
388: the values of $P(bimodal)$ are low for the BC03 \bv, \vi, and \bi\ colors
389: distributions.
390: This result for the BC03 models is not surprising, due to their lack of
391: detailed HB morphology modeling, which is the main cause of the wavy
392: MC relations for these colors. In contrast, the \jk\ colors
393: from the BC03 models are almost always bimodal, a result of some
394: non-linearity in their MC relation unrelated to HB morphology.
395: %
396: On the other hand, the Lee02 models, where nonlinear effects in the
397: MC relation are stronger with respect to other models, generally
398: predict higher $P(bimodal)$ values.\footnote{We chose the 14 Gyr
399: Lee02 models specifically because the ``wavy'' feature is more pronounced;
400: this allows us to highlight better the influence of such features
401: on the color distributions. The MC relations for the Lee02 preferred
402: 12~Gyr models give results more similar to the SPoT ones.}.
403:
404: By making a cross-check of the results based on these sets of SSP
405: models with the ones based on the SPoT models, we find that no one
406: color is completely unaffected by MC projection bias.
407: However, in almost all cases the \vh\ and \vk\ colors are predicted to be
408: less affected by this bias. Thus, these mixed optical--IR colors should be preferred
409: for GC studies, since, in normal galaxies, a bimodal distribution in these
410: colors is more likely linked to an underlying bimodal metallicity distribution.
411:
412:
413: \section{Conclusions}
414:
415: In this work we have performed a series of numerical experiments to
416: simulate the properties of GC populations observed in different
417: photometric colors. Our aim was to study how the nonlinear behavior of
418: the MC relations affect a unimodal (Gaussian) metallicity distribution
419: when it is projected into various optical and near-IR color distributions.
420: %%
421: By using the MC relations from the SPoT models, we have found that
422: {\it a unimodal metallicity distribution can be projected into a
423: bimodal color distribution in almost any of the colors considered here,
424: depending on the properties of the metallicity distribution, on the
425: particular color index, and on the photometric uncertainty of the sample.}
426:
427: This result is due to the fact that all the MC relations are by and large
428: nonlinear. To reduce the possibility of this bias in real data, and thus
429: help ensure that an observed bimodal color distribution is due
430: to a bimodal metallicity, one should choose a color whose MC relation
431: is nearly linear. Since, for the grid of colors that we have
432: considered here, there is no such ``unbiased'' color, the best colors
433: to use are those that are most robust against this effect.
434: Using the SPoT models, we have concluded that optical--to--near-IR
435: colors are the best choices to disclose real bimodal
436: metallicity distributions.
437:
438: In order to assess model systematics and make more firm conclusions,
439: we have also investigated several other sets of stellar population models.
440: As a general result, the differences existing
441: between model predictions do not allow us to pick any color index as
442: safely unaffected by the metallicity projection bias. However, all
443: models considered here, including the SPoT ones, predict that the bias
444: effect is reduced for \vk, \vh, and similar colors.
445: %%
446: One other result of these simulations is that photometric uncertainties
447: can affect, in surprising ways, the probability of obtaining a bimodal
448: color distribution from the KMM algorithm. Thus, decreasing the statistical
449: errors in real color data can help to avoid false detection of
450: significant bimodality.
451:
452: Further information on metallicity bimodality can of course come from the
453: analysis of spectroscopic data for a significant number of GCs in galaxies
454: with observed color bimodality. However, such observations are time consuming,
455: and only feasible for relatively nearby objects.
456: In addition, certain spectroscopic indices may themselves be affected by
457: similar nonlinear relations with metallicity.
458:
459: In conclusion, we confirm \vh\ and \vk\ as good colors to reveal (nearly)
460: unbiased bimodal metallicity distributions in extragalactic GC systems.
461: Future data on large GC samples in individual galaxies, including
462: optical and near-IR photometry, as well as spectroscopy, coupled with
463: further advances in stellar population modeling, should finally resolve
464: this issue. Until that time, the interpretation of bimodal color
465: distributions will remain, at least in part, ambiguous.
466:
467:
468: \acknowledgments
469:
470: We thank the anonymous referee for helping us to improve this paper
471: with useful suggestions. We would like to thank Eric Peng, Pat C{\^
472: o}t{\' e}, and Gabriella Raimondo for useful comments. This research
473: was supported by the NASA grant AR-10642, and the paper was completed
474: under the sponsorship of a INAF-OA Teramo grant.
475:
476:
477: \clearpage
478:
479:
480: \begin{figure}
481: \epsscale{1.} \plotone{fig1.eps}
482: \caption{The Teramo-SPoT models compared to observational data. The
483: models refer to three different ages: 5 Gyr (dashed lines), 11 Gyr
484: (dotted lines) and 13 Gyr (solid lines, reference models). The gray
485: dots mark Galactic GC data.
486: \label{cmspot}}
487: \end{figure}
488:
489:
490: \begin{figure}
491: \epsscale{1.1}
492: \plottwo{fig2a.eps}{fig2b.eps}
493: \caption{\textit{Left panel:} The color histograms obtained from a
494: unimodal (Gaussian) metallicity distribution with mean
495: [Fe/H]$\,=-0.65$, $\sigma_{\rm [Fe/H]}=0.5$ (shown in the upper left
496: panel). We find that all the optical and the \jk\ color distributions
497: are bimodal based on the KMM algorithm, while for the other colors a
498: double Gaussian distribution does not significantly improve the fit to
499: the data. \textit{Right panel:} Simulation of a bimodal metallicity
500: distribution (solid lines), chosen to match the Galactic GC
501: distribution. The observed color histograms for the Galactic GCs
502: are also shown with dotted lines (the observed metallicity
503: distribution is shown in the upper left panel with dotted histogram).
504: The parameters used for the simulated distribution are
505: [Fe/H]$_{low}=-1.59$, $\sigma_{[Fe/H],low}=0.30$ dex, and
506: [Fe/H]$_{high}=-0.55$, $\sigma_{[Fe/H],high}=0.27$ dex, with a
507: photometric uncertainty of 10\%, and the N$_{low}$ to N$_{high}$ ratio
508: is 2.
509: \label{bimodal}}
510: \end{figure}
511:
512: \begin{figure}
513: \epsscale{1.}
514: \plotone{fig3.eps}
515: \caption{The probability $P(bimodal)$ of having a bimodal color distribution
516: starting from a unimodal Gaussian metallicity distribution is
517: shown for different color indices and various metallicity
518: distributions. Left/middle/right panels refer to
519: simulations with $\sigma=0.25,0.5,0.75$ dex, respectively (see upper
520: labels). Different rows refer to different mean metallicities, as
521: labeled. High\,(low) values of $P(bimodal)$ mean that the color distribution
522: is significantly bimodal\,(unimodal). Filled dots mark numerical experiments without any
523: color scatter, and open circles show simulations including a 10\% color
524: scatter. Although all the simulated metallicity distributions are unimodal,
525: about 45\% of these color distributions are found to be bimodal.
526: The four panels with the ``\textsf{S}'' label refer to the simulations
527: that best match with observed GC color ranges.
528: \label{plotspot}}
529: \end{figure}
530:
531:
532: \begin{figure}
533: \epsscale{1.}
534: \plottwo{fig4a.eps}{fig4b.eps}
535: \plottwo{fig4c.eps}{fig4d.eps}
536: \caption{Same as Figure \ref{plotspot}, but for MC relations taken from
537: other sets of SSP models (labeled at the top of each set of panels).
538: Only the results for the simulations matching with observed GC colors are shown.
539: \label{plotabsel}}
540: \end{figure}
541:
542:
543:
544: \clearpage
545: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
546: %%%%%%%%%%%TABLES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
547: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
548:
549: \input{tab1}
550:
551: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
552: %%%%%%%%%%%BIBLIO %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
553: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
554:
555:
556: \bibliographystyle{apj}
557: \bibliography{cantiello_jun07}
558:
559: \end{document}
560:
561: