ar
X
iv
:a
str
o-
ph
/0
30
15
40
v1
2
7
Ja
n
20
03
To appear in Annual Reviews of Astronomy & Astrophysics
Embedded Clusters in Molecular Clouds
Charles J. Lada
Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge,
Massachusetts 02138; email: clada@cfa.harvard.edu
Elizabeth A. Lada
Department of Astronomy, University of Florida, Gainesville, Florida 32611; email:
lada@astro.ufl.edu
ABSTRACT
Stellar clusters are born embedded within giant molecular clouds (GMCs)
and during their formation and early evolution are often only visible at infrared
wavelengths, being heavily obscured by dust. Over the last 15 years advances
in infrared detection capabilities have enabled the first systematic studies of em-
bedded clusters in galactic molecular clouds. In this article we review the current
state of empirical knowledge concerning these extremely young protocluster sys-
tems. From a survey of the literature we compile the first extensive catalog of
galactic embedded cluster properties. We use the catalog to construct the mass
function and estimate the birthrate for embedded clusters within ∼ 2 Kpc of the
Sun. We find that the embedded cluster birthrate exceeds that of visible open
clusters by an order of magnitude or more indicating a high infant mortality rate
for protocluster systems. Less than 4-7% of embedded clusters survive emer-
gence from molecular clouds to become bound clusters of Pleiades age. The vast
majority (90%) of stars that form in embedded clusters form in rich clusters of
100 or more members with masses in excess of 50M⊙. Moreover, observations of
nearby cloud complexes indicate that embedded clusters account for a significant
(70-90%) fraction of all stars formed in GMCs. We review the role of embedded
clusters in investigating the nature of the IMF which, in one nearby example, has
been measured over the entire range of stellar and substellar mass, from OB stars
to subsellar objects near the deuterium burning limit. We also review the role
embedded clusters play in the investigation of circumstellar disk evolution and
the important constraints they provide for understanding the origin of planetary
systems. Finally, we discuss current ideas concerning the origin and dynamical

– 2 –
evolution of embedded clusters and the implications for the formation of bound
open clusters.
Contents
1 INTRODUCTION 3
2 EMBEDDED CLUSTERS: BASIC OBSERVATIONAL DATA 6
2.1 Definitions & Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Identification & Surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 The Embedded Cluster Catalog . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 The Embedded Cluster Mass Function . . . . . . . . . . . . . . . . . . . . . 9
2.5 Birthrates and Star Formation . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.6 Association with Molecular Gas and Dust . . . . . . . . . . . . . . . . . . . 15
2.7 Internal Structure and Mass Segregation . . . . . . . . . . . . . . . . . . . . 17
2.8 Ages and Age Spreads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3 EMBEDDED CLUSTERS AND THE INITIAL MASS FUNCTION 23
3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Methodology: From Luminosity to Mass Functions . . . . . . . . . . . . . . 27
3.2.1 Modelling the Luminosity Function . . . . . . . . . . . . . . . . . . . 27
3.2.2 Individual Stellar Masses from the HR Diagram . . . . . . . . . . . . 29
3.2.3 General Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 The IMF of the Trapezium Cluster from OB Stars to Brown Dwarfs . . . . . 32
3.4 Comparison With Other Embedded Clusters: A Universal IMF? . . . . . . . 39
4 LABORATORIES FOR STAR AND PLANET FORMATION 45
4.1 Protostars and Outflows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

– 3 –
4.2 Circumstellar-Protoplanetary Disks . . . . . . . . . . . . . . . . . . . . . . . 47
4.3 Brown Dwarfs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.4 Binary Stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5 ORIGIN AND DYNAMICAL EVOLUTION 53
5.1 Formation of Embedded Clusters . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2 Emergence from Molecular Clouds: Dynamical Evolution & Infant Mortality 55
6 CONCLUDING REMARKS 59
1. INTRODUCTION
Stellar clusters have been long recognized as important laboratories for astrophysical
research. Their study has played an important role in developing an understanding of the
universe. For example, clusters contain statistically significant samples of stars spanning a
wide range of stellar mass within a relatively small volume of space. Since stars in such
groups share the common heritage of being formed more or less simultaneously from the
same progenitor molecular cloud, observations of cluster color-magnitude (CM) diagrams
can be, and indeed, have been used to provide classical tests of stellar evolution theory.
Moreover, clusters offer the smallest physical scale over which a meaningful determination of
the stellar Initial Mass Function (IMF) can be made. Because a cluster is held together by
the mutual gravitational attraction of its individual members, its evolution is determined by
Newton’s laws of motion and gravity. In many body systems these interactions are inherently
complex and thus clusters are also important testbeds for studies of stellar dynamics. The
spatial distribution of clusters has also played a vital role in our understanding of galactic
structure. The distribution of globular clusters, for example, was critical for determining
the location of the galactic center, establishing the existence of a galactic halo and setting
the overall scale of the Galaxy. Young open clusters have provided an important tracer of
recent star formation in galaxies and of spiral structure in galactic disks. Such clusters are
also of interest for understanding the origin of the solar system, since the presence of rare
short-lived radio nuclides in meteoritic samples has long suggested that the Sun itself was
formed in near proximity to a massive star, and thus most likely within in relatively rich
cluster.
Little is known or understood about the origin of clusters. Globular clusters in the

– 4 –
Galaxy were formed billions of years ago. Because they are not being formed in the Milky
Way in the present epoch of galactic history, direct empirical study of their formation pro-
cess is not possible (except perhaps in certain extragalactic systems and at cosmological dis-
tances). On the other hand, open clusters appear to be continuously forming in the galactic
disk and, in principle, direct study of the physical processes leading to their formation is
possible. However, such studies have been seriously hampered by the fact that galactic clus-
ters form in giant molecular clouds (GMCs) and during their formation and earliest stages
of evolution are completely embedded in molecular gas and dust, and thus obscured from
view. Given the constraints imposed by traditional techniques of optical astronomy, direct
observation and study of young embedded clusters had been extremely difficult, if not im-
possible. However, during the last two decades the development of infrared astronomy and,
more recently, infrared array detectors, has dramatically improved this situation. Figure 1
shows optical and infrared images of the southern embedded cluster RCW 38 and amply
illustrates the power of infrared imaging for detecting such heavily obscured young clusters.
The deployment of infrared imaging cameras and spectrometers on optical and infrared
optimized telescopes has provided astronomers the ability to survey and systematically study
embedded clusters within molecular clouds. Almost immediately such studies indicated that
rich embedded clusters were surprisingly numerous and that a significant fraction, if not the
vast majority, of all stars may form in such systems. Consequently, it is now recognized that
embedded clusters may be basic units of star formation and their study can directly address a
number of fundamental astrophysical problems. These include the issues of cluster formation
and early evolution as well as the more general problems of the origin and early evolution
of stars and planetary systems. Because most stars in the galactic disk may originate in
embedded clusters, these systems must play a critical role in understanding the origins of
some of the most fundamental properties of the galactic stellar population, such as the form
and universality of the stellar IMF and the frequencies of stellar and planetary companions.
The purpose of this review is to summarize the current status of observational knowl-
edge concerning young embedded clusters in the Galaxy. We will consider both embedded
and partially embedded clusters. The embedded phase of cluster evolution appears to last
between 2-3 Myrs and clusters with ages greater than 5 Myrs are rarely associated with
molecular gas (Leisawitz, Bash & Thaddeus 1989) therefore, this review deals with clusters
whose ages are typically between 0.5-3 million years. Particular emphasis will be placed on
embedded clusters within ∼ 2 Kpc of the sun since this presents a sample which is most sta-
tistically complete and for which the most detailed observational data are available. Previous
reviews of embedded clusters, some with slightly different emphasis, can be found in various
conference proceedings (e.g., Lada & Lada 1991; Zinnecker, McCaughrean & Wilking, 1993;
Lada 1998; Clarke, Bonnell & Hillenbrand 2000, Elmegreen et al. 2000, Lada et al. 2002).

– 5 –
Fig. 1.— Optical (top) and Infrared (bottom) images of the RCW 38 region obtained with
the ESO VLT. The infrared observations reveal a rich embedded cluster otherwise invisible
at optical wavelengths. Figure courtesy of J. Alves.

– 6 –
2. EMBEDDED CLUSTERS: BASIC OBSERVATIONAL DATA
2.1. Definitions & Terminology
For the purposes of this review we consider clusters to be groups of stars which are
physically related and whose observed stellar mass volume density would be sufficiently
large, if in a state of virial equilibrium, to render the group stable against tidal disruption
by the galaxy (i.e., ρ∗ ≥ 0.1 M⊙ pc−3; Bok 1934), and by passing interstellar clouds (i.e.,
ρ∗ ≥ 1.0 M⊙ pc−3; Spitzer 1958). Furthermore we adopt the additional criterion (e.g., Adams
& Myers 2001) that the cluster consist of enough members to insure that its evaporation
time (i.e., the time it takes for internal stellar encounters to eject all its members) be greater
than 108 yrs, the typical lifetime of open clusters in the field. The evaporation time, τev, for
a stellar system in virial equilibrium, is of order τev ≈ 102τrelax, where the relaxation time is
roughly τrelax ≈ 0.1NlnN τcross and τcross is the dynamical crossing time of the system and N the
number of stars it contains (Binney & Tremaine 1987). The typical crossing time in open
clusters is of order 106 yrs, so if such a cluster is to survive disintegration by evaporation
for 108 yrs, its relaxation time must be comparable or to or greater than its crossing time or
0.1N
lnN ≈ 1. This condition is met when N ≈ 35. Therefore, for this review, we define a stellar
cluster as: a group of 35 or more physically related stars whose stellar mass density exceeds
1.0 M⊙ pc−3.
With our definition we distinguish clusters from multiple star systems, such as small (N
< 6) hierarchical multiples and binaries which are relatively stable systems and small multiple
systems of the Trapezium type, which are inherently unstable (Ambartsumian 1954; Allen
& Poveda 1974). We also distinguish clusters from stellar associations, which we define as
loose groups of physically related stars whose stellar space density is considerably below the
tidal stability limit of 1 M⊙ yr−3 (Blaauw 1964).
Clusters, as defined above, can be classified into two environmental classes depending
on their association with interstellar matter. Exposed clusters are clusters with little or
no interstellar matter within their boundaries. Almost all clusters found in standard open
cluster catalogs (e.g., Lynga 1987) fall into this category. Embedded clusters are clusters
which are fully or partially embedded in interstellar gas and dust. They are frequently
completely invisible at optical wavelengths and best detected in the infrared. These clusters
are the youngest known stellar systems and can also be considered protoclusters, since upon
emergence from molecular clouds they will become exposed clusters. A similar classification
can be applied to associations.
Our definition of a cluster includes stellar systems of two dynamical types or states.
Bound clusters are systems whose total energy (kinetic + potential) is negative. When de-

– 7 –
termining the total energy we include contributions from any interstellar material contained
within the boundaries of the cluster. We define a classical open cluster as a bound, exposed
cluster, such as the Pleiades, which can live to be at least 108 yrs in the vicinity of the sun.
Unbound clusters are systems whose total energy is positive. That is, unbound clusters are
clusters of 35 or more stars whose space densities exceed 1 M⊙ pc−3 but whose internal stellar
motions are too large to be gravitationally confined by the stellar and non-stellar material
within the boundaries of the cluster.
2.2. Identification & Surveys
Infrared surveys of molecular clouds are necessary to reveal embedded clusters, since
many if not all of their members will be heavily obscured. The initial identification of an
embedded cluster is typically made by a survey at a single infrared wavelength (e.g., 2.2µm
or K-band). The existence of a cluster is established by an excess density of stars over the
background. In general the ease of identifying a cluster depends sensitively on the richness
of the cluster, the apparent brightness of its members, its angular size or compactness, its
location in the Galactic plane and the amount of obscuration in its direction. For example,
it would be particularly difficult to recognize a spatially extended, poor cluster of faint stars
located in a direction where there is a high background of infrared sources, (e.g., l = 0.0, b
= 0.0).
Identification of the individual members of a cluster is considerably more difficult than
establishing its existence. In particular, for most clusters the source density of intrinsi-
cally faint members is usually only comparable to or even significantly less than that of
background/foreground field objects. In such circumstances cluster membership can be de-
termined only on a statistical basis, by comparison with star counts in nearby control fields
off the cluster. However, determining whether or not a specific star in the region is a cluster
member or not is not generally possible from a star counting survey alone. In situations
where field star contamination is non-negligible, other independent information (e.g., proper
motions, spectra, multi-wavelength photometry) is required to determine membership of
individual stars.
The first deeply embedded cluster identified in a molecular cloud was uncovered in near-
infrared surveys of the Ophiuchi dark cloud first made nearly thirty years ago using single-
channel infrared photometers (Grasdalen, Strom & Strom 1974; Wilking & Lada 1983).
However, it wasn’t until the deployment of infrared imaging cameras in the late 1980s,
that large numbers of embedded clusters were identified and studied. In a search of the
astronomical literature since 1988 we have found that well over a hundred such clusters have

– 8 –
been observed both nearby the sun (e.g., Eiroa & Casali 1992) and at the distant reaches of
the galaxy (e.g., Santos et al. 2000). To date embedded clusters have been discovered using
three basic observational approaches: 1) case studies of individual star forming regions, such
as for example, NGC 2282 (Horner, Lada & Lada 1997), LKHα 101 (Barsony, Schombert &
Kis-Halas 1991), and NGC 281 (Megeath & Wilson 1997), 2) systematic surveys of various
signposts of star formation, such as outflows (Hodapp 1994), luminous IRAS sources (e.g.,
Carpenter et al. 1993), and Herbig AeBe stars (Testi, Natta & Palla 1998), and 3) systematic
surveys of individual molecular cloud complexes (e.g., Lada et al. 1991b; Carpenter Snell &
Schloerb 1995; Phelps & Lada 1997; Carpenter, Heyer & Snell 2000; Carpenter 2000). To
date most known embedded clusters have been found in surveys of star formation signposts
(2), in particular the Hodapp (1994) survey of outflows has had by far the most prolific
success rate. In the near future, we expect surveys conducted using the data generated by
the all sky near-infrared surveys (i.e., DENIS and 2MASS) will likely provide the the most
systematic and complete inventory of the embedded cluster population of the Galaxy.
2.3. The Embedded Cluster Catalog
We have compiled a catalog of embedded clusters within ∼ 2 Kpc of the Sun. The
catalog is based on a search of the astronomical literature since 1988. This search produced
information on well over 100 clusters, most of which were identified in various systematic
surveys (e.g, Lada et al. 1991b; Hodapp 1994: Carpenter Heyer & Snell 2000). From
this list we selected 76 clusters which met the following criteria: 1) evidence for embedded
nature by association with a molecular cloud, HII region or some significant degree of optical
obscuration or infrared extinction, 2) identification of 35 or more members above field star
background within the cluster field, and 3) location within ∼ 2 Kpc of the Sun. Due to large
distance uncertainties of regions slightly beyond 2.0 Kpc, such as the W3 molecular clouds (
1.8–2.4 Kpc) we have included clusters with published distance estimates of up to 2.4 Kpc.
Our catalog of nearby embedded clusters is presented in Table 1 which lists the cluster name,
approximate location, distance, radius, number of members, and absolute magnitude limits
of the corresponding imaging observations. These data were compiled from the references
listed in the last column.
Given the heterogeneous nature of the observations from which this sample is drawn,
this catalog cannot be considered complete. In particular, southern hemisphere regions, such
as the Vela complex are not well represented since little observational data exists for this
portion of the Galaxy. In addition at least 24 additional clusters have been identified in
the Rosette GMC (Phelps & Lada 1997), North American and Pelican nebula (Cambresy,

– 9 –
Beichman & Cutri 2002) and Cygnus X region (Dutra & Bica 2001) but no properties
have been presented in the literature for them and they are not included in our cluster
catalog. Also there is a general incompleteness for the more distant clusters due to sensitivity
limitations. We estimate later, that this 2 Kpc sample is complete to only factors of 3-4.
Although not complete, the catalog is, however, likely representative of the basic statistical
properties of embedded clusters within ∼ 2 Kpc. This is because a significant portion of
the catalog is comprised of clusters drawn from systematic imaging surveys of individual
cloud complexes and a survey (Hodapp 1994) which is reasonably complete for clusters
associated with outflows, a primary tracer of very recent star formation activity in molecular
clouds. Moreover, the subset of clusters found in systematic surveys of nearby GMCs (Orion,
Monoceros and Perseus) is also likely to be reasonably complete for clusters with 35 or more
members.
2.4. The Embedded Cluster Mass Function
Masses were derived for each cluster in the catalog by assuming a universal IMF (initial
mass function) for all the clusters. We adopted the IMF of the Trapezium cluster that
was derived by Muench et al. (2002) from modeling of the cluster’s K-band luminosity
function (KLF). We then used the KLF models of Muench et al. (2002) to predict infrared
source counts as a function of differing limiting magnitudes for two model clusters whose
ages correspond to that of the Trapezium (0.8 Myr) and IC 348 (2 Myr) clusters. This
was necessary to attempt to account for the expected luminosity evolution of the PMS
populations of embedded clusters (see discussion below, Section 4). A conversion factor
from total source counts (for a given limiting magnitude) to total mass was then determined
for each synthetic cluster. The infrared source counts listed for each observed cluster in the
catalog were adjusted for distance and variable detection limits and then directly compared
with the two model predictions. In most cases the near-IR limits are faint enough (i.e.,
the IMF is reasonably sampled) that the both models yielded cluster masses that agree
extremely well for the two different ages. Given that no age information is available for the
bulk of the clusters in the catalog, we adopted a conversion factor that was the average of
the Trapezium and the IC 348 cluster ages. Additionally, we assumed that all clusters have
an average extinction of 0.5 magnitudes in the K band. The masses we have derived are
probably uncertain to less than a factor of 2 for most clusters.
The derived cluster masses in our sample range from about 20 to 1100 M⊙. In Figure 2
(left panel) we present the embedded cluster mass distribution function (ECMDF) for all
the clusters in our sample. The ECMDF was derived by summing individual embedded

– 10 –
Fig. 2.— The embedded cluster mass distribution function (ECMDF) for the entire embed-
ded cluster catalog is displayed in the left panel. This plot traces the distribution of total
cluster mass (i.e, N ×MEC) as a function of log mass (log(MEC)). The ECMDF is found to
be flat for clusters with masses between ∼ 50-1000 M⊙. This corresponds to an embedded
cluster mass function with a spectral index of -2 (i.e., dN/dM ∝ M−2). The right panel
compares the ECMDF for the entire catalog with those of two cluster subsamples which are
believed to be more complete at the lowest masses. The dotted line is the Hodapp outflow
sample and the dashed line the sample of all known embedded clusters within 500 pc of the
Sun. The ECMDFs all appear to decline below 50 M⊙.

– 11 –
cluster masses (Mec) in evenly spaced logarithmic mass bins, 0.5 dex in width, beginning at
Log(Mec)=1.2. (The boundaries of the bins were selected to insure that the least populated
bin would have more than one object.) The ECMDF is equal to Mec× dN/dlogMec, and
thus differs by a factor of Mec from the mass function (dN/dlogMec) of embedded clusters.
The histogram with the solid line represents the ECMDF for the entire cluster catalog
(i.e. for clusters having N∗ >35 and D < 2.4 Kpc). The mass distribution function displays
two potentially significant features. First, the function is relatively flat over a range spanning
at least an order of magnitude in cluster mass (i.e., 50 ≤ Mec ≤ 1000 M⊙). This indicates
that, even though rare, 1000 M⊙ clusters contribute a significant fraction of the total stellar
mass, the same as for the more numerous 50–100 M⊙ clusters. Moreover, more than 90%
of the stars in clusters are found in clusters with masses in excess of 50 M⊙ corresponding
to populations in excess of 100 members. The flat mass distribution corresponds to an
embedded cluster mass spectrum (dN/dMec) with a spectral index of -2 over the same range.
This value is quite similar to the spectral index (-1.7) typically derived for the mass spectrum
of dense molecular cloud cores (e.g. Lada, Bally & Stark. 1991c). The fact that the embedded
cluster mass spectrum closely resembles that of dense cloud cores is very interesting and
perhaps suggests that a uniform star formation efficiency characterizes most cluster forming
dense cores. The index for the mass spectrum of embedded clusters is also essentially the
same as that (-1.5 to -2) of classical open clusters (e.g., van den Berg & Lafontaine 1984;
Elmegreen & Efremov 1997).
The second important feature in the ECMDF is the apparent drop off in the lowest mass
bin (∼ 20-50 M⊙). Given that our cluster catalog only included clusters with more than 35
stars, it is likely that we will be considerably more incomplete for clusters in the 20 to 50
M⊙ range than for the higher mass clusters. To test the significance of this fall off to low
cluster masses we consider the mass distribution function of a subset of clusters drawn from a
sample of local clouds where observations are reasonably complete. These were selected from
systematic large scale NIR surveys of 4 molecular clouds (L1630, L1641, Perseus and Mon
R2) without applying any lower limit to the size of the cluster population. Therefore, this
sample should be sensitive to the full mass range of clusters in these representative GMCs.
This local molecular cloud sample is plotted as a dashed line in the right panel of Figure 2.
While the statistical errors due to the small sample size are large, the local sample confirms
that there is indeed a drop off in total cluster mass for the lowest mass clusters. As a further
check, we also plot the ECMDF for the Hodapp (1994) sample, again without applying any
lower limit to the richness of the cluster. We choose the Hodapp sample since it is selected
from a complete sample of outflows, indicative of very young stellar objects, and should not
be biased to any particular mass range of clusters. All three samples are consistent with a
fall off in the cluster mass spectrum below about 50 M⊙. Even if the cluster samples are not

– 12 –
formally complete they should be representative of the total local cluster population within
2 Kpc. Therefore, we conclude that the drop off in the ECMDF at masses less than 50 M⊙
is significant. Consequently, there appears to be a characteristic cluster mass (50 M⊙) above
which the bulk of the star forming activity in clusters is occurring. Recently, Adams and
Myers (2000) suggested, based on dynamical modeling of open clusters and knowledge of the
cluster formation rate, that most clustered star formation occurs in clusters with between
10 and 100 stars. However, our results imply that no more than about 10% of all stars
are formed in such small clusters. The discrepancy results from Adams & Myers use of the
Battinelli & Capuzzo-Dolcetta (1991) catalog of open clusters which undercounts clusters
with ages less than 3 Myrs and underestimates the cluster formation rate as discussed below.
Using the masses in Table 1 we can estimate the contribution to the star formation rate
made by embedded clusters. Because of the incompleteness of our sample, this estimate
necessarily will be a lower limit. To minimize the effect of incompleteness we can calculate
the star formation rate for the local (d < 500 pc) subset of clusters for which we are likely to
be reasonably complete. For this subsample we calculate a local star formation rate of ≥ 1-
3×10−9 M⊙ yr−1 pc−2 assuming typical embedded cluster ages of ∼1-2 Myrs. This rate is in
reasonable agreement with the local star formation rate derived from field stars by Miller and
Scalo (1979) of between 3-7 ×10−9 M⊙ yr−1 pc−2. This suggests that embedded clusters may
account for a large fraction of all star formation occurring locally as has been suggested by
other considerations (Lada et al. 1991b; Carpenter 2000). Extending our sample to 1 and 2
Kpc gives star formation rates of 1-0.7 ×10−9 M⊙ yr−1 pc−2, respectively (for τage ∼ 1 Myr).
The systematic drop in the star formation rate with distance likely reflects progressively
more incomplete cluster surveys as we move to greater distances. If we assume that we are
nearly complete for the local 0.5 Kpc sample, then the drop in the calculated birthrates
would imply that we are incomplete by factors of at least 3 to 4 for the 1 to 2 Kpc samples.
2.5. Birthrates and Star Formation
The embedded cluster catalog can be used to estimate a lower limit to the birthrate of
embedded clusters in molecular clouds. Early estimates of the embedded cluster birthrate,
based primarily on the number of clusters in the Orion cloud complex, found the rate to
be extremely high compared to the birthrate of classical open clusters suggesting that only
a small fraction of embedded clusters survived emergence from molecular clouds to become
classical open clusters (Lada & Lada 1991). Our more extensive embedded cluster cata-
log with an order of magnitude more clusters, allows for a straightforward but much more
meaningful estimate of this important formation rate. For (53) clusters within 2.0 Kpc we

– 13 –
estimate the formation rate to be between 2-4 clusters Myr−1 Kpc−2 for assumed average
embedded cluster ages of 2 and 1 Myrs, respectively. Although this rate is a lower limit, it
is a factor of 8–16 times that (0.25 Myr−1 Kpc−2) estimated for classical open clusters by
Elmegreen & Clemens (1985) and 5-9 times that (0.45 Myr−1 Kpc−2) estimated by Battinelli
& Capuzzo-Dolcetta (1991) for a more complete open cluster sample within 2 Kpc of the
Sun. This difference in birthrates between embedded and open clusters represents an enor-
mous discrepancy and is of fundamental significance for understanding cluster formation and
evolution.
By combining our embedded cluster catalog with the open cluster catalog of Battinelli
& Capuzzo-Dolcetta (1991) we can examine the age distribution of all clusters, open and
embedded, within 2 Kpc of the Sun. The Battinelli & Cappuzzo-Dolcetta catalog contains
about 100 classical open clusters and is thought to be complete out to a distance of 2 Kpc
from the Sun for clusters with MV < -4.5. In Figure 3 we plot the distribution of ages of
all known clusters both embedded and open within 2 Kpc. Embedded clusters populate the
lowest age bin. We have included only those embedded clusters with masses greater than
150 M⊙ to correspond to the magnitude-limited selection of Battinelli & Cappuzzo-Dolcetta.
This represents roughly one-third of our sample of clusters with published distances of 2 Kpc
or less. The average mass of these embedded clusters is 500 M⊙, the same as that estimated
for the open cluster sample by Battinelli & Cappuzzo-Dolcetta (1991). The number of
clusters is found to be roughly constant as a function of age for at least 100 Myr. In Figure 3
we also compare the merged cluster age distribution with the expected age distribution
for a constant rate of cluster formation. Our prediction also includes an adjustment for the
expected luminosity fading of clusters below the detection limits following the perscription of
Battinelli & Cappuzzo-Dolcetta (1991). There is a large and increasing discrepancy between
the expected and observed numbers. These distributions clearly confirm earlier speculations
that the vast majority of embedded clusters do not survive emergence from molecular clouds
as identifiable systems for periods even as long as 10 Myr. Figure 3 suggests an extremely
high infant mortality rate for clusters. Less than ∼ 4% of the clusters formed in molecular
clouds are able to reach ages beyond 100 Myr in the solar neighborhood, less than 10%
survive longer than 10 Myr. Indeed, most clusters may dissolve well before they reach an
age of 10 Myr. If we consider our entire sample of embedded clusters we predict, after
similarly adjusing for fading, that at least 4100 clusters would be detected within 2 Kpc
of the Sun with ages less than 300 Myr. The WEBDA open cluster catalog lists roughly
300 open clusters of this age or less within 2 Kpc, suggesting that only about 7% of all
embedded clusters survive to Pleiades age. It is likely that only the most massive clusters
in our catalog are candidates for long term survival. Roughly 7% of embedded clusters in
our catalog have masses in excess of 500 M⊙, and this likely represents a lower limit to the

– 15 –
mass of an embedded cluster that can evolve to a Pleiades-like system. Moreover, Figure 3
also indicates that the disruption rate for bound clusters between 10–100 Myrs of age is
significant, probably due to encounters with GMCs. Many of the observed open clusters in
this age range may also not be presently bound (Battinelli & Capuzzo-Dolcetta 1991).
The discovery of large numbers of embedded clusters coupled with the high birthrates
and star formation rates we have inferred for them from our analysis of the data in Table 1,
suggests that such clusters may account for a significant fraction of all star formation in the
Galaxy. However, because of the incompleteness of our sample it is difficult to produce an
accurate estimate of the actual fraction of stars born in embedded clusters from statistical
analysis of the data in our catalog. The best estimates of this quantity are derived from
systematic, large scale surveys of individual GMCs. The first systematic attempt to obtain
an inventory of high and low mass YSOs in a single GMC was made by Lada et al. (1991b)
who performed an extensive near-infrared imaging survey of the central regions ( 1 square
degree) of the L1630 GMC in Orion. Their survey produced the unexpected result that
the vast majority (60-90%) of the YSOs and star formation in that cloud occurred within
a few (3) rich clusters with little activity in the vast molecular cloud regions outside these
clusters. A subsequent survey by Carpenter (2000) using the 2MASS database to investigate
the distribution of young stars in 4 nearby molecular clouds, including L1630 produced
similar results with estimates of 50-100% of the clouds’ embedded populations be confined
to embedded clusters. In both studies the lower limits were derived with no correction for
field star contamination which is substantial. Consequently, it is likely that the fraction of
stars formed in clusters is very high (70-90%). Subsequent near-infrared surveys of L1630
(Li, Evans & Lada 1997) as well as other molecular clouds such as Mon OB1 (Lada, Young
& Greene 1993), the Rosette (Phelps & Lada 1997) and Gem OB1 (Carpenter, Snell &
Schloerb 1995) have yielded similar findings suggesting that formation in clusters may be
the dominant mode of formation for stars of all masses in GMCs and that embedded clusters
may be the fundamental units of star formation in GMCs. Since GMCs account for almost
all star formation in the Galaxy, most field stars in the Galactic disk may also have originated
in embedded clusters.
2.6. Association with Molecular Gas and Dust
The intimate physical association with interstellar gas and dust is the defining char-
acteristic of embedded clusters. Embedded clusters can either be partially (i.e., AV ∼ 1–5
mag.) or deeply (i.e., AV ∼ 5–100 mag) immersed in cold dense molecular material or hot
dusty HII regions. The degree of their embeddedness in molecular gas is related to their evo-

– 16 –
lutionary state. The least evolved and youngest embedded clusters (e.g., NGC 2024, NGC
1333, Ophiuchi, MonR2, and Serpens) are found in massive dense molecular cores, while the
most evolved (e.g., the Trapezium, NGC 3603, IC 348) within HII regions and reflection neb-
ulae or at the edge of molecular clouds. Our present understanding of the relation of dense
cores and embedded clusters is largely guided by the coordinated surveys of such clouds as
L1630 (Orion B), Gem OB1 and the Rosette (Mon OB2). These are the clouds for which
the most systematic and complete surveys for both embedded clusters and dense molecular
material exist (Lada 1992; Carpenter, Snell & Schloerb 1996; Phelps and Lada 1997). These
studies all show that embedded clusters are physically associated with the most massive
(100–1000 M⊙) and dense (n(H2) ∼ 104−5 cm−3) cores within the clouds. These cores have
sizes (diameters) typically on the order of 0.5-1 pc. The typical star formation efficiencies
range between 10-30% for these systems. The gas densities correspond to mass densities of
103−4 M⊙ pc−3 suggesting that clusters with central densities of a few times 103 M⊙ pc−3
can readily form from them.
Typically less than 10% of the area and mass of a GMC is in the form of dense gas. This
gas is non-uniformly distributed through the cloud within numerous discrete and localized
cores. These cores range in size between about 0.1 - 2 pc and in mass between a few solar
masses to up to a thousand solar masses. The largest cores which spawn clusters are highly
localized and occupy only a very small fraction (a few %) of the area of a GMC. Numerous
studies have indicated that the mass spectrum (dNdm) of dense molecular cloud cores is a
power-law with an index of α ∼ -1.7 (e.g, Lada, Bally & Stark 1991, Blitz 1993, Kramer et
al. 1998). For such a power-law index, most of the mass of dense gas in a cloud will be found
in its most massive cores, even though low mass cores outnumber high mass cores. Stars
form in dense gas and it is not surprising therefore that a high fraction of all stars form in
highly localized rich clusters, since most of a cloud’s dense gas is contained in its localized
massive cores. Moreover, as discussed earlier, the mass spectrum of cores is very similar to
that of both embedded and classical open clusters.
Not all massive dense cores in molecular clouds are presently forming clusters (e.g.,
Lada 1992). However, in the L1630 cloud, the cores with clusters appear to contain more
gas at very high density (n(H2) > 105 cm−3) and to be more highly clumped or structured
than those cores without clusters (Lada, Evans & Falgarone 1997). Whether this difference
in physical properties is a cause or a result of the formation of a cluster in a massive core
is unclear. Studies of the distribution of dust continuum emission in the Ophiuchi (Motte,
Andre & Neri 1998), Serpens (Testi & Sargent 1998) and the NGC 2068/2071 (Motte et al.
2001) cluster forming cores reveal numerous small scale (∼ 5000 AU) clumps whose mass
spectra are characterized by power-law slopes steeper than those of cloud cores but very
similar to those which characterize the stellar IMF (see below). This would suggest that

– 19 –
Fig. 4.— Contour map of the surface density of J-band infrared sources in the partially
embedded cluster NGC 2264. This is an example of a cluster that displays a hierarchical
structure.

– 23 –
clusters (including the Trapezium, IC 348, NGC 2264 and Rho Ophiuchi), Palla and Stahler
(2000) produced intriguing evidence for a strongly time dependent star formation rate in
these regions. Using a consistent analysis and a single set of PMS tracks, they found that
star formation appeared to be accelerating with time, with the star formation rate reaching
its peak in the last 1-2 Myr in all clusters still associated with significant molecular gas.
However, it is difficult to evaluate the significance of age spreads and distributions
estimated from CMDs of embedded clusters because differential extinction, source variability,
infrared excess, binarity, and contamination by field stars can contribute significantly to the
intrinsic scatter in the diagram (e.g., Hartmann 2001). The uncertainties due to such factors
as variability, infrared excess and extinction are expected to be greater for younger clusters.
Figure 6 shows the CMD obtained for NGC 2362, a 5 Myr old, exposed, open cluster where
such uncertainties should be minimized (Moitinho et al. 2001). The PMS of this cluster is
very well defined and relatively narrow indicating a clear upper limit to its age spread of
< 3 Myr. In this cluster, where the total and differential extinction are barely measurable,
and stellar activity associated with the youngest stars minimal, the CMD indicates a simple
star formation history characterized by cluster formation in a rapid, coeval burst of activity
less than 3 Myr in duration. These observations also may suggest that a significant portion
of the observed scatter in the CMDs of other younger and embedded clusters is due to
factors other than age. Unfortunately, it is presently not possible to determine whether
the large spreads in embedded cluster CMDs result from a wide variety of gestation times,
accelerating star formation or other factors. Since the number of embedded clusters with
age determinations is small, a systematic and detailed examination of a larger sample of
embedded and young open cluster CMDs would be useful in resolving this issue. At present,
self-consistent determinations of the mean ages and, in particular, the relative mean ages
of clusters may be the most robust information about star formation histories that can be
extracted from CMD or HRD analysis.
3. EMBEDDED CLUSTERS AND THE INITIAL MASS FUNCTION
3.1. Background
A fundamental consequence of the theory of stellar structure and evolution is that, once
formed, the subsequent life history of a star is essentially predetermined by one parameter,
its birth mass. Consequently, detailed knowledge of the initial distribution of stellar masses
at birth (i.e., the IMF) and how this quantity varies through time and space is necessary
to predict and understand the evolution of stellar systems, such as galaxies and clusters.
Detailed knowledge of the IMF and its spatial and temporal variations is also particularly

– 25 –
important for understanding the process of star formation, since it is the mysterious physics
of this process that controls the conversion of interstellar matter into stars. Unfortunately,
stellar evolution theory is unable to predict the form of the IMF. This quantity must be
derived from observations. However this is not a straightforward exercise, since stellar mass
is not itself an observable quantity. Stellar radiant flux or luminosity is the most readily
observed property of a star. Determination of stellar masses therefore requires a transfor-
mation of stellar luminosities into stellar masses which in turn requires knowledge of stellar
evolutionary states.
Numerous techniques have been employed in an attempt to determine the IMF both for
the galactic field star population and in open clusters. These techniques and results have
been extensively reviewed in the literature (e.g., Scalo 1978, 1986; Gilmore & Howell 1998;
Meyer et al. 2000; Kroupa 2002). IMFs derived from these studies appear to exhibit two
similar general properties. First, for stars more massive than the sun the IMF has a nearly
power-law form with the number of stars increasing as the stellar mass decreases. If we adopt
the classical definition that the IMF (ξ(logm∗)) is the number of stars formed per unit volume
per unit logarithmic mass interval, the slope at any point is then: β ≡ ∂logξ(logm∗)/∂logm∗,
and β ≈ –1.3 for masses greater than one solar mass (e.g., Massey 1998). This is very similar
to the value (–1.35) originally derived for field stars by Salpeter (1955). Second, the IMF
breaks and flattens near but slightly below 1 M⊙ , departing significantly from a Salpeter
slope. At the lowest masses (i.e., 0.5 - 0.1 M⊙ ), however, there is considerable debate
concerning whether the IMF declines, rises or is flat and whether or not it extends smoothly
below the hydrogen burning limit (HBL) to substellar masses.
However, IMF determinations for local field stars and in open clusters are hampered
by a number of serious difficulties. To deduce the IMF for field stars requires compilation
of a volume limited sample of nearby stars. This in turn requires accurate distance mea-
surements, usually parallaxes, for all stars in the sample. To obtain the necessary complete
sample to as low a mass as possible, necessitates that this volume be limited to stars rel-
atively nearby the sun (d ∼ 5-25 pc), because of the extreme faintness of the lowest mass
stars and the limitations inherent in the distance determinations. Such samples suffer from
incompletness for both the highest mass stars, due to their rarity and complete absence in
the solar neighborhood, and the lowest mass stars due to their faintness. Moreover, such
samples contain stars formed over a time interval encompassing billions of years (essentially
the age of the galactic disk). Therefore, the mass function derived directly from observa-
tions of field stars is a present day mass function (PDMF) and must be corrected for the
loss of higher mass stars due to stellar evolution in order to derive the IMF of the sample.
This, in turn, requires the assumptions of both a star formation rate, usually taken to be
constant, and a time independent functional form of the IMF. The standard final product is

– 28 –
bolometric luminosity function of a cluster would be most desirable for comparison with
theoretical predictions (e.g., Lada & Wilking 1984; Fletcher & Stahler 1994a,b), obtaining
the multi-wavelength observations necessary to do so would require prohibitive amounts
of observing time on telescopes both on the ground and in space. On the other hand, the
monochromatic brightness of a star is its most basic observable property and infrared cameras
enable the simultaneous measurement of the monochromatic brightnesses of hundreds of
stars. Thus, complete luminosity functions, which span the entire range of stellar mass,
can be readily constructed for embedded stellar clusters with small investments of telescope
time. The monochromatic (e.g., K band) luminosity function of a cluster, dNdmK , is defined as
the number of cluster stars per unit magnitude interval and is the product of the underlying
mass function and the derivative of the appropriate mass-luminosity relation (MLR):
dN
dmK
=
dN
dlogM∗
×
dlogM∗
dmK
(1)
where mk is the apparent stellar (K) magnitude, and M∗ is the stellar mass. The first term
on the right hand side of the equation is the underlying stellar mass function and the second
term the derivative of the MLR. With knowledge of the MLR (and bolometric corrections)
this equation can be inverted to derive the underlying mass function from the observed
luminosity function of a cluster whose distance is known. This method is essentially that
originally employed by Salpeter (1955) to derive the field star IMF. However, unlike main
sequence field stars, PMS stars, which account for most of the stars in the an embedded
cluster, cannot be characterized by a unique MLR. Indeed, the MLR for PMS stars is a
function of time. Moreover, for embedded clusters the duration of star formation can be
a significant fraction of the cluster’s age. Consequently, to invert the equation and derive
the mass function one must model the luminosity function of the cluster and this requires
knowledge of both the star formation history (i.e., age and age spread) of the cluster as well as
the time-varying PMS mass-luminosity relation. This presents the two major disadvantages
for this technique. First, a priori knowledge of the age or star formation history of the
cluster is required and this typically can be derived by placing cluster stars on an HRD.
However, this in turn requires additional observations such as multi-wavelength photometry
or spectroscopy of a representative sample of the cluster members. Second, PMS models
must be employed to determine the time varying mass-luminosity relation. The accuracy
of the derived IMF therefore directly depends on the accuracy of the adopted PMS models
which may be inherently uncertain, particularly for the youngest clusters (τ < 106 yrs) and
lowest mass objects (m < 0.08 M⊙ ). In addition, most PMS models predict bolometric
luminosities as a function of mass and time, and thus bolometric corrections must be used
to transform the theoretical predictions to monochromatic fluxes and magnitudes.

– 31 –
Fig. 7.— Comparison of theoretical predictions for the luminosities and effective temper-
atures of million year old PMS stars as a function of mass from a suite of standard PMS
models. The predicted K magnitudes of such stars (top) are in excellent agreement over the
entire mass spectrum, while the predicted effective temperatures (bottom) are in relatively
poor agreement.

– 32 –
the vast majority of binaries are not equal mass (brightness) systems this contribution is
typically small (0.1 - 0.2 magnitudes; e.g., see Simon et al. 1995) compared to the typical
bin sizes (0.5 mag) used to construct the infrared luminosity functions. Second, the presence
of unresolved binaries can result in an underestimate of the numbers of low mass stars in a
cluster compared to that expected for a system of stars in which all binaries are resolved,
since companion stars are not directly observed or counted (Kroupa et al. 1991). Thus, the
IMFs that are derived are system or primary star IMFs. Whether or not such a primary
star IMF should be adjusted by adding in the masses of companion stars depends on the
question being considered. For example, for comparisons with IMFs derived for field stars
as well as open and globular clusters, the primary star IMF is the appropriate IMF to use.
If one desires to exactly weigh the amount of interstellar medium transformed into stars
by the star formation process, then a primary + companion star IMF would be the more
appropriate mass function to consider. Unfortunately, the IMF of companion stars is not
very well known or constrained by existing observational data and determination of a total
IMF including primary and companion stars is not presently possible. Finally, since the
IMF is a statistical property of an ensemble of stars, it can only be meaningfully derived
over a mass interval which is statistically well sampled by observations. The richness of the
observed cluster thus sets a basic limit on the level of uncertainty in any derived IMF.
3.3. The IMF of the Trapezium Cluster from OB Stars to Brown Dwarfs
The Trapezium cluster in Orion is the best studied of all embedded clusters. First
identified by Trumpler (1931), the Trapezium cluster is a rich cluster of faint (mostly PMS)
stars embedded within the Great Orion Nebula with an age of about 106 yrs (Prosser et
al. 1994; Hillenbrand 1997). The cluster is approximately 0.3 - 0.4 pc in diameter (e.g.,
Lada et al. 2000) and contains approximately 700 stars (Hillenbrand & Carpenter 2000;
Muench et al 2002). It is thought to be the highly concentrated core of the more extended
Orion Nebula Cluster (ONC) which contains nearly 2000 stars spread over a region roughly
4 pc in extent (e.g., Hillenbrand & Hartmann 1998). At its center is the famous Trapezium,
a close grouping of four OB stars which excite the nebula. It is a superb target for IMF
studies because of its youth, richness, compactness, location in front of and partially within
an opaque molecular cloud, and its proximity to the sun (∼ 450 pc). These factors combine
to enable a statistically significant sampling of the IMF from OB stars to substellar objects
near the deuterium burning limit ( 0.01 M⊙ ) with minimal field star contamination. Indeed,
this cluster is particularly well suited for investigating the substellar portion of the IMF and
determining the initial distribution of masses for freely floating brown dwarfs. Deep infrared
surveys of this cluster have been performed using the HST (Luhman et al 2000), the Keck

– 33 –
Telescope (Hillenbrand & Carpenter 2000), UKIRT (Lucas & Roche 2000) and the NTT
(Muench et al 2002) and have produced infrared luminosity functions and mass functions
which sample well into the substellar mass range.
Figure 8 shows a three-color infrared image of the cluster resulting from the NTT
survey. Muench et al. (2002) used this data along with observations of the same region
obtained with a 1.2 meter telescope to recover the brighter stars typically saturated in deep
exposures with the larger telescopes and they produced a complete sampling of the K-band
(2.2 µm) luminosity function (KLF) of this cluster spanning the mass range from OB stars to
substellar objects near the deuterium burning limit. Figure 9 shows the field-star corrected,
complete, extinction-limited, KLF derived from the Muench et al. study. It counts all stars
within a cloud depth of 17 magnitudes of visual extinction with luminosities corresponding
to million year old objects with masses ∼ 0.010 – 0.015 M⊙ and greater and is representative
of the infrared luminosity functions obtained in all similarly sensitive investigations of this
cluster. In particular, the KLF is found to rise steadily from the brightest stars to around
mK ∼ 11 -12 mag where it flattens before clearly falling again to fainter magnitudes. A clear
secondary peak is present at approximately 15th magnitude, which is well into the brown
dwarf luminosity range. At lower luminosities the KLF rapidly drops off.
Muench et al. derived the IMF of the Trapezium cluster by using a suite of Monte Carlo
calculations to model the cluster’s KLF. The observed shape of a cluster luminosity func-
tion depends on three parameters: the ages of the cluster stars, the cluster mass-luminosity
relation, and the underlying IMF (i.e., Equation 1). With the assumptions of a fixed age
distribution, derived from the spectroscopic study of the cluster by Hillenbrand (1997), a
composite theoretical mass-luminosity relation adopted from published PMS calculations
(i.e., Bernaconi 1996; Burrows et al. 1997; D’Antona & Mazzitelli 1997; Schaller 1992),
and an empirical set of bolometric corrections, Muench et al. varied the functional form of
the underlying IMF to construct a series of synthetic KLFs. These synthetic KLFs were
then compared to the observed Trapezium KLF in a Chi-Squared minimization procedure
to produce a best-fit IMF. As part of the modeling procedure, the synthetic KLFs were
statistically corrected for both variable extinction and infrared excess using Monte Carlo
probability functions for these quantities derived directly from multi-color (JHK) observa-
tions of the cluster.
The best-fit synthetic KLF is plotted in Figure 9. The corresponding underlying mass
function is displayed in Figure 10 in the form of a histogram of binned masses of the stars
in the best-fit synthetic cluster. This model mass function represents the IMF of the young
Trapezium cluster. The main characteristics of this IMF are: 1) the sharp power-law rise of
the IMF from about 10 M⊙ (OB stars) to 0.6 M⊙ (dwarf stars) with a slope (i.e., β = –1.2)

– 34 –
Fig. 8.— Optical (top) and deep JHK infrared (bottom) images of the Trapezium cluster
in Orion obtained with the NASA HST and the ESO VLT, respectively. The infrared
observations taken from Muench et al. 2002

– 36 –
similar to that of Salpeter (1955), 2) the break from the single power-law rise at 0.6 M⊙
followed by a flattening and slow rise reaching a peak at about 0.1 M⊙, near the hydrogen
burning limit, 3) the immediate steep decline into the substellar or brown dwarf regime and
finally 4) the prominent secondary peak near 0.015 M⊙ or 15 MJ (Jupiter masses) followed
by a very rapid decline to lower masses beyond the deuterium burning limit (at ∼ 10 MJ ).
The most significant characteristic of this IMF is the broad peak, extending roughly
from 0.6 to 0.1 M⊙. This structure clearly demonstrates that there is a characteristic mass
produced by the star formation process in Orion. That is, the typical outcome of the star
formation process in this cluster is a star with a mass between 0.1 and 0.6 M⊙. The process
produces relatively few high mass stars and relatively few substellar objects. Indeed, no
more than ∼ 22% of all the objects formed in the cluster are freely floating brown dwarfs.
The overall continuity of the IMF from OB stars to low mass stars and across the hydrogen
burning limit strongly suggests that the star formation process has no knowledge of the
physics of hydrogen burning. Substellar objects are produced naturally as part of the same
physical process that produces OB stars (see also Najita, Tiede & Carr 2000; Muench et al.
2001).
In this respect the secondary peak at 0.015 M⊙ is intriguing. The existence of such a
peak may imply a secondary formation mechanism for the lower mass brown dwarfs, similar
to suggestions recently advanced by Reipurth and Clarke (2000) and thus is potentially
very important. However, the significance that should be attached to this feature depends
on the accuracy of the adopted mass-luminosity relations for substellar objects used in the
modelling. These MLRs may be considerably more uncertain than those of PMS stars.
Indeed, observations of an apparent deficit of stars in the M6-M8 range of spectral types in
a number of open clusters are suggestive of the existence of a previously unknown opacity
feature in the MLR for such cool stars (Dobbie et al. 2002). The presence of such a feature
could produce a peak in the luminosity function and a corresponding artificial peak in the
derived mass function if not included in the theoretical MLRs (e.g., Kroupa Tout & Gilmore
1990, 1993). Given that this spectral type range corresponds to the temperature range
predicted for young low mass brown dwarfs, it is quite plausible that the secondary feature
in the derived IMF is artificial and does not represent a true feature in the underlying IMF.
Clearly more data, both observational and theoretical is needed to assess the reality and
significance of this intriguing feature.
Figure 11 shows a comparison of Trapezium IMFs recently derived from a number of
different deep infrared imaging surveys using a variety of methods (Lucas & Roche 2000;
Hillenbrand and Carpenter 2000; Muench et al. 2000; Luhman et al. 2000; Muench et
al. 2003). The general agreement of the derived IMFs is impressive for m∗ > 0.015 M⊙.

– 37 –
1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5 -2.0
Log Mass (solar masses)
0.0
1.0
2.0
Lo
g
N
(lo
g m
) +
C
on
sta
nt
Trapezium Cluster Initial Mass Function
Fig. 10.— The IMF derived for the Trapezium cluster from Monte Carlo modelling of its
luminosity function (Muench et al. 2002). This plot displays the binned mass function of
the synthetic cluster whose luminosity function was found to best fit the observed KLF of
the Trapezium cluster. (See Figure 9) A vertical dashed line marks the approximate location
of the hydrogen burning limit (HBL). The derived IMF displays a broad peak between 0.1
- 0.6 M⊙ and extends deep into the substellar mass regime. The secondary peak is located
near 0.015 M⊙ or 15 MJ . It corresponds to the bump in the KLF at K ∼ 15.5 magnitudes
seen in Figure 9 and may be an artifact of the adopted substellar MLR.

– 39 –
The fundamental features (1-3) described above are evident in all the IMFs. In the region
of the secondary peak (4), the agreement is less impressive, likely reflecting the inherent
uncertainties in the modeling at the lowest substellar masses. Nonetheless, there is general
agreement that the IMF turns over and falls off below the hydrogen burning limit and
into the substellar mass regime. However, the precise details, such as the steepness of
the falloff and the amplitude of the secondary peak, remain somewhat uncertain. Other
differences in details between the various IMFs likely result from the uncertainties inherent
in the different techniques used in the IMF determinations and provide some measure of
the overall uncertainty in our present ability to measure the exact form of the IMF in this
cluster. Clearly, however, these infrared studies of the Trapezium cluster have established
the fundamental properties of its IMF.
The derived IMF of the Trapezuim cluster spans a significantly greater range of mass
than any previous IMF determination whether for field stars or other clusters (e.g., Kroupa
2002). Its statistically meaningful extension to substellar masses and the clear demonstration
of a turnover near the HBL represents an important advance in IMF studies.
For masses in excess of the HBL the IMF for the Trapezium is in good agreement with
the most recent determinations for field stars (Kroupa 2002). This is to some extent both
remarkable and surprising since the field star IMF is averaged over billions of years of galactic
history, assuming a constant star formation rate, over the age of the Galaxy, and over stars
originating from very different locations of galactic space. The Trapezium cluster, on the
other hand, was formed within the last million years in a region considerably less than a
parsec in extent. Moreover, there is evidence that this region is not yet finished producing
stars as significant star formation appears to be continuing in the molecular cloud behind
the cluster (Lada et al 2000). Taken at face value this agreement suggests that the IMF and
the star formation process that produces it is very robust, at least for stellar mass objects.
3.4. Comparison With Other Embedded Clusters: A Universal IMF?
Although few other embedded clusters have been as completely studied as the Trapez-
ium, the luminosity and mass functions of a number of such clusters have been investigated
in various levels of detail. These include such clusters as IC348 (Lada & Lada 1995; Luhman
et al. 1998; Najita, Tiede & Carr 2000; Muench et al. 2003), NGC 1333 (Aspin, Sandell &
Russell 1994; Lada, Alves & Lada 1996), NGC 2264 (Lada, Young & Greene 1993), NGC
2024 (Lada et al. 1991b; Comeron et al. 1996; Meyer 1996), Rho Ophiuchi (Lada and
Wilking 1984; Comeron et al. 1993; Green and Meyer 1995; Bontemps et al. 2001), Serpens
(Eiroa & Casali 1992; Giovannetti et al. 1998), M17 (Lada et al. 1991a), W3 (Megeath et

– 40 –
al. 1996), and NGC 3603 (Eisenhauer et al. 1998; Brandl et al. 1999; Nurnberger & Petr-
Gotzens 2002). Due varying distances, sizes, sensitivities, methodologies, etc., these KLFs
and corresponding IMFs were not uniformly sampled nor investigated using a common sys-
tematic approach. As a result only limited conclusions can be drawn from comparison of all
these results with each other and with the IMF derived for the Trapezium cluster. On the
other hand, when homogeneous data is analyzed with similar methodology more meaningful
comparisons of embedded cluster IMFs are possible (e.g., Lada et al. 1996; Luhman et al.
2000; Muench et al. 2002).
The first conclusion that can be drawn from such studies is that the KLF for embedded
clusters is not a universal function and statistically significantly variations are present in
observed clusters (e.g., Lada, Alves & Lada 1996). This is illustrated in Figure 12 which
displays the KLFs of the Trapezium, IC 348 and NGC 2362 clusters. Such variations in the
cluster luminosity functions are not unexpected for embedded clusters which consist mostly
of PMS stars. Even if such clusters were characterized by a universal IMF, they would ex-
perience significant luminosity evolution as their PMS population evolved and collectively
approached the main sequence. This luminosity evolution would be particularly rapid during
the first 5 Myrs of a cluster’s existence when the luminosities of its PMS stars experience
their most rapid declines. For example, Figure 13 shows the expected KLFs for three dif-
ferent aged synthetic clusters with identical IMFs. These KLFs were calculated from Monte
Carlo simulations by Muench et al. (2000). The systematic evolution of the KLFs to lower
brightness is clearly evident and certainly would be significant enough to be observable.
Indeed, modeling of the KLFs of clusters such as IC 348, NGC 2362 and the Trapezium,
indicates that observed differences in their KLFs can be explained by the expected lumi-
nosity evolution in clusters that have very similar underlying mass functions (Lada & Lada
1995; Muench et al. 2002; Alves et al. 2003). This can also be inferred empirically, without
resorting to modeling. Comparison of the KLFs of young clusters of known age shows that
clusters of similar age display KLFs of very similar shape while clusters of differing age show
the greatest variation of KLF forms (Lada, Alves & Lada 1996; Alves et al. 2003).
To illustrate this further we display in Figure 14 the background corrected KLF and
corresponding IMF derived for the IC 348 cluster from Monte Carlo modeling (Muench et
al. 2003). IC 348 is the best studied embedded cluster after the Trapezium. Although it is
not as rich and suffers appreciably more background contamination, this cluster is closer to
the solar system, older (τ ∼ 2-3 Myr.) and more evolved than the Trapezium cluster (Lada
& Lada 1995; Herbig 1998). This places additional constraints for IMF modeling making
the cluster a good candidate for comparison with the Trapezium. As seen in the figure,
the IMF for this cluster derived from KLF modeling compares very well with that of the
Trapezium over the entire range of mass (m∗ >∼ 15MJ) over which it was determined. In

– 42 –
Fig. 13.— Model luminosity functions (KLFs) of synthetic clusters of differing age but
with the same underlying IMF (that of the Trapezium cluster). Young clusters experience
systematic luminosity evolution, gradually becoming fainter with time as PMS stars within
them approach the main sequence. Figure prepared by August Muench.

– 43 –
Fig. 14.— The KLF and derived IMF for IC 348. The left panel shows the observed
background corrected KLF (histogram) along with the best fit synthetic KLFs (filled and
open circles) corresponding to the underlying model IMFs (filled and open circles) shown
in the right hand panel. Also plotted is the IMF derived for the Trapezuim cluster. From
Muench et al. (2003).
particular, the IMF displays a broad peak between 0.1 - 0.5 M⊙ with a clear turn down near
the HBL. This indicates that there is a characteristic mass produced by the star formation
process in IC 348 and this mass is essentially the same as that suggested for star formation
in the Trapezium. Moreover, the size of the substellar population of IC 348 is relatively well
determined and constitutes only about∼ 25 % of the cluster membership, again similar to the
Trapezium. Because of poor statistics, Muench et al. did not fit the KLF in the luminosity
range that corresponds to the location of the secondary peak in the substellar IMF of the
Trapezium. However examination of the IC 348 KLF shows a marginally significant bump
at the luminosity corresponding to that mass. The relatively poor statistics in the IC 348
KLF at these faint magnitudes is due to a combination of the fact that IC 348 is not as
rich as the Trapezium and suffers more contamination from background stars since it is not

– 45 –
(∼ 0.6 M⊙ ) of the corresponding inflection point in the Trapezium, open cluster and field
star IMFs and suggests a relative deficit of lower mass stars in this rich O cluster. However,
it is likely that the numbers of solar mass stars in R136 has been underestimated due to
the severe crowding in the cluster center. It is still quite possible that even in this cluster,
which contains 1000 O stars, the underlying IMF is characterized by the same universal
form as that derived for the Trapezium. A more significant indication of a departure from a
universal IMF may be present in recent observations of the nearby Taurus clouds. Luhman
(2000) has found a significant (factor of 2) deficit of substellar mass objects in this region
relative to the embedded population in the Trapezium and IC 348 clusters. The embedded
population of the Taurus clouds is an embedded T association consisting of isolated stars
and small loose groupings of stars formed over a relatively large area. These conditions
are decidedly different than those which characterize embedded cluster formation. Since for
Taurus the IMF above the HBL appears to be similar to that of clusters and the field (e.g.,
Kenyon & Hartmann 1995), the finding of a deficit of brown dwarf stars may indicate that
the substellar IMF is less robust than the stellar IMF and thus may be a sensitive function
of formation environment and/or initial conditions. However, more observations would be
necessary to test the significance of this possibility.
4. LABORATORIES FOR STAR AND PLANET FORMATION
Nearly half a century ago, Walker’s (1956) observations of the partially embedded cluster
NGC 2264 showed that its late type (i.e, F and later) stars were characterized by subgiant
luminosities which placed them well above the main sequence on the HRD. Thus, these
observations empirically established the pre-main sequence, pre-hydrogen burning, nature of
young low mass stars and provided the critical data needed to test and constrain the theory of
PMS evolution (e.g., Hayashi 1966). Thirty years later infrared observations of the embedded
cluster in Ophiuchus enabled the first systematic classification of infrared protostars and
young stellar objects based on emergent stellar energy distributions (Wilking & Lada 1983;
Lada & Wilking 1984, Lada 1987). Such observations were very influential in constructing
the early framework for a theoretical understanding low mass star formation (e.g., Shu,
Adams & Lizano 1987). Today, embedded clusters continue to play an important role for
the development and testing of theories dealing with the formation and early evolution of
both stars and planetary systems.

– 48 –
Fig. 15.— Disk Fraction as a function of cluster age for a sample of young clusters with
consistently determined mean ages. The disk fraction is initially very high, but then rapidly
drops with cluster age suggesting maximum disk lifetimes of less than 6 Myrs in young
clusters (Haisch et al. 2001a)

– 50 –
a rapid decline in the CDF with cluster age. Half of the disks in a cluster appear to be lost
within only 2-3 Myrs and essentially all the disks are gone in about 5-6 Myrs. Moreover,
observations also indicate that disk lifetimes are also functions of stellar mass with disks
around higher mass stars evolving more rapidly than disks around low mass stars (Haisch,
Lada & Lada 2001b) Such a rapid timescale for disk evolution places stringent constraints
on the timescale for building planets, particularly giant gaseous planets.
However, the JHKL observations of Haisch et al (2001a) trace the infrared excess arising
from small (micron-sized), hot (900 K) dust grains located in the inner regions of a circum-
stellar disk (∼ 0.25 AU). It is possible for substantial amounts of material to be still present
in the disk if either this material is purely gaseous or if the disk has a large inner hole. The
former situation could arise if significant grain growth occurs due dust settling and rapid
grain coagulation and growth, as is the expected first step in the formation of planetesimals.
However, any turbulence in the disk would be expected to keep detectable amounts of small
dust grains suspended within the gas (Ruden 1999), so dust should still remain a good tracer
of total disk mass as the disks evolve. Indeed, recent observations of dust and H2 emission
in more evolved and much less massive debris disks appears to support this assumption (Thi
et al. 2001) The latter possibility could occur if protoplanetary disks evolved large inner
holes as they aged. It has been noted that in the Taurus clouds, near-infrared L-band excess
is closely correlated with millimeter-continuum emission which traces outer disk material
suggesting that the evolution of the inner and outer disks is homologous and occurs on the
same timescale (Haisch et al. 2001b). Similarly, a recent millimeter continuum survey of 4
young embedded clusters (NGC 2071, NGC 2068, NGC 1333, and IC 348) has found that
the variation in the fraction of detected millimeter sources from cluster to cluster is similar
to the variation in the fraction of near-IR excess sources (Lada, Haisch and Beckwith 2003).
This implies that the lifetimes of the inner disk, as detected by near-IR excess, and the
outer disk, as detected at millimeter wavelengths, are coupled for embedded clusters as they
were for the Taurus population. These observations strengthen the previous L band excess
determination of a short disk lifetime of < 6 million years, and further suggest that the
lifetime for massive outer disks may be only as long as 3 million years. Indeed, the dearth of
strong millimeter continuum emission from the disk population of the very young Trapezium
cluster implies even more rapid outer disk evolution (Mundy, Looney & Lada 1995; Bally et
al. 1998; Lada 1999). It may also be possible that disk lifetimes depend on internal cluster
environment (e.g., the degree to which a cluster produces O stars). Systematic surveys of a
larger population of clusters are needed to assess this possibility. In particular, mid-infrared
surveys that will be carried out by the SIRTF mission, should more definitively address this
question.

– 51 –
4.3. Brown Dwarfs
As discussed earlier in this review, embedded clusters contain significant populations
of substellar objects. Indeed, the Trapezium cluster alone contains nearly as many brown
dwarfs as have been identified in the galactic field to this point in time. However, in the
two clusters (Trapezium and IC 348) with the most complete substellar census, substellar
objects account for only about 20-25% of the total cluster population. Thus, if embedded
clusters such as these, supplied the galactic field population, then we would expect to find
only 1 brown dwarf for every 3-4 stars in a typical volume of Galactic space.
Embedded cluster research has provided one of the most interesting discoveries concern-
ing the nature of brown dwarfs. Examination of young brown dwarfs and brown dwarf can-
didates in a number of embedded clusters, including IC 348, Ophiuchus and the Trapezium
and in other star forming regions, such as the Chamaeleon Clouds, has produced evidence
that some brown dwarfs emit excess emission at near- and mid-infrared wavelengths sim-
ilar to that emitted by stars with circumstellar disks (e.g., Wilking et al. 1999, Luhman
1999; Nata & Testi 2001; Natta et al. 2002). Moreover, deep near-infrared images of the
Trapezium have enabled the first statistically significant determination of the disk frequency
for an embedded substellar population. Muench et al. (2001) found a very high fraction
(65%) of the substellar objects in the cluster to display infrared excess emission at 2 µm
suggestive of the presence of circumstellar disks. Moreover, Muench et al (2001) discovered
that about 20% of the substellar population were optical proplyds on HST archive images,
independently confirming the presence of circumstellar disks around a significant fraction
of the substellar objects in the cluster. Because observations of 2 µm excess undercounts
disk bearing stars, the actual fraction of substellar objects with disks is likely greater than
65% and thus very similar to the disk fraction (80%) observed for the stellar population of
the cluster (Lada et al. 2000). Like stars, brown dwarfs appear to be formed surrounded
by disks and as a result possess the ability to form planetary systems. The detection of
infrared excess around the faintest sources in the KLF of the Trapezium cluster also conclu-
sively established the nature of these sources as young objects and cluster members and thus
confirmed their status as bona fide substellar objects. Furthermore, the detection of a high
disk fraction (similar to that of the stellar population) coupled with the smooth continuity
of the IMF across the hydrogen burning limit provides strong evidence that freely floating
brown dwarfs are a natural product of the star formation process in embedded clusters. The
formation process for brown dwarfs is essentially identical to that of stars.

– 52 –
4.4. Binary Stars
Since most field stars appear to be binary systems understanding the origin of binary
star systems is of fundamental importance for developing a general theory of star formation.
Clusters are important laboratories for investigating binary formation and evolution. In this
context it is interesting that the binary fractions of well studied embedded clusters such as
the Trapezium (Prosser et al. 1994; Petr et al. 1998; Simon, Close & Beck 1999), IC 348
(Duchene, Bouvier & Simon 1999) and Rho Ophiuchi (Simon et al. 1995) are found to be
indistinguishable from that of the galactic field. On the other hand, it is well established
that the binary fraction of the embedded population in the Taurus-Auriga association is
significantly (a factor of 2) in excess of that of the galactic field (e.g., Leinert et al. 1993;
Ghez, Neugebauer, & Matthews 1993; Duchene 1999). This difference in binary fraction
provides an important clue relating to the origins of galactic field stars. Specifically, it
supports the notion that most field stars originated in embedded clusters rather than in
embedded associations such as the Taurus-Auriga clouds.
However, it has been suggested that the Taurus-Auriga binary fraction may represent
the initial binary fraction even for stars that form in embedded clusters. Kroupa (1995a,b)
performed N-body experiments to simulate the evolution of a binary population in a young
cluster. These experiments began with clusters containing 100% binaries and showed that
stellar encounters could disrupt binaries and reduce the overall binary fraction with time.
Moreover, Kroupa found that the field star binary population could be produced by such a
model, if most stars formed in what he identified as a dominant-mode cluster, a cluster with
roughly 200 systems and a half mass radius of 0.8 pc. The fact that such systems are rela-
tively common (e.g., Table 1) lends support to this notion. Furthermore, the concept that
disruption of multiple systems can occur in clusters appears to be supported by the obser-
vation of a deficit of wide binaries in the Trapezium cluster (Scally, Clarke & McCaughrean
1999). However, the binary fraction of embedded clusters is not significantly different from
that of much older open clusters (Patience & Duchene 2001). This indicates that any evo-
lution of the binary fraction must have occurred on timescales of order 1 Myr or less. Such
rapid evolution in the binary population can occur for a very dense embedded cluster such
as the Trapezium (Kroupa, Petr & McCaughrean 1999, Kroupa Aarseth, & Hurley 2001).
It is not clear, however, whether binaries can be as efficiently disrupted in the lower density
clusters which account for most star formation. The question of a universal initial binary
fraction remains open. Star formation may produce a variety of outcomes for the emerging
binary fraction due to a corresponding variety of initial conditions. Determinations of the
binary properties of the protostellar populations in embedded clusters would provide an im-
portant test of this question. If there is a universal initial binary fraction then essentially all
protostellar objects must be nascent binary systems. High resolution infrared imaging and

– 53 –
spectroscopic monitoring of Class I sources in embedded clusters could resolve this issue.
5. ORIGIN AND DYNAMICAL EVOLUTION
5.1. Formation of Embedded Clusters
To understand how an embedded cluster forms we must understand two basic physical
processes: 1) the formation of a massive, dense core in a GMC and 2) the subsequent
development of stars from dense gas in the core. Molecular clouds form from the turbulent,
diffuse and atomic interstellar medium by a physical process or collection of processes that
are far from understood. Overall this process likely involves the complex interplay of such
things as spiral density waves, supernova explosions, the galactic dynamo, phase transitions,
and various types of instabilities (e.g, thermal, gravitational, magneto-hydrodynamic, etc.)
(e.g., Elmegreen 1991, 1993). The vast majority of GMCs are observed to contain dense
gas and signposts of star formation suggesting that the formation of dense cores and then
stars proceeds very rapidly after the cloud has formed from the diffuse interstellar medium.
The GMCs that form from the ISM are gravitationally bound entities with highly supersonic
and turbulent velocity fields. The turbulent dissipation timescales for GMCs are thought to
be shorter than the cloud lifetime suggesting that on global scales the clouds are stabilized
against collapse by internal turbulent pressure. Numerical simulations (e.g., Klessen, Heitsch
& Mac Low 2000) suggest that under such conditions supersonic turbulent flows can collide,
shock and dissipate energy. Under the right conditions these collisions can produce dense
cores which are gravitationally unstable and decouple from the overall turbulent flow. The
largest and most massive of these fragments are then the potential sites of cluster formation.
The second step of the cluster formation process, the rapid evolution of dense gas in
a massive core to form stars, likely involves the continued dissipation of turbulence in the
dense gas which is followed by, fragmentation, gravitational instability and the formation
of protostellar seeds which grow by accreting their infalling envelopes and then perhaps
other surrounding dense gas from the general potential in which they are embedded (e.g.,
see reviews by Clarke et al. 2001, Elmegreen et al. 2001). This scenario appears to be
quite different than that which has successfully explained the formation of isolated low mass
stars from individual low mass cores. Such solitary stars form from initially turbulent,
magnetically supported, dense cores which evolve through ambipolar diffusion of magnetic
fields to be dynamically unstable and then collapse from the inside out (Shu, Adams &
Lizano 1987). The cores which form isolated low mass stars in this manner have sizes that
are considerably larger than the separation of stars in an embedded cluster. Evidently
protostellar cores in clusters must have smaller radii than those which form in isolation.

– 54 –
This suggests that cluster forming cores must experience significant fragmentation in their
evolution to form stars. The physical mechanism that produces this fragmentation is not
well understood. This process likely involves progressive cooling of a marginally stable or
collapsing massive core which continuously reduces the Jean’s mass. In turbulent dense
cores this cooling takes place as a result of dissipation or loss of turbulence. An elegant
possibility to account for fragmentation was proposed by Myers (1998) for the case of MHD
turbulence. If the ionization rate in a massive core is low enough (i.e., the extinction is
high enough that cosmic rays are the sole source of ionization) then MHD waves greater
than a certain frequency cannot couple well to the neutral gas. This corresponds to a
cutoff wavelength, below which turbulence can no longer be sustained (e.g., Mouschovias
1991). This situation can lead to the formation of a matrix of critically stable Bonnor-Ebert
condensations or kernels confined by the pressure in the surrounding gas. Myers (1998) finds
that for typical conditions the sizes of these kernels can be comparable to the separation
of stars in embedded clusters. Fragmentation can also be produced in the turbulent decay
process as flows collide and shock, creating density enhancements, which if massive enough
can become gravitationally bound and separate from the general turbulent velocity field
(e.g., Klessen & Burkert 2000, 2001).
Once these fragments or kernels become gravitationally unstable they collapse, and
gaining mass through infall of surrounding material become protostars. However, the rates
at which protostellar condensations grow must vary significantly within the cluster. This is
because the star formation process must produce a range of stellar and substellar masses
spanning three orders of magnitude within a timescale of only a few (1-2) million years
in order to reproduce the stellar IMF. As they move through the cluster core protostellar
fragments also accrete additional material from the reservoir of residual gas not locked up
in other protostellar objects (Bonnell et al. 2001a). Since all these stellar embryos share
a common envelope, a process of competitive accretion begins with initially more massive
protostellar clumps or clumps closer to the center of the cluster experiencing higher accretion
rates. The process is highly nonlinear and, even for a cluster with initially equal mass
protostellar fragments, can lead to the development of a protostellar mass spectrum similar
to that of the stellar IMF (Bonnell et al. 2001b; Klessen 2001). In this picture the more
massive stars tend to be formed in the central regions of the cluster leading to some degree
of primordial mass segregation. It is also possible for protostellar fragments in the dense
inner regions of the cluster to collide and coalesce leading to the production of very massive
stars (Bonnell, Bate & Zinnecker 1998). It otherwise would be difficult to built up a massive
star from general accretion since radiation pressure from embryonic stars more massive than
about 10 M⊙ can reverse infall and stunt the growth of the star (e.g., Adams, Lada & Shu
1987).

– 57 –
cluster). There exist two important dynamical regimes for τgr corresponding to explosive
(τgr << τcross) and adiabatic (τgr >> τcross) gas removal times. Typical embedded clus-
ters are characterized by τcross ∼ 1 Myr. Clusters that form O stars will likely remove any
residual gas on a timescale shorter than this dynamical time. O stars can quickly ionize and
heat surrounding gas to temperatures of 104 K causing an abrupt increase in pressure which
results in rapid expansion of the gas. The expansion velocities are on the order of the sound
speed in the hot gas and for the dimensions of embedded clusters correspond to gas removal
timescales that can be as short as a few times 104 years. The dynamical response of the stars
which are left behind after such explosive gas removal will depend on the SFE achieved by
the core at the moment of gas dispersal. The condition for the cluster to remain bound in
the face of rapid gas removal is that the escape speed from the cluster, Vesc ≈ (2GM/R)0.5,
is less than σ, the instantaneous velocity dispersion of the embedded stars at the time of
gas dispersal. Thus a bound group will emerge only if the SFE is greater than 50% (Wilk-
ing & Lada 1983). Consequently, the fact that the SFEs of embedded clusters are always
observed to be less than 50% is critically important for understanding their dynamical evolu-
tion. Apparently, it is very difficult for embedded clusters to evolve to bound open clusters,
particularly if they form with O stars.
However, classical open clusters, like the Pleaides, do exist in sufficient numbers that
at least some embedded clusters with SFEs less than 50% must have evolved to become
relatively long-lived, bound systems. For slow gas removal times, τgr > τcross, clusters
even with low SFEs can adiabatically adjust and expand to new states of virial equilibrium
and remain bound. The fact that clusters older than about 5 Myr are observed to rarely
be associated with molecular gas suggests that τgr < 5 Myr. This is close enough to the
crossing timescales that numerical calculations are necessary to investigate the response of
clusters to this slow gas removal. Moreover, to produce a bound group which is stable
against galactic tides and the tidal forces of its parental GMC requires additional stringent
constraints on the initial conditions prevailing in the cluster forming cloud core and on τgr
(Lada et al. 1984). To evolve to a bound cluster like the Pleiades, the typical embedded
cluster would have to have a gas removal time of at least a few (3-4) crossing times, which
corresponds to a few million years for typical conditions. This timescale is of the same order
as the cluster formation or gestation time and would require a cluster to be losing mass while
simultaneously forming stars. Outflows generated by low mass stars can remove gas during
the star formation process and Matzner & McKee (2000) have shown that such outflows
can completely disrupt cluster forming cores for SFEs in the 30-50% range. Whether such
outflow driven mass dispersal can occur over as long a timescale as required is not clear.
This depends on the detailed star formation history of the cluster, in particular the star
formation rate and its variation in time and neither of these quantities is sufficiently well

– 59 –
The production of a bound cluster from a dense cloud core clearly requires very special
physical conditions. This must be a rare occurrence. The observed low SFEs for embedded
clusters can account for the high infant mortality rate of clusters inferred from the relatively
large numbers and high birthrates of embedded clusters compared to classical open clus-
ters. Most (∼ 90-95%) embedded clusters must emerge from molecular clouds as unbound
systems. Only the most massive (MEC ≥ 500 M⊙) embedded clusters survive emergence
from molecular clouds to become open clusters. Thus, although most stars form in embed-
ded clusters, these stellar systems evolve to become the members of unbound associations,
not bound clusters. However, bound classical clusters form at a sufficiently high rate that
on average, each OB association (and GMC complex) probably produces one such system
(Elmegreen & Clemens 1985) accounting for about 10% of all stars formed within the Galaxy
(Roberts 1957; Adams & Myers 2001).
6. CONCLUDING REMARKS
The discovery of large numbers of embedded clusters in molecular clouds over the last
fifteen years has lead to the realization that these young protoclusters are responsible for
a significant fraction of all star formation currently occurring in the Galaxy. Embedded
clusters may very well be the fundamental units of star formation in GMCs. Conceived in
the mysterious physical process that transforms diffuse interstellar matter into massive and
dense molecular cloud cores, embedded clusters are born at a rate that significantly exceeds
that estimated for classical open clusters. Evidently, the vast majority of embedded clusters
do not survive their emergence from molecular clouds as bound stellar systems. Their high
infant mortality rate is mostly the result of the low to modest star formation efficiency
and rapid gas dispersal which characterizes their birth. There are more than 20 embedded
clusters formed for every cluster born that ultimately evolves into a long-lived system like
the Pleiades. As the primary sites of star birth in molecular clouds, embedded clusters are
important laboratories for studying the origin and early evolution of stars and planetary
systems. The fundamental properties of the Galactic stellar population, such as its IMF, its
stellar multiplicity, and the frequency of planetary systems within it, are forged in embedded
clusters.
Although, observations over the last fifteen years have clearly established the the central
role of embedded clusters in the star formation process, the fundamental parameters of these
extremely young clusters are still very poorly constrained. In particular, the overall census
of embedded clusters is far from complete, even within 1 Kpc of the Sun. In addition, very
little information concerning the ages of embedded clusters exists. Accurate information

– 60 –
concerning the spatial sizes, number of members, masses, and distances for most embedded
clusters is also lacking. Nonetheless, despite these deficiencies, studies of individual embed-
ded clusters have provided new insights into fundamental astrophysical problems, such as
determining the functional form and universality of the IMF, the frequency and lifetimes of
protoplanetary disks, and the ubiquity and nature of brown dwarfs. However, to date only a
small number of such clusters have been studied in any detail. It remains to be determined
whether the trends determined for this small sample are representative of the majority of
embedded clusters and star formation events in the Galaxy as a whole. For example, does
the IMF of the Trapezium cluster truly represent a universal IMF? Is the fraction of freely
floating brown dwarfs always approximately 20-25% of a cluster population? Is the circum-
stellar disk lifetime the same in all clusters? Other important questions also remain open.
Do the progenitors of bound open clusters ever contain O stars? How frequently do O stars
form in embedded clusters? Is the primordial binary fraction the same for stars formed in
and outside rich clusters? What is the most massive embedded cluster that can be formed
from a GMC? How many such clusters exist in the Galaxy? What is the actual number of
poor (N∗< 35) embedded clusters or stellar aggregates formed and what is the fraction of all
stars produced in such groups? Is the process that produces embedded clusters in any way
related to that responsible for the formation of globular clusters?
Resolving these issues will require an extensive effort in both observation and theory.
Prospects for progress continue to be bright due to the development of important new ob-
servational capabilities. These include, wide field infrared imaging and multi-object spec-
troscopy using large, ground-based telescopes, airborne and space-based infrared imaging
and spectroscopy provided by missions such as SOFIA, SIRTF and NGST, and NIR all
sky surveys such as 2MASS and DENIS. Practically everything we know about embedded
clusters we have learned in the last fifteen years. It is difficult to predict but exciting to
contemplate what will be learned in the next fifteen years as a result of these new capabilities
and the continued dedicated efforts of astronomers who work on these problems. However,
whatever the outcome of such research, there can be little doubt that the result of these
efforts will be to enrich our understanding of the star and planet formation process in the
universe.
ACKNOWLEDGEMENTS
We are grateful to Joa˜o Alves, Richard Elston and August Muench for assistance in
preparation of figures. We thank also August Muench and Richard Elston for many useful
comments and criticisms of earlier versions of this manuscript. EAL acknowledges support
from a Presidential Early Career Award for Scientists and Engineers (NSF AST 9733367)
to the University of Florida. Finally, we gratefully acknowledge the efforts of the many

– 61 –
researchers whose contributions have made this such an interesting and stimulating subject
to review.
REFERENCES
Andre, P & Montmerle, T. 1994, Ap.J., 420:837-862
Andre, P, Ward-Thompson, D & Barsony, M. 1993, Ap.J., 406:122-141
Aspin, C & Barsony, M.1994, A.&A., 288:849-859
Aspin, C, Sandell, G & Russell APG. 1994, A.&A.S., 106:165-198
Adams, JD, Stauffer, JR, Monet, DG, Skrutskie, MF & Beichman, CA. 2001, Ap.J.,
121:2053-2064
Adams, FC, Lada, CJ & Shu, FH. 1987, Ap.J., 312:788-806
Adams, FC & Myers, P. 2001, Ap.J., 533:744-753
Allen, C & Poveda, A. 1974, in The Stability of the Solar System and of Small Stellar
Systems, ed. Y. Kozai, pp. 239-246. Dodrecht: Kluwer
Alves, JF, Lada, CJ, Lada, EA, Muench, AA & Moitinho, A. 2003, A.&A., in press.
Ambartsumian, VA. 1954, Contrib. Byurakan Obs., 15:3
Bally, J, Devine, D, & Reipurth, B. 1996, Ap.J.L., 473:L49-L53
Bally, J, Testi, L, Sargent, A, & Carlstrom, J. 1998, A.J., 116:854-859
Baraffe, I, Chabrier, G. Allard, F & Hauschildt, PH. 2002, A.&A., 382:563-572
Barrado y Navascues, D, Bouvier, J, Stauffer, JR, Lodieu, N & McCaughrean, MM. 2002,
A.&A., in press.
Barsony, M, Schombert, JM & Kis-Halas, K. 1991, A.J., 379:221-231
Battinelli, P. & Capuzzo-Dolcetta 1991, M.N.R.A.S., 249:76-83
Bernasconi, PA., 1996, A&AS, 120: 57-
Binney, J & Tremaine, S. 1987, Galactic Dyanmics, Princeton, Princton University Press.

– 63 –
Clarke, CJ, Bonnell, IA & Hillenbrand, LA. 2000, in Protostars and Planets IV, ed. V
Mannings, AP Boss, & SS Russell, pp. 151-177, Tucson, University of Arizona Press
Comeron, F, Rieke, GH, Burrows, A, & Rieke, MJ. 1993, Ap.J., 416: 185-203
Comeron, F, Rieke, GH & Rieke, MJ. 1996, Ap.J., 473: 294-303
D’Antona, F & Mazzitelli, I. 1997, Mem.Soc.Astron.Italiana, 68: 807-
Davis, CJ, Matthews, HE, Ray, TP, Dent, WRF & Richer, JS 1999, M.N.R.A.S., 309-141-152
Dobbie, PD, Pinfield, DJ, Jameson, RF & Hodgkin, ST. 2002, M.N.R.A.S., in press.
Duchene, G. 1999, A.&A., 341:547-552
Duchene, G, Bouvier, J & Simon, T. 1999, A.&A., 343:831-840
Dutra, CM & Bica, E. 2001, A.&A., 376:434-440
Eiroa, C, & Casali, MM. 1992, A.&A., 262:468-482
Eisenhauer, F, Quirrenbach, H, Zinnecker, H & Genzel, R. 1998, Ap.J., 498:278-292
Elmegreen, BG. 1991, in The Physics of Star Formation and Early Stellar Evolution, eds.
CJ Lada & ND Kylafis, pp, 35-60, Dodrecht: Kluwer
Elmegreen, BG. 1983, M.N.R.A.S., 203:1011-1020
Elmegreen, BG & Clemens, C. 1985, Ap.J., 294:523-532
Elmegreen, BG & Efremov YN. 1997, Ap.J., 480:235-245
Elmegreen, BG, Efremov, Y, Pudritz, RE & Zinnecker, H. 2000, in Protostars and Planets
IV, ed. V Mannings, AP Boss, & SS Russell, pp. 179-215, Tucson: University of
Arizona Press
Elson, R, Fall, SM, Freeman, KC. 1987, Ap.J., 323:54-78
Fletcher, A & Stahler, S. 1994a,Ap.J., 435:313-328
Fletcher, A & Stahler, S. 1994b,Ap.J., 435:329-338
Gehz, AM, Neugegauer, G & Matthews, K. 1993, A.J, 106:2005-
Geyer, MP, & Burkert, A. 2001, M.N.R.A.S., 323:988-994

– 64 –
Giovannetti, P, Caux, E, Nadeau, D & Monin, J-L 1998, A&A, 330:990-998
Gilmore, G & Howell, D. 1998 The Initial Mass Function, San Francisco: The Astronomical
Society of the Pacific
Goodwin, SP. 1997, M.N.R.A.S., 284:785-802
Grasdalen, G, Strom SE & Strom, KM. 1973, Ap.J.L., 184:L53-L57
Greene, TP, Wilking, BA, Andre, P, Young, ET & Lada, CJ. 1994, Ap.J., 434:614-626
Greene, TP & Lada, CJ. 1996,A.J., 112:2184-2221
Greene, TP & Lada, CJ. 1997,A.J., 114:2157-2165
Greene, TP & Lada, CJ. 2002,A.J., 124:2185-2193
Greene, TP & Meyer, MR 1995, Ap.J., 450:233-244
Haisch, KE, Lada, EA & Lada CJ. 2000, A.J., 120:1396-1409
Haisch, KE, Lada, EA & Lada CJ. 2001b, A.J., 121:2065-2074
Haisch, KE, Lada, EA, & Lada, CJ. 2001, Ap.J.L., 553:L153-L156
Haisch, KE, Lada, EA, Pina, RK, Telesco, CM & Lada, CJ. 2001c, A.J., 121:1512-1521
Hayashi, C. 1966, A.R.A.A., 4:171:1191
Hartmann, L. 2001, A.J., 121:130-1039
Herbig, GH. 1998, Ap.J., 497:736-758.
Herbig, GH, & Dahm, SE. 2002,A.J., 123:304-327
Hillenbrand, LA. 1997, A.J., 113:1733-1768
Hillenbrand, LA & Carpenter, JM. 2000, Ap.J., 540:236-254
Hillenbrand, LA. & Hartmann L. 1998, Ap.J., 492:540-553
Hills, JG. 1980, Ap.J., 240:242-245
Hodapp, K. 1994, Ap.J.S., 94:615-649
Hodapp, K & Rayner, 1991, A.J., 102:1108-1117

– 65 –
Horner, DJ, Lada, EA & Lada CJ. 1997,A.J., 113:1788-1798
Howard, EM, Pipher, JL & Forrest, WJ. 1997, Ap.J., 481:327-342
Hurt, RL & Barsony M. 1996, Ap.J.L., 460:L45-L48
Jiang, Z, Yao, Y, Yang, J, et al. 2002, Ap.J., 577:245-259
Kenyon, SJ & Gomez M. 2001, A.J., 121:2673-2680
Kenyon, SJ & Hartmann, L. 1995, Ap.J.Supp., 101:117-171
Kenyon, SJ, Lada, EA & Barsony, M. 1998, A.J.,115:252-262
Klessen, RS. 2001, Ap.J., 556:837-846
Klessen, RS & Burkert, A. 2000, Ap.J.S., 128:287-319
Klessen, RS & Burkert, A. 2000, Ap.J., 549:386-401
Klessen, RS, Heitsch, F & Mac Low, M-M. 2000, Ap.J., 535:887-906
Knee, LBG & Sandell, G. 2000, A.&A., 361:671-684
Kramer, C, Stutzki, J, Rohrig, R & Corneliussen, U. 1998, A.&A., 329:249-264
Kroupa, P. 1995, M.N.R.A.S, 277:1491:1506
Kroupa, P. 1995b, M.N.R.A.S, 277:1507:1521
Kroupa, P. 2002, Science, 295:82-91
Kroupa, P, Aarseth, S & Hurley, J. 2001, M.N.R.A.S., 321:699-712
Kroupa, P, & Boily, CM. 2002, M.N.R.A.S., in press.
Kroupa, P, Petr, M & McCaughrean, MJ. 1999, New Ast., 4, 495-501
Kroupa, P, Tout, CA & Gilmore, G. 1990, M.N.R.A.S., 244:76-85
Kroupa, P, Tout, CA, & Gilmore, G. 1991, M.N.R.A.S., 251:293-302
Kroupa, P, Tout, CA & Gilmore, G. 1993, M.N.R.A.S., 262:545-587
Lada, CJ. 1987, in Star Forming Regions, IAU Symposium No. 115, eds. M. Peimbert & J.
Jugaku, pp. 1-17, Dodrecht: Reidel

– 67 –
Luhman, KL. 2000, Ap.J., 544:1044-1055
Luhman, KL, Rieke, G, Lada, CJ, & Lada, EA. 1998, Ap.J., 507:347-369
Luhman, KL, Rieke, GH, Young, ET, Cotera, AS, Chen, H, Rieke MJ, Schneider, G &
Thompson, RI. 2000, Ap.J., 540: 1016-1040.
Lynden-Bell, D & Pringle, J. 1974, M.N.R.A.S., 168:603-637
Lynga, G. 1987, Catalogue of Open Cluster Data, 5th edition, Stellar Data Center, Observa-
toire de Strasbourg
Marschall, LA,van Altena, WF & Chiu, L-TG. 1982, 87:1497-1506
Matzner, CD & McKee, CF. 2000,Ap.J., 545:364-378
McCaughrean, MJ, Rayner, JT & Zinnecker, H. 1994, Ap.J.L., 436:L189-L192
Megeath, ST. 1996, A&A, 311: 134-144
Megeath, ST, Herter, T, Beichman, C, et al. 1996, A&A, 307:775-790
Megeath, ST & Wilson, TL. 1997, A.J., 114:1106-1126
Meyer, MR. 1996, unpublished PhD Thesis, University of Massachusetts
Meyer, MR, Adams, FC, Hillenbrand, LA & Carpenter, J. 2000, in Protostars and Planets
IV, ed. V Mannings, AP Boss, & SS Russell, pp. 121-149, Tucson: University of
Arizona Press.
Moitinho, A, Alves, J, Huelamo, N & Lada, CJ. 2001, Ap.J.L., 563:L73-76
Motte, F, Andre, P & Neri, R. 1998, A.&A., 336:150-172
Motte, F, Andre, P, Ward-Thompson, D & Bontemps, S. 2001, A.&A., 372:41-44
Mouschovias, TCh. 1991, Ap.J., 373:169-186
Muench, AA, Lada, EA & Lada, CJ. 2000, Ap.J., 553: 338-371
Muench AA, Alves, JF, Lada, CJ, & Lada, EA. 2001, Ap.J.L., 558:L51-L54
Muench, AA, Lada, EA, Lada, CJ & Alves, JF. 2002, Ap.J., 573: 366-393
Muench, AA, Lada, EA. Lada, CJ, Elston RJ, et al. 2003, A.J., in press.

– 70 –
van den Berg, S & Lafontaine, A. 1984, A.J., 89:1822-1824
Veschueren, W & David, M. 1989, A.&A., 219:105-120
Walker, MF. 1956, Ap.J.Supp., 2:365-387
Warin, S, Castets, A, Langer WD, Wilson, RW, & Pagani, L. 1996, A.&A., 306:935-946
Wilking, BA, Greene, TP & Meyer, MR. 1999,A.J., 117:469-482
Wilking, BA & Lada, CJ 1983, Ap.J., 274:698-716
Wilking, BA, Lada, CJ & Young, ET. 1989, Ap.J., 340: 823-852
Wilking, BA, McCaughrean, MJ, Burton, MG, Giblin, T, Rayner, JT & Zinnecker, H.
1997,A.J., 114:2029-2042
Whitworth, A. 1979, M.N.R.A.S., 186:59-67
Wolf, G, Lada, CJ & Bally, J. 1990, A.J., 100:1892-1902
Wood, K, Lada, CJ, Bjorkman, JE, Kenyon, SJ, Whitney, B. & Wolff, MJ. 2001, Ap.J.,
567:1183-1191
Zinnecker, H., McCaughrean, M. & Wilking, BA. 1993, in Protostars and Planets III, eds.
EH. Levy & JI. Lunine, pp. 429-495, Tucson: University of Arozona Press
This preprint was prepared with the AAS LATEX macros v5.0.

– 71 –
Table 1: Catalog of Embedded Clusters
EC Name RA Dec Distance Size N∗ K Mass Ref
(J2000) (J2000) (pc) (pc) (limit) M⊙
1 NGC281W 00:52:23.7 +56:33:45 2100 231 18.0 130 1,2
2 NGC 281E 00:54:14.7 +56:33:22 2100 88 17.0 57 1
3 01546+6319 01 58:19.8 +63:33:59 2400 0.54 54 17.5 35 3
4 02044+6031 02:08:04.7 +60:46:02 2400 0.73 147 17.5 94 3
5 02048+5957 02:08:27.0 +60:11:46 2400 0.56 58 17.5 37 3
6 02054+6011 02:09:01.3 +60:25:16 2400 0.59 70 17.5 45 3
7 02175+5845 02:21:07.7 +58:59:06 2400 0.73 109 17.5 70
8 IC 1805W 02:25:14.5 +61:27:00 2300 79 17.0 57 1
9 02232+6138 02:27:04.1 +61:52:22 2400 0.91 205 17.5 130 3
10 02245+6115 02:28:21.5 +61:28:29 2400 0.64 121 17.5 77 3
11 02407+6047 02:44:37.8 +60:59:53 2400 0.46 50 17.5 32 3
12 02461+6147 02:50:09.2 +61:59:58 2400 0.72 115 17.5 73 3
13 02484+6022 02:52:18.7 +60:34:59 2400 0.62 86 17.5 55 3
14 02497+6217 02:53:43.2 +62:29:23 2400 0.38 36 17.5 23 3
15 AFGL 4029 03:01:32.3 +60:29:12 2200 173 16.5 140 1
16 02541+6208 02:58:13.2 +62:20:29 2400 0.45 40 17.5 26 3
17 W3IRS5 02:25:40.6 +62:05:52 2400 87 17.0 64 4
18 02570+6028 03:01:00.7 +60:40:20 2400 0.62 78 17.5 50 3
19 02575+6017 03:01:29.2 +60:29:12 2400 1 240 17.5 150 3
20 02593+6016 03:03:17.9 +60:27:52 2400 0.62 88 17.5 56 3
21 AFGL 437 03:07:25.6 +58:30:52 2000 122 17.0 79 1
22 AFGL 490 03:27:38.7 +58:46:58 900 45 16.5 25 1
23 NGC 1333 03:32:08.1 +31:31:03 318 0.49 143 14.5 79 5
24 IC 348 03:44:21.5 +32:10:16 320 300 15.0 160 6,7
25 LKHalpha 101 04:30:14.4 +35:16:25 800 150 15.0 98 8,9
26 S242 05:52:12.9 +26:59:33 2100 96 16.5 81 1

– 73 –
Table 1: Catalog of Embedded Clusters (Continued)
EC Name RA Dec Distance Size N∗ K Mass Ref
(J2000) (J2000) (pc) (pc) (limit) M⊙
53 S 287 C 06:59:36.6 – 04:40:22 1400 50 16.5 31 1
54 L 1654 06:59:41.7 – 07:46:29 1100 415 17.0 230 1
55 BIP 14 08:15:14.8 – 04:04:41 1400 98 16.5 61 1
56 RCW38 08:59:05.4 – 47:30:42 1700 1300 18.0 730 1
57 NGC 3576 11:11:57.0 – 61:18:54 2400 51 13.0 720 22
58 Rho Oph 16:27:01.6 – 24:36:41 125 100 14.0 53 23
58 NGC 6334I 17:20:53.0 – 35:46:57 1700 0.6 93 16.0 78 24
60 Trifid/ M20 18:02 23.0 – 23:01:48 1600 85 14.3 190 25
61 M16/NGC6611 18:18 48.0 – 13:47:00 1800 300 14.0 960 26
62 M17 18:20 26.0 – 16:10:36 1800 100 12.8 890 27
63 NGC 6530/M8 18:07:51.9 – 24:19:32 1800 100+ 28
64 MWC 297 18:27:39.5 – 03:49:52 450 0.5 37 16.7 20 15
65 Serpens SVS2 18:29:56.8 +01:14:46 250 51 15.5 27 1,29
66 S87E 19:46:19.9 +24:35:24 2110 101 15.2 180 30
67 R CrA 19:01:53.9 – 36:57:09 700 40 16.5 22 1,31
68 S88B 19:46:47.0 +25:12:43 2000 98 15.5 120 1
69 S 106 20:27:25.0 +37:21:40 600 0.3 160 14.0 120 32
70 W 75 N 20:38:37.4 +42:37:56 2000 130 16.5 99 1
71 L1228 20:57:13.0 +77:35:46 150 47 18.0 25 1
72 IC 5146 21:02:36.3 +47:27:59 1200 100 33
73 L988 e 21:03:57.5 +50:14:38 700 46 14.5 32 1
74 LKHalpha234 21:43:02.2 +66:06:29 1000 139 17.0 76 1
75 Cep A 22:56:19.0 +62:01:57 700 580 17.0 310 1
76 Cep C 23:05:48.8 +62:30:02 750 110 16.5 60 1
References: 1-Hodapp 1994; 2-Carpenter et al. 1993; 3-Carpenter, Heyer & Snell 2000; 4-
Megeath et al. 1996; 5-Lada, Alves & Lada 1996; 6-Muench et al. 2003; 7-Lada & Lada
1995; 8-Aspin & Barsony 1994; 9-Barsony et al. 1991; 10-Strom, Strom & Merrill 1993;
11-Carpenter 2000; 12-Hillenbrand & Carpenter 2000; 13-Muench et al. 2002; 14-Lada E. et
al. 1991; 15-Testi, Palla & Natta 1998; 16-Horner, Lada & Lada 1997; 17-Carpenter et al.
1997; 18-Carpenter, Snell & Schloerb 1995; 19-Marshal, van Altena & Chiu 1992; 20-Lada,
Young & Greene 1993; 21-Alves 2002, private comm.; 22-Persi et al. 1994; 23-Kenyon, Lada
& Barsony 1998; 24-Tapia et al. 1996; 25-Rho et al. 2001; 26-Chini, Kruegel & Wargau 1992;
27-Lada C. et al. 1991; 28- van den Ancker et al. 1997; 29-Eiroa & Casali 1992; 30-Chen et
al. 2002; 31-Wilking et al 1997; 32 Hodapp & Rayner 1991; 33-Herbig & Dahm 2002.

This figure "figure1.jpg" is available in "jpg"
format from:
http://arXiv.org/ps/astro-ph/0301540v1

This figure "figure5.jpg" is available in "jpg"
format from:
http://arXiv.org/ps/astro-ph/0301540v1

This figure "figure8.jpg" is available in "jpg"
format from:
http://arXiv.org/ps/astro-ph/0301540v1