A&A 441, 117-129 (2005)
DOI: 10.1051/0004-6361:20042241
H. J. G. L. M. Lamers 1,2 - M. Gieles1 - N. Bastian 1 - H. Baumgardt3 - N. V. Kharchenko4,5,6 - S. Portegies Zwart7,8
1 - Astronomical Institute, Utrecht University,
Princetonplein 5, 3584CC Utrecht, The Netherlands
2 - SRON Laboratory for Space Research, Sorbonnelaan 2,
3584CC Utrecht, The Netherlands
3 - Sternwarte, University of Bonn, Auf dem Hügel 71,
53121 Bonn, Germany
4 - Astrophysikalisches Institut Potsdam, An der Sternwarte 16,
14482 Potsdam, Germany
5 - Astronomisches Rechen-Institut, Mönchhofstraße 12-14,
69120 Heidelberg, Germany
6 - Main Astronomical Observatory, 27 Academica Zabolotnogo Str.,
03680 Kiev, Ukraine
7 - Astronomical Institute, University of Amsterdam,
Kruislaan 403, 1098SJ Amsterdam, The Netherlands
8 - Informatics Institute, University of Amsterdam,
Kruislaan 403, 1098SJ Amsterdam, The Netherlands
Received 23 October 2004 / Accepted 26 May 2005
Abstract
We present a simple analytical description of the disruption of star
clusters in a tidal field.
The cluster disruption time, defined as
,
depends on the mass M of the cluster as
with
for clusters in a
tidal field, as shown by
empirical studies of cluster samples in different galaxies
and by N-body simulations.
Using this simple description we derive an analytic expression
for the way in which the mass of a cluster decreases with time due to
stellar evolution and disruption.
The result agrees very well with those of detailed N-body simulations for clusters in the tidal field of our galaxy.
The analytic expression can be used to predict the mass and age
histograms of surviving clusters for any cluster initial mass function
and any cluster formation history.
The method is applied to explain the age distribution of the
open clusters in the solar neighbourhood within 600 pc, based on a new
cluster sample that appears to be unbiased
within a distance of about 1 kpc. From a comparison between the
observed and predicted age distributions in the age range between 10 Myr to
3 Gyr we find the following results:
(1) The disruption time of a 104
cluster in the solar
neighbourhood is about
Gyr.
This is a factor of 5 shorter than that derived from N-body simulations
of clusters in the tidal field of the galaxy. Possible reasons for
this discrepancy are discussed.
(2) The present star formation rate in bound clusters
within 600 pc of the Sun is
Myr-1,
which corresponds to a surface star formation rate of bound clusters
of
yr-1 pc-2.
(3) The age distribution of open clusters shows a bump between 0.26 and 0.6 Gyr when the cluster formation rate was 2.5 times higher
than before and after.
(4) The present star formation rate in bound
clusters is about half that derived from the
study of embedded clusters.
The difference suggests that about half of the clusters in the solar
neighbourhood become unbound within about 10 Myr.
(5) The most massive clusters within 600 pc had an initial
mass of about
.
This is in agreement with
the statistically expected value based on a cluster initial mass function
with a slope of -2, even if the physical upper mass limit for cluster
formation is as high as 106
.
Key words: Galaxy: globular clusters: general - Galaxy: halo - Galaxy: kinematics and dynamics - Galaxy: open clusters and associations: general - Galaxy: solar neighbourhood - galaxies: star clusters
Bound star clusters
in a tidal field lose mass due to
internal effects, i.e. mass loss by stellar evolution, and by the external
effect of tidal stripping. The combination of these effects results
in a decreasing mass of the cluster until the cluster is destroyed
completely. The time scale of this disruption depends on the initial
conditions of the cluster, e.g. the stellar initial mass function and
its concentration, and on the tidal forces experienced by the cluster
during its galactic orbit.
The disruption of star clusters determines the mass and age distributions of the existing clusters. Therefore any study of the cluster formation history of a galaxy has to take into account the disruption of clusters.
The age and mass distribution of the initial cluster population is described by the cluster formation rate as a function of time, CFR(t), and the cluster initial mass function, CIMF. The distributions of the present observable star clusters is modified because: (a) the disruption time of the clusters depends on the initial mass, (b) the mass of each cluster decreases with time, (c) the clusters fade as they age due to stellar evolution. A (simple) description of these three effects would facilitate the studies of samples of star clusters. The purpose of this paper is to provide such a simple description and show how it can be used in the analysis of star cluster samples.
The structure of the paper is as follows:
In Sect. 2 we discuss the arguments that the cluster disruption time
depends on its mass as a power law of the type
.
In Sect. 3 we calculate the evolution of a
cluster in terms of its decreasing mass due to stellar evolution and disruption.
In Sect. 4 we will show that the results agree very well with those
of N-body simulations.
In Sect. 5 we predict
the mass and age distributions of cluster samples for
various cluster formation histories.
In Sect. 6 we apply the method to the age distribution
of open clusters in the solar neighbourhood, using the new cluster sample
from Kharchenko et al. (2005). We derive the
disruption time of open clusters in the solar neighbourhood as well
as the cluster formation rate and the star formation rate.
The discussion and conclusions are given in Sect. 7.
Boutloukos & Lamers (2003, hereafter BL03) have studied the mass and age
distributions of magnitude-limited cluster samples in selected
regions in four galaxies, and concluded that these distributions
can be explained if the disruption time of clusters depends on the initial
mass
of the clusters as
,
with
for clusters in very different local environments.
Baumgardt & Makino (2003, hereafter BM03) have calculated a
grid of N-body simulations of clusters
in the tidal field of our galaxy for different initial masses
and initial concentration factors in circular and elliptical
orbits at various galactocentric distances. They take into account
mass loss by stellar evolution and by tidal relaxation. Their
calculations show that the disruption time of a cluster, defined as
the time when 5% of the initial number of stars remain in the
cluster, scales with the
half mass relaxation time
and the clusters crossing time
as
with x=0.82 for clusters with an initial dimensionless depth
W0=7 (which is a measure of the concentration index of the cluster,
see King 1966)
and x=0.75 for less concentrated clusters with W0=5.
BM03 and Gieles et al. (2004) have shown that for all the
models of BM03 the disruption time can be expressed as
a function of the initial cluster mass as
The discussion above has concentrated on the comparison between the
disruption time of clusters of different initial masses, i.e.
,
within one galactic environment. We have not yet
discussed how the mass of an individual cluster decreases with time. This
is the topic of the next section.
The mass of a cluster decreases due to stellar evolution and tidal disruption. We will describe the evolution of the bound mass, using analytic expressions for the mass loss from the cluster by stellar evolution and by tidal effects.
The mass loss from clusters due to stellar evolution has been
calculated for cluster evolution models by several groups,
e.g. Bruzual & Charlot (1993) and the Starburst99 models by
Leitherer et al. (1999). We adopt the
GALEV models for single stellar populations (Schulz et al. 2002;
Anders & Fritze-v. Alvensleben 2003).
These models contain stars in the mass range of
0.15 < M* < 85 ,
distributed over this mass range with either
the Salpeter or Scalo mass function.
We adopt the models with the Salpeter mass
function because deep photometry of clusters in the LMC shows that the
cluster IMF is a powerlaw with a slope of about -2.35 down to at least 0.6
(de Marchi 2003). Lower mass stars hardly contribute to
the luminosity at ages less than 10 Gyr, but may contribute
significantly to the cluster mass. The GALEV models are based
on stellar evolution tracks from the Padova group, which include mass
loss and overshooting (Bertelli et al. 1994; Girardi et al. 2000).
Lamers (2005) has shown that the fraction of the initial
cluster mass, Mi, that is lost by stellar evolution in the GALEV models,
i.e.
where
is the mass lost by stellar evolution,
can be approximated very accurately by a function of the form
In this paper we use the symbol
to describe the
fraction of the mass of a cluster with initial mass
that is still
bound at age t.
We define
Table 1:
Approximations to the mass lost by stellar evolution
for
cluster models with a Salpeter IMF of
,
and
0.0004<Z<0.05.
We describe the decreasing mass of a bound cluster that survived infant
mortality (
yrs) as a function of time.
Let us define
as the mass of a cluster of initial mass
and age t, and
as the fraction of the initial
mass that is
still in the cluster. The decrease of mass due to both stellar evolution
and disruption can then be described as
Equation (4)
can easily be solved numerically.
It turns out that the mass decrease of a cluster can be
approximated very accurately by the following formula
![]() |
Figure 1:
The predicted decrease in cluster mass due to stellar evolution and
disruption for Z=0.020 and four values of the initial cluster
masses: 103, 104, 105 and
![]() ![]() ![]() |
Open with DEXTER |
These equations imply that the total disruption time of a cluster
with an initial mass of 104 ,
which was referred to as t4 in BL03 and in Lamers et al. (2005), is related to t0 by
Equation (8) shows that
is approximately
proportional to
,
which was the relation adopted in the study of the
disruption times of clusters in different galaxies by BL03 and Lamers et al. (2005).
![]() |
Figure 2:
Comparison of the decrease of the cluster mass with
time between the results of the N-body simulations
by BM03 ( left) and our description ( right) for clusters of different
initial numbers of stars, ![]() ![]() ![]() |
Open with DEXTER |
We can compare our analytic expression for the decreasing mass of a cluster with the results of N-body simulations. We adopt the simulations by BM03 who calculated the fate of clusters under various conditions. Before making this comparison, we would like to point out that:
BM03 give the function
for their models of clusters of different
initial numbers of stars,
,
and with an
initial concentration described by W0=7, in circular
orbits around the galactic center at a distance of 8.5 kpc.
(See BM03 Fig. 6.) Their results show that
decreases almost linearly with time, and that the function plotted
against
is about the same for all clusters.
We compare their results with our analytic expression Eq. (6).
The left hand panels of Fig. 2
show the normalized decrease in cluster
mass,
,
of the models by BM03.
The top figure gives the mass as a function of time, both on a linear
scale. However, since the
mass of clusters will decrease several orders of magnitudes before they are
disrupted, we also plot the logarithm of
as a function of the logarithm of
in the lower figure.
The right hand panels of Fig. 2 show
the results of our models. For a fair comparison between our
calculations and those of BM03, we plot
as a
function of
.
The
agreement between the predictions of the N-body calculations and
our description is very good, both on a linear and
logarithmic scale. The very small difference of less than 5 percent
during the early phase of the most massive models is probably due to the
difference in the adopted mass loss by stellar evolution.
(The plotted relations are to the first order corrected for the effects of
stellar evolution. However because the cluster evolution is caused by
the combination of stellar evolution and tidal effects, the corrected
relations still bear the imprint of the adopted stellar evolution.)
Figure 3 shows the direct comparison between the
N-body prediction for a cluster of initial mass
(
Ni=16 384) with a concentration of W0=7 and our analytic
solution. For the analytic solution we adopted the same mass and
the parameter
Myr was chosen in such a way that the 95% disruption time is 5.7 Gyr, very close to that of the N-body simulation.
The times are normalized to the time when
of the cluster mass is
gone. We see that the prediction by the N-body calculations (full line) and the analytic expression (dotted line) are
very similar, apart from a small offset of the timescale.
If we normalize the timescale of the analytic solution to the time
when 96.5% of the initial mass is gone (dashed line), the agreement
is almost perfect. The difference in timescale is due to the fact that in
the N-body simulations there is a preference for the low mass stars
to be ejected from the clusters, whereas in the analytical solution
stars of all masses are lost. Apart from this small difference, the
analytical solution with t0 as a free parameter describes the
decrease of the mass of clusters surprisingly accurately.
![]() |
Figure 3:
Comparison between the mass decrease of a cluster of
![]() ![]() |
Open with DEXTER |
We conclude that our analytic description of the decreasing mass of a cluster, with the adjustable free parameter t0, is very similar to the one derived by N-body simulations. This result is not trivial.
The values of
that results from the N-body simulations by BM03
depend on the initial number of stars as
,
in
excellent agreement with the dependence derived empirically
by BL03 (see Sect. 2).
However this does not automatically imply that the decrease
of mass with time of each individual cluster can be described by
a function that depends on the
mass of that cluster
to the same power
(Eq. (5)).
However the good agreement between the simulations by
BM03 and our result from the analytic description shows that it does
depend on the mass in this way.
So, not only does the total lifetime of all clusters depend on their
initial mass as
,
but also at every
moment during the lifetime of a cluster the exponential disruption time,
,
scales with the mass to the same power,
0.62. This is a consequence of the fact that the timescale of the
disruption process depends on both the half mass relaxation time,
and the crossing time
as about
,
with
for
models with a concentration parameter W0=5.0 and x=0.82if W0=7.0 (BM03). For both sets of models the disruption time
scales with mass as M0.62 (Gieles et al. 2004).
It is the continuous adjustment of the half mass
relaxation time and the crossing time to the changing conditions
of the cluster that results in an exponential disruption time that
varies during the life of a cluster as the present mass
.
Using the expression for the decreasing mass of clusters, Eq. (6), we can predict the mass and age distribution of cluster samples (for open clusters as well as globular clusters) for any adopted cluster formation rate, CFR(t), and cluster initial mass function, CIMF.
Suppose that the CIMF is a power law with a slope
(Zhang & Fall 1999; Larsen 2002; Bik et al. 2003; de Grijs et al. 2003) in the range of
,
and
,
then the
number of clusters with initial mass
formed at time t will be
Given the initial mass distribution of the clusters, their formation rate, CFR(t), and an expression for the way in which the mass of each cluster changes with time (Eq. (6)), we can calculate the present distribution of existing clusters as a function of age or as a function of their mass.
If
is the number of clusters
of mass M and age t in Nr
-1 yr-1,
then
and
are related by the conservation of the
numbers of clusters
![]() |
Figure 4:
The changes in the mass distribution (Eq. (15)) of
a sample of clusters as a function of their age, in the case where
stellar evolution can be neglected. We adopted a cluster
initial mass function in the range of
![]() ![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 5:
The age distribution of cluster samples formed at a
constant formation rate, in cases where
stellar evolution can be neglected. We adopted a cluster
initial mass function in the range of
![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
The mass and age distribution of cluster samples N(M,t) depends on the stellar evolution and on disruption.
To obtain insight into the effect of disruption on the evolution of a
cluster sample we first consider a simplified case when mass loss by
stellar evolution is neglected, i.e.
.
In that case the function
N(M,t)/S(t) depends only on the slope
of the cluster IMF,
the mass-dependence
of the disruption and on
the ratio
.
Figure 4 shows the shape
of
N((M,t)/S(t). For very young ages
or very long disruption time (
)
the distribution
is the initial CIMF with slope
.
For strong disruption, i.e.
,
the distribution of the low mass clusters becomes flatter and
approaches a power law of the type
.
This distribution is similar to the one predicted by BL03
for instantaneous disruption, except that the transition
between the two slopes is gradual, whereas it shows a sharp kink
for models with instantaneous disruption.
Figure 5 shows the normalized age distribution
,
for cluster samples
formed at a constant cluster formation rate in the mass range of
for various values of
,
when mass loss by stellar evolution
can be ignored. The distribution is flat for young clusters
at a value of
and curves down to older clusters, approaching a slope
.
This was predicted by BL03 for instantaneous disruption. The distribution drops to
zero at the age at which the most massive clusters are disrupted,
i.e. when
(Eq. (16)), which is at
for
and
.
Ideally one would like to compare the predictions with complete (or at least unbiased) samples of clusters with known masses and ages. Unfortunately this is not possible at the moment, because samples of clusters in external galaxies are usually magnitude-limited. (The method for determining the disruption times from magnitude limited cluster samples with gradual disruption will be described by Lamers (2005) and applied to the cluster sample in M 51 by Gieles et al. (2005a).) Samples of open clusters in the solar neighbourhood are unbiased, but only the cluster ages have been determined systematically and not the cluster masses. We will compare our predictions for the age distribution to the sample of open clusters in the solar neighbourhood.
Kharchenko et al. (2005) published a catalogue of
astrophysical data of 520 galactic open clusters (COCD = Catalogue of
Open Cluster Data) in the wider
neighbourhood of the Sun with the values of angular sizes of cluster
cores and coronae, heliocentric distances d, E(B-V), mean proper
motions, radial velocities and ages. These parameters have been
determined by homogeneous methods and algorithms including a careful procedure
of cluster member selection. The basis of this study is the ASCC-2.5
- All-Sky Compiled Catalogue of about 2.5 million stars (Kharchenko
2001) down to
(completeness limit at
),
with compiled proper motions and B,V magnitudes based on
the Tycho-2 data and supplemented with
data sets, as well
as with some ground-based catalogues
.
Cluster membership is based on a combined probability which takes into
account kinematic (proper motion), photometric and spatial
selection criteria (see Kharchenko et al. 2004 for details). For
stars within a circle with a cluster radius the membership probability
is calculated as the measure of a deviation
either from the cluster mean proper motion (kinematical probability),
or from the Main Sequence edges (photometric probability). Stars
deviating from the reference values by less than one
(rms)
are classified as most probable cluster members (1
-members,
i.e., with a membership probability
%). Those falling in
semi-intervals [1
,2
)
or [2
,3
)
are
considered as possible members (P=14-61%) or possible field stars
(P=1-14%), respectively. Stars with deviations larger than
3
are regarded as definite field stars (P < 1%). As a
rule, all cluster parameters were determined from the data of the
most probable cluster members. Cluster ages were determined with an
isochrone-based procedure which provides a uniform age scale
(see Kharchenko et al. 2005 for details). Thus the COCD is the most
homogeneous and most complete catalogue of open clusters in the solar
neighbourhood available.
![]() |
Figure 6:
Top: the surface density distribution projected onto the
Galactic plane of open clusters
in the solar neighbourhood from the homogeneous Kharchenko et al. (2005)
catalogue. Error bars indicate 1![]() ![]() ![]() |
Open with DEXTER |
Figure 6 shows the distance distribution
of the density of clusters projected onto the
Galactic plane, in number per pc2.
Within the statistical uncertainty the surface density is constant
up to at least 600 pc, and possibly even up to 1 kpc.
The lower part of the figure shows the ratio between old (
yr) and young (
yr) clusters as a function of
distance. Up to a distance of about 1 kpc there is no
significant change in this ratio within the statistical uncertainty.
This is important for our study because it shows that the age
distribution of open clusters within about 1 kpc is not affected by
detection limits.
Figure 7 shows the age distribution
in number per year of the 114 clusters within 600 pc in the Kharchenko
et al. (2005) sample.
The effect of binning is demonstrated by plotting two sets of data,
where the bins have been shifted by 0.1 dex relative to one another.
This distribution decreases with age, apart from a small
local maximum around
.
The distribution
at young ages is sensitive to the choice of the age-bins and shows a
significant scatter. The steep slope at
demonstrates that cluster disruption is important.
![]() |
Figure 7:
The age histogram in units of number per
year, in logarithmic age-bins of 0.2 dex,
of 114 open clusters
within d<600 pc from Kharchenko et al. (2005).
The
distributions are plotted for two sets of bins, shifted by 0.1 dex,
with and without squares respectively. The error bars indicate the
1![]() ![]() ![]() |
Open with DEXTER |
For the determination of the cluster formation rate and the disruption times we need an estimate of the minimum mass of the clusters in the Kharchenko et al. (2005) sample. This catalogue does not list the mass of the clusters, but it can be estimated roughly from the age, distance, extinction and the number of stars of each cluster. We have estimated the lower mass limit of the Kharchenko et al. cluster sample in the following way.
To estimate the sensitivity of the cluster masses to the
adopted stellar lower mass limit we also determined the
masses of the clusters in the Kharchenko et al. (2005) catalogue within 600 pc in the same way as described above
but with an adopted minimum stellar mass of
instead of 0.15
.
The resulting cluster masses are about 80%
of those for
,
so in that case the
minimum lower mass limit of the Kharchenko sample would be about 80
.
The limiting cluster mass of 80
corresponds to about 140 stars.
![]() |
Figure 8: The mass-versus-age diagram of 114 clusters of the Kharchenko et al. (2005) catalogue within a distance of 600 pc. The mass is derived from the number of main sequence stars with V<11.50. |
Open with DEXTER |
![]() |
Figure 9:
The mass histogram of 114 clusters of the Kharchenko et al. (2005) catalogue within a distance of 600 pc. The steep edge at
the low mass side suggests that the sample is complete for clusters
with a mass
![]() |
Open with DEXTER |
Figure 10 shows the fits to the data
for
and various values of
,
based on the method
described in Sect. 5.1.
For the top figure
we adopted a constant cluster formation rate and a CIMF with
and a mass upper limit of
.
This latter choice
agrees with the steep decrease around
.
We assumed a minimum detectable cluster mass of 100
.
The predicted distributions are
normalized to the data point at
,
which is one
of the most accurate data points. The best fit is reached for
Myr. (The low datapoint at
and the subsequent high point at 7.5
may be due to the adopted binning; see Fig. 7.)
![]() |
Figure 10:
Comparison between observed and
predicted age histogram of clusters in the solar neighbourhood
within d<600 pc. The data are fitted to predicted
relations based on our analytical expression of the cluster
disruption with various values of
![]() ![]() ![]() ![]() |
Open with DEXTER |
The distributions for
Myr
underpredict the observed numbers at
and overpredicts the numbers at old ages.
The distribution for
Myr overpredicts the distribution at
young ages. We have applied a
test to express the goodness of
the fit. For this test we only considered the data at
and
,
i.e.
we excluded the bump at
which
will be discussed below. We also excluded the data younger than 10 Myr, because they may be affected by infant mortality
. The criterion
results in the best estimate of
Myr.
The data of the fits are given in the top line of Table 2.
To investigate the effect of the adopted lower mass limit,
we have also compared the observed age distribution with that
predicted for
and
.
This last value is derived from the
Kharchenko et al. cluster sample if the stellar lower mass limit of 0.25
is adopted instead of 0.15
.
This comparison is very similar to the one in the top
panel of Fig. 10 and is not shown here.
The resulting data are listed in
the second line of Table 2. The best fit is at
a slightly longer disruption time of
Myr.
Combining the two values of t0, derived for
and 100
,
we conclude that
Myr (see bottom line of
Table 2).
The derived value of t0 implies a total disruption
time of a 104
cluster of
Gyr (Eq. (11)).
This empirically-derived disruption time of open clusters is
about a factor of 5 smaller
than the value predicted by the N-body simulations of BM03, who predict a
disruption time of 6.3 Gyr for a 104
cluster with an
initial concentration factor of W0=5.0 and 5.9 Gyr if W0=7.0.
The N-body simulations of clusters by Portegies Zwart et al.
(1998) gave about twice as short disruption times as those of BM03 (see also Lamers et al. 2005). This is still
longer than the empirically derived disruption time for open clusters
in the solar neighbourhood.
We consider three possible reasons for this discrepancy: (a) the clusters do not start with the initial concentration factors adopted in the simulations by BM03 and (b) the presence of another mechanism (apart from the tidal field) that contributes to the destruction of clusters in the solar neighbourhood.
BM03 assumed in their N-body simulations that the clusters initially fill their tidal radius with a concentration factor of W0=5.0 or 7.0, defined by King (1966). These are the values suggested by the current density profiles of the globular clusters. However, open clusters are much less centrally condensed than globular clusters. Moreover they may not fill their tidal radius when they are formed. If the clusters are smaller than their tidal radius the internal relaxation will be faster than predicted and so the disruption might be faster. We suggest that open clusters are formed so far out of equilibrium that they lose a substantial fraction of their mass within a few crossing times. Most clusters will then disperse completely, in agreement with the high infant mortality rate. The surviving clusters might then dissolve along the lines predicted by the N-body simulations, but on a faster time scale. N-body simulations of clusters with various initial concentration factors and various initial radii are needed to test this suggestion.
The disruption times calculated by BM03 is an upper limit because the
values are calculated for tidal disruption in a smooth
tidal field without other destruction mechanisms.
The destruction of open clusters by encounters with giant
molecular clouds (GMCs) has been proposed by several authors, e.g. Oort
(1958) and Terlevich (1987).
The problem with this explanation is that disruption by a GMC
is expected to result in a mass dependence disruption
of
with
(e.g. Spitzer 1987), whereas the observed age distribution
agrees perfectly with
predicted for tidal effects.
On the other hand, massive clusters are
likely to have larger tidal radii and hence more interactions
than low mass clusters and will therefore
be more susceptible to the influence of GMCs. This might soften the
mass dependence of the disruption time to
.
Table 2: The disruption time and the formation rate of clusters within 600 pc from the Sun for two assumed values of the lower mass limit of clusters in the Kharchenko et al. (2005) sample.
The vertical shift of the predicted relative to the observed age
distribution in the top panel of Fig. 10
gives the CFR.
We find a CFR of
clusters per Myr. (This is about
twice as high as the value derived by Battinelli & Capuzzo-Dolcetta
(1991) based on the Lyngå (1987) catalogue of open clusters brighter
than Mv=-4.5.) Our value of the CFR corresponds
to a starformation rate in clusters of
/Myr within a region of 600 pc from the
Sun for an CIMF with a slope of
and an adopted
lower cluster mass limit of
.
This
corresponds to a surface formation rate of the galactic disk near the
Sun of
yr-1 pc-2.
This value can be compared with the present total star formation rate in
the galactic disk near the Sun.
Lada & Lada (2003) derived a SFR of 7 to
yr-1 pc-2 from
embedded clusters in the solar neighbourhood. This value is a factor
of 1.3 to 1.9 higher than the value derived from the clusters in the Kharchenko
sample. The difference is most likely due to the fact that many of the
star clusters formed in embedded clouds will be dispersed within 10 Myr. At later ages they would not be recognized as clusters.
Thus, if the sample of embedded stars studied by Lada & Lada (2003) is
complete, the infant mortality rate of clusters in the solar neighbourhood
is about 40 percent.
We can also estimate the infant mortality rate of the clusters from
the data in the Kharchenko et al. (2005) catalogue. The mean cluster formation
rate in the age bin of
is
clusters/yr, whereas it has dropped to
clusters/yr at
.
This indicates a survival rate of about 40% and an infant mortality rate of about 60%. The same rates are found if we
compare the mean value of the cluster formation rates in the age bin
of
with that of
.
So, the comparison of the cluster formation
rate with the star formation rate of
Lada & Lada (2003) and the comparison between the formation
rates of the youngest to the slightly older clusters both suggest
an infant mortality rate of about 50%.
The predictions shown in the top panel of Fig. 10
and discussed above
do not explain the bump around
which is higher than any of the distributions for a constant
cluster formation rate. Thus, the cluster sample of
Kharchenko et al. (2005) suggests that there was an increased cluster
formation rate around that time. We have modelled this with
Eq. (15) for several non-constant CFRs. The best fit is shown
in the lower part of Fig. 10, which was
calculated for
Myr and with a CFR that is increased by 0.40 dex between 250 and 600 Myr ago. The fit matches the data
well. This suggests that the cluster formation rate was a factor
of 2.5 higher during this age range.
Taking into account this burst we find that the mean CFR within 600 pc
from the Sun during the last Gyr in the solar neighbourhood was
/Myr which corresponds to a surface formation
rate of
yr-1 pc-2.
Zaritzky & Harris (2005) found a peak near 400 Myr in the SFR of the SMC. During this peak the SFR was at least twice as high as the quiescent SFR of the SMC. They attribute this peak to the perigalactic passage of the SMC and the Galaxy. Possibly this passage also triggered the increased SFR in the galactic disk.
De la Fuente Marcos & de la Fuente Marcos (2004) have studied the star formation history in the solar neighbourhood, based on various open cluster catalogues. They identified five bursts at 0.35, 0.70, 1.13, 1.50 and 1.93 Gyrs respectively. The burst that we found between 0.25 and 0.6 Gyr may correspond to the one of 0.35 or, more likely, to a combination of those at 0.35 and 0.70 Gyr found by de la Fuente Marcos & de la Fuente Marcos (2004). Our analysis of the Kharchenko et al. (2005) cluster sample does not confirm the other bursts.
The theoretical fits of the predicted age distribution to the
observed one in Fig. 10 shows that
the best fit is reached if the
maximum mass of the clusters formed within 600 pc of the Sun during
the last few Gyr was about 105 .
This upper limit explains the
steep drop in the age distribution at
Gyr.
To check the robustness of this conclusion we have also
calculated models with higher mass upper limits for the clusters.
For instance, if the upper mass limit was 106
then
we would expect the following predicted and observed numbers of
clusters in the oldest
logarithmic agebins, indicated in a vector (
;
nr predicted, nr observed): (8.9, 9.1; 5.1, 6), (9.1, 9.3;
4.1, 3), (9.2, 9.4; 3.3, 2), (9.3, 9.5; 2.9, 1). In these
oldest agebins the number of observed clusters is within the
statistical
uncertainty of the number of predicted clusters.
This means that we cannot exclude the possibility that the maximum
cluster mass was higher than 105
and possibly
as high as 106
.
An alternative way to consider this point is to find the clusters in
the Kharchenko sample with the highest initial mass. We use the
estimate of the present cluster masses, derived from the number of
member stars, as described in Sect. 6.1, and then applied Eq. (7) to convert the present mass into the initial
mass with a disruption parameter
Myr.
We find that the two initially most massive clusters within
a distance of 600 pc,
i.e. the ones with
= {9.2, 3.5} and {9.4, 2.8} in Fig. 8,
had an initial mass of
and
respectively. For a CIMF with
the
number of clusters decreases with mass as
,
so the
expected number of clusters initially more massive than e.g. 105
will be smaller than 1.
The observed mass upper limit is probably determined by statistical
effects (Hunter et al. 2003; Gieles et al. 2005b). Its expected value can be
estimated from the number of observed clusters within 600 pc. For a
CIMF with a slope of -2 the maximum mass expected in the sample is
roughly
,
where N=114 is the
number of observed clusters and
.
So we
expect a maximum initial mass, set by statistical effects,
of about
.
(In reality
it should be slightly higher because the observed number of clusters
is already affected by disruption.) This agrees quite well with the
maximum initial mass of about
derived in the previous
paragraph.
We have derived a simple analytical expression for the mass loss from
star clusters due to stellar evolution and disruption as a function of time,
Eq. (6). This expression agrees very well with results of N-body simulations of clusters in the tidal field of our galaxy.
The expression is derived for mass loss by stellar evolution using the
cluster evolution models (Schulz et al. 2002; Anders &
Fritze-v. Alvensleben 2003) but can easily be applied to other cluster
evolution models, provided that the mass loss due to stellar evolution
can be expressed by an analytic approximation (e.g. of the type
proposed in Sect. 3.1). Our analytical expression for the mass loss
from star clusters is different to the one by Vesperini & Heggie
(1997) because they assumed that tidal effects decrease the mass
of a cluster linearly with time.
Our method is based on the fact that the disruption time of clusters,
defined as
,
depends on the mass M as
with
for
disruption by two body relaxation in a tidal field.
This dependence was found
both empirically from a study of cluster samples in four galaxies by BL03 and by Lamers et al. (2005), and
theoretically from N-body-simulations by Baumgardt (2001) and BM03.
The description contains a normalization parameter,
,
that depends on the
environment in the parent galaxy of the clusters.
This parameter can vary by more than an order of magnitude between
cluster samples of different galaxies, as shown by BL03. Simple theoretical
predictions, numerical simulations and
empirical determinations of cluster disruption in a few galaxies
showed that
,
where
is the ambient density in the galaxy at the
location of the clusters, and
Myr
(Lamers et al. 2005).
The value of t0 may be shorter in interacting galaxies, as
suggested by the study of the star cluster sample of M 51 by Gieles et al. (2005a).
Using our description of mass loss from clusters by stellar
evolution and disruption, we can predict the
resulting present day mass, M, and age distributions in terms of
of the surviving
clusters for different cluster initial mass functions and different
cluster formation histories. This is given by Eq. (15).
An integration of
over mass for different ages gives the
expected age distribution of the surviving clusters. An integration
of
over age for different masses gives the expected mass
distribution. This method can be applied to any cluster formation
history.
A comparison between predicted and observed distributions
of selected cluster samples can be used to derive the basic properties
of that cluster population, such as the cluster formation rate,
the cluster IMF, and the disruption parameter
.
To demonstrate this method we have applied it to the
age distribution of open clusters in the solar neighbourhood
within 600 pc based on the new cluster catalogue by
Kharchenko et al. (2005). Tests showed that the cluster sample in
this catalogue is unbiased up to a distance of at least 600 pc
and possibly 1 kpc. The predicted age distribution
agrees very well with the empirical one
(Fig. 10). The main uncertainty in the derived
disruption time
is the unknown lower mass detection limit
of this cluster sample. We estimated this mass limit from the given
number of cluster members brighter than V=11.5 magn. and found it to
be between 80 and 100
.
With these values, the fit of the
predicted age distribution to the observed one shows that the
disruption time
is
3.3+1.4-1.0 Myr, which corresponds to
a disruption time of a 104
cluster of
Gyr. This is
about a factor of 5 shorter than predicted by N-body simulations of
clusters in the tidal field of the solar neighbourhood (BM03). The difference
is possibly due to the fact that BM03 adopted a rather high initial central
concentration of the clusters that is more applicable to globular
clusters than to open clusters. Moreover encounters with giant
molecular clouds may also shorten the lifetime of open clusters.
The present star formation rate in the solar neighbourhood within 600 pc, derived in this
paper, is
/Myr for the last few Gyrs.
This corresponds to a surface formation rate of
yr-1 pc-2, which is about a factor of 0.5 to 0.7 smaller than
derived from the formation rate of stars in embedded clusters
(Lada & Lada 2003). This suggests that a considerable fraction of the
embedded clusters will be dispersed before reaching an age of several Myr.
The observed age distribution clearly shows evidence for a bump in the cluster formation rate between 0.25 and 6 Gyr, when the formation rate was a factor 2.5 higher than before and after. This corresponds to two of the bumps in the star formation rate derived from cluster samples by de la Fuente Marcos & de la Fuente Marcos (2005). The observed bump might be due to an encounter between the SMC and the Galaxy (e.g. Zaritsky & Harris 2005).
Although the upper mass limit of the observed clusters in the solar
neighbourhood is less than about 104 ,
we show that this
does not exclude the possible formation of higher mass clusters. Given
a cluster IMF with a slope of -2, the absence of clusters more
massive than 104
is in agreement with the statistical
uncertainty, even if the real upper limit for the initial mass
was as high as 106
.
Acknowledgements
We thank R.-D. Scholz for his collaboration in the production of the new open cluster catalogue used in this study. Paul Hodge and Anil Seth are acknowledged for useful discussions and suggestions. We thank the kind referee, Douglas Heggie, for useful suggestions and critical comments that resulted in an improvement of this paper. This work is supported by a grant from the Netherlands Research School for Astronomy (NOVA) to HJGLML and by a grant from the University of Washington in Seattle where a major part of this research was carried out when the first author was on sabbatical. N.V. Kharchenko acknowledges financial support by the DFG grant 436 RUS 113/757/0-1, and the RFBR grant 03-02-04028.