NGC 1866: a workbench for stellar evolution
Extracted source text ยท arXiv astro-ph/0202128
% AA vers. 4.01, LaTeX class for Astronomy & Astrophysics
%
%(c) Springer-Verlag HD
%-----------------------------------------------------------------------
%\documentclass[referee]{aa} % for a referee version
\documentclass{aa}
\usepackage{graphicx}
\usepackage{natbib}
\begin{document}
%\thesaurus{ 08(08.05.3;08.09.3;08.08.1) }
\title{NGC 1866: a workbench for stellar evolution}
\author{R. Barmina, L. Girardi, and
C. Chiosi}
\institute{
Department of Astronomy, University of Padova,
Vicolo dell'Osservatorio 2, 35122 Padova, Italy }
\offprints{Roberto Barmina \\
e-mail: barmina@usm.uni-muenchen.de}
\date{Received {August, 2001}. Accepted {} }
\authorrunning{Barmina et al.}
\titlerunning{NGC 1866: a workbench for stellar evolution}
%\maketitle
\abstract{
NGC 1866 is a young, rich star cluster in the Large Magellanic Cloud.
Since the cluster is
very well populated both in the main sequence and post main sequence
stages, thus providing us with a statistically complete
sample of objects throughout the various evolutionary phases of
intermediate mass stars,
it represents a good laboratory
for testing stellar evolutionary models.
More precisely, NGC 1866 can be used to discriminate among
classical stellar models, in which the extension of the
convective regions is fixed
by the classical Schwarzschild criterion, from models with
overshooting, in which an ``extra-mixing'' is considered to take
place beyond the classical limit of the convective zone.
Addressing this subject in a recent work, \citet{Te99} reached the
conclusion that the classical scheme for the treatment
of convection represents a good and sufficient approximation for
convective interiors.
Using their own data, we repeat here the analysis. First we revise
the procedure followed by \citet{Te99} to correct the data for
completeness,
second we calculate new stellar models with updated physical input
for both evolutionary schemes, finally we present many simulations
of the colour-magnitude diagrams and
luminosity functions of the cluster using the ratio of the
integrated luminosity
function of main sequence stars to the number of
giants as the normalization factor of the simulations. We also take
into account several possible physical agents that could alter the
color-magnitude diagram and the luminosity function: they are unresolved
binary stars, dispersion in the age, stochastic effects in the initial
mass function. Their effect is analyzed separately, with the conclusion
that binary stars have the largest impact.
The main result of this study is that
the convective overshoot hypothesis (together with a suitable
percentage of
unresolved binaries) is really needed to fully match the whole
pattern of data. The main drawback of the
classical models is that they cannot reproduce the correct ratio of
main sequence to post-main sequence stars.
%Adopting a metallicity $Z\simeq 0.008$ and
%an intrinsic distance modulus $(m-M)_{\rm 0}=18.5$, the
%estimated NGC~1866 age is about 140 -- 160 Myr.
\keywords{Stars: evolution -- Stars: interiors --
Stars: Hertzsprung-Russell
diagram}
}
\maketitle
\section{Introduction}\label{introd}
In the context of stellar theories, it has long been acknowledged that the
penetration of convective elements beyond the classical limit set by
the Schwarzschild criterion could produce non-negligible effects on the
stellar structure.
The young star cluster NGC 1866 in the Large Magellanic Cloud (LMC)
is considered as an ideal laboratory for testing stellar models,
especially concerning the extension of convection in the
interiors of real stars, and discriminating among classical and so-called
overshooting models.
This cluster, well populated both in main sequence and giants
stars, represents a statistically complete sample of objects,
contrary to the young clusters of the Milky Way (that are
relatively poor, especially in
giant stars). Furthermore, NGC 1866 is young enough ($\simeq$100 Myr) to
possess stars which
develop a convective core in the phase of central H-burning
($M_{\rm TO}=4 - 5$ M$_{\odot}$), thus providing us the way
to test the extension of overshooting. The theory of stellar structure
predicts that convective cores in main sequence stars set in starting
from initial masses greater than
$M \simeq 1.1$ M$_{\odot}$.
For these reasons NGC 1866 has been the subject of several analyses,
aiming at testing how far convective elements overshoot from the core
into the surrounding stable layers.
\citet{BM83}, comparing observational and
synthetic colour-magnitude diagrams (CMD) of NGC 1866, obtained the
best fit to the data simulating a cluster of about $86$ Myr:
the general features of the CMD were well-fitted on the whole, but the
authors argued that the classical stellar models they used predicted
a number of red giants larger than observed, and a smaller number of
main sequence stars in turn as compared to the observations.
The same authors suggested that a more careful treatment of core convection
(i.e. larger convective cores) could remove the discrepancy.
A number of subsequent studies confirmed this idea and demonstrated
that the inclusion of overshooting in the description of convective motions
could reproduce the correct ratio $N_{\rm PMS}/N_{\rm MS}$.
\citet{C89a}, in particular, clearly showed that the overshooting scheme,
by reducing the ratio $t_{\rm He}/t_{\rm H}$, offers a good and simple
solution to the problem. This conclusion was also reinforced by
\citet{L91}, who analyzed the CMD of NGC 1866 by means of overshooting and
classical models \footnote{It is worth recalling here that models of
intermediate mass stars calculated with the
classical Schwarzschild criterion are known to develop during
the core He-burning
phase the so-called He-semi-convective instability, i.e. a region surrounding
the fully convective core in which the condition of neutrality
$\nabla_R=\nabla_A$ is maintained by suitably modifying the profile of
chemical composition. This is the physical analog of what happens in massive
stars during the core H-burning phase, which develop the so-called
H-semi-convective instability. More details on the physical
origins of both H- and He-semi-convection,
can be found in \citet{C92}. Let it suffice here to mention
that in intermediate-mass stars
the effects of semiconvection are negligible, if compared to the effects
of convective overshooting. Hereinafter this type of
stellar models are
referred to as the classical, semi-convective models.}.
Over the years,
sophisticated formulations of convective overshooting have been elaborated
that are based on turbulence theories \citep{Cloutman80,Xiong80} and
fluid hydro-dynamics \citep{Canuto91,Canuto96,UnnoKondo89}. However, the
ballistic approach at the problem proposed by
\citet{Br81} turns out to be fully adequate and it has been
proved to best reproduce the
numerical results of laboratory fluid-dynamics simulations
\citep{Zahn91}. The \citet{Br81} algorithm adopts a
non-local treatment of convection in the context of the mixing-length
theory (MLT) by \citet{Bohm55}: it looks for the layers where the
velocity of convective elements (accelerated by the buoyancy force
in the formally unstable regions) becomes zero in the surrounding
stable regions, then adopts a suitable temperature stratification in the
overshooting regions, and finally assumes straight mixing over-there.
Since the \citet{Br81} formalism makes use of the MLT, it expresses
the mean free
path of the convective elements as $l=\Lambda_{\rm c} \times H_{\rm p}$ where
$H_{\rm p}$ is the local pressure scale height.
Although some criticism has been advanced by \citet{Renzini87}
--who erroneously
concludes that the ``miscellaneous'' of local and non local
formalisms leads to overestimating the overshooting distance-- we remind
the reader that
overshooting is simply a logical consequence of the inertia principle,
so that neglecting its existence would not be physically sound.
It is worth recalling that convective
overshooting is quite common
in nature \citep{Deardorff69}, and it has been demonstrated
in a number of studies, including numerical simulations
\citep{Freytag96}, that
the penetration depth of convective elements into a formally
stable region represents a non-negligible fraction of the size of
the unstable zone.
In addition to this, there are a number of astrophysical situations
in which the hypothesis of substantial convective overshooting
has been found to offer better and more elegant solutions than
other explanations
\citep[see][]{Be86,C92}. Among others, it suffices to recall here
the so-called mass discrepancy of Cepheid stars \citep{Bw93,C92}.
Despite this, it has been often argued that
unresolved binaries could mimic the effects of convective overshooting
and easily
account for the low ratio of red giant to main sequence stars
observed in the young LMC clusters.
This hypothesis has been investigated by many authors,
both in open clusters of the Milky Way \citep[see][]{Ca94} and
young clusters of the
LMC \citep{C89a,C89b,VaC92} with somewhat contrasting results.
In their study, \citet{Te99} emphasize the role
of unresolved
binaries in solving the problem, reaching
the conclusion that overshooting is not needed. They
obtain agreement with the observational data only by
introducing a fraction of about $30\,\%$ of binary stars
in a population of $\simeq 100$ Myr,
and using the classical semi-convective models of \citet{Dominguez99}.
In contrast, models computed with ``enlarged'' convective cores
(i.e. cores in which convection extends beyond
the Schwarzschild criterion to simulate overshooting),
lead to worse fits of the observational data, especially when
binary stars are included.
Unfortunately, \citet{Te99} adopt a
wrong procedure to correct the observational star counts for photometric
completeness (see the section below),
thus obtaining differential ($N_{\rm MS}$, shortly indicated with
DLF) and integrated luminosity functions
($\Sigma N_{\rm MS}$, shortly referred to as ILF) for the
main sequence stars of NGC 1866 that are inconsistent.
Furthermore, \citet{Te99} fail to make use of the ILF normalized
to the number of evolved stars ($\Sigma N_{\rm MS}/N_{\rm PMS}$,
hereinafter indicated as N-ILF) introduced by \citet{C89a},
which has been proved
to be the only way to effectively discriminate
between the two different evolutionary schemes. It is worth reminding the
reader that the N-ILF is by definition proportional to the lifetime ratio
$t_{\rm H}/t_{\rm He}$.
In this paper, first we correct the analysis of the observational data
and second we show that the conclusions reached by \citet{Te99} entirely
depend on the kind of diagnostic they have adopted. Exploring
the relative effects of core overshoot and unresolved
binary star frequency on intermediate age cluster CMD's will be
a side product of our analysis.
Starting from the same observational data used by
\cite{Te99}\footnote{All the photometric data used in this study were
kindly provided by \cite{Te99} already reduced and calibrated.},
and correctly applying the completeness
correction, in this paper we shall discriminate --by means
of simulated CMDs and LFs of
NGC 1866-- between classical semi-convective and
overshooting models.
After a brief description of the observational data (Sect.
\ref{sec:data}), in Sect. \ref{sec:tracks} we
compare the stellar models calculated
according to the two schemes, with particular attention to
the values of critical masses and lifetime ratios. These latter are
indeed the crucial point to test the validity of stellar models.
In Sect.
\ref{sec:isochrones} we describe the isochrones and the methods used
to simulate CMDs and LFs. In Sect.
\ref{sec:simulations} we compare the results of these simulations with
their observational counterparts. In Section \ref{why:testa} we
thoroughly
discuss the reasons why our conclusions differ from those by \citet{Te99}.
Finally, in Sect. \ref{sec:conclusions}
we summarize the main results of this study.
%%%%%%%%%%%%%%%%Figure 1
\begin{figure}%%%%%%%%%[b]
\resizebox{\hsize}{!}{\includegraphics{fig1.eps}}
\caption{CMD for stars in the adopted sample,
after statistical de-contamination \citep{Te99}}
\label{fig:1866_grid}
\end{figure}
\section{Observational data}\label{sec:data}
The observational data \citep{Te99} have been obtained with the
$2.2$ m ESO telescope at La Silla, Chile, on 1993 January 27. Two fields
have been observed in $B$ and $V$ pass-bands: the
first field is centered on the cluster, and the second one is
located at $\sim 6^\prime$ from it. The second field is
utilized for statistical field subtraction: the corrected and calibrated
CMD of \citet{Te99} is presented in Fig. \ref{fig:1866_grid}, in which
the main sequence stars are separated from
giants by dashed lines.
%%%%%%%%%%%%%%Table 1 %%%%%%%completeness factors%
\begin{table*}
\caption{\small Luminosity functions for main sequence stars
in NGC~1866 as adopted in \citet{Te99}: differential and
integrated
luminosity functions before the completeness
correction (columns 2 ad 3), the completeness factors (column 4),
differential and integrated luminosity functions after the
completeness
correction (column 5 and 6).
The last three columns present the corrected
luminosity functions as obtained independently by us, starting from the
original \citet{Te99} data.}
\label{tab:factors}
\begin{center}
%\vspace{9pt}
\begin{tabular}{c|rrcrr|rrr}
\hline\noalign{\smallskip}
\multicolumn{1}{c|}{} & \multicolumn{5}{c|}{ \citet{Te99}} &
\multicolumn{3}{c}{ This work} \\
\hline\noalign{\smallskip}
(1) & (2) & (3) & (4) & (5) & (6) & (7) & (8) & (9) \\
\noalign{\smallskip}\hline\noalign{\smallskip}
$m_{\rm V}$ & $N_{\rm MS}$ & $\Sigma N_{\rm MS}$ & $\lambda_{\rm c}$
& $(N_{\rm MS})_{\rm corr}$ & $\Sigma (N_{\rm MS})_{\rm corr}$
& $N_{\rm MS}$
& $(N_{\rm MS})_{\rm corr}$ & $\Sigma (N_{\rm MS})_{\rm corr}$\\
\noalign{\smallskip}\hline\noalign{\smallskip}
16.5 & 0 & 0 & 1.000 & 0 & 0 & 3 & 3 & 3 \\
16.7 & 2 & 2 & 1.000 & 2 & 2 & 3 & 3 & 6 \\
16.9 & 9 & 11 & 1.000 & 9 & 11 & 10 & 10 & 16 \\
17.1 & 17 & 28 & 1.000 & 17 & 28 & 20 & 20 & 36 \\
17.3 & 37 & 65 & 1.000 & 37 & 65 & 37 & 37 & 73 \\
17.5 & 29 & 94 & 1.000 & 29 & 94 & 33 & 33 & 106 \\
17.7 & 56 & 150 & 1.000 & 56 & 150 & 59 & 59 & 165 \\
17.9 & 55 & 205 & 1.000 & 55 & 205 & 61 & 62 & 227 \\
18.1 & 69 & 274 & 0.996 & 69 & 275 & 74 & 74 & 301 \\
18.3 & 90 & 364 & 0.992 & 90 & 367 & 84 & 85 & 386 \\
18.5 & 116 & 480 & 0.988 & 117 & 486 & 129 & 131 & 517 \\
18.7 & 135 & 615 & 0.981 & 137 & 627 & 144 & 146 & 663 \\
18.9 & 160 & 775 & 0.975 & 163 & 795 & 167 & 172 & 835 \\
19.1 & 164 & 939 & 0.966 & 168 & 972 & 174 & 180 & 1015 \\
19.3 & 188 & 1127 & 0.958 & 194 & 1177 & 208 & 217 & 1232 \\
19.5 & 215 & 1342 & 0.950 & 222 & 1413 & 227 & 238 & 1470 \\
19.7 & 266 & 1608 & 0.939 & 279 & 1712 & 276 & 292 & 1762 \\
19.9 & 250 & 1858 & 0.929 & 263 & 2001 & 256 & 274 & 2036 \\
20.1 & 274 & 2132 & 0.916 & 292 & 2327 & 274 & 297 & 2333 \\
20.3 & 323 & 2455 & 0.903 & 352 & 2720 & 313 & 346 & 2679 \\
20.5 & 303 & 2758 & 0.889 & 332 & 3101 & 316 & 351 & 3030 \\
20.7 & 322 & 3080 & 0.874 & 355 & 3524 & 299 & 334 & 3364 \\
20.9 & 302 & 3382 & 0.859 & 336 & 3939 & 295 & 331 & 3695 \\
21.1 & 290 & 3672 & 0.842 & 323 & 4362 & 282 & 318 & 4013 \\
21.3 & 316 & 3988 & 0.823 & 358 & 4843 & 289 & 332 & 4345 \\
21.5 & 261 & 4249 & 0.803 & 299 & 5290 & 238 & 278 & 4623 \\
21.7 & 86 & 4335 & 0.776 & 100 & 5587 & 88 & 106 & 4729 \\
21.9 & 16 & 4351 & 0.744 & 29 & 5850 & 36 & 50 & 4779 \\
\noalign{\smallskip}\hline\noalign{\smallskip}
\end{tabular}
\end{center}
\end{table*}
\subsection{Correction for completeness}
In their analysis, \citet{Te99} divide the sample in six
concentric rings centered on the cluster,
and calculate the completeness factors, $\lambda^i_{\rm c}$,
ring by ring. Since $\lambda^i_{\rm c}$ is equal to the number
ratio of artificial stars added and recovered in a given
magnitude bin, they should be used to correct the
DLF -- i.e.\ the number of stars per
magnitude bin -- separately for each ring $i$:
\begin{equation}
(N^i)_{\rm corr} = N^i / \lambda^i_{\rm c}
\label{eq:compl}
\end{equation}
where $N^i$ and $(N^i)_{\rm corr}$ stand for the observed
and corrected numbers, respectively.
However, a substantial error is made by \citet{Te99}:
they use the same $\lambda_{\rm c}$ factors to
{\em correct both the number of stars
and the ILF at each magnitude bin}. This
clearly produces the correct DLF $(N^i)_{\rm corr}$, but
over-estimates the ILF and N-ILF
at fainter magnitude bins.
This can be appreciated by looking at the first 6 columns
of Table \ref{tab:factors}, where we
present the original star counts and the luminosity functions used by
\citet{Te99}\footnote{These data have been kindly provided by
V.\ Testa.}. In particular, it can be noticed that, starting from
the uncorrected DLF and ILF of columns 2 and 3, the same ``average''
$\lambda_{\rm c}$ factors (column 4) have been used to derive
the corrected DLF shown in column 5 (i.e.\
$(N_{\rm MS})_{\rm corr} = N_{\rm MS} / \lambda_{\rm c}$), and the
corrected ILF shown in column 6 (i.e.\
$\Sigma(N_{\rm MS})_{\rm corr} = \Sigma N_{\rm MS} / \lambda_{\rm c}$).
The corrected DLF and ILF obtained in such a way are no longer
consistent with each other.
Noticing this,
we decided to completely re-derive the luminosity functions
for NGC~1866,
starting from the statistically decontaminated
photometric data of \citet{Te99} (see
Fig.~\ref{fig:1866_grid}), and the completeness factors presented in
their Fig.~6. In short, we have obtained the
DLF for each annulus around NGC~1866,
and corrected them using the corresponding completeness factors
separately, according to Eq.~(\ref{eq:compl}).
Then, the DLFs for annulus 2 to 5
were added
\begin{equation}
(N_{\rm MS})_{\rm corr} = \sum_{i=2}^5 (N^i_{\rm MS})_{\rm corr}
\end{equation}
producing the total DLF presented in column 8 of
Table \ref{tab:factors}. (For the sake of comparison, column 7 of the
same table presents the counts obtained by adding stars from
annulus 2 to 5, without applying the completeness correction.)
>From this corrected DLF we derive the total ILF of
column 9 of the same table, by simply summing up the number of stars
above each magnitude bin. As can be readily seen, our final
numbers for the DLF are roughly consistent with those of
\citet{Te99}, but the two ILFs are sizeably different, especially for
the faintest magnitude bins.
There is a final remark to be made for the sake of clarity. Comparing
columns (2) and (7), which contain the rough counts in the various magnitude
bins, a marginal disagreement is evident: our counts
do not
exactly coincide with those by \citet{Te99}. The difference is due to the
slightly different criteria adopted to define the numbers $N_{MS}$ of stars
per magnitude
interval. \citet{Te99} follow a complicated scheme in which (i) the mean
location of the MS together with its color dispersion $\sigma$
are derived as a
function of the magnitude, and (ii)
in every magnitude bin only stars falling within 7$\sigma$ from the
mean MS are taken into account.
In contrast, we prefer to consider all stars falling within the
considered magnitude interval. As a consequence of this, the $N_{MS}$ of
\citet{Te99} are slightly smaller than ours because stars far away from
the mean MS are neglected.
In Fig. \ref{fig:conf_corrections} we compare our results
for the corrected ILF (continuous line) with those obtained
by \citet{Te99} (dashed line). It is important to stress here that
one of reasons why our results differ from those by \citet{Te99},
resides in the different (and erroneous) way of applying the
completeness corrections used by them.
%%%%%%%%%%%Figure 2
\begin{figure}%%%%%%[h]
%\resizebox{\hsize}{!}{\includegraphics{blank.ps}}
\resizebox{\hsize}{!}{\includegraphics{conf_corrections.eps}}
\caption{Integrated Luminosity Function of the main sequence stars of
NGC 1866 before (dotted lines) and after (thick line) correcting
for photometric completeness. The dashed line shows the
same luminosity function calculated by \citet{Te99}}
\label{fig:conf_corrections}
\end{figure}
Finally, we summarize in Table \ref{tab:counts} the total
number of main sequence
($\Sigma N_{\rm MS})$ and post main sequence ($N_{\rm PMS})$ stars
before and after
the completeness correction is applied.
The number of post main sequence stars amounts to $N_{\rm PMS}=100$.
\section{Stellar tracks}\label{sec:tracks}
We have calculated two sets of stellar tracks with the classical
semi-convective scheme and
initial chemical composition [$Y$=0.240,$Z$=0.004] and
[$Y$=0.250,$Z$=0.008],
and one set of models with convective overshooting and composition
[$Y$=0.240,$Z$=0.004]. The initial masses of the models range
from $1.0$ to $8$ M$_{\odot}$. The input physics is the
same as in \citet{Gi00} and subsequently updated by \citet{Sa00}, to whom
the reader should refer for all details.
For purposes of comparison, whenever required we also
utilize the stellar models by \citet{Sa00}
calculated with overshooting and chemical composition
[$Y$=0.250,$Z$=0.008].
%%%%%%%%%%%%%%Table 2
\begin{table}%%%%%[h]
\caption{\small Star counts for main sequence ($\Sigma N_{\rm MS}$)
and post main sequence stars ($N_{\rm PMS}$), before and after
completeness correction.}\label{tab:counts}
\begin{center}
\begin{tabular}{l|cc}
\hline\noalign{\smallskip}
& $\Sigma N_{\rm MS}$ & $N_{\rm PMS}$\\
\noalign{\smallskip}\hline\noalign{\smallskip}
observed & 4395 & 100\\
corrected & 4779 & 100\\
\noalign{\smallskip}\hline
\end{tabular}
\end{center}
\end{table}
\subsection{Prescriptions for semi-convection and overshooting}
The algorithm dealing with semi-convection in classical models
strictly follows the recipes
given by \citet{Al93} to whom the reader should refer for all details.
Convective overshooting is based on the formulation by \citet{Br81} and the
more recent revision by \citet{Be85} and \citet{Al93}. Suffice it to recall
here that the overshooting parameter $\Lambda$ is chosen to be
$\Lambda_{\rm c}$=0.5 for core convection and
$\Lambda_{\rm e}$=0.7 for envelope
convection \citep[see][]{Al93}.
For the sake of clarity and to avoid misunderstanding, we remind the
reader that $\Lambda_{\rm c}$=0.5 in the formalism of \citet{Br81} is
equivalent to $\Lambda_{\rm c}$=0.25 in the description of
\citet{MaederMey91}.
More precisely, while \citet{Br81} define the overshooting
distance as the path
traveled by a convective element across the Schwarzschild border (half
distance beneath and half above the border), \citet{MaederMey91}
define it only as the distance above the border.
Furthermore, the choice of $\Lambda_{\rm c}$=0.5
we have adopted is the same as in previous studies aimed at
generating stellar models best suited to interpret observational data of
star clusters.
To mention a few we recall \citet{Al93}, \citet{Br93a},
\citet{MaederMey91}, \citet{Meynet94}, \citet{C92}, \citet{Chiosi99} and
references therein. Exploring the effects of core overshooting for other
choices of $\Lambda_{\rm c}$ is beyond the scope of this study.
%%%%%%%%%%%Figure 3
\begin{figure}%%%%%[b]
%\resizebox{\hsize}{!}{\includegraphics{blank.ps}}
\resizebox{\hsize}{!}{\includegraphics{conf1_schema_im8.eps}}
\caption{\small The stellar tracks for the $2.5$ e $4$ M$_{\odot}$ stars
with composition [$Y$=0.250, $Z$=0.008]. The solid lines are the stellar
models with convective overshooting, whereas the dotted lines are the
models with the classical extension of the convective core}
\label{fig:conf_tracks}
\end{figure}
\subsection{Critical masses $M_{\rm HeF}$, $M_{\rm up}$}
As long ago pointed out by
\citet{C89a}, see also \citet{C92}, in the presence of core overshooting
the minimum initial masses,
below which core He-flash
and core C-ignition in highly degenerate material occur, $M_{\rm HeF}$
and $M_{\rm up}$ respectively, get smaller.
Table \ref{tab:critical_masses} summarizes the results obtained
for classical semi-convective and overshooting models with metallicity
$Z$=0.004 and $Z$=0.008. The same result holds good for both values of
$Z$.
%%%%%%%%%%%%%%%Table 3
\begin{table}%%%%%[h]
\caption{Critical masses $M_{\rm HeF}$ e $M_{\rm up}$ --
in units of M$_{\odot}$ -- for classical semi-convective and
overshooting models.}
\label{tab:critical_masses}
\begin{center}
\begin{tabular}{c | c c}
\hline\noalign{\smallskip}
& $M_{\rm HeF}$ & $M_{\rm up}$\\
\noalign{\smallskip}\hline\noalign{\smallskip}
classical & $2.2 - 2.5$ & $>8$\\
overshooting & $1.8 - 2.0$ & $\simeq 7$\\
\noalign{\smallskip}\hline
\end{tabular}
\end{center}
\end{table}
%%%%%%%Table 4 tempi classici %%%
\begin{table}%%%%%%[h]
\caption{Lifetimes for classical models with chemical composition
[$Y$=0.250,$Z$=0.008]. The lifetimes are in Gyr.}
\label{tab:classic_times}
\begin{center}
\begin{tabular}{c c c c}
\hline\noalign{\smallskip}
$M/M_{\odot}$ & $t_{\rm H}$ & $t_{\rm He}$ & $t_{\rm He}/t_{\rm H}$\\
\noalign{\smallskip}\hline\noalign{\smallskip}
2.5 & 0.4094 & 0.2436 & 0.595\\
3.5 & 0.1769 & 0.1293 & 0.389\\
4.0 & 0.1271 & 0.0688 & 0.340\\
5.0 & 0.0775 & 0.0219 & 0.283\\
6.0 & 0.0524 & 0.0126 & 0.241\\
7.0 & 0.0385 & 0.0079 & 0.208\\
8.0 & 0.0299 & 0.0056 & 0.189\\
\noalign{\smallskip}\hline
\end{tabular}
\end{center}
\end{table}
%%%%%%%%%Table 5 tempi overshooting %%%
\begin{table}%%%%%%%[h]
\caption{The same as in Table~\protect\ref{tab:classic_times} but for the
overshooting models.}
\label{tab:overs_times}
\begin{center}
\begin{tabular}{c c c c}
\hline\noalign{\smallskip}
$M/M_{\odot}$ & $t_{\rm H}$ & $t_{\rm He}$ & $t_{\rm He}/t_{\rm H}$\\
\noalign{\smallskip}\hline\noalign{\smallskip}
2.5 & 0.5438 & 0.1376 & 0.253\\
3.5 & 0.2308 & 0.0380 & 0.165\\
4.0 & 0.1674 & 0.0237 & 0.141\\
5.0 & 0.0999 & 0.0116 & 0.116\\
6.0 & 0.0672 & 0.0065 & 0.096\\
7.0 & 0.0487 & 0.0043 & 0.090\\
8.0 & 0.0376 & 0.0032 & 0.085\\
\noalign{\smallskip}\hline
\end{tabular}
\end{center}
\end{table}
\subsection{Lifetimes and lifetime ratios}
Because of the increased size of the convective cores, the central
H-burning
lifetime $t_{\rm H}$ in the overshooting models is $30\,\%$ longer than in
classical ones, this
effect depending on the parameter $\Lambda_{\rm c}$ \citep{Br81}. The
``over-luminosity'' caused by overshooting during the core H-burning
phase (see Fig. \ref{fig:conf_tracks})
still remains during the He-burning phase (for stars with initial mass
greater
than $M_{\rm HeF}$): consequently, the lifetime of the He-burning phase
($t_{\rm He}$) gets shorter by about $40 - 45\,\%$. This, combined with
the
longer $t_{\rm H}$, results in a strong decrease ($\sim 55 - 60\,\%$)
of the ratio $t_{\rm He} / t_{\rm H}$.
Tables \ref{tab:classic_times} and \ref{tab:overs_times} summarize the
lifetimes for some of the $Z$=0.008 models calculated with the classical
and overshooting scheme, respectively.
\section{Isochrones and synthetic CMDs}\label{sec:isochrones}
\subsection{Isochrones}
For the above four groups of evolutionary tracks, we
have derived the isochrones, adopting the algorithm
of ``equivalent evolutionary points'' developed by \citet{Be94}. The age
ranges from about $0.05$ to $10$ Gyr.
Isochrones are calculated at $\Delta \log t=0.01$ intervals: this means
that any two consecutive isochrones differ by $\sim 2\,\%$ in their ages.
Along each isochrone we give: the initial and actual stellar
masses, the logarithm of surface luminosity, the
effective temperature and surface
gravity, the absolute bolometric magnitude, and the absolute magnitudes
in the $UBVRIJHK$ pass-bands. Finally we list the indefinite integral
over the initial mass $M$ of the initial mass function (IMF). For any
other detail see \citet{Gi00}.
Theoretical luminosities and effective temperatures are translated into
magnitudes and colours in the Johnson-Cousins system, using the
conversions of \citet{Be94}, which provides colours and
bolometric corrections as a function of the effective temperature and
gravity. The \citet{Be94} conversions are based on the
library of synthetic spectra obtained
from theoretical models of stellar atmospheres by \citet{Ku92}.
We only note here, as illustrated in Fig. \ref{fig:iso_conf_cl_ov}, that
an isochrone of the overshooting scheme runs at
higher luminosities as compared to a ``classical'' isochrone of the same
age, that is, in the presence of overshooting isochrones
of older ages are required to fit the observational data of a cluster.
%%%%%%%%%%%%%%Figure 4 confronto due isocrone %%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}%%%%%[h]
%\resizebox{\hsize}{!}{\includegraphics{blank.ps}}
\resizebox{\hsize}{!}{\includegraphics{iso_conf_cl_ov.eps}}
\caption{Isochrones of same age ($200$ Myr) and chemical composition,
[$Y$=0.250,$Z$=0.008] calculated with classical
(dashed line) and overshooting
models (solid line)}\label{fig:iso_conf_cl_ov}
\end{figure}
\subsection{Synthetic CMDs}
The synthetic CMDs are constructed by means of a
Monte Carlo algorithm, which randomly distributes stars along a
given isochrone according
to evolutionary lifetimes and IMF. The code takes into account the
photometric errors affecting the observational data
(derived from \citet{Te99} tables), by adding an artificial
dispersion to the theoretical magnitudes and colours, and generates
numerical simulations of CMDs allowing for effects due to:
age spread of the cluster population, different slopes of the IMF, and
presence of a certain fraction of unresolved binary stars with mass
ratio ranging in a given interval.
Since any plausible IMF preferentially populates the main sequence
region of an isochrone, in order to reproduce the right proportion
of main sequence to post main sequence stars --according
to the lifetimes in the various evolutionary phases-- we populate our
synthetic CMDs until an assigned number of post main
sequence stars $N_{\rm PMS}$ is matched: in particular
in each simulation presented here, we impose that
the total number of red giant stars present in the data is matched. In
such a way $N_{\rm PMS}$ is taken as the ``normalization'' parameter
of the simulations \citep[see][]{C89a,V91,L91}.
\subsection{Luminosity functions}
Special care is paid to the integrated number of stars from
the tip of main sequence band down to the current magnitude
interval, normalized to the number of red giant stars (i.e. the
N-ILF).
As pointed out by
\citet{C89a}, the N-ILF
can be directly compared with the ratio of lifetimes in the core H- and
He-burning phases, $t_{\rm H}$ and $t_{\rm He}$, respectively.
At fixed $N_{\rm PMS}$, the ratio in each simulation
$\Sigma N_{\rm MS}/N_{\rm PMS}$ depends on the age
--an older population clearly has a larger $N_{\rm PMS}$-- and
the theoretical model in use. As already
noted in previous sections, the size of the convective cores strongly
affects the ratio $t_{\rm He}/t_{\rm H}$, and in turn, the ratio
$\Sigma N_{\rm MS}/N_{\rm PMS}$. Thus the N-ILF, when compared to the
observational counterpart, reliably discriminates classical from
overshooting models.
Concluding this section, it is worth remarking that reproducing
synthetic CMDs with an assigned number of post main sequence stars
is by far preferable to simulating
CMDs with a fixed total number of stars brighter than
a certain magnitude. As a matter of fact, the
distribution of stars along the main sequence is driven almost
exclusively by
the IMF, evolutionary effects playing only a marginal role.
Therefore,
the LF (either DLF or ILF) of the main sequence stars,
if not normalized to the number of giants,
would only reflect the underlying
IMF, and would not be
affected by the effects of overshooting we want to test.
\section{The simulations}\label{sec:simulations}
In this section we present the CMDs and LFs resulting
from our simulations, and compare them with the observational ones, both
for classical semi-convective and overshooting models.
\subsection{Adopted parameters}
Before presenting the simulations, we briefly summarize the values
adopted for parameters such as reddening, metallicity, and distance
modulus.
\subsubsection{IMF}
All the simulations are calculated adopting the \citet{Salp55} law
$dN \propto M^{- \alpha}dM$ for the IMF and assuming,
unless otherwise specified, the parameter $\alpha =2.35$.
\subsubsection{Metallicity}
Rough estimates of the metallicity of NGC 1866 derived from the
$BVI_{\rm c}$ colors of the Cepheids \citep{Cald85,Feast89} yield
[Fe/H]$=-0.1\pm0.3$
\citep[see also ][]{Bw93}, finally the recent analysis
by \citet{Te99} yielded [Fe/H]$\simeq -0.35$. This roughly corresponds to
$Z=0.009$. Therefore, improving upon the metallicity used in the old
studies by \citet{C89a} and \citet{Broca89} who adopted
isochrones with [$Y$=0.280,$Z$=0.020] in the present simulations we
assume the composition [$Y$=0.250,$Z$=0.008].
\subsubsection{Distance modulus}
The distance modulus to the LMC is a hotly debated issue: we summarize
here some of the most recent determinations of this parameter.
Cepheid stars with {\it Hipparcos} parallaxes indicate
$(m-M)_{\rm 0}\sim18.6$ \citep{Feast00}.
The analysis of the expanding ring around
SN1987A yields $(m-M)_{\rm 0}=18.58 \pm 0.05$ \citep{Panagia98}.
The simultaneous study of the Cepheids stars and CMD of NGC 1866 by
\citet{Bw93}
yields $(m-M)_{\rm 0}=18.51\pm0.21$
In a recent study on young LMC clusters \citet{Keller00} utilize the
value $(m-M)_{\rm 0}=18.45$.
We adopt here $(m-M)_{\rm 0}=18.5$, as the
``classical'' value for the LMC distance modulus \citep{Westerlund97}.
Passing from $(m-M)_{\rm 0}=18.5$ to $(m-M)_{\rm 0}=18.6$ would not
significantly change the results of the present analysis.
\subsubsection{Reddening}
The study of the dust distribution across the LMC by means of
photometric and
spectroscopic data gives reddening maps from which one derives an
estimate of the mean value $E_{B-V}=0.16$ \citep{Oes96}, with a maximum
value of 0.29 (reached in the region of {\it 30 Dor}) and a minimum of
0.06 \citep{Z97}. While \citet{Te99}
assume $E_{B-V}$ in the range between $0.06$ to $0.10$, we prefer to
adopt the value
$0.10$ which, taking the total visual extinction $A_{V}$ to be
$3.1\times E_{B-V}$ \citep{Rieke85}, corresponds to $A_{V}=0.31$.
\subsection{Classical semi-convective models}
In Fig. \ref{fig:classic_single_CM}
is shown a synthetic CMD for the age of
76 Myr, and turn-off mass $M_{\rm TO}=5M_{\odot}$.
The termination magnitude of the
main sequence band fits the
observational data.
However, the luminosity interval spanned by the blue loop does not agree
with the observed CMD. Theoretical luminosities are indeed brighter than
observed. Contrary-wise, if we impose that the mean luminosity of
red giants is matched, the age increases to $110$ Myr but
the observational termination magnitude of the main sequence
is about $1$ mag brighter than the theoretical one. In Fig.
\ref{fig:classic_single_ILF} we present
the comparison between the observational (dashed line) and the
theoretical N-ILF
(solid line) for the population of $76$ Myr in
Fig. \ref{fig:classic_single_CM}. The stellar models in use yield a ratio
$\Sigma N_{\rm MS}/N_{\rm PMS}$ lower than the observational one.
Since $N_{\rm PMS}$ is taken equal to the number of red giants detected
in NGC 1866, this implies that the classical semi-convective models
produce more giants than observed (a long known result).
%%%%%%%%%%%%%%%Figure 5
\begin{figure}%%%%%%[h]
\resizebox{\hsize}{!}{\includegraphics{fig5.eps}}
\caption{\small Synthetic CMD constructed with
classical
models. The age is $76$ Myr, the chemical composition is
[$Y$=0.250,$Z$=0.008].}\label{fig:classic_single_CM}
\end{figure}
%%%%%%%%%%%%%%%Figure 6
\begin{figure}%%%%%%[h]
%\resizebox{\hsize}{!}{\includegraphics{blank.ps}}
%\resizebox{\hsize}{!}{\includegraphics{ILF_sing_cl_7.88_18.6.eps}}
\resizebox{\hsize}{!}{\includegraphics{ILF_sing_cl_7.88.eps}}
\caption{\small Observational (dashed line) N-ILF ---with the poissonian
error bars--- and theoretical N-ILF (thick line) for the
case of classical models shown in
Fig. \protect{\ref{fig:classic_single_CM}}. The age is
of $76$ Myr and the chemical composition is [$Y$=0.250,$Z$=0.008]
}\label{fig:classic_single_ILF}
\end{figure}
%%%%%%%%%%%%%%%%%Figure 7
\begin{figure}%%%%%[h]
\resizebox{\hsize}{!}{\includegraphics{fig7.eps}}
\caption{\small Synthetic CMD for the age of $160$
Myr, obtained from models with overshooting and chemical
composition [$Y$=0.250,$Z$=0.008]}\label{fig:over_single_CM}
\end{figure}
%%%%%%%%%%%%%Figure 8
\begin{figure}%%%%%[h]
%\resizebox{\hsize}{!}{\includegraphics{blank.ps}}
%\resizebox{\hsize}{!}{\includegraphics{ILF_sing_ov_8.20_18.6.eps}}
\resizebox{\hsize}{!}{\includegraphics{ILF_sing_ov_8.20.eps}}
\caption{\small Observational (dashed line) N-ILF ---with the poissonian
error bars--- and theoretical N-ILF (thick line) for the
case of $160$ Myr shown in Fig.
\protect{\ref{fig:over_single_CM}}
}\label{fig:over_single_ILF}
\end{figure}
\subsection{Models with overshooting}
As already anticipated commenting Fig. \ref{fig:iso_conf_cl_ov}, the
longer
lifetime and brighter luminosities of stellar models with overshooting
impose older isochrones to be adopted in the CMD simulations.
Figs. \ref{fig:over_single_CM} and \ref{fig:over_single_ILF}
illustrate the synthetic CMD and the N-ILF, respectively,
and compare them
with their observational counterparts for a population $160$ Myr
old and turn-off mass of $M_{\rm TO}=4.1M_{\odot}$.
Although a larger ratio
$\Sigma N_{\rm MS}/N_{\rm PMS}$ is
obtained as compared with the classical case, still the agreement
between theory and observations is not fully satisfactory:
indeed, while the region of the blue loop is well fitted,
the main sequence termination magnitude
is about $1$ mag fainter than observed.
Alternatively, if a younger age ($ \simeq 100$ Myr) is adopted
to match the main sequence termination luminosity,
theory and observations would disagree in the morphology of the blue
loop. Furthermore, it would predict too a high number
of main sequence stars ($\Sigma N_{\rm MS} \simeq 10000$) compared to
the observed number
of red giants ($N_{\rm PMS} = 100$), i.e.
the ratio $\Sigma N_{\rm MS}/N_{\rm PMS}$ would be too high.
\subsection{Including binary stars}
The effect of unresolved binary stars (UBS) on the CMD of stellar
clusters has been investigated in a
number of studies \citep[see][]{C89a,V91}.
The main effect of UBS on the main sequence morphology
would be the appearance of a second sequence,
running parallel to the main sequence, but systematically
cooler and with brighter luminosities ($\simeq 0.7$ mag for binary stars
of equal mass).
It goes without saying that in order to detect
this secondary sequence the number of UBS must be significant.
The percentage of UBS
in open clusters of the Milky Way seems to amount to about
30\,\% -- 50\,\% of
the cluster population \citep{Mermi-Mayor89,Ca94}. Likely
the same percentage of UBS could exist also in young clusters
of the LMC.
Strong observational evidence of this comes
from the recent study of \citet{El98} on NGC 1818, a young cluster
similar to NGC 1866. In the CMD obtained with $HST$ data, the
existence of a ``double'' main sequence is soon evident.
\citet{El98} explain this
feature by supposing the existence of a population of UBS amounting to
30--35\,\% of the total and with mass ratios in the range
$0.7$ to $1$. Systems with
mass ratios different from these cannot be excluded. In any case they
would be hardly distinguishable from single objects.
Basing on the study by \citet{El98}, we include in our
simulations the same fraction
of binary stars, to understand how they alter the above
conclusions both for classical and overshooting models. Unless
otherwise
specified, the mass ratio $q$ of the UBS included in the
simulations falls in the interval $0.7<q<1$ in agreement with
\citet{El98}\footnote{Detailed simulations of the CMD of NGC 1818,
using the prescription for the population of UBS by \citet{El98} can be
found
in \citet{Barmi01}. They are not shown here for the sake of brevity.}.
The presence of UBS, shifting the main sequence
termination toward brighter luminosities, requires older
ages. At the same time, the older
the isochrone, the fainter is the blue loop. Therefore, under the action
of the two effects, the luminosity gap between the main sequence
termination and red giants is reduced and the
overall morphology of the CMD is better reproduced.
%%%%%%%%%%%%%%%%%Figure 9
\begin{figure}%%%%%[h]
\resizebox{\hsize}{!}{\includegraphics{fig9.eps}}
\caption{\small Synthetic CMD constructed with
classical models, chemical composition
[$Y$=0.250,$Z$=0.008], and presence of UBS. The age is
$105$ Myr; the percentage of UBS amounts to $30\,\%$
}\label{fig:classic_binary_CM}
\end{figure}
%%%%%%%%%%%%%%%Figure 10
\begin{figure}%%%%%[h]
%\resizebox{\hsize}{!}{\includegraphics{blank.ps}}
%\resizebox{\hsize}{!}{\includegraphics{ILF_bin_cl_8.02_18.6.eps}}
\resizebox{\hsize}{!}{\includegraphics{ILF_bin_cl_8.02.eps}}
\caption{\small Observational N-ILF (dashed line) ---with the poissonian
error
bars--- and theoretical ILF (thick line) for the
case shown in Fig. \ref{fig:classic_binary_CM}: classical
models, age 105 Myr, and $30\,\%$ of UBS.}
\label{fig:classic_binary_ILF}
\end{figure}
\subsubsection{Binary stars in the classical picture}
In Fig. \ref{fig:classic_binary_CM} we present the synthetic CMD for the
age of $105$ Myr (turn-off mass $M_{\rm TO}=4.3$ M$_{\odot}$),
in which $30\,\%$ of UBS are included.
Compared to the results obtained simulating only single stars,
the agreement with the observed CMD has much improved,
but the theoretical N-ILF
(the thick line in Fig. \ref{fig:classic_binary_ILF}),
clearly suffers the same difficulty already encountered neglecting the
effects
of binary stars, i.e. too low a ratio $\Sigma N_{\rm MS}/N_{\rm PMS}$.
The same simulation has been repeated adopting the same percentage
of binaries ($30\,\%$) but supposing that the mass ratio is distributed
according to
a Gaussian curve, as suggested by \citet{Te99}.
No significant difference is noticed
both in the CMD morphology and N-ILF.
\subsubsection{Initial mass function}
{\it How do the above results depend on the particular slope
($\alpha =2.35$) adopted for the Salpeter IMF?}
To cast light on this topic we have repeated
our simulations varying $\alpha$ and keeping
constant all other parameters. The analysis shows that while
the CMD morphology is scarcely affected, the N-ILF is sensitive to
variations of $\alpha$.
As illustrated in Fig. \ref{fig:slopes_ILF},
the N-ILF gets steeper at increasing $\alpha$.
We note that, in order to
reach agreement between observational and theoretical N-ILFs using
the classical semi-convective
models, one should adopt $\alpha$=(3.5 -- 4), This would
favor the formation of low mass stars still located on the
main sequence. The problem is whether or not such high values of
$\alpha$ are acceptable. The present results
re-confirm what already pointed out in the earlier studies by
\citet{BM83} and \citet{C89a}.
%%%%%%%%%%%%%%%%%Figure 11
\begin{figure}%%%%%[h]
%\resizebox{\hsize}{!}{\includegraphics{blank.ps}}
\resizebox{\hsize}{!}{\includegraphics{ILF_slopes.eps}}
\caption{\small Comparison between the observational N-ILF (dashed line)
and
the theoretical ones (thick lines)
obtained in the classical picture for a population of
$105$ Myr and three different slopes for
the Salpeter IMF, namely $\alpha =3, \alpha =3.5, \alpha =4$. The
initial chemical composition, [$Y$=0.250,$Z$=0.008], the
number of post main sequence stars ($N_{\rm PMS}=100$), and
the fraction of binaries present in the sample ($30\,\%$) are
constant
in the three simulations}\label{fig:slopes_ILF}
\end{figure}
%%%%%%%%%%%%%%%%Figure 12
\begin{figure}%%%%%[h]
\resizebox{\hsize}{!}{\includegraphics{fig12.eps}}
\caption{\small Synthetic CMD for
overshooting models: age of $148$ Myr
$30\,\%$ of UBS, and chemical composition
[$Y$=0.250,$Z$=0.008]}\label{fig:over_binary_CM}
\end{figure}
%%%%%%%%%%%%%%%%%Figure 13
\begin{figure}%%%%%[h]
%\resizebox{\hsize}{!}{\includegraphics{blank.ps}}
%\resizebox{\hsize}{!}{\includegraphics{ILF_bin_ov_8.17_18.6.eps}}
\resizebox{\hsize}{!}{\includegraphics{ILF_bin_ov_8.17.eps}}
\caption{\small Observational N-ILF (dashed line) --with the Poissonian
error bars-- and theoretical N-ILF (thick line) for the $148$ Myr old
population shown in Fig. \ref{fig:over_binary_CM}: overshooting
models
and $30\,\%$ of UBS}\label{fig:over_binary_ILF}
\end{figure}
\subsubsection{Binary stars in the overshooting picture}
The synthetic CMD presented in Fig. \ref{fig:over_binary_CM} refers to
stellar models with convective overshooting and simulates
a population $148$ Myr old (turn-off mass
$M_{\rm TO}=4.2$ M$_{\odot}$)
containing $30\,\%$ of UBS.
We note that the luminosity gap between the red giant region and the main
sequence termination has now disappeared. The CMD of NGC 1866 is
perfectly reproduced. The most important result is, however, the N-ILF
shown in Fig. \ref{fig:over_binary_ILF}, which fully agrees with the
observations. In other words, the ratio $\Sigma N_{\rm MS}/N_{\rm PMS}$
predicted by stellar models with convective overshoot together with
the presence of a certain fraction of binary stars fully agrees
with that obtained from star counts.
%%%%%%%%%%%%Figure 14 effetti stocastici - 100 SIMULAZIONI %%%%%%%%%%%%%%%%%%%
\begin{figure}
\begin{minipage}{0.5\textwidth}
%\resizebox{\hsize}{!}{\includegraphics{blank.ps}}
%\resizebox{\hsize}{!}{\includegraphics{ILF_deviations_cl_8.02_18.6.eps}}
\resizebox{\hsize}{!}{\includegraphics{ILF_deviations_cl_8.02.eps}}
\end{minipage}
\begin{minipage}{0.5\textwidth}
%\resizebox{\hsize}{!}{\includegraphics{blank.ps}}
%\resizebox{\hsize}{!}{\includegraphics{ILF_deviations_ov_8.17_18.6.eps}}
\resizebox{\hsize}{!}{\includegraphics{ILF_deviations_ov_8.17.eps}}
\end{minipage}
\caption{Mean N-ILF (thick line) and related $2 \sigma$ levels of
expectancy (dotted lines) resulting from 100 simulations of
the same population including $30\,\%$ of UBS.
The chemical composition is [$Y$=0.250,$Z$=0.008].
With the classical scheme (top panel) and age of 105 Myr
the observational N-ILF
(dashed line) is steeper than the theoretical one; for the
overshooting scheme (bottom panel) and age of 148 Myr,
the observational N-ILF
lies inside the $\pm 2 \sigma$ range around the theoretical
prediction
}\label{fig:ILF_deviations}
\end{figure}
\subsection{Effects of stochastic fluctuations in the IMF}
The discussion of the ILF we have presented so far, refers to a
particular
simulation of the CMD, which does not explicitly take into account
possible
effects of stochastic nature caused by the finite number of stars in a
cluster. It is worth recalling here that the concept itself of IMF
allows for important fluctuations to occur around the ideal distribution
of stars as a function of the mass.
The problem may be particularly severe for $N_{\rm PMS}$ which in most
cases
is a small number. In other words the question to be addressed is whether
stochastic effects may change $N_{\rm PMS}$ with respect to
$\Sigma N_{\rm MS}$ or
conversely, whether $N_{\rm PMS}$ which has been used as the key
constraint to
our simulated CMDs and ILFs is generated by a unique population of main
sequence stars or if many others are equally eligible.
To cast light on this issue, we have performed many simulations of
our reference CMD keeping constant all parameters and letting the
stochastic
nature of the IMF develop thanks to the Monte Carlo technique we have
adopted.
To this aim, 100 simulations of the same CMD are calculated in order to
get the mean N-ILF and its standard deviation $\sigma$.
For the sake of brevity, we illustrate here the results only for
the cases
of classical models (Fig. \ref{fig:ILF_deviations}a)
and overshooting models (Fig. \ref{fig:ILF_deviations}b),
in which $30\,\%$ of binary stars are included.
No variations to the
previous conclusions seem to appear. With classical semi-convective
models the observed N-ILF (dashed line)
runs above the $\pm 2 \sigma$ interval (dotted lines) around the mean
N-ILF (thick line). In contrast, with the overshooting models
theoretical and observational N-ILFs agree at the $\pm 2 \sigma$ level.
Considering the errors affecting the observational N-ILF, the
agreement is extremely good.
\subsection{Age dispersion}
There is a final effect to be considered, i.e. the possibility that stars
in
the cluster are born over a finite time interval, i.e. that a
significant age spread exists.
Starting from our best-fit CMD, obtained from overshooting models with
age of
$\simeq 148$ Myr and including $30\,\%$ of binary stars, adding an
age spread of about $20$ Myr (about 10\,\%)
would further improve the CMD, especially in
the region of red giant stars. This final CMD is presented in Fig.
\ref{fig:HR_age_spread}b, and compared with that of NGC 1866
(Fig. \ref{fig:HR_age_spread}a). This small spread in the age
while giving the final make-up of the CMD does not affect the N-ILF,
and the picture outlined in Fig. \ref{fig:ILF_deviations}b still holds.
%%%%%%%%%%%%%%%%%Figure 15
\begin{figure}%%%%%[!t]
\begin{minipage}{0.5\textwidth}
\resizebox{\hsize}{!}{\includegraphics{fig15a.eps}}
\end{minipage}
%\hfill
\begin{minipage}{0.5\textwidth}
\resizebox{\hsize}{!}{\includegraphics{fig15b.eps}}
\end{minipage}
\caption{\small Comparison between the CMD
of NGC 1866 \citep{Te99} (top panel) and our best-fit case
(bottom panel),
characterized by stellar models with convective overshooting,
age in the interval 140--160 Myr, $30\,\%$ of UBS.
The initial chemical composition is [$Y$=0.250,$Z$=0.008]. The
intrinsic distance modulus is $(m-M)_{\rm 0}=18.5$.}
\label{fig:HR_age_spread}
\end{figure}
\subsection{Major uncertainties}
Looking at CMD of Figs. \ref{fig:HR_age_spread}b and
\ref{fig:HR_age_spread}a, we see that while overall
agreement is reached especially as far as
the magnitudes and colors of the main
sequence stars, and
the mean luminosity of the stars in the blue loop are concerned,
there is still
a marginal disagreement in the colour range
spanned by the red giant stars, i.e. the extension of the blue loop.
This, however, is
known to depend on several factors: (i) the overshooting parameters
$\Lambda_{\rm c}$ assumed for the nuclear region
\citep[see][]{C89a} and
the one adopted for the envelope, $\Lambda_{\rm e}$ \citep[see][]{Al91};
(ii) the rate of the nuclear reaction
$^{12}{\rm C}(\alpha, \gamma)^{16}{\rm O}$ --which determines
the abundances of $^{12}{\rm C}$ and $^{16}{\rm O}$ at the end of
central
He-burning phase-- the higher the rate, the bluer
the loop \citep{Be85}.
{\it It is worth noting, however, that these
uncertainties on the extension of the blue loop do not affect the
star counts at the base of the ILF.}
On the observational side, apart from photometric errors, completeness
correction, and fraction of UBS that have already been taken into
account, one has to
check whether the results could depend in a crucial way on the particular
calibration adopted to convert luminosities and effective
temperatures into magnitudes and
colours of the $UBVRI$ system.
To clarify this point, we have also used the empirical
conversions by \citet{Alo99}, who present tabulations
of colours and bolometric corrections as a function of the
effective temperature
and metallicity [Fe/H] of the stars. However, comparing isochrones
calculated with this transformation and those obtained with the
\citet{Be94} no appreciable difference is noticed.
Some uncertainty could be related to the particular choice for the
percentage of UBS ($30\,\%$) and their mass ratios $0.7 < q < 1$ we
have adopted. {\it How could a change in
these parameters affect our conclusions? } To cast light on the
issue
we have performed additional simulations with different percentages and
mass ratios. Increasing the
binary fraction up to about $40\,\%$, and extending
the mass ratio interval down to $0.5 - 0.6$, the CMD and N-ILF are still
in agreement with the cluster data. However, assuming an even larger
percentage of
binary stars ($>40 - 45 \,\%$), the CMD gets worse as
the main sequence band spreads too much toward red colors.
\section{Why do we disagree with \citet{Te99}? } \label{why:testa}
\label{sec:why}
The results of our analysis wholly disagree with those by
\citet{Te99}, despite the fact that we are using the same observational
data
for NGC 1866. Their main conclusion is indeed that classical stellar
models
without overshooting fit the CMD and ILF of NGC 1866 at the age of 100
Myr
provided that a population of binary stars amounting to about
$30\,\%$ of the total is included. In contrast our main conclusion is
that even
including the same percentage of binary stars only models with
overshooting simultaneously reproduce the CMD, the ILF, and the correct
ratio $\Sigma N_{\rm MS}/N_{\rm PMS}$. The age now is about 150 Myr.
{\it What is the cause of the opposite conclusions?} There are several
factors concurring to the final result:
(i) The different procedure applied to correct star counts for
photometric
completeness. The ILF of \citet{Te99} is over-estimated due to the
the fact that instead of summing up the DLF corrected for
completeness in every magnitude bin, they apply the correction procedure
to every step of the ILF.
(ii) The different normalization procedures of the simulations.
In the present study, the normalization parameter of
the simulated CMD and ILF is
the total (observed) number of red giants $N_{\rm PMS}$. It is worth
recalling
that the N-ILF is given by $\Sigma N_{\rm MS}/N_{\rm PMS}$. The
advantage of this type of normalization has already been explained by
\citet{C89a} and briefly summarized in previous sections.
In contrast \citet{Te99} assume that each simulation is complete
as soon as a given total number of stars with magnitude brighter than
$M_V=2.6$ is reached. They justify this choice
by saying that their main sequence is well defined and complete
up to three magnitudes fainter than the turn-off, and assume the total
number of stars observed in this magnitude interval as the
{\it normalization parameter}
of the simulations instead of the giant (post main sequence) stars.
(iii) This choice of the normalization factor is, however, not physically
sound because a deeply observed, well populated main sequence does not tell
much about the inner structure, evolutionary rates, lifetimes etc. of its
stars.
Indeed the distribution of the stars along the main sequence almost
exclusively depends on the IMF.
Therefore, the choice of a ``normalization parameter''
insensitive to the size of the convective core simply precludes all
chances
of discriminating between the two evolutionary schemes under examination.
Simulations of CMDs and ILFs based on this criterion may fail to
reproduce the
correct ratio $\Sigma N_{\rm MS}/N_{\rm PMS}$.
(iv) In relation to this, is also the fact that the total number of
stars brighter than $M_V=2.6$ on which the simulations are normalized
is determined by the procedure of
correcting for photometric incompleteness. We have already argued that the
\citet{Te99} method is inconsistent, and overestimates the total
number of stars in the main sequence. Since their simulated
CMDs and ILF do actually depend on this parameter, there is no way
of becoming aware of the internal inconsistency.
Considering all the differences between the present and \citet{Te99}
analyses, we should recognize that {\em
the contrasting results are mainly due to the
different way of normalizing the ILF: we choose a normalization factor
(number of giants) that is sensitive to overshooting,
whereas Testa et al.'s (1999) choice (stars brighter than $M_V=2.6$) is
essentially insensitive to this effect}.
In fact, no clear discrimination between classical and overshooting
models is expected with the latter method.
(v)
To better understand this point of controversy
we have repeated our simulations strictly following the procedure
described
by \citet{Te99}: each one contains the same total number of stars
brighter than $V=22$ as the observational data, i.e.\ 4879.
The results are displayed in Fig.~\ref{fig:tuttotesta},
limited to the few cases we have already examined. The layout of
Fig.~\ref{fig:tuttotesta} is organized in vertical and horizontal groups
of panels. Panels in the same vertical row correspond to the
case as indicated by the top heading of the row. Panels in horizontal
rows
show different types of LFs as indicated: the DLF in the bottom, the
total
ILF (main sequence and giant stars) in the middle panels, and the ILF
for the sole main sequence stars in the top panels.
The results can be commented as follows:
%%%%%%%%%%%%%%%%Figure 16
\begin{figure*}%%%%%%%%%[!t]
\begin{center}
\begin{minipage}{0.80\textwidth}
%\resizebox{\hsize}{!}{\includegraphics{blank.ps}}
%\resizebox{\hsize}{!}{\includegraphics{tuttotesta.ps}}
\resizebox{\hsize}{!}{\includegraphics{summary.eps}}
\end{minipage}
\end{center}
\caption{\small Summary of the four cases under consideration:
the heading of each column lists the type of stellar model
(classical or overshooting), the percentage of unresolved binaries,
the age, and the number of giant stars.
The three horizontal rows display the LFs as shown in
\citet{Te99}:
the top row is the integrated LF for the sole main sequence stars in the
simulation, the middle
row is the integrated LF for all stars in the simulation, finally the
bottom row is the DLF for the whole sample.
In each panel the solid line is the theoretical LF or DLF,
whereas the dashed line is the same but for the
observational data.
Each simulation contains a {\it total number of stars } brighter than
$V=22$ equal to $4879$}
\label{fig:tuttotesta}
\end{figure*}
\begin{itemize}
\item The DLFs in the bottom horizontal row make evident the
luminosity gap between the main sequence termination and
the giant stars, located at $M_{V}\simeq-2$.
On the observational side, the gap is almost invisible, whereas
on the theoretical side, the inclusion of binaries wipes out the gap
in both classical and overshooting models.
All models seem to reproduce the main sequence part of the DLF
($M_{V}\ga-1$) equally well.
\item The total ILFs, i.e. inclusive of the post main sequence stars,
are shown in the central horizontal row.
For the models without binaries, the luminosity interval between the main
sequence termination and the giant stars appears as a plateau in the
ILF at $M_{V}\simeq-2$. To the left of this plateau, the detailed
shape of the ILF is determined mainly by the location of
giant stars: a slight over luminosity of the giants (with respect to
observations) causes an apparent excess of stars at $M_{V}<-2$,
which can be noticed in the cases of classical models
with and without binaries, and overshooting models without binaries.
In the case of overshooting models with binaries, giant stars are
predicted at the right magnitude level, and no such discrepancy can be
noticed in the ILF.
But it is important to notice that the ``excess'' suggested by the
ILF does not necessarily correspond to a real excess of evolved stars.
This is evident if we look at
the total number of giants in each simulation (see the labels at the top
of the figure): in the case of classical models there is a real excess
of giants (158 and 189 depending on whether or not binaries are
included), which has no observational counterpart.
The situation is the opposite for models with convective overshoot
and no binary stars. Now the simulated number of giants is 48 (about
half of what observed).
Finally, models with overshooting and binaries predict the correct
number of post main sequence stars (120, to be compared with the
100 observed).
\item The ILF of main sequence stars (top horizontal row) does not
show substantial differences passing from one case to another.
This simply reflects the constancy of the IMF. There are small
differences at the turn-off level, however too small to be significant.
\end{itemize}
>From the plots in Fig.~\ref{fig:tuttotesta}, there is a hint that models
with overshooting and binary stars are in slightly better position,
because they fit well the observed LFs in all panels.
However, the most indicative result is not derived from the
shape of the different LFs, but comes from the expected number
of giants in the simulations: models with overshooting and binary stars
are {\em the only ones} to predict the observed number of giant stars.
In other words, with the kind of plot used by \citet{Te99},
one cannot pin down the correct value of the ratio
$\Sigma N_{\rm MS}/N_{\rm PMS}$. Finally, we like to call attention
to the fact
that {\it even with the \citet{Te99} method,
no strong indication is found that classical stellar models are to be
preferred.}
\section{Conclusions}\label{sec:conclusions}
The main goal of this study is to compare with observational data
the predictions for CMD and ILF obtained from two different types of
stellar models: the classical ones in which the extension of convective
zones is defined by the
Schwarzschild criterion (whenever required semi-convection is also taken
into account), and the models with overshooting (calculated
accordingly to the \citet{Br81} formalism), which allow the
penetration
of convective elements into the surrounding formally stable layers.
The ultimate goal is to cast light on which type of stellar models find
better correspondence with the physical structure of real stars.
The main result that bears very much on the
comparison with the observational data
is the net decrease
--about $55 - 60\,\%$-- of
the lifetime ratio $t_{\rm He}/t_{\rm H}$
passing from classical to overshooting models.
The analysis begins
with correcting in the proper way the star counts and the LFs
to get the LF of main sequence stars normalized
to the number of evolved stars (N-ILF). The N-ILF
yields the ratio
$\Sigma N_{\rm MS}/N_{\rm PMS}$ of main sequence to giant stars, and
directly measures the ratio $t_{\rm He}/t_{\rm H}$ \citep[see][]{C89a}.
Many simulations of the CMD and LF of NGC 1866 have
been performed, in which not only the effect of the different stellar
models but also those given by the presence of binary stars is taken
into account. The results can be summarized as follows:
\begin{itemize}
\item{
The observed ILF cannot be matched at all by the classical models,
showing too large a ratio of evolved to main sequence stars.
The reason for the discrepancy between our results and the conclusion
reached by
\citet{Te99} has already been discussed and will not be repeated here.}
\item{The situation is the opposite for the case of models with
overshooting.
It is slightly worse neglecting unresolved
binaries and fully satisfactory with binaries. }
\item{The presence of unresolved binaries in the simulations both for
classical and overshooting models, allows to simultaneously match
the luminosity of the main sequence termination and giant stars. }
\item{Binary stars act on the main sequence morphology in a way roughly
mimicking the effect of an enlarged H-burning core, i.e. shifting the
turn-off toward brighter luminosities and forcing us to use older
isochrones to fit the cluster data. However, they do not alter
the ratio $t_{\rm He}/t_{\rm H}$. }
\item{If all this can be taken as observational evidence for
more extended mixing in the cores of
hydrogen burning stars, the stellar models with convective
overshooting constitute the simplest solution. They are indeed able
to simultaneously match the CMD
and ILF of NGC 1866, thus re-confirming what already
found in previous studies
\citep{C89a,C89b,L91,V91,VaC92}.}
\item{
By adopting the intrinsic distance modulus $(m-M)_{\rm 0}$=$18.5$ and the
initial
chemical composition [$Y$=0.250,$Z$=0.008], our best-fit of the CMD and
ILF
of NGC~1866 is for an age in the range 140 to 160 Myr, turn-off mass
$M_{\rm TO}$=$4.2$ M$_{\odot}$, 30 to 40\,\% of binary stars
with mass ratio $q$ lying in
the interval $0.6<q<1$. }
\item{
Finally, it is expected that adopting slightly different values
for the distance modulus will not change the above conclusions.
Looking at the turn-off luminosity
and its relation with the age, we estimate that
$\Delta (m-M)_{\rm 0}$=$\pm0.1$ with respect to the adopted value
$(m-M)_{\rm 0}$=$18.5$ would change the age (and age range) by
$\Delta t \simeq \pm 10$
Myr. Small variations in the percentage of
binaries and/or age dispersion would immediately follow. }
\end{itemize}
New observational data
will certainly improve the exact values of age, metallicity, binary
star
percentage so that the present values ought to be considered as
provisional
estimates. Although the new generation telescopes could
make it possible to reach the
innermost regions of the cluster, scarcely or not at all affected by
field star contamination, thus providing us with better CMDs and ILFs,
it is unlikely that the present conclusions will change significantly.
{\it Real stars should indeed possess bigger convective cores
than predicted by the
classical theory of stellar evolution.}
\vspace{0.5truecm}
\rightline{ {\it The civil war strikes again ...} }
%%%%%%%%%%%%%%%%%%%%%%%Acknowledgments %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\vspace{0.5truecm}
{\it Acknowledgments.}
We like to thank V. Testa and O. Straniero for the helpful clarifications
about the observed luminosity functions of NGC 1866.
This study has been financed by the Italian
Ministry of Education, University, and Research (MIUR).
\begin{thebibliography}
\expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
\bibitem[{{Alongi} {et~al.}(1991){Alongi}, {Bertelli}, {Bressan}, \&
{Chiosi}}]{Al91}
{Alongi}, M., {Bertelli}, G., {Bressan}, A., \& {Chiosi}, C. 1991, A\&A, {\bf
244}, 95
\bibitem[{{Alongi} {et~al.}(1993){Alongi}, {Bertelli}, {Bressan}, {Chiosi},
{Fagotto}, {Greggio}, \& {Nasi}}]{Al93}
{Alongi}, M., {Bertelli}, G., {Bressan}, A., {Chiosi}, C., {Fagotto}, F.,
{Greggio}, L., \& {Nasi}, E. 1993, A\&AS, {\bf 97}, 851
\bibitem[{{Alonso} {et~al.}(1999){Alonso}, {Arribas}, \&
{Martinez-Roger}}]{Alo99}
{Alonso}, A., {Arribas}, S., \& {Martinez-Roger}, C. 1999, A\&AS, {\bf 140},
261
\bibitem[{{Barmina}(2001)}]{Barmi01}
{Barmina}, R. 2001, in Thesis for the Master Degree in Astronomy, University
of
Padova, Italy
\bibitem[{{Becker} \& {Mathews}(1983)}]{BM83}
{Becker}, S. \& {Mathews}, J. 1983, AJ, {\bf 270}, 155
\bibitem[{{Bertelli} {et~al.}(1985){Bertelli}, {Bressan}, \& {Chiosi}}]{Be85}
{Bertelli}, G., {Bressan}, A., \& {Chiosi}, C. 1985, A\&A, {\bf 150}, 33
\bibitem[{{Bertelli} {et~al.}(1986){Bertelli}, {Bressan}, {Chiosi}, \&
{Angerer}}]{Be86}
{Bertelli}, G., {Bressan}, A., {Chiosi}, C., \& {Angerer}, K. 1986, A\&AS, {\
bf
66}, 191
\bibitem[{{Bertelli} {et~al.}(1994){Bertelli}, {Bressan}, {Chiosi}, {Fagotto}
,
\& {Nasi}}]{Be94}
{Bertelli}, G., {Bressan}, A., {Chiosi}, C., {Fagotto}, F., \& {Nasi}, E. 199
4,
A\&AS, {\bf 106}, 275
\bibitem[{{Bertelli} {et~al.}(1993){Bertelli}, {Bressan}, {Chiosi}, {Mateo},
\&
{Wood}}]{Bw93}
{Bertelli}, G., {Bressan}, A., {Chiosi}, C., {Mateo}, M., \& {Wood}, P. 1993,
ApJ, {\bf 412}, 160
\bibitem[{{B\"ohm-Vitense}(1958)}]{Bohm55}
{B\"ohm-Vitense}, E. 1958, Z. Astroph., {\bf 46}, 108
\bibitem[{{Bressan} {et~al.}(1981){Bressan}, {Bertelli}, \& {Chiosi}}]{Br81}
{Bressan}, A., {Bertelli}, G., \& {Chiosi}, C. 1981, A\&A, {\bf 102}, 25
\bibitem[{{Bressan} {et~al.}(1993){Bressan}, {Fagotto}, {Bertelli}, \&
{Chiosi}}]{Br93a}
{Bressan}, A., {Fagotto}, F., {Bertelli}, G., \& {Chiosi}, C. 1993, A\&AS, {\
bf
100}, 647
\bibitem[{{Brocato} {et~al.}(1989){Brocato}, {Buonanno}, {Castellani}, \&
{Walker}}]{Broca89}
{Brocato}, E., {Buonanno}, R., {Castellani}, V., \& {Walker}, A. 1989, ApJS,
{\bf 71}, 25)
\bibitem[{{Caldwell} \& {Coulson}(1985)}]{Cald85}
{Caldwell}, J. \& {Coulson}, I. 1985, MNRAS, {\bf 212}, 879
\bibitem[{{Canuto} \& {Mazzitelli}(1991)}]{Canuto91}
{Canuto}, S. \& {Mazzitelli}, I. 1991, ApJ, {\bf 370}, 295
\bibitem[{{Canuto} {et~al.}(1996){Canuto}, {Goldman}, \&
{Mazzitelli}}]{Canuto96}
{Canuto}, V.~M., {Goldman}, I., \& {Mazzitelli}, I. 1996, ApJ, {\bf 473}, 550
\bibitem[{{Carraro} {et~al.}(1994){Carraro}, {Chiosi}, {Bressan}, \&
{Bertelli}}]{Ca94}
{Carraro}, G., {Chiosi}, C., {Bressan}, A., \& {Bertelli}, G. 1994, A\&AS, {\
bf
103}, 375
\bibitem[{{Chiosi}(1999)}]{Chiosi99}
{Chiosi}, C. 1999, in Stellar Structure: Theory and Test of Convective Energy
Transport, eds. Alvaro Gimenez, Edward F. Guinan, and Benjamin Montesinos,
ASP Conference Series, 173, ASP (San Francisco), 9
\bibitem[{{Chiosi} {et~al.}(1992){Chiosi}, {Bertelli}, \& {Bressan}}]{C92}
{Chiosi}, C., {Bertelli}, G., \& {Bressan}, A. 1992, ARA\&A, {\bf 30}, 235
\bibitem[{{Chiosi} {et~al.}(1989{\natexlab{a}}){Chiosi}, {Bertelli}, {Meylan}
,
\& {Ortolani}}]{C89a}
{Chiosi}, C., {Bertelli}, G., {Meylan}, G., \& {Ortolani}, S.
1989{\natexlab{a}}, A\&A, {\bf 219}, 167
\bibitem[{{Chiosi} {et~al.}(1989{\natexlab{b}}){Chiosi}, {Bertelli}, {Meylan}
,
\& {Ortolani}}]{C89b}
---. 1989{\natexlab{b}}, A\&AS, {\bf 78}, 89
\bibitem[{{Cloutman} \& {Whitaker}(1980)}]{Cloutman80}
{Cloutman}, L. \& {Whitaker}, R.~W. 1980, ApJ, {\bf 237}, 900
\bibitem[{{Deardorff} {et~al.}(1969){Deardorff}, {Willis}, \&
{Lilly}}]{Deardorff69}
{Deardorff}, J., {Willis}, G., \& {Lilly}, D. 1969, Fluid Mech., {\bf 35}, 7
\bibitem[{{Dominguez} {et~al.}(1999){Dominguez}, {Chieffi}, {Limongi}, \&
{Straniero}}]{Dominguez99}
{Dominguez}, I., {Chieffi}, A., {Limongi}, M., \& {Straniero}, O. 1999, ApJ,
{\bf 524}, 226
\bibitem[{{Elson} {et~al.}(1998){Elson}, {Sigurdsson}, {Davies}, {Hurley}, \&
{Gilmore}}]{El98}
{Elson}, A., {Sigurdsson}, S., {Davies}, M., {Hurley}, J., \& {Gilmore}, G.
1998, MNRAS, {\bf 300}, 857
\bibitem[{{Feast}(1989)}]{Feast89}
{Feast}, M. 1989, in The World of Galaxies, (ed.) {Corwin}, H.G. and and
{Bottinelli}, L., (NY: Springer), 118
\bibitem[{{Feast}(2000)}]{Feast00}
{Feast}, M. 2000, astro-ph/0010590
\bibitem[{{Freytag} {et~al.}(1996){Freytag}, {Ludwig}, \&
{Steffen}}]{Freytag96}
{Freytag}, B., {Ludwig}, H., \& {Steffen}, M. 1996, A\&A, {\bf 313}, 497
\bibitem[{{Girardi} {et~al.}(2000){Girardi}, {Bressan}, {Bertelli}, \&
{Chiosi}}]{Gi00}
{Girardi}, L., {Bressan}, A., {Bertelli}, G., \& {Chiosi}, C. 2000, A\&AS, {\
bf
141}, 371
\bibitem[{{Keller} {et~al.}(2000){Keller}, {Da Costa}, \& {Bessel}}]{Keller00
}
{Keller}, S., {Da Costa}, G., \& {Bessel}, M. 2000, astro-ph/0011285, to appe
ar
in Astronomical Journal
\bibitem[{{Kurucz}(1992)}]{Ku92}
{Kurucz}, R.~L. 1992, in IAU Symposium: The Stellar Populations of Galaxies,
{Barbuy}, B. and {Renzini}, A. (eds.). Dordrecht: Kluwer, Vol. 149, 225
\bibitem[{{Lattanzio} {et~al.}(1991){Lattanzio}, {Vallenari}, {Bertelli}, \&
{Chiosi}}]{L91}
{Lattanzio}, J., {Vallenari}, A., {Bertelli}, G., \& {Chiosi}, C. 1991, A\&A,
{\bf 250}, 340
\bibitem[{{Maeder} \& {Meynet}(1991)}]{MaederMey91}
{Maeder}, A. \& {Meynet}, G. 1991, A\&AS, {\bf 89}, 451
\bibitem[{{Mermilliod} \& {Mayor}(1989)}]{Mermi-Mayor89}
{Mermilliod}, J. \& {Mayor}, M. 1989, A\&A, {\bf 219}, 15
\bibitem[{{Meynet} {et~al.}(1994){Meynet}, {Maeder}, {Schaller}, {Schaerer},
\&
{Charbonnel}}]{Meynet94}
{Meynet}, G., {Maeder}, A., {Schaller}, G., {Schaerer}, D., \& {Charbonnel},
C.
1994, A\&AS, {\bf 103}, 97
\bibitem[{{Oestreicher} \& {Schmidt-Kaler}(1996)}]{Oes96}
{Oestreicher}, M. \& {Schmidt-Kaler}, T. 1996, A\&AS, {\bf 117}, 303
\bibitem[{{Panagia}(1998)}]{Panagia98}
{Panagia}, N. 1998, MmSAI, {\bf 69}, 225
\bibitem[{{Renzini}(1987)}]{Renzini87}
{Renzini}, A. 1987, A\&A, {\bf 188}, 49
\bibitem[{{Rieke} \& {Lebofsky}(1985)}]{Rieke85}
{Rieke}, G.~H. \& {Lebofsky}, M. 1985, ApJ, {\bf 288}, 618
\bibitem[{{Salasnich} {et~al.}(2000){Salasnich}, {Girardi}, {Weiss}, \&
{Chiosi}}]{Sa00}
{Salasnich}, B., {Girardi}, L., {Weiss}, A., \& {Chiosi}, C. 2000, A\&A, {\bf
361}, 1023
\bibitem[{{Salpeter}(1955)}]{Salp55}
{Salpeter}, E. 1955, ApJ, {\bf 121}, 161
\bibitem[{{Testa} {et~al.}(1999){Testa}, {Ferraro}, {Chieffi}, {Straniero},
{Limongi}, \& {Fusi Pecci}}]{Te99}
{Testa}, V., {Ferraro}, F., {Chieffi}, A., {Straniero}, O., {Limongi}, M., \&
{Fusi Pecci}, F. 1999, AJ, {\bf 118}, 2839
\bibitem[{{Unno} \& {Kondo}(1989)}]{UnnoKondo89}
{Unno}, W. \& {Kondo}, M. 1989, PASJ, {\bf 41}, 197
\bibitem[{{Vallenari} {et~al.}(1992){Vallenari}, {Chiosi}, {Bertelli},
{Meylan}, \& {Ortolani}}]{VaC92}
{Vallenari}, A., {Chiosi}, C., {Bertelli}, G., {Meylan}, G., \& {Ortolani}, S
.
1992, AJ, {\bf 104}, 1100
\bibitem[{{Vallenari} {et~al.}(1991){Vallenari}, {Chiosi}, {Bertelli},
{Meylan}, \& {Ortolani}}]{V91}
{Vallenari}, A., {Chiosi}, C., {Bertelli}, G., {Meylan}, S., \& {Ortolani}, S
.
1991, A\&AS, {\bf 87}, 517
\bibitem[{{Westerlund}(1997)}]{Westerlund97}
{Westerlund}, B. 1997, in The Magellanic Clouds (Cambridge: Cambridge
University Press)
\bibitem[{{Xiong}(1980)}]{Xiong80}
{Xiong}, D.~R. 1980, ChA, {\bf 4}, 234
\bibitem[{{Zahn}(1991)}]{Zahn91}
{Zahn}, J. 1991, A\&A, {\bf 252}, 179
\bibitem[{{Zaritsky} {et~al.}(1997){Zaritsky}, {Harris}, \& {Thompson}}]{Z97}
{Zaritsky}, D., {Harris}, J., \& {Thompson}, I. 1997, AJ, {\bf 114}, 1933
\end{thebibliography}
%\bibliographystyle{apj}
%\bibliography{biblio}
\end{document}