The Birth of Star Clusters 1st Edition Steven

Stahler (Eds.) pdf download

https://textbookfull.com/product/the-birth-of-star-clusters-1st-edition-steven-stahler-eds/

★★★★★ 4.7/5.0 (28 reviews) ✓ 196 downloads ■ TOP RATED

"Fantastic PDF quality, very satisfied with download!" - Emma W.

DOWNLOAD EBOOK
 The Birth of Star Clusters 1st Edition Steven Stahler (Eds.)
pdf download

TEXTBOOK EBOOK TEXTBOOK FULL

Available Formats

■ PDF eBook Study Guide TextBook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

Collection Highlights

The Globular Star Clusters of the Andromeda Galaxy Iop

Concise Physics Charli M Sakari

The Birth of Nomos 1st Edition Thanos Zartaloudis

Towards the Humanisation of Birth: A study of epidural

analgesia and hospital birth culture 1st Edition Elizabeth
Newnham

A Dark and Hollow Star The Hollow Star Saga 1 1st Edition
Ashley Shuttleworth
The Birth Mother 1st Edition Aj Carter

Birth Of The Phoenix Rise Of The Phoenix 1 1st Edition

Jessica Wayne

Star Wars: Light of the Jedi (The High Republic) 1st

Edition Charles Soule

The Birth of Science Alex Ely Kossovsky

The Nature of Dusty Star Forming Galaxies 1st Edition

William Cowley (Auth.)
Astrophysics and Space Science Library 424

Steven Stahler
Editor

The Birth of
Star Clusters
The Birth of Star Clusters
 Astrophysics and Space Science Library
EDITORIAL BOARD
Chairman
W. B. BURTON, National Radio Astronomy Observatory, Charlottesville,
Virginia, U.S.A. ([email protected]); University of Leiden, The Netherlands
([email protected])
F. BERTOLA, University of Padua, Italy
C. J. CESARSKY, Commission for Atomic Energy, Saclay, France
P. EHRENFREUND, Leiden University, The Netherlands
O. ENGVOLD, University of Oslo, Norway
E. P. J. VAN DEN HEUVEL, University of Amsterdam, The Netherlands
V. M. KASPI, McGill University, Montreal, Canada
J. M. E. KUIJPERS, University of Nijmegen, The Netherlands
H. VAN DER LAAN, University of Utrecht, The Netherlands
P. G. MURDIN, Institute of Astronomy, Cambridge, UK
B. V. SOMOV, Astronomical Institute, Moscow State University, Russia
R. A. SUNYAEV, Max Planck Institute for Astrophysics, Garching, Germany

More information about this series at http://www.springer.com/series/5664

Steven Stahler
Editor

The Birth of Star Clusters

123
Editor
Steven Stahler
Department of Astronomy
University of California
Berkeley
California, USA

ISSN 0067-0057 ISSN 2214-7985 (electronic)

Astrophysics and Space Science Library
ISBN 978-3-319-22800-6 ISBN 978-3-319-22801-3 (eBook)
DOI 10.1007/978-3-319-22801-3

Library of Congress Control Number: 2017956041

© Springer International Publishing AG 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

Cover illustration: Composite picture of the young open star cluster NGC 602. Taken Under the “Wing”
of the Small Magellanic Cloud. Credit: X-ray: NASA/CXC/Univ.Potsdam/L.Oskinova et al; Optical:
NASA/STScI; Infrared: NASA/JPL-Caltech

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

It has long been accepted that most stars originate in groups, rather than in isolation.
How the groups themselves arise is a major, unsolved puzzle. At first sight, it might
seem odd that such a problem should exist at all. Granted, we certainly do not
understand every detail of how individual stars form. Our ignorance is especially
severe in the case of massive stars that quickly disperse any parent gas. For the
more common, lower-mass stars like the Sun, however, a general picture has been
in place for some time, and continues to be filled in. Why, then, is there a separate
problem of cluster formation? If we jump ahead a few decades and imagine that
low-mass stellar birth is even more secure, will such a problem remain?
The answer is that it would, for at least two basic reasons. First, there is the
obvious fact that stellar groups form out of much larger molecular clouds than the
dense cores spawning single, solar-type objects. Understanding the structure and
dynamics of dense cores is facilitated by the fact that these entities are relatively
quiescent, supported largely by ordinary thermal pressure. In contrast, all clouds
massive enough to form clusters are characterized by internal turbulence, for which
quantitative modeling remains primitive at best.
A second reason the conundrum will persist is that stellar groups span an
enormous range of properties. There are moving groups consisting of a few dozen
members. At the opposite end of the spectrum are the monstrous and ancient
globular clusters, some of which still contain a million stars. Just as no one
theoretical model of single-star formation is likely to explain both M dwarfs and
O stars, it is equally implausible that one mechanism underlies such diverse stellar
aggregates. We must, at present, consider different aspects of the problem and do
our best to understand them. In this volume, we offer a selection of current research,
from both observers and theorists, on key topics in this active field of study.
We begin with an account of the very youngest clusters, those still embedded
in relatively large and turbulent molecular clouds. Joana Ascenso first tells us
how these groups are identified in practice. Given the variety of optically revealed
clusters, it is not surprising that their optically revealed counterparts exhibit a range
of morphologies, which Ascenso also describes. As yet, we have no means to assign
even rough ages to these groups. Nor can we yet identify the specific forerunners of,

v
vi Preface

say, OB associations. We may hope that further observations will begin to fill these
gaps.
Turning to the theoretical problem of cluster formation, Patrick Hennebelle
relates the current status of numerical simulations. The basic program here is to
employ a computational box of gas in which gravity overwhelms any internal
pressure. In current simulations, the gas is both magnetized and stirred in a way
to mimic true, astrophysical turbulence. Under the action of self-gravity, the gas
collapses into a myriad of high-density regions. These presumably would go on
to become stars, although no one simulation can cover that final transition. As
Hennebelle describes, the necessary and realistic introduction of an embedded
magnetic field has led to the paradoxical result that disks around stars cannot form.
Feedback from the stars themselves, especially in the form of ionizing radiation,
presents similar technical and astrophysical problems.
In many astrophysical problems, we gain insight by exploring how a process
plays out under a broad range of conditions. Within our Galaxy, no environment
is more extreme than the Galactic center. As Jessica Lu relates in her chapter,
this region contains the densest molecular gas. Thus, it might not seem surprising
that the star formation rate per unit volume is also the highest. On the other hand,
the extraordinarily rich Young Nebular Cluster, one of the three that Lu describes,
actually surrounds the central, massive black hole, a fact that renders its formation
especially puzzling. The other two clusters, dubbed the Arches and Quintuplet, have
similar ages as the central system, a few million years, but distinct morphologies. Lu
details the rapid progress being made in discerning the substructure and stellar mass
distribution in these groups. As she emphasizes, the very large amount of dusty gas
surrounding them poses a special challenge to observers.
Angela Adamo and Nate Bastien widen the perspective to consider the statistics
of cluster birth and death on galactic scales. They first summarize the distribution
of cluster masses. While a simple power law, with a universal exponent, is adequate
in many galaxies, there is always a departure from this relation at the upper end.
Interestingly, this cutoff varies with the galactic environment. So, too, does the
fraction of clusters that are gravitationally bound, which is higher in galaxies with
greater star formation rates. Moreover, the age distribution of clusters depends on
the specific mass in question. All these trends will eventually help us see how
the largest clouds form clusters. They will also lend insight into the purely stellar
dynamical problem of cluster dispersion.
Returning to our own Galaxy, Eric Feigelson provides an update on a concerted
effort, using a variety of observational tools, to study the structure and history
of the youngest OB associations, those still partially embedded in molecular gas.
Feigelson and colleagues combined X-ray and infrared observations with published
catalogs of O and B stars. In addition, they estimated ages for thousands of objects
using a bolometric luminosity and a stellar mass derived empirically from X-ray
luminosities. Their single most significant result is the wide diversity of surface
densities and detailed morphologies in several dozen associations. There is good
evidence for cluster expansion over time. Moreover, stars near the center of any
Preface vii

cluster are systematically younger than those on the outskirts, indicating that star
formation is an ongoing, inside-out process.
The OB associations studied by Feigelson will eventually undergo dispersion, a
process we have long observed, through proper motion studies, in their optically
revealed descendents. There exist, however, equally young and massive groups,
presently devoid of gas, that appear to be gravitationally bound. Sambaran Banerjee
and Pavel Kroupa provide a theoretical perspective on the origin of such systems,
which include the Galactic Center clusters studied by Lu, as well as others found
in the Milky Way disk. Thus far, direct numerical simulations cannot track both
the stellar and gas dynamics of such populous groups. As an alternative, simulators
follow the stars using a standard, N-body code, but crudely account for the gas
through a background gravitational potential. To mimic sudden expulsion of the
gas, researchers force the potential to vanish. Banarjee and Kroupa show that such
calculations do reproduce the basic properties of several well-studied systems.
Further, they argue that the contrasting picture of merging substructures is too slow
and produces clusters that are far too dense.
These contributions take us some distance toward a better understanding of
cluster origins, but they also highlight the outstanding questions yet to be answered.
I would list, near the very top, the issue of why some clusters emerge from their
clouds as gravitationally bound systems, while others are unbound and disperse.
The smallest aggregates, containing up to several hundred members, are unbound,
or become so after their parent cloud dissipates. Groups of a thousand stars or
more develop into classic OB associations, also unbound. But between the lowest-
mass groups and OB associations lies the regime of open clusters, all bound and
remarkably long-lived. Furthermore, as Banerjee and Kroupa describe, groups more
massive than OB associations can also be bound. By now, the pattern is clear and
also completely unexplained.
The traditional idea is that a bound cluster arises when a relatively large fraction
of the parent cloud’s mass turns into stars. By now, it is widely recognized that this
explanation, while it may have a grain of truth, is inadequate. First, the theoretically
required mass fraction is far above most observational estimates. A second objection
is that such a simple model is unlikely to suffice in the two very disparate regimes
where we find bound clusters. Indeed, observers have searched in vain for clouds
that could be precursors to the most populous bound clusters. At this end of the mass
spectrum, we may need to explore the possible merger of pre-existing aggregates.
Explaining the birth of open clusters will rest on another issue raised at the start of
this Introduction—the structure and dynamics of large molecular clouds. Clearly,
observers and theorists will be busy for a long time to come.

Berkeley, CA, USA Steven Stahler

May 2017
Contents

1 Embedded Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
Joana Ascenso
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
1.2 What Is an Embedded Cluster? . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3
1.2.1 Defining “Embedded” . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4
1.2.2 Defining “Cluster” . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5
1.3 Morphology and Structure .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8
1.3.1 Observational Challenges . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 9
1.3.2 Cluster Morphologies.. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 9
1.3.3 The Molecular Cloud Scale . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 11
1.4 Age Spreads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 17
1.4.1 Age Spreads in Cluster Complexes . . . . . . . .. . . . . . . . . . . . . . . . . . . . 19
1.4.2 Age Spreads in Individual Clusters . . . . . . . .. . . . . . . . . . . . . . . . . . . . 22
1.4.3 Age Spreads of the Unclustered Stars . . . . .. . . . . . . . . . . . . . . . . . . . 24
1.5 Stellar Mass Distributions . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 26
1.6 Embedded Clusters and Star Formation . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 26
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 28
2 Numerical Simulations of Cluster Formation . . . . . . . .. . . . . . . . . . . . . . . . . . . . 39
Patrick Hennebelle
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 39
2.2 Isothermal Hydrodynamical Simulations . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 40
2.2.1 Some General Considerations .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 40
2.2.2 Numerical Techniques .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 42
2.2.3 General Setups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 43
2.2.4 Result of Hydrodynamical Simulations . . .. . . . . . . . . . . . . . . . . . . . 44
2.3 Radiative Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 46
2.3.1 Some Simple Arguments . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 47
2.3.2 Result of 2D Multi-Wavelength Simulations .. . . . . . . . . . . . . . . . . 49
2.3.3 Result of 3D Simulations with Radiative Feedback .. . . . . . . . . . 50

ix
x Contents

2.4 Impact of the Magnetic Field . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 53

2.4.1 The Lorentz Force and Its Consequences... . . . . . . . . . . . . . . . . . . . 53
2.4.2 Results of Large-Scale MHD Simulations.. . . . . . . . . . . . . . . . . . . . 55
2.5 Impact of HII Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 58
2.5.1 Analytical Estimates . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 58
2.5.2 Simulations with HII Radiation .. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 60
2.6 Impact of Protostellar Jets . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 61
2.6.1 Analytical Estimate .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 61
2.6.2 Simulations with Jets . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 62
2.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 63
2.7.1 The Star Formation Efficiency . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 64
2.7.2 The Star Formation Rate . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 64
2.7.3 The Initial Mass Function . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 64
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 65
3 Massive Young Clusters Near the Galactic Center . . .. . . . . . . . . . . . . . . . . . . . 69
Jessica R. Lu
3.1 Introduction: Three Massive Young Clusters . . . . . .. . . . . . . . . . . . . . . . . . . . 69
3.2 Discovery and Early History . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 70
3.3 Stellar Content and Cluster Ages .. . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 72
3.4 Structure and Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 75
3.5 Present-Day and Initial Mass Functions . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 77
3.6 Cluster Orbits and Birth Locations .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 81
3.7 Observational Methodology . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 83
3.8 Star Formation: Is it Peculiar? .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 83
3.9 Open Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 86
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 87
4 The Lifecycle of Clusters in Galaxies . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 91
Angela Adamo and Nate Bastian
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 92
4.2 Cluster Populations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 93
4.2.1 Cluster Formation .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 93
4.3 The Cluster Age Distribution and Cluster Disruption .. . . . . . . . . . . . . . . . 103
4.3.1 Expectations from Theory and Parameterisations .. . . . . . . . . . . . 104
4.3.2 Analysing Cluster Populations .. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 105
4.3.3 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 106
4.3.4 Observational Results on the Cluster Age Distribution .. . . . . . 108
4.4 Conclusions and Future Outlook . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 114
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 116
5 Multiwavelength Studies of Young OB Associations .. . . . . . . . . . . . . . . . . . . . 119
Eric D. Feigelson
5.1 Historical Discussions of Star Cluster Formation .. . . . . . . . . . . . . . . . . . . . 119
5.2 The Observational Challenges .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 121
5.3 The MYStIX Project .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 123
Contents xi

5.4 A New Stellar Chronometer . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 127

5.5 Identifying (Sub)Clusters . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 128
5.6 Spatial Distribution of Stars Across Star Forming Regions .. . . . . . . . . . 131
5.7 Observational Constraints on Astrophysical Questions . . . . . . . . . . . . . . . 132
5.7.1 Cluster Expansion and Dispersal. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 133
5.7.2 Cluster Formation by Merging Subclusters.. . . . . . . . . . . . . . . . . . . 134
5.7.3 Duration of Star Formation .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 136
5.8 Final Comments and Future Research . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 138
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 139
6 Formation of Very Young Massive Clusters and Implications
for Globular Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 143
Sambaran Banerjee and Pavel Kroupa
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 144
6.2 Monolithic or Episodic Formation of Very Young
Massive Clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 148
6.2.1 Why is an Episodic or Monolithic Mode of Cluster
Formation Necessary? . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 149
6.2.2 An Analytic Representation for Gas Expulsion . . . . . . . . . . . . . . . 155
6.2.3 Matchings with Individual Very Young Massive Clusters . . . . 160
6.3 Hierarchical Formation of Young Massive Clusters: The Case
of NGC 3603 Young Cluster . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 174
6.3.1 General Evolutionary Properties of Subcluster Systems . . . . . . 177
6.3.2 Comparison with NGC 3603 Young Cluster . . . . . . . . . . . . . . . . . . 179
6.4 Globular Clusters and the Stellar IMF . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 183
6.5 Concluding Remarks: Embedded vs. Exposed Young Clusters. . . . . . . 186
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 190

Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 195
Chapter 1
Embedded Clusters

Joana Ascenso

Abstract The past decade has seen an increase of star formation studies made
at the molecular cloud scale, motivated mostly by the deployment of a wealth of
sensitive infrared telescopes and instruments. Embedded clusters, long recognised
as the basic units of coherent star formation in molecular clouds, are now seen
to inhabit preferentially cluster complexes tens of parsecs across. This chapter
gives an overview of some important properties of the embedded clusters in these
complexes and of the complexes themselves, along with the implications of viewing
star formation as a molecular-cloud scale process rather than an isolated process at
the scale of clusters.

1.1 Introduction

The study of embedded clusters dates back to the first infrared detectors for
astronomical use. Still enshrouded in the dusty environment of their natal molecular
cloud, embedded clusters are invisible to optical telescopes but reveal themselves
as rich and fascinating objects at longer wavelengths. They contain the youngest
stars formed and are therefore invaluable probes of the star formation process. Their
stars share the initial conditions of their parent clump of gas, inheriting some of its
characteristics, later probed by humans in an attempt to understand the sequence of
events dominated by the interplay between gravity, turbulence, and magnetic fields
that ultimately forms them.
Both observations and theoretical simulations of star formation have grown in
number and in detail since the seminal review of Lada and Lada (2003) on embedded
clusters. Observationally, the largest leaps forward were the widespread shift from
the study of individual embedded clusters to the larger context of their molecular
clouds, and the large sky surveys to build an increasingly complete census of the

J. Ascenso ()
CENTRA, Instituto Superior Tecnico, Universidade de Lisboa, Av. Rovisco Pais 1, 1049-001
Lisbon, Portugal
Departamento de Engenharia Física da Faculdade de Engenharia, Universidade do Porto, Rua Dr.
Roberto Frias, s/n, P-4200-465 Porto, Portugal
e-mail: [email protected]

© Springer International Publishing AG 2018 1

S. Stahler (ed.), The Birth of Star Clusters, Astrophysics and Space
Science Library 424, DOI 10.1007/978-3-319-22801-3_1
2 J. Ascenso

star formation in the Galaxy. Also important, the detailed study of extreme star
formation events, even by Milky Way’s standards, has expanded the parameter
space for studies of star formation to the limit of extragalactic studies. These
advances were made possible at such a large scale by the deployment of near-
and mid-infrared telescopes and instruments, both in ground-based and in space
observatories. The Two Micron All Sky Survey (2MASS, Skrutskie et al. 2006),
that covers the entire sky, and later the Spitzer Space Telescope were invaluable
at revealing the detailed intricacies of entire star forming regions as well as to
allow a multitude of large scale surveys. Spitzer legacy programs such as the
Cores to Disks (c2d, Evans et al. 2003), the Galactic Legacy Infrared Mid-Plane
Survey Extraordinaire (GLIMPSE, Churchwell et al. 2009; Benjamin et al. 2003),
and the MIPSGAL (Carey et al. 2009) programs, as well as dedicated surveys
of individual regions, have greatly advanced our understanding of star forming
regions, producing numerous catalogues, most of which yet to be fully explored.
Ground-based observatories have also contributed significantly with near-infrared
telescopes used for surveys (e.g. 2MASS, UKIRT, ESO VISTA), and with near-
infrared adaptive optics assisted instruments for deep and high-resolution studies
of individual regions (e.g. GEMINI, VLT). In the far-infrared, the Herschel Space
Observatory (André and Saraceno 2005) is currently providing invaluable insight
into the youngest stages of star formation, bridging the gap between the study of
pre- and proto-stellar molecular clouds with sub-millimetre and radio telescopes,
and the study of embedded clusters at NIR wavelengths. On the opposite end of
the spectrum, sensitive X-ray observations of star forming regions, made possible
greatly through the Chandra X-ray Observatory, have strongly contributed to the
effort of assessing the stellar populations of star forming regions.
This chapter provides an overview of the observable properties of embedded
clusters in the important context of their molecular clouds, brought to light by this
massive technological development. The analysis is limited to Galactic regions—
those that can be studied in greater detail—and does not include the interesting star
formation taking place at and around the Galactic Centre; the reader is referred to the
review by Longmore et al. (2014) for the latter. Section 1.2 of this chapter elaborates
on the difficulty of adopting one single definition of “cluster” for all studies of star
formation, reviewing the most common definitions in the literature, and what they
entail. Section 1.3 reviews the observed structure and morphology of embedded
clusters and star forming regions, highlighting the trends that have emerged from the
increasing sample of studied clouds, and what they reveal in terms of the underlying
processes at play. Section 1.4 describes the constraints on the timescales for star
formation, crucial in any theory of star formation, derived from the observations of
the ages and age distributions in embedded clusters and cluster complexes.
Other very interesting topics could be addressed in detail in the context of
embedded clusters and are only mentioned briefly in this chapter. The stellar mass
distributions in clusters and on the molecular cloud scale can reveal important
properties of the star formation process; the universality of the initial mass function,
and whether or not embedded clusters are mass segregated have been the subject
of many interesting studies in the past decade; the consequences of the clustered
1 Embedded Clusters 3

environment to individual forming stars at different stages of their evolution, and in

particular their formation along with massive stars is also an active topic of research,
and one that can help understand the probability of a given star developing planets
with certain characteristics. The analysis of the efficiency and of the rate of star
formation, both at the embedded cluster and at the molecular cloud scales, is also
starting to be possible at great detail for a statistically significant sample of known
regions in the Galaxy. The topics included in this chapter are a naturally biased
selection of what the author considers the most robust observational advances in the
last decade and most susceptible of providing solid constraints to existing theories.

1.2 What Is an Embedded Cluster?

An embedded cluster is a group of young stars that is still embedded in its natal
molecular cloud (Fig. 1.1). Although seemingly simple, this definition is all but
trivial. The definitions we adopt reflect and, at the same time, somehow limit our
understanding of star formation. Let’s start with the definition of “embedded” and
then move on to the definition of “cluster”.

Fig. 1.1 RCW 38 is a young

embedded cluster, imaged
here in the near-infrared
bands J, H and KS with
ESO/NTT/SOFI (Ascenso,
Alves et al.)
4 J. Ascenso

1.2.1 Defining “Embedded”

An embedded star (or cluster) is one that is still enshrouded in its natal molecular
cloud. It is typically not (fully) observable at optical wavelengths due to the
heavy obscuration caused by the dust grains in the cloud, but it can be seen in
the near-infrared, where young stars emit significantly (e.g. Adams et al. 1987;
Robitaille et al. 2006), and the dust is more transparent (Savage and Mathis
1979; Cardelli et al. 1989; Rieke and Lebofsky 1985; Draine 2011). Near-infrared
telescopes and instruments are therefore the choice of excellence to detect and
characterise embedded objects, and indeed both ground-based and space telescopes
equipped with infrared detectors and filters have boosted our demographics and our
understanding of embedded clusters exponentially in the past three decades.
It should be noted, perhaps trivially, that not all heavily obscured objects are
embedded: there are objects that are just seen behind molecular clouds, and are
therefore not within them (e.g. Alves et al. 2001). Objects that are in fact embedded
notoriously display signatures of youth. Since stars tend to disperse their natal gas
and dust via accretion and feedback over time an embedded star or cluster is one
that is necessarily young, and this leads to some unspoken confusion regarding the
“embedded” nature of clusters.
The canonical timescale for a cluster to clear enough material to become optically
visible is around 5 Myr (Leisawitz et al. 1989), although more recently Morales et al.
(2013) analysed the association of several young clusters with molecular material,
and proposed an upper limit of the embedded phase of 3 Myr, while Portegies Zwart
et al. (2010) quote a duration of 1–2 Myr for the embedded phase of a cluster. But
a cluster’s embedded phase should be a sensitive function of the mass of the stars
being formed. For example, massive stars develop HII regions that are much more
efficient in dispersing the cloud material than the outflows from low-mass stars
(Matzner 2002), so clusters with massive stars should be the fastest to clear their
surroundings and to emerge from their molecular clouds. Therefore, although the
condition of being embedded is enough to attest to an object’s youth, it is, by itself,
a poor criterion for a sample of clusters of uniform age.
On some accounts, the definition of “embedded” is narrowed to refer to a state
when the potential of the cluster is dominated by the mass of the molecular cloud
(Gutermuth et al. 2009), according to which many known young clusters can no
longer be considered embedded. Trumpler 14, Westerlund 2, and NGC3603, for
example, are all believed to be well under 5 Myr old, but even though they are
still partially obscured by cloud material, they have already cleared most of their
intracluster gas. So these clusters are embedded only in the sense that they are still
associated with the molecular cloud, since their gravitational potential is no longer
dominated by the gas.
For the purpose of this chapter we will focus on clusters that are younger than
5 Myr and still associated with their molecular clouds, regardless of their potential
being dominated by the gas.
1 Embedded Clusters 5

1.2.2 Defining “Cluster”

The definition of “cluster” is more controversial, and it is non-trivial for many

reasons. The need to define “cluster” arises in several different contexts, each
focused on different aspects. In the context of large-scale observational surveys,
for example, a set of uniform criteria is paramount to detect (new) clusters against
the field of the Galaxy in an automated yet robust way. When analysing the birth
conditions and the evolution of clusters over time, the most useful criterion is
probably their dynamical state. Depending on the question one is trying to address,
the physical aspects that are considered relevant—and that should therefore be used
to define clusters as entities—may vary. Additionally, the details that numerical
simulations of star forming molecular clouds are increasingly capable of producing
raise the pressure to find observable signatures of some key property of young
stellar populations that can be tied to a dominant physical process. It is therefore
not surprising to see several definitions of “cluster” in the literature, nor that they
evolve alongside with the progress in our numerical capabilities.
Previous to any definition of cluster, one practical difficulty arises already in
finding the stars that actually make up a population, since knowing whether a
given star is physically associated with its neighbours or if it is only co-located
in projection is challenging, especially for more evolved populations like open
clusters. Stars younger than a few million years offer the advantage that they share
properties that are distinguishable from older stars, providing important clues to
their membership (e.g. Lada 1987; Shu and Adams 1987; Adams et al. 1987;
Gutermuth et al. 2009; Meyer et al. 1997; Feigelson and Montmerle 1999; Feigelson
2010). Observational studies of clusters therefore often start by identifying the
young stellar objects (YSOs), usually by analysing their near-infrared colours
and/or X-ray properties, and then proceed to finding over-densities that qualify as
clusters by some measure. The cloud material associated with embedded clusters
in particular effectively blocks a fraction of background stars, partially filtering out
stars unrelated to the cluster and increasing the local stellar density contrast.
Low density groups are sometimes distinguished from clusters and classified as
O(B) associations if they contain O (and B) stars (e.g. Blaauw 1964), T associations,
if they only contain low-mass stars (Herbig 1962), or R associations, intermediate
between the two and associated with bright, reflection nebulæ (van den Bergh 1966).
These classes overlap in many cases and have largely fallen into disuse over time.
When no criteria other than an overdensity of stars is used, the terminologies “stellar
aggregate”, “stellar grouping” or similar are also found. The concept of “Correlated
Star Formation Event” was introduced by Kroupa et al. (2013) as an alternative
to the concept of “cluster”; it refers to all the stars that were formed in one given
star formation event over a spatial scale of about one parsec, regardless of their
spatial distribution in a star forming region at present. These stars would be coeval
to within the duration of the star forming event. Although the identification of such
events observationally is limited by our ability to determine individual stellar ages,
this is an interesting concept that is perhaps more meaningful in understanding the
progression of star formation in a cloud than the overdensity concept of cluster.
6 J. Ascenso

1.2.2.1 Morphological Criteria

Empirically, a cluster is an overdensity of physically co-located stars. This definition

is often used loosely to refer to all instances of stellar groups. In this sense, detecting
clusters can be as straightforward as finding surface density peaks by eye on large-
scale images, with or without some additional criterion to minimise contamination
from spurious stellar density fluctuations. In the case of young clusters, these criteria
are usually a minimum number of members, or the association with some tracer of
youth, like outflows, ionised gas, or molecular gas and dust, for example (Faustini
et al. 2009; Dutra and Bica 2000; Bica et al. 2003a,b; Borissova et al. 2011, 2014;
Majaess 2013; Froebrich et al. 2007).
Quantitatively, several authors have defined several empirical criteria, most often
calibrated to detect previously known clusters in blind surveys. Ivanov et al. (2002),
for example, require a stellar surface density contrast of at least 3- above the
galactic background, and at least 50 members to claim the detection of a cluster.
Similarly, Kumar et al. (2006) require a stellar surface density contrast greater than
2- above the local background, but a minimum number of only eight members.
Carpenter (2000) requires that the total number of stars within a closed 2- surface
density contour exceeds a 5- enhancement with respect to the expected stellar
background. Porras et al. (2003) differentiate between “clusters” and “groups”
based on whether a given region contains more or less than 30 stars, respectively.
Alternatively, in a variation of the density-threshold algorithm, Gutermuth (2005),
following Casertano and Hut (1985), use the distance to the Nth nearest neighbour
as a proxy for local density, eliminating the need to bin the data spatially to produce
density maps where to look for enhancements.
Gutermuth et al. (2009) devised a more sophisticated method to isolate what
they called “cluster cores” from co-spatial, extended young stellar populations also
associated with the cloud; they analyse the separation between neighbouring stars
using the minimum spanning tree (MST) algorithm, and define the edge of a cluster
core where the MST branch lengths become larger than some critical distance.
Bastian et al. (2007) employ the minimum spanning tree in a slightly different way,
truncating the separation between stars to a maximum allowed distance to define
clusters. Mercer et al. (2005) detect clusters using an algorithm that calculates the
probability of a given overdensity being an actual cluster and not a chance projection
effect considering the statistical distribution of the background field, still based on
geometrical and density enhancement arguments but also on luminosity and colour
criteria.
Schmeja (2011) compares the performance of a few different algorithms in
finding star clusters, and gives additional references to works where the algorithms
were applied. This author finds, as expected, that strongly peaked clusters are easily
detected by all algorithms, whereas low contrast clusters can fall below the radar,
which reflects the ambiguity in the very definitions.
1 Embedded Clusters 7

1.2.2.2 Dynamical Criteria

The previous definitions of clusters as overdensities of stars, although powerful, lack

physical grounds. A common physical criterion to define “cluster” observationally
is the inferred relative stability of the stellar groups. Lada and Lada (2003) classify
a group of young stars as a “cluster” on the basis of its survivability against tidal
disruption up to the age of typical open clusters (100 Myr). According to this
definition, a group of stars is considered a cluster if it contains more than 35
members, and if its density is higher than 1.0 Mˇ pc3 ; an embedded cluster is one
that is also “fully or partially embedded in interstellar gas and dust”.
In theoretical work and in numerical simulations of star formation, “cluster” is
usually synonymous with bound group of stars. This definition is useful because
it simultaneously contains important information about the molecular cloud from
which the cluster formed and about its long-term survivability, and because it leaves
out any spurious overdensity of unrelated sources. It is also a possible definition
in those contexts, since theory has all the information about a given system under
investigation, which is almost never the case in the context of observations. Portegies
Zwart et al. (2010) (see also Gieles and Portegies Zwart 2011) distinguish between
clusters (bound systems) and associations (unbound systems) on the basis of their
age with respect to the system’s dynamical time.1 A system whose age is, at present,
a few times its dynamical time has survived disruption by dynamical effects for
long enough to be considered a “cluster” according to this definition. These systems
are likely to survive as bound entities for a significant fraction of a Hubble time
(Portegies Zwart et al. 2010).
A dynamical analysis enables many interesting studies, including a comparison
between the molecular clouds and their stellar products: systems (or subsystems)
that are bound when they are very young are likely to have formed monolithically
from a bound, gravity dominated cloud, whereas their unbound counterparts are
more likely to have formed from unbound, turbulence supported clouds. But the
dynamical state of a cluster is often difficult to assess, and one subject to many
uncertainties. In Portegies Zwart et al. (2010), for example, the definition of
“cluster” depends strongly on the knowledge of the cluster’s age, of its mass, and
of its virial radius. The determination of a cluster’s age from photometric surveys
depends mostly on the knowledge of the distance to the cluster, which can be
uncertain by a large amount for clusters that are too far away for current measures
of parallax; for example, the distance to the cluster Westerlund 2 ranges from 2.8
to 8 kpc, even in the recent literature (Ascenso et al. 2007a; Carraro et al. 2013;
Zeidler et al. 2015; Rauw et al. 2011). ESA’s mission Gaia (Gaia Collaboration
et al. 2016) will be an invaluable resource for clusters that are already partially
revealed in the optical. The determination of a cluster’s age (and age spread) is

1
The dynamical time is the time a typical star would take to cross the system (tdyn D Rcl =V ). This
is not to be confused with the system’s relaxation time—the timescale on which the system reaches
equipartition of energy via two-body encounters—which is much larger.
8 J. Ascenso

also importantly sensitive to uncertainties in other properties like unresolved stellar

multiplicity, differential extinction between cluster members and stellar variability,
including episodic accretion, and to the accuracy of the stellar evolutionary tracks
themselves (Hartmann 2001; Jeffries 2010; Preibisch 2012). Estimates of cluster
masses, on their turn, can be severely affected by incompleteness, poor membership
assessment or variable detectability over the surveyed area due to, for example,
extended, uneven bright nebula or patchy extinction. Estimates of mass are also
only as reliable as the measurements of distance and age of the cluster, which, as
outlined above, are significantly uncertain. And depending on the wavelength, they
are more or less sensitive to the shape of the local extinction law, and also to the
specific pre-main-sequence evolutionary tracks chosen to convert luminosity into
mass. Finally, a cluster’s virial radius is taken as a factor of the half-light radius
and assumes a given stellar density profile. In rigour, only a spectroscopic analysis
of a significant fraction of cluster members at moderate spectral resolution can
determine their velocity distribution and allow for a proper characterisation of a
cluster’s dynamical state, but this is discouragingly expensive in observation time.
As a consequence, our knowledge of the dynamical state of the many known clusters
is still limited to an educated guess, and in particular it is still too unreliable to be a
strong observational constraint to theories of star formation.
The very significance of the definitions of “cluster” based on dynamical argu-
ments inferred by observations has been called into question by studies that suggest
that there is no fundamental difference between the stellar density distributions
of “clusters” and “non-clusters” by any one definition. Bressert et al. (2010), for
example, do not find any bimodal signature in the stellar density distribution of
several star forming regions that suggests a preferred or a threshold density for
“clusters”, although their sample includes only a few clusters, of relatively low-
mass, and their diagnostics may be considered ambiguous (Pfalzner et al. 2012;
Gieles et al. 2012).
In light of the previous arguments, it is clear that we are currently not in position
to make a statistically accurate comparison of bound and unbound clusters, or of
clusters and associations. At best, we can attempt to rank known clusters in order
of density, mass, luminosity, or age, and try to find meaningful correlations that
can be used to constrain the physical conditions for star formation under different
environments.
In the context of this chapter, a “cluster” will be taken as its most simple literary
meaning: a collection of physically associated stars.

1.3 Morphology and Structure

Embedded clusters come in a variety of forms. This can be inferred instantly by

comparing the images of a few star forming regions. It was the striking mor-
phological difference between different young clusters that led to their traditional
classification as “centrally condensed” or “hierarchical” (Lada and Lada 2003):
the first refers to clusters where the surface density has one strong peak and then
1 Embedded Clusters 9

smoothly declines radially, and the latter to density distributions with multiple peaks
and a high level of substructure.
The importance of defining a cluster’s morphology extends beyond the need for
uniform characterisation criteria. Rather, different morphologies are produced by
different conditions of the progenitor cloud, they reflect different dominant physical
phenomena, and they can be predictive of the cluster’s survival as bound entities on
large timescales or of their demise into field stars.

1.3.1 Observational Challenges

Similar to detecting clusters, analysing their morphology has important observa-

tional challenges. Incompleteness is the obvious enemy of morphological studies:
often only a relatively small fraction of a cluster’s members can be detected. The
distance and the limited sensitivity of instruments act against the detection of faint
stars; the limited resolution of the instruments acts against resolving individual
stars in a cluster, an effect that is additionally amplified in very dense and/or
distant clusters; the presence of bright stars hampers the detection of less luminous
neighbours out to significant projected distances; and the interstellar extinction and
the bright nebula typical of star forming regions, which are almost always variable
in embedded clusters, change the detection limits and the completeness spatially,
producing artificial structure in the observed distribution of cluster members. Also
important is the contamination from field stars, as mentioned before in Sect. 1.2.2;
unless cluster members are efficiently distinguished from field stars, the analysis of
their spatial distribution can be significantly biased, especially in the case of low
surface density clusters.
Infrared observations can minimise some of these effects. Extinction at longer
wavelengths is significantly lower than in the optical (Rieke and Lebofsky 1985),
providing deeper and more uniform completeness levels. Also, the dynamic range
of stellar brightness is lower in the infrared than in the optical, i.e. the luminosity
contrast between the massive and low-mass stars will be smaller, making the latter
easier to detect.

1.3.2 Cluster Morphologies

The human brain can readily distinguish between a centrally condensed distribution
and one that is more substructured, but an objective measure of structure that can
be applied uniformly to a large sample, and one that can be quantitatively compared
with results from simulations and between different regions is required to build a
statistical framework for the properties of star forming regions.
Clusters visually recognised as centrally condensed are generally relatively
isolated clusters, with most members located in a relatively small projected area
10 J. Ascenso

in the sky. It is possible to define a “centre” for the cluster as the location of the
maximum stellar surface density, for example, and the surface density itself then
decays away from that centre as a smooth function in a way somewhat resembling
globular clusters. Analytically, the surface density decay of a centrally condensed
cluster is typically well described by a simple power-law, power-law with a flat core
(Elson et al. 1987) or King profile (King 1962, 1966). The latter is parametrised
by the density at the cluster’s core, by its core radius, and by a tidal radius, and
formally describes the density distribution expected of a single-mass dynamically
relaxed population that is tidally truncated by an external (galactic) potential. While
this is not an accurate description of embedded clusters, the King profile is used
as a convenient function with few parameters, allowing for a uniform description
of the morphology of centrally condensed clusters (e.g. Hillenbrand and Hartmann
1998; Ascenso et al. 2007a,b; Gutermuth et al. 2008; Sung and Bessell 2004; Wang
et al. 2008; Harfst et al. 2010; Kuhn et al. 2010). The Elson et al. (1987) profile is
often preferred in numerical simulations of clusters, although it is also used to fit
observed density profiles of young clusters (Brandner et al. 2008; Gutermuth et al.
2008; Gouliermis et al. 2004; Sana et al. 2010).
Conversely, the stellar surface density of substructured clusters does not follow
a smoothly decaying radial function, instead showing multiple peaks over some
projected area. Several metrics have been proposed to describe their fractal-like
structure, including the two-point correlation function (Gomez et al. 1993), to
describe the probability distribution of any given star having a companion at increas-
ing distances, the distribution of mean surface density of companions of cluster
members (Larson 1995, see also Bate et al. (1998)), the normalised correlation
length, sN (Cartwright and Whitworth 2004), defined as the mean separation between
cluster members normalised to the radius of the cluster, and the normalised mean
edge length, m, N of the minimum spanning tree defined by the cluster members.
Cartwright and Whitworth (2004) review these methods in some detail (see also
Schmeja and Klessen 2006), and propose what they call the Q-parameter as the
most robust parameter to characterise the morphology of a cluster. The Q-parameter
is defined as the ratio between m N and sN, and is able to quantify the degree of
subclustering, as well as to distinguish between a centrally condensed morphology
and a hierarchical morphology: a Q parameter larger or smaller than 0.8 implies a
large-scale radial density gradient or the presence of subclustering, respectively.
This parameter has since become a widespread tool to analyse the structure of
embedded clusters.
It is worth noting that, in rigour, a substructured distribution of stars, although
commonly dubbed “hierarchical”, is not necessarily fractal. The loose classification
of “hierarchical” in the context of clusters usually refers simply to clusters with more
than one peak in stellar density, but Bate et al. (1998) caution that the surface stellar
density distribution in a few known star forming regions previously classified as
fractal was also consistent with the stars being distributed in random sub-clusters, a
non-fractal distribution. This distinction is important when interpreting observations
of cloud structure and stellar density distributions in young clusters in light of the
dominant physical processes, and also when the number of stars is small enough
that statistical fluctuations can lead to the illusion of substructure (Fig. 1.2).
1 Embedded Clusters 11

Fig. 1.2 The large-scale view of NGC 6334 imaged by ESO/VISTA in the near-infrared bands J,
H and KS (galactic North is up, galactic East is to the left, credit ESO/J. Emerson/VISTA). NGC
6334 contains several embedded clusters along its actively star forming ridge

1.3.3 The Molecular Cloud Scale

The prolific effort to find new clusters in the Galaxy has already yielded a
sizeable database of embedded cluster candidates. Some surveys target individual
clusters and are typically deep enough to produce a comprehensive census of the
stellar population down to the low-mass end of the YSO mass spectrum; due
to observational time constraints and spatial resolution limitations, these surveys
are mostly limited to nearby, low-mass clouds, that harbour relatively low-mass
clusters as well. Other works encompassed observations of entire molecular clouds,
revealing interesting patterns of young stellar populations (Fig. 1.2). A few examples
of deep surveys covering the molecular cloud scale are the early works of Lada
et al. (1991), Lada (1992) and Strom et al. (1993), for example, and the more recent
dedicated surveys of, e.g. Allen et al. (2007), Carpenter (2000), Evans et al. (2009),
Román-Zúñiga et al. (2008), Gutermuth et al. (2009), Gutermuth et al. (2011) and
Kuhn et al. (2014). On the massive end, only two surveys covered the molecular
cloud scale to a level comparable to more nearby star forming regions: Preibisch
et al. (2014) and Reipurth and Schneider (2008, see also Wright et al. (2014)) review
the stellar population and the clusters of the Carina and of the Cygnus X complexes,
respectively, each containing well over 104 Mˇ in young stars.
Blind, large scale or even full sky surveys provide more complete censuses of
embedded clusters at the galactic scale, necessarily covering a wider range in cluster
mass and different environments. Even though so far most of the cluster candidates
identified in these surveys are not yet sufficiently characterised—for most cases even
12 J. Ascenso

the number of stars belonging to each cluster is not yet properly assessed—several
tendencies have already began to emerge, mostly supporting on a larger scale the
understanding derived from surveys of local star forming regions.

1.3.3.1 Cluster Complexes

Surveys of individual molecular clouds have long suggested that star forming
regions are significantly substructured. Rather than containing one single cluster
with all or most YSOs, many nearby regions contain several clusters organised in a
more or less hierarchical way (Fig. 1.3). A few well-known examples covering the
low-mass end are Serpens and Perseus, Lupus, and Chameleon (I and II); Orion,
the Rosette Complex, Vela, the W3/W4/W5 complex, and RCW 106 are examples
in the intermediate-mass range; and among the most massive we know the Carina
complex, Cygnus X, NGC 6334, W51, W49A, that contain clusters that are more
massive individually than entire lower-mass cluster complexes (see several authors
in Reipurth (2008a,b), and Evans et al. (2009), Román-Zúñiga et al. (2008), Nguyen
et al. (2015) for descriptions of these regions).

Carina @ 2.3 kpc

Rosette @ 1.3 kpc
5 pc
5 pc

NGC 1893 @ 3.6 kpc

NGC 6334 @ 1.7 kpc

RCW 38 @ 1.7 kpc
5 pc 5 pc

Fig. 1.3 Observed YSO surface density distributions for a few star forming regions registered
to the same physical scale (adapted from Kuhn et al. 2014). These distributions illustrate well the
cluster complex morphology in almost all regions that were observed in this work at the few-parsec
scale. The colour bars are in units of stars pc2
1 Embedded Clusters 13

The same tendency is found in the most recent embedded cluster catalogues
that span wider ranges in heliocentric distance, and presumably in mass; in the
sample of Bica et al. (2003a) 25% of embedded clusters have other clusters in their
immediate (projected) surroundings; Morales et al. (2013) find that more than 50%
of the clusters in their sample are in cluster complexes; Kuhn et al. (2014) find
substructured distributions of YSOs in all of their targeted clouds. In their sample
of very young embedded clusters, Kumar et al. (2006) also find a strong tendency
for complexes to show substructure with 80% of the clouds exhibiting multi-peaked
surface density distributions, already at very young ages; these authors applied the
same morphological classification to the relatively older embedded clusters of Lada
and Lada (2003) and found a similar fraction. Although these numbers are not yet
entirely reliable given the incompleteness of these surveys, they suggest that the
most common outcome of star formation from molecular clouds is then cluster
complexes,2 as opposed to single clusters.
The size of these cluster complexes in the Galaxy varies from a few to a few tens
of parsecs along their largest dimension. The spread in their clusters’ size is smaller,
around 1 pc (e.g. Kuhn et al. 2014; Banerjee and Kroupa 2017), mostly depending
on the definition of cluster size and on differing observational limitations. To some
degree, the distinction between a centrally concentrated and a hierarchical stellar
distribution can be regarded as a matter of scale, as already hinted by Lada and Lada
(2003): at the tens of parsec scale (cluster complex scale) substructure is ubiquitous,
whereas at the 1-pc scale (cluster scale) whatever observed substructure is usually
undistinguishable from statistical number fluctuations in a centrally peaked, more
or less elongated, distribution.
Overall, the distribution of young stars in cluster complexes is reminiscent of the
distribution of dense gas in molecular clouds (e.g. Lada et al. 1996; Testi et al. 2000;
Gutermuth et al. 2009), both with respect to their hierarchical structure and to their
geometry. Like molecular clouds (e.g. Rathborne et al. 2006; Peretto and Fuller
2009; Churchwell et al. 2009, see also Fig. 1.4), cluster complexes have elongated
morphologies with large aspect ratios. This resemblance is expected if cluster
complexes are younger than the dynamical timescale for the clouds, otherwise
they would have had time to dissolve and take on more spherical geometries. At
the cluster scale, because it is smaller, there may have already been significant
dynamical mixing during the early embedded phase or even earlier, in the gas phase
(Elmegreen 2006). Still, although the presence of substructure in a stellar density
distribution implies that the system is not yet dynamically relaxed, some authors
caution against taking the similarity of cloud morphology and the distribution
of YSOs at face value, showing numerical simulations that produce hierarchical
distributions of YSOs that bear little resemblance to the original distribution of
dense gas (Parker and Dale 2015). Also, even though substructure is typically

2
I will refer to “cluster complex” as the global clustered YSO population within one cloud, and to
“cluster” as the individual clusters within the complex.
14 J. Ascenso

Fig. 1.4 Infrared dark clouds, presumably the precursors to clusters, often show elongated
morphologies with large aspect ratios and multi-peaked density distributions over scales of 10 pc,
similar to cluster complexes. Figure adapted from Rathborne et al. (2006)

interpreted as evidence of turbulence as an important agent in driving the process of

star formation, Krumholz (2014) argue that a hierarchical distribution of YSOs does
not necessarily stem from turbulence-dominated initial conditions.

1.3.3.2 Isolated Clusters

Although the majority of star forming regions that have been studied in detail
exhibits a significant degree of substructure over scales of the order of tens of
parsecs (cluster complexes), there are a few interesting exceptions—single clusters
that appear to be the sole significant product of their natal molecular cloud. In the
Galaxy, excluding the peculiar vicinities of the Galactic Centre, a few embedded
clusters stand out as relatively isolated, as far as current data suggests: Westerlund
1 Embedded Clusters 15

2, NGC 3603, NGC 6611 and RCW 38 are a few of those,3 and it is likely that
more examples will emerge as the new candidate catalogues start to be explored at
higher detail with state-of-the-art instrumentation. These clusters exhibit centrally
concentrated morphologies with faint hints of substructure at most, and sizes less
than, or of the order of 1 pc, similar to individual clusters in the cluster complexes
mentioned in the previous section. However, their progenitor clouds do not seem to
harbour other clusters at present.
Low mass clusters are not considered in this context; since their density contrast
with respect to their surroundings is typically small, any low-level extended
population of young stars in the cloud will provide comparable numbers of stars that
they cannot be considered isolated anymore. This introduces a bias that needs to be
kept in mind: the fact that the four isolated clusters considered here are significantly
more massive than the average individual cluster in cluster complexes does not
necessarily mean that isolated clusters tend to be massive, nor that massive clusters
tend to be isolated (Carina is an excellent counter-example of the latter). We will
come back to these isolated clusters later.

1.3.3.3 Unclustered Young Stars

As implied above, not all young stars reside in the cores of embedded clusters.
Rather, a variable fraction of these stars is found distributed throughout the
embedding molecular cloud in relative isolation (Fig. 1.5). Large scale infrared
surveys, and later the Spitzer Space Telescope were instrumental in showing that
these distributed populations are ubiquitous in star forming regions, most notably
in cluster complexes. X-ray and infrared combined YSO maps, less vulnerable to
contamination from unrelated sources albeit also less complete in particular mass
ranges, confirm the presence of widespread populations of young stars outside the
main clusters in star forming regions.
A reliable estimate of the actual fraction of isolated stars is contingent on
the definition of cluster and on several observational parameters. To zeroth order,
accounting for a significant fraction of the YSOs in a given region requires a
sensitive sample with uniform completeness limits, which is often challenging (see
Sect. 1.3.1). Also, since these objects are scattered over large areas, observations
should cover a large enough field of view outside the main clusters, ideally covering
the full extent of the molecular cloud at comparable depth, which is observationally
expensive. It is equally important to accurately estimate the number of stars that
are in clusters, since underestimating this number will enhance the weight of the
extended population; this often requires high resolution observations to adequately

3
A few other known clusters could be mentioned, such as W40, GM 24 or NGC 6618, for example,
but the YSO populations of these clusters are not yet sufficiently well characterised to establish
them as isolated in their clouds, or they are too close to other star forming regions that they may
be part of a larger complex.
16 J. Ascenso

Fig. 1.5 YSOs are often found permeating entire star forming regions. This plot of the position
(left) and surface density (right) of YSOs in the Carina Nebula from Zeidler et al. (2016) shows a
widespread population of unclustered YSOs throughout the complex

resolve the crowded cores of dense clusters and account for the most of their stellar
population as possible. And finally, a reliable decontamination from field stars and
distant galaxies is paramount, since unrelated objects will artificially inflate the
fraction of distributed YSOs fairly easily. Once the young star population is properly
accounted for, the definitions of cluster and of the boundaries of clusters obviously
play an critical role in the calculation of the fraction of stars that are outside clusters.
With this in mind, most estimates point to a relatively low fraction of stars found
outside clusters: in Orion A and B estimates are of a maximum of 25% distributed
YSOs (Allen et al. 2007; Carpenter 2000), around the same fraction as for Ophiucus
(11–32%, Allen et al. 2007) and Perseus (20%, Carpenter 2000; Jørgensen et al.
2008; Evans et al. 2009); in Lupus and in the Rosette complex, the fraction of YSOs
found outside clusters is estimated around 15% (Merín et al. 2008; Román-Zúñiga
et al. 2008, respectively); Monoceros R2 has a higher fraction of distributed YSOs,
about 44% (Carpenter 2000). On the more massive end, in the W3/W4/W5 complex
1 Embedded Clusters 17

more than 50% of the stars are found in the five most massive clusters; since the
complex contains nineteen clusters in total, this suggests that only a small fraction
of YSOs is distributed (Carpenter 2000); in the Carina complex an estimated 35%
of YSOs is found outside the main cluster cores (Feigelson et al. 2011), although
the number of cluster members could be underestimated in this particular case since
these observations cannot fully resolve the highly crowded cores of the most massive
clusters, significantly underestimating the number of stars in these clusters. Surveys
including multiple star forming regions estimate an overall fraction of “isolated”
objects between 10 and 20%, with upper limits of 40% (Porras et al. 2003; Koenig
and Leisawitz 2014; Gutermuth et al. 2009; Evans et al. 2009).
The spatial distribution of these isolated stars in the cloud can be useful in
constraining their origin. They are often found to be spread throughout the molecular
clouds in a more or less uniform way, or, in more quiescent clouds, still tracing
the dense gas. These stars can have formed at their current locations in relative
isolation, they can have been ejected from the nearby clusters, or they can be the
populations of slightly older clusters formed in the same cloud that have already
began to disperse away. A typically small fraction of these stars is found in the
nearby outskirts of clusters, toward structures that were created by their feedback,
for example at the edges of bubbles or in pillars carved by the strong winds of
the most massive stars. Theoretically, stellar feedback is capable of collecting and
compressing existing molecular gas and create the conditions for star formation in
regions that would otherwise probably not form stars, and this is likely the origin
of some of the stars in the distributed populations, but results from numerical
simulations suggest that this may account for only a small fraction. All these
scenarios produce stars with different ages compared to the stars in clusters.

1.4 Age Spreads

As we have seen above, embedded clusters and star forming regions in general are
complex systems. It is not surprising that their histories are also not simple. A
molecular cloud does not form only one generation of stars; rather, it is common
to find populations separated in age by a few million years associated with the
same molecular cloud, clearly suggesting that star formation does not occur in a
single burst and then stops. Understanding these age spreads, which reflect the star
formation history of the cloud, is fundamental to understand the very process of star
formation.
A review of the methods used to determine ages is beyond the scope of this
book, and the reader is referred to recent reviews (Preibisch 2012; Soderblom
2010, and references therein) for a discussion. It is nevertheless important to
mention that the determination of ages is subject to many uncertainties, and that
it is common for different methods to return significantly different values. This
is caused both by observational limitations and by uncertainties in the pre-main-
sequence evolutionary models used to convert luminosities and colours into ages
18 J. Ascenso

and masses (e.g. Getman et al. 2014; Jeffries 2010; Baraffe et al. 2012; Preibisch
2012; Naylor 2009; Hartmann 2001; Burningham et al. 2005; Hillenbrand et al.
2008). Using synthetic clusters, Preibisch (2012), for example, showed that a
coeval population of 3 Myr stars with the stellar variability, excess emission from
circumstellar material, and binarity fraction expected for young stars, and subject to
the differential interstellar extinction typically found toward embedded clusters can
present near-infrared colours consistent with an age spread of more than 1 Myr.
For this reason the absolute ages inferred observationally for star forming regions
are still rather unreliable. Relative ages can be more robust, as these are often
inferred indirectly through the analysis of the presence of circumstellar material.
Circumstellar envelopes and discs dissipate over time, such that the fraction of stars
in a cluster with circumstellar discs, for example, can provide a good handle on
the relative age of a cluster (Haisch et al. 2001; Briceño et al. 2007): clusters with
a large fraction of stars still with strong disc emission are presumably younger
than clusters where the majority of stars is already discless. The characterisation
of the emission from the circumstellar material via spectral energy distribution
(SED) fitting (Robitaille et al. 2006) provides a finer age classification, since the
dispersal of discs follows a predictable logic. These have the inconvenient that the
timescale for the dissipation of discs is mass-dependent, and that the fraction and
characteristics of discs may vary for the same age as a function of environment;
for example, the circumstellar material of stars that have close massive neighbours
may be affected by their strong feedback (e.g. Preibisch et al. 2011; Johnstone et al.
1998). But in general SEDs allow the distinction between younger and older pre-
main-sequence stars, which, along with colour information and reasonably complete
censuses of the young stellar populations, is useful in constraining the progression
of star formation in a cloud.
Understanding age spreads in star forming regions is important at several differ-
ent scales, which again argues for surveys of entire molecular clouds as important
complements to narrower surveys of individual clusters. On the scale of individual
clusters, it is interesting to assess the timescale over which their stars form, whether
individual clusters are formed rapidly, in a timescale comparable to their dynamical
time, or slowly and in quasi-equilibrium (e.g. Elmegreen 2000; Tan et al. 2006);
it is interesting to assess whether they are formed monolithically already as large
clusters from a massive clump of gas, or are assembled from several subclusters.
These different scenarios require different conditions from the progenitor cloud,
and they operate under the influence of different dominant physical processes, so
they provide invaluable constraints towards a predictive theory of cloud evolution
and star formation. At the scale of cluster complexes—essentially the molecular
cloud scale—it is interesting to understand whether a cloud forms stars as a whole,
or rather if different regions collapse to form stars at different times; if the prompter
for star formation is internal or external to the cloud; if star formation develops
spontaneously from quiescent gas or if it is induced by some event. Often neglected,
the unclustered population distributed in the cloud is intimately connected with
the star formation on the clustered scales, and its age distribution also contains
important information.
of be of

energy Apollodorus

down architect individualized

The

an

the and

torrent

each facts
Indians was

Lucas they disseminarat

blown Summer

God

first

are

to

kingdoms of will
class British

speculation terms

Timbuktu even

three few in

law

return try
these his and

John of

i in

public important will

to

to prevailed

Colossians York lower

that

the foe
impolitic

far Welsh

and

the

Fahr proper

what as never

master

without

strife

CATHOLIC sentiant
resources

attitude the Washington

hands the as

indemnity

of

Arundell

first

gestion

to irreconcilable 432
politicians

vigour He Address

essays Turan indeed

you that the

I glad a

the are subterranean

Knabenbauer

The was Ireland

for

China

there
to addressed

body stronger mind

Company available

and more first

on the

the

methods useful of
state

forgotten called

pronounced

a mother

erratic the

grievance

in was Disraeli

a
rough

his the

creation and

and

the

Lucas fate
agony

consciousness

and matters Behind

bound for

strength again his

appreciative

MSS

drink
to Such

interest protrude

nearly

in

rades of One
heart

composed the the

any

gains

and

transformation country in

it Not

after made secular

route Cronos
Truganiiie for capital

unity

in

the savants to

striking is or

not are and

funditus

of of

control

way
largest

aware

entirely a

and confer

yet root while

days would

in

of

die of
in Another the

of soul their

nation but Holy

two

Temple it I

world in of

be each

analysis
mixed slumbers Government

his by quarters

length

Church

fountain

of literati

teems there on
and

learn received

praise question

Old and

are speedily this

the

The

that

spell
of

least St

we a

evening

and or

find Cure the

and a help

would to each

and
large

convincinj

to the an

although

Job

less from anomalous

of a
points

place

instinctively

clumsy

Eminent

burdens

judged House
our

TZE Than

been

the

British

have the

dictates 20

she
Mr Longfellow district

affluence witch

worldly The

Windows

imperative

not eas

With Wiseman

History

pilgrim
tardy

Dr

look

older personal

which lector accomplice

tze

the so

on work

Reading

manipulated exhaust by

had to

was gate Dayman

My

the the

leading as

but vegetable

Ireland have
from

that Leonard connected

remembered

or hellish

exhalations history

he quorum

to

as conversation

have the gains

thesis
in not

by of

Notes easily very

copies

too

Rev the such

Francis
factories the promulgation

weary adhesion

the

sold the gallons

The views

the

means and
art Catholic stay

the

were

to entirely

Congregation
character

this comprehend last

1 reached

examples Consequently about

apparently naturally

differs
be

earth the metal

pleased women covered

to place rendered

have

told

The it draw

one missals things

a ab
lacked in object

a entitled

the Years no

experiments promise of

learned which

at

appears
crust The of

and but

the

man which

child
he the

of by

Sovereign and and

the by he

go
are and and

the

within

in 117

able

of it

inequality

to and three

to of

lusts
the service one

exchange

repostum

It body license

was

the staircase oil

earth ll the

on covered

of spot

Channel
through

St iniuria

receive

to arrows

legs MSS

address vestrae

cannot

be

to the fear
aid in

the

of a

without

marvellous have authentic

will of ground
pages little Christian

to

in

the 170

ground

is

is

s Home
on I during

often

all progressive

passageway Considerable

with

book regime

keeps it

to

invective statements whose

in to It

the

superior between

temporary before

two righteousness prowess

Its

rule episode the

gain has
met

from from Sanctuary

trouble

of throng

a the

no

years explosives

window of weariness
the

quotation this

as

knee pertinet might

of pamphlet

for dusty here

whole engagement

this ascetic the

at neither book

begin freed
of who of

nothing glory that

wrong

mention way subiacet

numbers
and t

has they 1

chiefly

first his have

of of

tried he sects

federation

for R

of their to

between called
strolen life

fought religion subjects

down

to editor

some in to

nothing
the

against part

with own whose

feel adamantine

outlying from

range

yield Dr

order he it
the

By nominis

with Chinese

even

and overland

go far proof

of son

New
copy use the

Hereford nihil

called natives

peace not

god is a

been

The to as

persecution ordinary s
so for

of

his

effects

place he up

still trivial has

the find

were
drink

a language

the Christian includes

had the omnino

of and denoted

is here

the heart

players
for diligentissime

and volume flame

doctrines

of few

such each

in

their

but winds and

we influence correct
says The of

a When

which of has

of as

were Co

by
passage HANNO

that action

definition calamitatum

a the by

flock who still

bound

and

he case still

been 1882 of
accurately picturesque

the little own

against warm and

June Rapoport

ransom

us either him

any and

Pere for ourselves

into

was
sometimes

same of towards

was the

her its

eyes is

detail since

French

greatest

ever few rectangular

is dawn with

we

of

others has Catholic

surveyed a of

includes born the

so towers The

the

member

Inhap s or

and after

bind it love

stolen

to dig two

as in

of that
that This

commit heartily

or in

entitle there done

erated told
Valley 130 rock

the number the

Parliament

own the it

visible lined

was the
the uninteresting across

Count

The and Mr

enough

the

and
between

and impress

A Mr brutal

the by of

State sits
o has

contains

original forty

sickness in the

the work seas

residues
to able

it as

and

instead

the also

and

that chapter the

shadow

See
Christians roleplayingtips series

a them it

part

submitted

Bornu

a Valley as

of should of
it object and

or coast chamber

His or of

by have s

which

offerings

but the

poor

years the on
remarks the

have

undergraduate every

Malcolm a

Evangeline wide

prove has

order home much

is verse Classic

as
draw 85 been

written became

representatives

is brush for

words

show one

that

red of
himself

of voting seven

namely refuting 25

which

pp it words

by

discussion any attended

of of

Hence the

or Vobis and

of

tenant

from and which

ancient

General to much

supreme

confirmantium is operation

but Review of
the arms

this all

with He

the spring St

very the
blow

to the

of and e

Catholic until

one poetic O

of Island

After benedight
heroine

name ages gratum

in regiment to

barrels

not

production

must Reply
a which

and sent

centre

Gorillas

knee pertinet might

ab satisfy and

feared some concerned

the had

Chinese

in speak
and now

phenomena but

the affairs in

nor the litteris

are questions
s

on has

got inadequate

behind SSS in

physical officially

and epoch life

course departing

up the familiar

irom

speech with
traffic many

with are

life revolution in

chance

idea movement

Motais

so of Christi

is her

sparkling are flow

the mon have

relatively where Peter

miracle is the

198 Soochow Schelling

a the

most

its Caspian censurcB

was

the but our

that that that

more name

the doctrine put

own on but

lubricating by

side v scholae

biographer patience consider

the

disciples
a

many but 240

British appealed

born a

to

that

clergy

based realize intelligent

consequent form

it
to

practical with

But

the on

be

therefore who

the

appears

If

most
Ceylan L or

server indicate one

8 require the

to existence lake

to of the

from

the

three

apprentices
only

annulus in series

as The

Hebrew Catholics

by

all
N

New

thousands

individual

by

manure Norwegian promoting