David Stevenson the Complex Lives of Star Clusters Astronomers’ Universe

Home , R Scuti, SN 2007bi

David Stevenson The Complex Lives of Star Clusters Astronomers’ Universe

More information about this series at http://www.springer.com/series/6960

David Stevenson

The Complex Lives of Star Clusters David Stevenson Sherwood , UK

ISSN 1614-659X ISSN 2197-6651 (electronic) Astronomers’ Universe ISBN 978-3-319-14233-3 ISBN 978-3-319-14234-0 (eBook) DOI 10.1007/978-3-319-14234-0

Library of Congress Control Number: 2015936043

Springer Cham Heidelberg New York Dordrecht London © Springer International Publishing Switzerland 2015 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifi cally the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfi lms 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 specifi c 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.

Cover photo of the Orion Nebula M42 courtesy of NASA Spitzer

Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media (www.springer.com) For my late mother, Margaret Weir Smith Stevenson

Preface

Stars are fairly social beasts. They are mostly born in extended, open families and then stay together for most of their childhood. Low and intermediate mass stars tend to wander with increasing time, but many stars stay with their siblings right through to the moments they die. Of the types of relationships stars are born into, some are very tight and others much more relaxed. The densest associations are globular clusters, while the loosest are stellar associations. For some considerable time, it was thought that these stellar groupings were distinct. However , as further evidence has come to light, it has become clear that the boundaries between these cluster groupings and the smallest dwarf galaxies have become blurred. Meanwhile, new types of clusters have emerged as technology has improved and more observations have been carried out. This book documents many of these discoveries, as well as new findings about the structure, origin and evolution of star clusters. For example, globular clusters were once thought to host a single population of stars. However, many appear to hold two or more generations of stars. How exactly is this possible, given the apparently small mass of these clusters? Are all globular clusters really star clusters, or are some the denuded remains of something much grander? As a result of containing dense assemblages of stars, clusters also allow relatively rare events to transpire that seemingly mock conventional rules of stellar evolution and engagement. Fairly massive stars can emerge from populations of low mass objects as a result of collisions or downright theft, while planetary systems might form and then degrade as stars harass one another. Some astronomers have come to question whether planetary systems could exist that might host life. Are conditions within clusters too severe to permit the acquisition of life, in particular intelligent life?

vii viii Preface

By considering the fates of stars, stellar partnerships and the star cities in which they reside as a whole, the outcome for life emerges—at least in outline. Many of these questions, and thus the story as a whole, will only be resolved with further observation and a considerable amount of deliberation. Remember that as you are reading much of the current work is still contentious and clouded in considerable uncertainty. It will remain so until further observations are made.

Sherwood, UK David Stevenson About the Author

David Stevenson was born in Paisley, Scotland, in 1968, and stud- ied molecular biology at Glasgow University and a PhD in genetics from the University of Cambridge. After a stint in academia, he became a teacher, but continued to write science articles for various magazines including Sky & Telescope. His publications with Springer include Extreme Explosions (2014) and “Under a Crimson Sun” (2014), and he is at work on a new book about weather and exoplanets. Despite a background in Biology, David’s father inspired his interest in astronomy from an early age. Exploring the crossroads of different scientific disciplines is a key interest and motivation in both teaching and writing. David lives in Nottingham, UK, with his wife and family.

Contents

1. Initial Observations ...... 1 Introduction ...... 1 Historical Observations ...... 1 Open, Globular or Simply Associated? ...... 3 The Hertzsprung-Russell and Color-Magnitude Diagrams ...... 5 Luminosity Classes of Stars ...... 11 A Schism Between Open and Globular Clusters: Observational Bias? ...... 14 How Star Clusters Revealed the Structure of the Milky Way ...... 20 Conclusions ...... 23

2. Adventures in Stellar Evolution ...... 25 Introduction ...... 25 Star Formation: Standard Monolithic Models ...... 25 Discs and Jets ...... 27 Cracks Within the Monolith ...... 28 Realistic Models for the Formation of Star Clusters ...... 30 Starburst Formation of Clusters ...... 32 Cluster Formation and the Evolution of Galaxies ...... 32 The Lives of Stars ...... 36 Brief Lives: An Overview of the Lives of Massive Stars ...... 36 Intermediate Mass Stars ...... 43 Sun-Like Stars ...... 48 The Fate of the Smallest Stars in the Universe ...... 51 Corpses ...... 53 Cosmic Recycling ...... 54 Conclusions ...... 55

xi xii Contents

3. Variable Stars ...... 57 Introduction ...... 57 Low and Intermediate Mass Stars from Birth Through Middle Age and Death ...... 58 The Music of the Stars ...... 61 A Varying Journey Through Time ...... 62 T Tauri Stars ...... 63 Delta Scuti Stars ...... 63 Exit Stage Left ...... 65 Semi-regular Variable Stars ...... 66 The Horizontal Branch ...... 67 W Virginis Stars...... 72 Mira Variables ...... 73 OH/IR Variables ...... 75 RV Tauri Stars ...... 76 ZZ Ceti Stars ...... 77 Young, Fickle and Massive ...... 80 Herbig Ae/Be stars ...... 80 Be Stars ...... 83 Supergiant Variable Stars ...... 87 Luminous Blue Variables and Wolf-Rayet Stars ...... 90 Cataclysmic Variables ...... 91 Conclusions ...... 98

4. Globular Cluster Formation ...... 101 Introduction ...... 101 The Stars of Globular Clusters ...... 101 Cluster Formation: A Reprise ...... 102 What Do Observations of Globular Clusters Tell Us About How They Formed? ...... 103 Secret Agents: Dwarf Galaxies Masquerading as Star Clusters ...... 105 Omega Centauri and Its Kin ...... 105 The Problem ...... 108 A Question of Mass ...... 112 A Helium Clue ...... 115 Evidence from the Physical Distribution of Stars ...... 119 Do Observations of Young Globular Clusters Back Up This Model? ...... 121 Contents xiii

Salt in the Diet ...... 124 Do Helium-Rich Main Sequence Stars Become Helium-Rich Geriatrics? ...... 125 A (Somewhat Silly) Gedankenexperiment ...... 129 A Summary: A Confusing Picture Painted with Salt ...... 130 Have Globular Clusters Been Consigned to the Dustbin of History? ...... 131 Conclusions ...... 134

5. Open Clusters ...... 137 Introduction ...... 137 The Structure of Open Clusters ...... 137 Old Yet Open? ...... 140 Classiﬁ cation of Open Clusters ...... 140 How Birth Determines Life ...... 142 How Open Clusters Come Apart ...... 144 Location, Location, Location ...... 148 Planets in Globular Clusters ...... 151 Conclusions ...... 152

6. Stellar Soap Operas ...... 155 Introduction ...... 155 Binary Star Systems ...... 155 General Principles ...... 157 Young Clusters ...... 159 Lurid Marriages and Messy Divorces ...... 161 Two Routes to Blue ...... 165 Beyond the Blue: The Twisted Fates of Cluster Stars ...... 170 W Ursa Majoris Stars ...... 179 How Binary Stars Can Affect One Another: SN 1993J ...... 182 How Westerlund-1 Solved the Puzzle of Magnetars ...... 184 Pair Instability: The Unfolding Stories of SN 2006gy and SN 2007bi ...... 189 The Likely Tale of a Massive Straggler ...... 200 X-Ray Binary Systems ...... 202 The Universe’s Loneliest Supernovae ...... 213 Type Ia Supernovae and Beyond ...... 215 Conclusions ...... 218 xiv Contents

7. The Complex Lives of Globular Clusters ...... 221 Introduction ...... 221 Speed, Distance and Crossing Time ...... 221 Violent Relaxation ...... 225 The Two-Body Relaxation Time ...... 227 Disc Shocking ...... 231 Core Collapse ...... 233 The Large Magellanic Cloud: Snapshots of Creation ...... 237 Ring Clusters ...... 240 Hodge 11 ...... 241 M33: A Brief Tale of Two Clusters ...... 242 Core Collapse in M33 ...... 243 The Young Clusters of M82: MGG 9 and 11, Density and Fate ...... 245 Giant Elliptical Galaxies ...... 250 Adrift in a Sea of Galaxies ...... 253 Cluster Evaporation ...... 257 Multiple Populations of Stars: An Afterthought ...... 261 Conclusions ...... 264

8. From Science Fiction to the Reality of Planets in Star Clusters ...... 267 Introduction ...... 267 Living Worlds ...... 267 Along Came a Spider: What Life (Appears) to Need to Arise ...... 268 Limited Clues from an Earthly Tree ...... 271 Energy, Entropy and Evolution ...... 272 Capturing Energy ...... 275 Is Life on Earth a Reasonable Model for Life Elsewhere? .. 281 The Galaxy’s Oldest Planet? ...... 285 A Planet Pair for Kapteyn’s Star ...... 288 Visions of Heaven: The Artistic and Visionary View from the Surface of a Cluster World ...... 291 Planets in the Open Cluster, M67 ...... 293 The Fate of M4’s Pulsar Planet ...... 293 Conclusions ...... 294 Contents xv

9. Milkomeda and the Fate of the Milky Way ...... 297 Introduction ...... 297 The Inevitability of Collisions in the Local Group ...... 298 Low Metallicity High Velocity Clouds and Star Formation ...... 300 Harassment and Merging Between M33 and M31 ...... 302 The Fate of M31 and the Milky Way’s Dwarf Satellites ..... 304 The Grand Collision ...... 307 The Fate of Milkomeda ...... 313 Galactic Dissolution ...... 316 Conclusions ...... 320

Glossary ...... 323

Further Reading ...... 335

Index ...... 341 1. Initial Observations

Introduction

Star clusters are glorious objects to look at. The closest, such as the youthful Pleiades or the more mature Hyades, form little villages of light woven amongst the constellations and naturally attract the eye. With binoculars the six or seven stars of the Pleiades that are visible to the naked eye resolve into dozens of stars encased in a wispy nebulosity of gas and dust that shimmers with a pallid blue glow. It is only when the full glare of a moderate or larger scope is brought upon these objects that their full glory emerges. Hundreds of stars, the majority far dimmer than the Sun, appear. Through the steady increase in our ability to resolve the stars we have begun to understand how these hypnotic objects have emerged from the cold darkness of space. It is this story that The Complex Lives of Star Clusters explores.

Historical Observations

Star clusters have been known since antiquity; after all many are easy to spot by casual observation of the sky. The first recorded, systematic cataloging of these stellar aggregates was done by done by Galileo Galilei in 1610. Galileo turned his attention to the faint patches of fuzziness recorded by Aristotle, the Persian astronomer Al-Sufi and others. Galileo resolved the peach-fuzz of light into dozens of stars and developed the idea that the Milky Way was a vast metropolis of stars. It was the progressive development of the telescope, as well as the burgeoning curiosity of an increasingly affluent (and influential) academia that exposed the true number, distance and scale of star clusters. The nearby open clusters Pleiades and Hyades were obviously observed since the dawn of mankind, but it wasn’t until 1665 that

Abraham Ihle discovered the first globular cluster, later named M22. However, Ihle only had access to a telescope with a small aperture. Consequently, individual stars within this globular cluster were not resolved and the object was assumed to be some form of nebula. Over a century later Charles Messier observed M4 and recognized that the “nebulosity” was really a tightly knit cabal of stars. Using his larger aperture telescope, Messier could resolve hundreds of stars. In the intervening decades increasing numbers of open clusters were cataloged but the true nature of them was not fully understood. In 1767 an English naturalist, the Reverend John Michell, calculated the probability that the spatial associations of light in the Pleiades might be chance alignments. Michell surmised that the chance these were a fluke alignment was one in 496,000: these associations had to be real, physically bound groupings of stars. Between 1774 and 1781 Messier published his catalog of comet-like objects—nebulous associations of light that were, in some cases, only barely resolved in his telescopes. William Herschel then examined many of these, teasing the light into further agglomerations of stars. Although we now reject Herschel’s idea that star clusters were the handiwork of gravity pulling stars together, his observations provide the framework for later classification schemes. Towards the end of the nineteenth century astronomers worked with instruments vastly superior to those commonplace in the seventeenth and eighteenth centuries. It was apparent at the higher resolutions that the globular clusters formed some sort of spherical cloud centered towards the constellation of Sagittarius, while the open clusters lay in all directions in the plane in which the solar system was located. As technology advanced the science of astrometry developed with improvements in the measurement of proper stellar motion. In 1943, nearly 170 years after Michell deduced the likelihood that the members of the Pleiades were physically associated, Adriaan van Maanen measured the proper motion of its visible stellar population. Comparing the motion of individual stars within the cluster to the mean motion of the cluster as a whole across the sky, van Maanen confirmed that they were physically associated and moving together as a coherent object. Initial Observations 3

In the 1940s it was certainly not clear how these associations were produced and whether the stars formed at the same time. The basic idea that star formation commenced when a cloud of gas and dust collapsed was well-established. However, despite this understanding, without an ability to deduce the ages of the stars within the visible association, it could not be clear that the stars had comparable ages. This understanding would come soon afterwards when astronomers began to interpret the pattern of stars in the color-magnitude diagram—a more useful and quantitative form of the Hertzsprung-Russell diagram developed three decades earlier.

Open, Globular or Simply Associated?

So, what exactly constitutes a star cluster, and what is the difference between a globular and an open cluster? These are seemingly trivial questions and the former is rather easy to answer. The second question seemed easy to address but as with many areas of science, the more you look the more complicated the picture becomes. Let’s start with the most diffuse affiliations—the stellar associations. These are loose groups of stars that began life in a single cloud of gas and move through space together. They are typically separated by 1 or more light years and are not bound together by gravity. These are simply cars on a race track, moving at a similar pace. Sooner or later the individual members of the association will fall out with one another and drift apart. There are many well- known examples of associations—perhaps most notably the Orion OB1 association, which includes the stars of Orion’s belt and numerous other, nearby, luminous O or B-class stars, as well as a multitude of much dimmer objects. Associations have lifetimes measured in millions or tens of millions of years. Their undoing is a combination of internal strife such as supernovae blowing material out of the cluster and physically removing stars, as well as the more insidious pull of all the other stars and matter within the galaxy. The Sun may well have begun life in some loose and, long ago, abandoned association. The one, famous, exception to the rule that stellar associations lead short lives is the Plough. 4 The Complex Lives of Star Clusters

This asterism is a collection of fairly luminous stars that are physically associated with one another, but that are (somewhat) more than 200 million years old. Moving up the ladder somewhat leads to open clusters. As with associations, there may be hundreds or thousands of stars present, but in this case gravity does hold them at least loosely together. Like those in an association, the members of this grouping of stars have approximately the same age. Associations may contain one or more open star clusters as part of their make-up, much like villages congregating along a river valley. Within an open cluster there are a few hundred to around 10,000 stars, which are located inside a space of 10–20 light years in diameter. They are usually found with a distinct and considerably denser core, measuring a few light years across. This is surrounded by a more diffuse halo of stars. Within the central core of an open cluster stars may be separated by less than 1 light year or roughly the distance between the Sun and its distant cloud of comets—the Oort cloud. Although close, this is still around nine trillion kilometers between each star, so the chance of collisions will be rare. As the stars within open clusters are only loosely bound by gravity, tides generated by any nearby gas clouds or other groups of stars will tend to pull stars out of the cluster (see Chap. 5 ). Consequently, open clusters have relatively short life-spans that are measured in tens or hundreds of millions of years, the exact figure depending on the mass and density of the cluster—and its location within the galaxy. Naturally, there are exceptions to this rule but these are rare. The most prominent exception is an ancient cluster called Berkley 17 (Be17). This ancient open cluster is likely 8 billion years old. Globular clusters are larger, with anything between around 50,000 to a few million stars crammed into a sphere perhaps 100–200 light years across. The stellar density rises towards the cluster core just as it does in an open cluster. However, with so many stars, within the central 10–20 light years, stars are crammed so tightly that the stars nestle less than one tenth of a light year apart. In some, the star-star distance is comparable to the width of our Solar System. As a result, in these clusters in particular, stars frequently interact and perhaps, on occasion, directly collide (Fig . 1.1 ). Initial Observations 5

FIG. 1.1 Zooming in on the heart of 47 Tucanae. In the first ground-based photo the globular cluster is a largely featureless and impenetrable blob of stars. In the left most Hubble image, the cluster is resolved into a structured mass, through which there is a hint of space between each star. But only in the zoomed in image do stars completely separate revealing space between them. Credits: ground-based (left-most ) image ESA/Hubble (Davide De Martin), the ESA/ESO/NASA Photoshop FITS Liberator & Digitized Sky Survey 2; Hubble ( central and right images ) NASA, ESA, and G. Meylan (Ecole Polytechnique Federale de Lausanne)

The Hertzsprung-Russell and Color- Magnitude Diagrams

The Hertzsprung-Russell diagram compares two fundamental properties of stars: their color—and hence surface temperature— and their absolute magnitude, or luminosity. As astronomical and photographic technology improved various filters became available that restricted the light entering the telescope’s viewer to specific wavelengths or narrow groups of wavelengths. This allowed astronomers to compare the brightness of stars at each, selected waveband. By doing this, astronomers can accurately categorize the brightness, or more precisely, the luminosity, of each star that was observed in a narrow part of the spectrum (Fig. 1.2). Astronomers observe that stars of different colors are often under-luminous in the visible part of the spectrum—particularly that which our eyes are most sensitive to. What we perceive 6 The Complex Lives of Star Clusters

-10 Io Ia -5 Ib II 0 III IV

+5 V

Absolute Magnitude +10 VI VII +15

OBAFGKML Spectral Class

FIG. 1.2 The HR diagram in a more modern guise showing luminosity classes I through to VIII. Io—Hypergiants, Ia—Bright Supergiants, Ib—Dim- mer Supergiants, II—Bright Giants , III—Red Giants , IV—Sub-Giants, V— Main Sequence, VI—Sub-dwarfs, VII—White Dwarfs. As well as inclusion of terms like hypergiant and sub-dwarf, spectral classes L and T (showing lithium and methane respectively) have been added. These are primarily populated by brown dwarfs (“failed stars”) although a few bona fide hydrogen burning red dwarfs lie at the brighter end of class L—and the odd red giant (Class T is omitted) through our eyes is a surprisingly poor guide to how bright or how colorful a star truly is. In dim light, cells in our eyes called rod cells, because of their shape, collect the vast majority of the visual information. Unfortunately, rod cells collect visual information principally in the green part of the spectrum. Rod cells are also hardwired in twos or threes to the central nervous system, which allows them to pool information. On the plus side, this allows them to detect stimuli at very low light intensities by triggering impulses to the brain. On the down side, they cannot resolve small, closely spaced objects very well. This neatly explains why a telescope is so much better at resolving stars in a globular cluster than human eyes are. Initial Observations 7

Rod cells are also distributed away from the primary focal point of our eyes—a place called the fovea. So, if you look at a faint star directly, it may appear to wink out. Avert your gaze slightly and the star magically appears again. At that key focal point, another type of cell, called a cone cell predominates. Cone cells come in three flavors with overlapping sensitivities: the so- called red, green and blue sensors. By a relatively obscure mechanism, the brain interprets the differing inputs of the cone cells to work out the color of an object. However adept these cells may be at distinguishing color, they are poor at working in dim light and this is their downfall. Each of these cells principally connects to your brain via single links to the optic nerve. Hence color sensa- tion is great but only if there is sufficient light to trigger the cone cells. As a consequence, our eyes poorly distinguish color in the night sky, except in the relatively rare situation where there is a bright star with a profoundly striking color. Moreover, as stars release radiation across a range of wavelengths, and this pattern broadly follows a so-called black-body curve, many different cones can be triggered at once giving a relatively bland perception of a star’s color. The color-magnitude diagram gets around this by quantify- ing the information from each star. A device called a Bolometer detects the intensity of radiation at two differing wavebands. Most commonly, these are U, B, V and I for ultraviolet, blue, visual and infrared. A waveband is a region of the electromagnetic spectrum over which light is sampled. For example U is the ultraviolet magnitude with a filter at 365 nm and a band-with of 68 nm, while B is blue magnitude with a filter at 440 nm and a band-with of 98 nm. The most commonly used is the B-V diagram, which has a comparison of the magnitude of a star at B and V, the “visual waveband” with a width of 98 nm. The V-waveband is broadly centered on the region of the spectrum most readily absorbed by our rod cells. Hot stars emit most of their light in the blue part of the spectrum, while cool stars emit most of this in the red end. As the color of each star gives the temperature, plotting V-B allows astronomers to plot temperature but in a way that allows the information to be accurately quantified (Fig. 1.3 ). The beauty of this measure is that it allows astronomers to compensate for gaps in knowledge. If you assume that every star 8 The Complex Lives of Star Clusters

FIG. 1.3 Standard photometric filters (or passbands) which allow light of particular wavelengths to pass. The intensity of light at each passband is taken for each star and the data can then be used to plot the star onto colour-magnitude diagrams which compare the intensity of light received at the U, B and V passbands. For example the bluish star Rigel has a B-V of –0.03 (its B magnitude is 0.09 and its V magnitude is 0.12; therefore, B-V = –0.03). The yellower Sun has a B-V of 0.656 in a cluster is physically close to the others in space then you can assume that the observed luminosity is simply a property of the star—rather than being affected by how far the stars is from you. Clearly, if the stars in a cluster were far apart then those which were furthest from us would appear to be fainter. Making the assumption of close proximity allows a simple plot of B-V (or U-B, etc) against the luminosity, or absolute magnitude. But if you are unsure about how far away each star is, then you can still plot the diagram. However, this time, if you plot B-V against V then your diagram will take into account differences in the brightness of the stars caused by variation in the distance to each star. This is because you are effectively plotting the temperature against the brightness at one waveband. Astronomers can then still get some very useful information from the graph without really being sure how far away each star is from one another and us. The only fly in the ointment is a phenomenon called interstellar extinction (Fig. 1.4 ). Imagine you are watching the Sun go down at the end of a hot summer’s day. The sunlight reddens as the Sun descends towards the horizon. This is because as the Sun sinks lower in the sky the Initial Observations 9

B-V 0.0 0.5 1.0 1.5 12

Asymptotic Red Giant Branch 14

Horizontal Branch Red Giant Branch 16

Blue Hook

18 Blue Stragglers Main Sequence Turn-off

20 Main Sequence

BAFGKM Spectral Class

FIG. 1.4 Basic features of a colour-magnitude diagram of a globular cluster. The large numbers of stars within the cluster allow processes like stellar evolution to be tracked light shines through an increasing depth of air. As it does so, more and more molecules and particles of dust and other material inter- cept its light. Blue light is preferentially scattered, while the red light shines through. The same phenomenon affects the light from distant stars with (in general) the effect increasing with distance as there is more space between us and the star with which to fill with dust. Therefore, stars further away from us appear redder as less and less of their blue light shines through the flotsam and jetsam of the galaxy. However, some simple mathematical tricks can correct for this and allow for relatively straightforward analysis of star color and luminosity. Nonetheless, with suitable tweaking, sufficient data can be acquired to accurately determine the properties of stars within clusters, no matter how far away they are. 10 The Complex Lives of Star Clusters

B-V 0.0 0.5 1.0 1.5 12

Horizontal Branch

18 Apparent Magnitude Apparent

B A F G K M Spectral Class

FIG. 1.5 The problems posed by real stars. At the blue end of the Horizon- tal branch and the red end of the main sequence and red giant branch the apparent luminosity of the stars doesn’t follow the trend of their neighbors. At the blue end the horizontal branch falls because most of the light is released in the ultraviolet. At the top of the red giant branch and base of the main sequence (not shown) the visible luminosity also appears to flatten out as most light is released in the infrared. The true luminosity (across the infrared to ultraviolet) is shown by the black lines

Yet, there is one problem with the color-magnitude diagrams that are most commonly used. Most astronomers use B-V (blue minus yellow-green) as the proxy measurement of surface temperature. As Fig. 1.5 shows, this grossly underestimates the true luminosity of the hottest and coolest stars. At the hot end of the stellar spectrum most stars are emitting ultraviolet. While at the cool end, most of the energy is released as infrared radiation. This has two, related but odd effects on plots of B-V versus luminosity, whether this is V or apparent or absolute magnitude. At the blue end of the horizontal branch—the region on the HR or color- magnitude diagram where stars are burning helium, the flat Initial Observations 11 horizontal branch becomes distinctly non-horizontal. The entire branch curves downwards towards the region populated by white dwarf stars. While the most extreme members may be intrinsically fainter than the cooler stars in this region, the majority are actually as bright as the cooler stars. The branch curves downwards because these stars are dimmer at visible wavelengths. However, when you measure the energy across all wavelengths these stars are just as luminous as the cooler members of the horizontal branch. Similarly, as you head up along the red giant branch—or down to the bottom of the main sequence, most of the energy is released in the infrared. Along the red giant branch this has the effect of flattening off what is really a rather nice straight upwards line. The luminosity appears to tail off, reaching a plateau, while in reality although there is less of an increase in visible light, the total amount of radiated energy is still increasing: the red giants are getting brighter as they continue their ascent.

Luminosity Classes of Stars

What are stars made of? How can we find out about their chemistry given that they are trillions of kilometers away? The key lies with light. When the light from the stars in a cluster is split into spectra dark bands of absorption—and in a few cases bright lines of emission—allow us to chemically fingerprint each object. A combination of luminosity, color and chemistry allows a precise determination of the stars that are present. Stars do not show continuous changes in luminosity and color but form distinct classes called luminosity classes. These are much like different animals and this is evident in the HR diagram of stars in the local vicinity of Earth—as well as in stars further afield. On Earth there are winged, warm-blooded animals we call birds but no winged humans or cold-blooded furry rabbits. Similarly there are no very large, hot stars—the largest are red and cool and there are no very luminous, yet small, red stars. This provides clues to the structure of stars. A highly luminous small, red star just wouldn’t work. With a low surface temperature it simply couldn’t radiate enough energy to be bright if it was small. The HR diagram 12 The Complex Lives of Star Clusters then provides those locations in which stars are stable (or mostly so) because the other areas simply aren’t populated with objects. If stars are only stable in certain structures then they must follow the same basic rules. Even if you don’t know for sure how a star works, by looking at a cluster made up of stars of different masses you can see some interesting but very basic processes at work. Of these, the most important is that massive stars live the short- est lives: more fuel does not mean a longer-burning fire. Similarly, stars appear to peel away from the main sequence, becoming both larger and cooler as they age. Similarly, the HR diagram reveals how stars evolve, a process that takes millions of years: this is particularly true when we look at stars in clusters. The evolution of even the most massive, short-lived star takes vastly longer than the age of a human being. Therefore, you cannot tell how one star will evolve simply by watching it grow old. But, as with humans, were you to take a snap shot of the population in a city you’d see people of all ages. Even without seeing the process of birth, aging, decrepitude and death you would be able to establish the basic ground rules of how people are born and eventually succumb to aging and pass away. Star clusters are the cities of the cosmos where stars enact their life cycles.

The Source of Elements: A Primer In the beginning there was only hydrogen and helium. The clouds that formed the galaxy were bereft of anything else, bar the barest whiff of lithium. However, when we look at the oldest stars in the galaxy we see the ashes from the first generation of stars that came before them. Population II stars are the galaxy’s senior citizens—and this is true of all galaxies we observe. They date back to around 12–13 billion years ago and contain only one hundredth to less than one ten thousandth the proportion of heavy elements found in the 4.6 billion year old Sun. To astronomers, these heavier elements are known as metals and include every other element in the Periodic Table.

(continued) Initial Observations 13

(continued)

The elements the Population II stars contain are revealing. Iron is generally scarce, carbon, oxygen and elements up to around the mass of calcium less so. But within this scheme there is considerable variation. The very oldest and metal- poor stars tend to be disproportionately poor in iron—sometimes to the point where it really isn’t detectable at all in the star’s spectrum. Carbon is sometimes present in relative excess. These differences, in the abundance of heavy elements, provide clues to what sorts of stars populated the early universe and how these contributed to the chemistry we see in stars today. Extremely massive stars, those with more than 130 times the mass of the Sun, should produce excessive quantities of iron. Stars which are less massive, down to around eight times the mass of the Sun, principally produce elements like oxygen, magnesium and sodium. Less massive stars produce carbon and oxygen. Any Population III star less than 0.7 solar masses will still be alive today—and, so far, have not been detected. As the Population II stars we see are not overly rich in iron, their immediate predecessors can’t have been extremely massive otherwise their chemical footprint would be visible. That is not to say that there were no extremely massive stars—only that they were not the most common—or indeed particularly abundant. Instead, stars similar in mass to the most massive of today (around 10–60 solar masses) seemed to contribute the greatest amount of pollution. It was their ashes that ended up in today’s Population II stars.

Where there are thousands of stars you can watch the stars carve out a path on the HR diagram. In essence the stars on the HR, or more precisely, the color-magnitude diagram becomes a vast stellar dot-to-dot diagram. By joining the dots the general flow of stellar history from birth to death becomes apparent. This can only be done with stars in clusters where there are sufficient numbers of objects to produce this pattern. 14 The Complex Lives of Star Clusters A Schism Between Open and Globular Clusters: Observational Bias?

When first discovered, it was clear that in the Milky Way globular clusters seemed to float high above the galactic disc, while the open clusters were found exclusively within the body of the galactic disc. Moreover, the open clusters consisted of primarily hot, young, blue stars, likely less than 1 billion years old—with many very young indeed. Some were still associated with the gas and dust they had formed from. The globular clusters by contrast were mostly yellow and red stars, likely much older. Although the principles of star formation and their power source were unknown, there were enough clues to point to a vast difference in age (Fig. 1.6 ).

FIG. 1.6 Open or globular cluster? A selection of images taken by the Hub- ble Space Telescope of different clusters, some classified as open and some as globular. The first, the Pleiades, is clearly open and the final, M80 is clearly globular. But what about the others? You can Google them to find out what the official line is with each. Credit: all images, HST, NASA Initial Observations 15

Finally, spectra of the stars in globular clusters showed weak absorption lines of elements other than hydrogen and helium. This indicated that they had very few heavy elements compared to the Sun. These distinctions led to the idea of different populations of stars: the ancient globulars and associated stars of the galactic halo were Population II, while the more metal-rich stars of the galactic disc were Population I. With those criteria everything seems fairly clear cut: there are clearly two populations of stars within the galaxy. However, things get murkier when we apply more recent, higher resolution optics to the populations. Within many galaxies there exist young star cities that have the characteristics of globular clusters—albeit very young ones. For example in the Large Magellanic Cloud there is a massive molecular cloud known as the Tarantula Nebula. At its heart are young clusters of stars. The most massive is known as 30 Doradus. This monstrosity likely contains upwards of 600,000– 1,000,000 stars, with over 1,000 massive stars clustered tightly at its heart. The distribution of the stars is strikingly similar to the ancient globular clusters of today. In the nearby galaxy M33 there is a very similar cluster, NGC 604, which bears at least a superficial similarity to 30 Doradus. NGC 604 forms the most prominent single feature within this otherwise small and unremarkable spiral galaxy (Fig. 1.7 ). Even more convincing are massive star clusters found in the starburst galaxies M82 and The Antennae. Here, several dense clusters of stars are found at various stages of formation and evolution. Once again, although these contain very un-globular cluster-like collections of young and massive stars, the distribution of these stars closely mirrors those seen in the globular clusters orbiting the cores of all observed massive galaxies. Finally, within the more placid confines of the Milky Way there are a few star clusters near to the galactic heart that also bear strong similarities to globular clusters, but much younger: Westerlund-1 and The Arches clusters show similarly dense collections of stars. Therefore, perhaps our classification scheme needs a bit of tweaking. When we look beyond our galactic shores we see other types of star cluster (Fig. 1.8 ). In the nearby gas-poor Lenticular, or SO, galaxy, NGC 1023 we see red, diffuse clusters with magnitudes and colors similar to globular clusters but with much larger radii. 16 The Complex Lives of Star Clusters

FIG. 1.7 NGC 1850 is a double cluster of stars within the Large Magellanic Cloud (LMC). Then majority of the stars belong to the larger foreground cluster which is several tens of millions of years old. Within here numerous red giant stars are visible. Towards the lower right is a tighter, smaller group of very young, massive stars. Although separated by tens of millions of years, the smaller cluster probably began life when supernovae from the background cluster triggered the collapse of nearby gas. Credit: R. Gilm- ozzi, Space Telescope Science Institute/European Space Agency; Shawn Ewald, JPL; and NASA

These have been dubbed “Faint Fuzzies” by discoverer Jean Brodie and co-worker Søren Larson (UCL, Santa Cruz) and they appear to be analogous to another group of faint globular-like star cities called extended clusters (ECs). These were identified by Avon Huxor (University of Hertfordshire, UK) and colleagues in the halo of M31, with further teams of researchers finding even more within the halos of nearby giant elliptical galaxies1 . These extended clusters have large, so-called half-light radii of around 30 parsecs (each parsec is about 3.26 light years, or the distance at which one astronomical unit subtends an angle of one

1 Recent research suggests that Faint Fuzzies appear to form when a small galaxy passes through the disc of a spiral. A shock wave passes across the disc triggering a wave of star formation. The spiral morphs into an SO galaxy surrounded by a disc of faint fuzzie clusters. Initial Observations 17

FIG. 1.8 Left , the first “Faint Fuzzy ” cluster discovered by Jean Brodie, near the SO galaxy NGC 1023 in 2000. For a astronomically convenient comparison, a normal globular cluster lies nearby to the right of the “Faint Fuzzy”. Faint Fuzzy clusters have similar overall diameters to globular clusters but are more diffuse in nature with broader half-light radii (see text). Image Credit: NASA/Søren Larsen . Right , Extended Clusters , captured by Avon Huxor and colleagues from the University of Hertfordshire, around nearby spiral galaxy, M31. These also have similar masses to globular clusters, but again are more spatially extended. Credit Nial Tanvir, Mike Irwin, and Avon Huxor (WASP consortium/Isaac Newton Group of Telescopes) arcsecond). The half-light radius demarcates the radius of an imaginary circle in which half the light of the star cluster is released. It is analogous to the half-mass radius, inside which half a cluster’s mass is found. For globular clusters, with predominantly very old, low mass stars, the two values, mass and light, are effectively equivalent as one solar mass liberates the equivalent light of one Sun. Most globular clusters have half-light radii of less than 10 parsecs. Their red colors put them at a likely comparable age to the standard globular cluster, but their large radii were at odds with a simple pairing. The extended clusters of Andromeda (M31) are all deep in the halo of the galaxy lying distant to 15,000 parsecs. The geographical location reinforced the view that these are ancient star systems. 18 The Complex Lives of Star Clusters

Meanwhile, at the other end of the density spectrum lie Ultra- Compact Dwarfs (or UCDs for short). These are as bright as dwarf elliptical galaxies but much more compact. However, like faint fuzzies these also have half-light radii of greater than 10 parsecs, but are two magnitudes brighter due to the high density of stars within them. Several of these were discovered in the vicinity of the Fornax galaxy cluster and until recently it remained unclear whether these were some sort of new galaxy or a much more massive globular-like cluster. When the extended clusters, globular clusters and UCDs are plotted on graphs of absolute magnitude against diameter (half- light radius) they appear to fall into distinct classes. This further suggests a distinct origin. Indeed Narae Hwang coined the term

10 9 8 7 w Centauri 6 5

Half-Light Radius 4

1 -4 -6 -8 -10 -12 -14 Absolute Magnitude

FIG. 1.9 Different kinds of large star cluster. UCDs are ultra-compact dwarf galaxies; EC/FFs are extended clusters or “Faint Fuzzies” and GCs are globular clusters. The position of Omega Centauri is indicated by a black cross. Lines indicate luminosity (compared with the Sun) per square parsec of sky. The grey triangle was formerly empty of clusters leading many to conclude that the three groups of clusters were distinct. However this area is now filling in with higher resolution studies indicating that all three cluster types are related, forming a continuous spectrum Initial Observations 19 the zone of avoidance to emphasize the apparent separation of the three cluster types (Fig. 1.9 ). Moreover, as recently as 2011 Jean Brodie (University of California, Santa Cruz) presented work suggesting that the UCDs in M87 were a distinct population of objects compared to the metal-poor globular clusters. Brodie and colleagues suggested that the brightest, red UCDs were the remnants of disrupted dwarf galaxies, while the fainter, blue ones were the stripped remains of globular clusters. Two years after Jean Brodie’s work, Duncan Forbes (University of Swinburne, Australia)—working with her and Jay Strader—published a far more complete analysis of these giant star systems. Using carefully calibrated distances they made a detailed comparison of all of the old dense star systems in the vicinity of a number of elliptical galaxies: NGC 4278 at 15.6 million parsecs; NGC 4649 at 17.3 million parsecs; and NGC 4697 at 11.4 million parsecs. At these distances the HST could resolves distances as small as 1–2 parsecs, easily enough to fully characterize the size and color of these systems. Forbes and colleagues identified a raft of old star clusters that have properties lying between these three (apparent) extremes. Many of these lie within Hwang’s “zone of avoidance”. Although Forbes’s data allows the most luminous red UCDs to have a different origin than the others, it no longer excludes a different origin for these star clusters to the globulars. With the zone of avoidance now filling in, it would appear that all, or most, of these clusters are related and form a nearly continuous spectrum of mass and density. The observational science is obviously evolving. The question to be resolved is in what sort of circumstances do these related, but distinct groups of star cluster form. As an analogy we can think of a population of humans. Some are small and stocky, while others are lean and thin. If you were to see the tall, thin humans in one country and the shorter, stocky individuals in another, you might imagine that they were unrelated. However, expand your view and see more of the population and soon it will become apparent that there is a continuum of humans from the stocky to the lean. It appears that this is true of star clusters too. What is less clear is why the Milky Way lacks these extremes: the extended clusters or the ultra compact dwarfs. It may simply be a question of circumstance. The Milky Way does contain some diffuse star clusters—the Palomar clusters that are described more 20 The Complex Lives of Star Clusters fully in Chap. 7 . However, these lack the scale of the extended clusters. And while the Milky Way has at least a semblance of the extended clusters, it completely lacks systems that are analogous to the dwarf spheroidals. It remains to be determined how the mechanisms that form globular clusters produce the apparent continuity in scale and mass of the globulars, themselves, and more widely the extremes—the extended clusters and the dwarf spheroidals. Chapter 5 will explore this more fully.

How Star Clusters Revealed the Structure of the Milky Way

Towards the close of the nineteenth century there had been many observations showing that globular clusters were preferentially located in one direction in the sky. What wasn’t known was the distance to these clusters. By 1914 several groups of stars had been identified that showed periodic, or regular and predictable, changes in their brightness. Of these, the Cepheid variables were best known. From observation it was clear that there was an explicit relationship between the period of pulsation of the Cepheid and its overall luminosity. The brighter the star is, the longer the period of pulsation. Therefore, if you know the period, from observation, you can tell how luminous the star should be. We also know how the light of an object decreases with distance. The luminosity of any light source, whether it’s a light bulb or a supernova, goes down with its distance from the observer by a mathematical law called the inverse-square law. In essence if you are twice as far from a star as another observer, the star will appear a quarter as bright. If you are five times further away, the star will appear one twenty-fifth as bright. Therefore, by comparing the expected luminosity of the star, from its pulsation, with the observed luminosity, the distance to the star can be determined. There are quite a few intellectual, but utterly sound, leaps in these two paragraphs. It was only through the genius of combining these that the scale of the galaxy—and sometime later the Universe— could and would become clear. Initial Observations 21

Table 1.1 Comparison between Cepheid Variable and RR Lyrae variable stars Luminosity Mass (com- (compared to Type of Star pared to Sun) Sun) Location Age Cepheid 3–9 solar 1,000–50,000 Galactic Less than 150 Variable masses Disc million Stars years RR Lyrae 0.5–0.9 solar 10–100 Galactic Up to 13 Stars masses Halo and billion Disc years RR Lyrae stars are far less massive, less luminous and much older than Cepheid Variable stars. Cepheid variables are exclusively conﬁ ned to the galactic disc where there is ongoing or recent star formation

Astronomer Harlow Shapley’s key contribution to our understanding of the universe in the early twentieth century was to combine these ideas, once he had identified suitable variable stars within the globular clusters. He then calculated the distances to each cluster, accordingly. Shapley also, quite reasonably assumed that the globular clusters were centered on the galactic heart. With these pieces of information the general dimensions of our galaxy became apparent. Shapley’s work not only confirmed that globular clusters were distributed in a spherical halo around the galactic heart, but he also determined the Sun’s distance to them. Thus Shapley confirmed the general scale of the galaxy, making it substantially larger than originally believed. He found the Milky Way’s distance across to be 300,000 light years, and while we now know that this distance is too large by a factor of three, it did expand the distances to the galaxy’s furthest margins (Table 1.1 ). Subsequently, two pieces of information were acquired that shrank Shapley’s universe somewhat. For one, Shapley assumed that the variable stars he observed were the luminous Cepheid variety. However, Shapley had in fact found a group of stars with a similar period to the Cepheids. These stars, RR Lyrae variables, were smaller with shorter periods. They were also considerably less bright, meaning that they were a lot closer than a Cepheid of similar apparent luminosity. Furthermore, the phenomenon of interstellar extinction—the dimming and reddening of light, which was discussed earlier, wasn’t taken into account. Thus stars 22 The Complex Lives of Star Clusters appeared dimmer and hence further away than they truly were. Both of these conspired to make the galaxy appear a lot larger than it actually was. Nonetheless, Shapley’s conclusions were broadly robust and were fundamental to later understanding of the scale of the universe. This is all the more remarkable—and somewhat (retrospectively) ironic: Shapley was a vehement opponent of the view that the distant “spiral nebulae” were in fact distinct galaxies, far removed from the Milky Way. In 1920 he vigorously debated the issue with Heber Curtis—a debate he lost. Several years later Edwin Hubble’s confirmed that these “spiral nebulae” were in fact separate galaxies. It was Hubble’s use of Cepheid variables that proved Shapley wrong: these spiral nebulae lay far from the Milky Way. Indeed the luminosity-period relationship showed that the Cepheids that Hubble observed were a good 20 times further from the Earth than Shapley’s RR Lyrae variables that he observed in the globular clusters. Today, we understand that the luminous portion of the Milky Way, its stars, are situated within a structure that is 100,000 light years wide. Inside this domain there is a thick disc that is 1.8 kpc (kiloparsecs) wide and swallows both the peanut-shaped bulge and the thin disc. Some globular clusters are associated with this thicker slice of stars. At the heart of the thick disc, along the entire diameter of the Milky Way, is a thin disc of Population I stars that is approximately 1 kpc thick. Within this structure lies our galaxy’s entire population of open star clusters. This thin disc contains 85 % of the galaxy’s stars, including the Sun. It is the site of our galaxy’s most prolific star formation, extending from the dark periphery of the galaxy into the heart of the galactic bulge. Around the entire width of the galaxy is a predominantly dark halo. Here are found the majority of the galaxy’s globular cluster stars, as well as a smattering of a few million stars that either formed alone, or have been torn from globular clusters or neighboring dwarf galaxies that have wandered too close to the Milky Way and been torn apart by its intense gravity. The bulge of our galaxy may be regarded as a scaled down version of an elliptical galaxy and almost certainly formed in a similar manner. During the first phase of the Milky Way’s formation, the bulge formed along with the Halo when matter was streaming in from all directions, colliding violently and generating Initial Observations 23 a highly prolific wave of star formation. As the Milky Way took shape, encounters with neighbors grew less frequent and the supply of gas rained in at a far more leisurely pace. During this phase the disc, first thick, then thin, took shape and the Milky Way assumed its current form. The bulge of our galaxy is arranged in the shape of a cosmic peanut with a central bar pointed roughly in the direction of the Sun. The origin of the bar is unclear, although the structure is fairly common in the local universe. Many bars appear to form spontaneously from the surrounding disc, although some—the Milky Way included—may have formed when a close neighbor (the Magellanic Clouds in our case) perturbed the movement of gas and dust in the disc, sending more of it cascading into the heart of our galaxy.

Conclusions

Star clusters form the center piece of several pieces of astronomical understanding. Globular clusters have been used to measure the size of the galaxy and constrain the size of the universe as it was understood 100 years ago. More recently, as our understanding of the processes by which stars evolve has increased, star clusters have been used to test our theories of these evolutionary steps. Moreover, as telescopic power has improved, the boundaries between different types of star cluster have blurred as very massive, yet young clusters of stars have been resolved in neighboring galaxies. More exciting still, is the discovery of star clusters that lack a counterpart within our galaxy. Why some lenticular galaxies, or indeed the Andromeda galaxy, play host to faint, globular- like clusters, but the Milky Way does not, is utterly unclear. In the next chapters we explore more about these ideas and see how star clusters have contributed not only to our understanding of how stars form and evolve, but also formed test-beds of our theories and challenged the processes by which we believe stars evolve. 2. Adventures in Stellar Evolution

Introduction

Stars are born cold and dark out of vast amounts of gas and dust that collect in oppressive clouds called nebulae. Approximately 200 years ago, Immanuel Kant and Pierre Simon Laplace indepen- dently reached the same conclusions about what happens next: gravity grabs the cloud and causes it to collapse into a spinning disc. At its heart most of the material collects into a star, while around it a wide but thin, low mass disc of gas and dust forms the womb in which the planets coalesce. Laplace and Kant’s model has held up remarkably well against contemporary observations. Amongst many telescopic observations, those of the Hubble Telescope of the Orion nebula’s central regions, close to the Trapezium cluster, reveal many small protostars, flanked by dark, dusty discs that are approximately the width of the solar system. It is here, and in places like it, that the galaxy’s next generation of stars will form. When Kant and Laplace dreamt of spinning discs of gas and dust it was unclear if stars were born alone or grew up in extended families that we see as clusters. After all, technology was just getting to the level where telescopes could discern individual stars within clusters. Although it was likely that stars formed within familial aggregations, the details of this fairly labyrinthine process took nearly 200 years to fully grasp. It is to this problem that we now turn our attention to.

Star Formation: Standard Monolithic Models

Imagine a cloud of gas and dust between 10 and 100 light years across. It is made up of trillions upon trillions of tons of hydrogen and helium gas, as well as a much smaller proportion of heavier

© Springer International Publishing Switzerland 2015 25 D. Stevenson, The Complex Lives of Star Clusters, Astronomers’ Universe, DOI 10.1007/978-3-319-14234-0_2 26 The Complex Lives of Star Clusters elements bonded together as dust. Within this cloud, particles are not distributed evenly, but instead will be collected into denser tendrils and sheets, interwoven with pockets of lower density. Imagine smoke from a fire, drifting lazily in the breeze. The overall density of this interstellar cloud is millions of times lower than we find within the atmosphere of our planet, with perhaps a hundred thousand trillion particles per cubic centimeter. Yet it is vastly denser than the sparsely populated particle brew of space as a whole. Within the cloud are denser cores of gas and dust, perhaps a light year across. It is here that a star begins its life. Given the right nudge this frigid stew of particles can begin to collapse under its own weight. In the simplest model gravity slips and slides molecules of gas and dust towards their common centre of gravity. Lacking any appreciable internal heat supply at this stage, the gases are cool and the galactic magnetic field has little to grab hold of. However, as the collapse continues friction begins to warm the gas. Once it is a few thousand degrees in temperature, molecules within the cloud first fragment, then begin to ionize. Hydrogen comes apart into its component nuclei and electron at temperatures above 4,000 K. With electrons free to roam around, the magnetic field of the galaxy suddenly has a hold on the gas and begins to resist its collapse. Not only does this hold back the hydrogen, but because the particles within the cloud are colliding, the other components of the soup become embroiled within the battle. As these too, ionize, the situation worsens and more particles become suspended be the magnetic field. However, if there is sufficient mass, gravity, with some inevitability, wins the battle. The inward pull drags particles across the magnetic field lines and the collapse continues. The release of further gravitational energy heats the gases until helium and other elements become fully ionized. Helium begins to come apart at temperatures of around 50,000 K, while many metals only fully ionize at temperatures of millions of degrees. Before this process is complete any deuterium is burnt to helium-3 at temperatures of around 700,000 K releasing further heat. Lithium follows at around 2 million Kelvin. These two elements comprise very little of the mass of the cloud and they can’t hold the collapsing protostar up against gravity for long. Adventures in Stellar Evolution 27 Discs and Jets

From the outside, the collapsing interior of the cloud with the spectacle of its internal machinations is completely invisible to our eyes. No visible light escapes through the wall of inky blackness. However, escaping microwave and infrared radiation waves map out the constellation of fragments that are forming into stars deep within. As a star approaches the main sequence, and its internal engine prepares to turn on, protostars can reveal themselves in differing ways. If the surface temperature exceeds 15,000 K a massive protostar will begin to shed copious amounts of ultraviolet light. This ionizes the surrounding nebula, releasing even greater quantities of heat, but also causing the nebula to fluoresce. Strong winds begin to carve out chasms around the stars, blowing away the dusty remains of the nebula. These vast, scarlet voids of light are known as emission nebulae, which mark out the locations of star birth in all spiral and irregular galaxies. However, not every star is born with such enthusiasm. Lower mass stars cannot directly ionize the gas as they don’t produce sufficient ultraviolet light. However, they can generate a light show by other means. For one, light from their surfaces can reflect off remaining portions of the gas generating reflection nebulae. But it is their jets that generate the brightest shows. Within the protostar the interstellar magnetic field is still present, compressed and potentially amplified by the motion of ionized gases within the star. As the protostar spins, the field is wound up liked a spring until it erupts from the poles of the star. Shooting outwards at the speed of light, the field pulls with it copious amounts of gas and dust that are blasted into the walls of the surrounding nebulae. In some instances these are directly visible as Herbig-Haro Objects—glowing blobs or jets of gas lying in opposite directions within 1 light year or so of a protostar. In others the jets are largely obscured and only visible in the infrared portion of the spectrum. With protostar development, time is of the essence. The most massive stars develop on timescales equivalent to the evolution 28 The Complex Lives of Star Clusters of modern man: less than 50–100,000 years. Stars with masses comparable to the Sun take approximately 30 million years to evolve from nebula to star. At the lower end of the main sequence, a red dwarf, with one tenth the mass of the Sun, will take a whopping 2 billion years to really settle in and stably burn its fuel. Thus, in that time, a nebula could, hypothetically at least, give rise to a hundred generations of massive stars while its most diminutive members are still gestating. There is a certain irony in these timescales. Most models (and observations) suggest that it is the smaller protostars that begin their collapse first, not the most massive stars. However, despite the head-start, the greater mass of the larger protostars ensures that these soon overtake their smaller brethren.

Cracks Within the Monolith

Can the monolithic model support the development of star clusters? Well, simply put, yes it could . Certainly, most open clusters, containing a few hundred to a few thousand stars could easily be produced in a single collapse. However, when the basic model is scaled up many interesting problems emerge (Fig . 2.1 ). For one, particularly large nebulae—something that was likely common earlier in the history of the universe when gas was more abundant—won’t collapse at once. The densest portions will collapse first, generating one wave of star formation, followed sometime later by neighboring regions. You can still see this effect in the Large Magellanic Cloud. The Tarantula nebula contains the dense, central cluster of stars—30 Doradus. Surrounding this is a void, blown clear of gas and dust by the strong stellar winds and radiation from the central cluster. Lying several light years further out is a distinct shell of stars, many very massive as well. These are somewhat younger than R136 at the heart of 30 Doradus and were undoubtedly spawned when the stellar froth from this cluster slammed into and compressed the surrounding gas. Lying further out still, you might expect even younger regions of star formation as winds from these compressed more distant gas. In somewhat older clusters, stars that have already formed and that are close to death, will unavoidably pollute less evolved, neighboring regions with gas recycled from their interiors. Thus, Adventures in Stellar Evolution 29

FIG. 2.1 Three scenarios for the formation of globular clusters. In a , a single giant molecular cloud with a million or so solar masses of material collapses to form a single cluster. In b , a single cloud collapses but massive stars form first towards the centre of the cloud and die as supernovae as further star formation is continuing. The debris change the composition of the subsequent stars and initiate further later rounds of star formation leading to hybrid clusters. Finally, in c , two giant molecular clouds with different chemical compositions collide as stars are forming within them. The merger of the two clouds produces hybrid clusters. Such a scenario appears to be playing out in the colliding Antennae galaxies Model A can account for some globular clusters but many others show distinct stellar populations spread over a 1–200 million year age gap long before the cloud is exhausted progressive generations of stars can form in fairly close geographical locations. This gives rise to clusters with fairly complex histories. Over the course of 10 million years, these different generations of stars will readily become scrambled through gravitational interactions. Finally, before star formation can complete, gravitational interactions within and between the cores inside the cloud, will cause some to be ejected, while others merge (Fig. 2.2 ). On an even larger scale, entirely separate collapsing clouds of gas might collide and merge, mixing gases of differing compositions. The end result of all of this activity is that the largest clouds, referred to as giant molecular clouds, will collapse in stages, generating waves 30 The Complex Lives of Star Clusters ab

FIG. 2.2 The effect of passing stars on developing binary systems. ( a ) Pro- tostars that are condensing within a nebula may be subject to frequent interactions with close neighbors. In some instances a passing protostar ( yellow ) may add momentum to both protostars in a developing binary system. This could eject the smaller protostar completely, if it gains enough momentum from the interloper (blue arrow in b ). In other cases the interloper is captured while the lighter protostar is ejected from the binary. This is particularly true of the interloping star is considerably more massive than the lighter, more vulnerable protostar of star formation. This can lead to stars that are dispersed in terms of time and space, but also, where circumstances might be more favor- able, single clusters of stars that vary in age. Clearly, the difference in age might not be large. But recall that in the time it takes low mass stars to form the cloud from which they are collapsing may have been thoroughly polluted by generation upon generation of massive stars. Star formation is a very messy business.

Realistic Models for the Formation of Star Clusters

Although we cannot see star clusters form, we can observe very young clusters where radiation and strong stellar winds are driving away the surrounding obscuration. We can also model star clusters on the most powerful supercomputers, then compare the results of the modeling with observations. What the models show is a demolition derby of protostars. Models by Matthew Bate (University of Exeter) and Ian Bonnell (University of St Andrews) show protostars forming within long tendrils of gas within the molecular cloud. These accelerate Adventures in Stellar Evolution 31 towards and away from one another, exchanging mass, energy and momentum. Some protostars violently merge while others are ejected in their entirety from the molecular cloud. This leads to the failure of the smallest protostars: these, now bereft of further fuel, fail to grow and become stunted brown dwarfs or rogue giant planets. Within the clouds the densest portions, unsurprisingly, become the most massive stars. Not only are they born with access to the greatest proportion of material, but they also can collect more through simple gravitationally-driven accretion of further gas and they can physically absorb other protostars and dense clumps of gas before they fire up their engines. Moreover, if the star cluster is dense enough they appear to be able to absorb other stars from the surrounding field. The end result is a general increase in the mass of stars towards the center of the youngest clusters. Some LMC clusters show more complex patters—particularly the giant 30 Doradus cluster. In these, though there certainly is a central concentration of mass, a few, rogue massive stars and binaries are found further out. This discrepancy could, however, be accommodated if we imagine the giant molecular cloud that spawned the central cluster also contained smaller, denser portions that collapsed separately to form these more distant stars. In other clusters, massive stars could be ejected by two different means. In some instances gravitational interactions between stars near the heart of the cluster could toss stars outwards (Fig. 2.2 ). Alternatively, a massive star could be tossed out of the cluster by a supernova explosion. In this scenario the star belonged to a binary system. When the most massive of the two stars exploded, the remaining star was given a sufficient kick from its lost orbital energy to cause it to be flung asunder. Of these, the first option might explain stars that could travel to their currently observed position within a few million years of the cluster forming. While the second mechanism might take a bit longer, it most likely would still require the star to reach its location within 10 million years of the cluster forming. If the massive stars are too distant or simply too young and moving too slowly to have reached their positions, they must have formed in more isolated dense clumps within the giant molecular cloud. 32 The Complex Lives of Star Clusters Starburst Formation of Clusters

M82 is 12 million light years away in the northern constellation of Ursa Major (the Big Dipper or Bear). Although M82 is an isolated galaxy, it displays very active star formation, pumping out stars at five times the rate of the Milky Way (around ten solar masses per year). Broad outflows of hydrogen gas are observed streaming away from the poles of the galaxy; while the side-on spiral arms are fes- tooned with clusters of massive stars. Numerous intense X-ray sources dot the arms, revealing the position of newly-formed, but hidden neutron stars and black holes. The reason for all this activity, in what is otherwise a fairly small spiral galaxy, is a recent encounter with its neighboring and much more massive partner, M81. As M81 streamed past M82, changing gravitational fields swept the smaller galaxy much like the wake of a boat passing a shore. Some of M81 and M82’s hydrogen was pulled out, forming a tenuous bridge between the two star systems. But much of M82’s gas was compressed and triggered a massive burst of star formation. Most of the star clusters are less than 20 million years old and home to a menagerie of massive stars. More interestingly, it is clear from the color and radii of these clusters that some contain the mass of a more familiar globular cluster but are built from thousands upon thousands of hot massive stars—as well as the more usual cohort of smaller (and still developing) furnaces. However, given the distances it is still very hard to resolve the individual stars—particularly those lighter and hence dimmer objects that hide within the glare of their more massive brethren. This is a shame, as hidden within the light of these young monstrosities is, perhaps, the truth of how our more ancient globular clusters came about.

Cluster Formation and the Evolution of Galaxies

Star clusters trace the formation of the galaxy as a whole. This is for simple and obvious reasons. The galaxy is simply a bag of stars: the leftovers from their formation and all the stellar excreta released Adventures in Stellar Evolution 33

Halo Thin disc

Thick disc

Galactic Bulge

Open cluster Globular cluster

FIG. 2.3 A simplified view of the Milky Way galaxy, showing the key regions discussed in this book. The Sun lies within a region of active star formation 3,000 light years thick, but 100,000 light years across, called the thin disc. Flanking this is a thicker region of much older stars, called the thick disc which is between 8 and 10 billion years old. This is about three times the width of the thin disc and about the same diameter. Within the central few thousand light years of the galaxy lies a region of stars 5–10 billion years old called the bulge—which in the case of the Milky way, is bar shaped . This entire structure is encased in a large diffuse halo of stars, clusters as well as diffuse gas and dark matter. The halo is between 11 and 13 billion years old and forms the most arcane portion of the galaxy during their lives and deaths. As such, by looking at the distribution of star clusters and their chemical make-up it is possible to see how the galaxy as a whole came together (Fig. 2.3 ). For instance globular clusters appear to mark the formation of the galaxy’s overall structure but the story isn’t as simple as it first appears. Globular clusters have different distances and orbits. Many, such as M22, follow the overall distribution of the halo— the vast sphere of stars that orbits the core of the galaxy in vast swooping orbits that take them lunging through the galactic disc. Others, such as 47 Tucanae, follow an orbit that is much more closely aligned with the so-called thick galactic disc. The thick disc is a wider zone of stars approximately 3,000 light years in depth comprised mostly of relatively old, metal-poor stars. Lying along the mid-riff of this thicker band of stars is the thin disc in which the most recent bouts of star formation have occurred. This 34 The Complex Lives of Star Clusters distribution suggests that 47 Tucs formed with the galactic disc rather than the halo. As astronomers assume that the halo formed first, globular clusters within the halo must have formed during this period, followed by those of the thick disc. Imagine the galaxy condensing out of a nearly spherical ball of gas. As the gas is settling into a disc shape waves of star formation are occurring. The globular clusters and perhaps isolated stars form when the gas is most spherical in nature: this is the present halo. Over time, through the action of gravity the ball flattens into a plump pancake of gas—the future thick disc. This will have picked up some gas expelled from dying stars in the halo. Over time, more and more enriched gas is left to fall into the current thin disc, while the thick disc and halo empty of most of the hydrogen-rich gas. Star formation ceases when the gas supply becomes too low. The dearth of globular clusters within the thin disc indicates that the conditions needed for their development were absent when the thin disc formed a few billion years later. This gave rise to the suggestion that something unique happened early in the evolution of the galaxy that made the formation of globulars happen here but only here—at least in significant numbers. Since all the galaxies we observe appear to have a similar population of these star clusters, it implies this unique event was universal in extent. But what was it? Could it simply be that there was enough gas early on but that this was depleted at later times, or was there something else about the universe as a whole? François Schweizer (Carnegie Observatories) suggested that the defining event in the formation of the galaxy’s oldest citizens was a period known as reionization. 380,000 years after the Big Bang the universe had cooled down to less than 3,000 K. At that point the sea of electrons and protons that made up the abundant supply of hydrogen plasma lost enough energy to allow them to bind together. When this happened photons of light from the Big Bang’s inferno could run free, liberated from the soup of particles that made up the universe. The universe rapidly went dark. The stygian night was shattered a few tens of millions of years later when the first stars formed from the collapsing clouds of gas. This process of re-ignition was clearly going to be patchy. When a star formed ultraviolet light and stellar winds began to heat up and reionize the gas, which was predominantly hydrogen Adventures in Stellar Evolution 35 and helium. During reionization, the electron orbiting the nucleus of the hydrogen and helium nuclei picks up enough energy to jump free once more. Around each star or star cluster a bubble of hot, fluorescing gas develops—something astronomers now refer to as HII regions. These pink bubbles of light would initially be scattered like islands in a dense, dark sea. Over the next few tens of millions of years these pockets grew in size as more stars formed and as those already senescing blew up in supernovae. As the pockets grew in size the pressure exerted by them on the surrounding gas might just have been enough to send the remaining gas collapsing into vast clouds that soon, thereafter, spawned the universe’s first generation of globular clusters. Since the process of reionization was universal—and limited to a window at most a billion or so years long—this would neatly explain the age distribution of the older generation of globular clusters. That they match the window of reionization is explained by the kick this process gave to star formation. The alternative view is somewhat more prosaic. This perspective sees the match between reionization and globular cluster formation as coincidental. In this theory, these early clusters simply formed during this stage in the universe’s life because this was when the supply of gas was greatest and when gravity was collapsing the majority of it into the structures that we now observe as galaxies. Once the vast halos of each galaxy had condensed out of the universal broth, the formation of globular clusters was an inevitability that was driven by the vast reservoir of material that could now collapse into stars. However, there is one more piece of the puzzle that might just reinforce the link between globular cluster formation and reionization: the chemical composition of the stars in these clusters. When astronomers look at the distribution of chemical elements heavier than oxygen, a large collection of these tell a simple tale. As we have already seen the majority of elements heavier than helium appear to come from one source, so-called Type II supernovae. The abundance of these elements hardly varies from one star cluster to another in the outer halo of the galaxy. This reinforces the view that the clusters found in the galactic halos all formed at a similar time. This is certainly suggestive. This chemical portrait implies that a combination of stellar winds, ultraviolet light and 36 The Complex Lives of Star Clusters supernova blast waves drove the formation of globular clusters during the epoch of reionization. Whatever the truth of the matter, analysis of the stars of the halo and disc gives at least a good impression of the events that led to the formation of the galaxy as a whole. The halo forms first, followed by the thick disc and bulge. Over time, the remaining gas and later dust settled into the remnant thin disc, where today’s stars form. Over time the quantity of gas and dust has decreased as more and more of it has become locked up in stellar remnants like white dwarfs. What remains has become enriched in heavier elements so that stars such as the Sun, which formed a few billion years after the globular clusters, have up to 1,000 times the quantity of these elements. Even younger stars, which are forming today, have up to three or four times the amount of these elements found in the Sun. This process of chemical enrichment will continue until the last stars form in a few trillion years and the galaxy’s supply of gas and dust is finally exhausted. By this time the Milky Way will have morphed into something much grander but perhaps isolated from the surrounding universe in a way we can hardly imagine now. The final chapter in the book will look more closely at the destiny of the galaxy and how it mirrors the ultimate fate of every star cluster contained within it today.

The Lives of Stars

In most cases how single stars evolve from cloud of gas to wither- ing remnant is fairly well understood. In order to paint a suitably colorful picture of stars in clusters we must look closely at how stars of different masses change, or evolve, from cradle to grave. It’s perhaps too much to look at all of the finer points of the processes. But, that said, there are some interesting moments that mark out the beginnings and ends of particular stages in their lives. It is to these that we now turn our attention.

Brief Lives: An Overview of the Lives of Massive Stars

Imagine a star with 10–40 times the mass of the Sun. It was born in less than 100,000 years—less time than it has taken modern man to go from cave to cul-de-sac. Its nuclear fires power through Adventures in Stellar Evolution 37 its massive reserve of fuel in less than 7 million years. In general the more massive the star is, the less time it takes to rip its store of hydrogen apart. Each of these stars is class O or very early (bluer) class B and releases prodigious quantities of energy in the ultraviolet portion of the electromagnetic spectrum. Along with their brilliance comes a powerful stellar wind. With increasing mass the strength of the wind picks up until it is blowing at over 1,000 km/s. These features drive powerful flows of gas away from their immediate surroundings and reveal the glaring blue light encased in a pink halo of dispersing, fluores- cent hydrogen gas. Around the universe, these HII regions luridly mark the sites of active star formation. This foundry is a violent place. The core of the star seethes at over 50 million Kelvin. Hydrogen is combined in fours with carbon to form isotopes of nitrogen and oxygen before a freshly minted nucleus of helium is spat out, restoring the original carbon nucleus. The process can then begin again. This “CN” cycle is highly efficient and drives the core of the star into violent convective motion. These motions bring fresh hydrogen down to the centre of the core and push the helium ash outwards, maintaining a fairly steady fire for the few million years that the star can sustain it. Of importance, these reactions branch at higher temperatures and produce small, yet significant amounts of neon and sodium. The relevance of these activities will become clear in Chap. 4 . As the supply of hydrogen wanes, the CN cycle starts misfiring. The reactions begin erratically producing more nitrogen at the expense of helium and carbon. This means that these aging massive stars begin to run out of carbon while slowly becoming more nitrogen rich. This is the prelude to the coming storm. When the supply of hydrogen has all but expired in the core, its massive weight causes it to collapse inwards over a few tens of thousands of years. The store of gravitational energy is converted to heat and the temperature soars. Hydrogen burns vigorously in a shell surrounding the shrinking helium-rich core and the energy from these fires causes the star to balloon outwards, cooling all the while. As the outer layers cool the star changes rapidly from blue to red, zipping across the HR diagram in the astronomical blink of an eye. Convection currents dig deeper and deeper down from the 38 The Complex Lives of Star Clusters outside of the star until they hit the helium- and nitrogen-enriched layers produced earlier. The star briefly burps these materials outwards into space before the helium in the core reaches 100 million Kelvin. This is an important indicator to astronomers on Earth that the star’s life is approaching the end. The star is now a red supergiant with a surface temperature of less than 3,500 K. Shortly thereafter, helium is fused in threes to make carbon and in fours to make oxygen. This phase lasts less than 1 million years, before this fuel is used up. Exhausted, the core collapses and heats once more, until its temperatures exceeds 600 million Kelvin. Carbon burns to make neon and magnesium, while the oxygen sits awaiting the next rise to over one billion Kelvin. At one billion degrees neon fuses to make more oxygen as well as a soup of mid-weight nuclei. All of this fuel is exhausted in a matter of months. When the carbon and neon are gone, in less than 10,000 years, oxygen is burnt to make silicon and sulfur in a matter of weeks. As the temperatures soar ever higher, nuclei begin to fall apart forming a rich brew of atomic building blocks. From this mess of nuclear scraps, iron begins to assemble at temperatures above 3,500 million Kelvin. The whole process takes about a day, but when it is over an object with the mass of 1.4 Suns lies inert and brooding at the star’s core. The word ‘inert’ is something of a misnomer: at temperatures over 3.5 billion degrees, nothing is inert. Nuclei are smashing together, falling apart and reassembling. The core resembles an episode of Scrapheap Challenge with nuclei coming and going as the star attempts to assemble something meaningful from a froth of radiation and particles. But in the end the game is lost and the iron the star has coagulated begins to catastrophically fall to pieces. Still desperate for energy, the core continues to collapse and heat up. The extra energy attempts to reassemble more complex nuclei but all that it succeeds in doing is ripping the iron core to pieces. In an instant, the core implodes at over 70,000 km/s and the dismembered pieces of nuclei fuse into one immense nucleus, capped by a thin crust of iron. This rapidly rotating ball seethes at over 500 billion Kelvin. Nearly mass-less particles stream away from the inferno and out of the doomed star, removing 99 % of the core’s energy. The remaining 1 % launches a massive shockwave, Adventures in Stellar Evolution 39 by an as yet mysterious mechanism. Within a day this has powered through the star, boiling and eviscerating it. The shockwave continues out into space at nearly 5 % of the speed of light, while the shredded remains of the star expand behind it. The star has morphed into a core-collapse supernova. Over the ensuing weeks radioactive elements—principally cobalt-56—power a declining display that can outshine a billion lesser stars. The core continues to cool, gyrating in space, as a neutron star or pulsar. The only difference between the two is whether the beams of radiation that the spinning neutron star generates from its poles are aimed at and thus visible from Earth (Fig. 2.4 ). There are numerous variations on this theme. Stars with masses above 25 times that of the Sun, will, if left alone to evolve, produce a black hole within them after the iron core collapses. In some instances, the inception of the black hole is marked by the appearance of two opposing and very brief jets of gamma rays: a gamma ray burst (Fig. 2.4 ). In stars with more than 30 times the mass of the Sun, the star also skips the red supergiant stage. These stars become first blue supergiants, or in the case of the largest stars, a type of violent, eruptive variable star called a luminous blue variable, or LBV for short. In many instances, the star generates very fierce winds that scatter most of the hydrogen-rich outer layers into space. What is left is a stellar core called a Wolf-Rayet, or WR star. These fiercely hot embers are rich in nitrogen, carbon or oxygen. The nitrogen-rich forms still cling onto some of their outer hydrogen-rich envelope, and these represent massive O-class stars that have just run out of hydrogen and are leaving the main sequence. As they continue to lose hydrogen, the helium core ignites, producing carbon and oxygen. As the incessant push of the star’s massive stellar wind drives these elements outwards, the nitrogen-rich Wolf-Rayet morphs into a carbon-rich, then oxygen-rich form. As violent stellar winds whittle away at the star, its mass decreases dramatically. A star born with 40 times the mass of the Sun may end up with only a quarter this value. Still more massive stars may lose proportionately more before death comes a-calling. With the most massive stars, on the whole they probably die most frequently as WR stars. Very few get to cling onto their outer layers. It is the exceptions to this rule that prove to be the most interesting.

FIG. 2.4 Different fates for massive stars. (a ) A star with 10–25 times the mass of the Sun becomes a red supergiant then explodes leaving a neutron star. (b ) A more massive star’s core collapses into a black hole and launches a gamma ray burst. (c ) A star with 100 times the Sun’s mass pulses violently before dying as a supernova. These stars should leave behind a black hole or an object called a magnetar. ( d ) A still more massive star with 130–260 times the Sun’s mass disintegrates altogether. See text for details Adventures in Stellar Evolution 41

A few, born with a mass equal or greater than 95 times the mass of the Sun, seem to occasionally break the rules. How these monsters come about may tell astronomers a lot about how the most massive stars are formed in star clusters. Chapter 6 looks at their violent lives in more detail. In these stellar giants the core may become unstable as its temperature soars above one billion degrees. Within the fusing mêlée of particles, oxygen ignites, releasing a torrent of energy. However, the gamma rays that are produced—and which are needed to hold the core up against gravity—morph transiently into pairs of electrons and their antimatter counterparts, positrons. This is known as pair-instability. Robbed of this crucial support, the core implodes. Oxygen fuses directly to iron and the resulting massive surge in energy production shocks the core driving a wave through the star. As the wave punches through the stellar mass, the hydrogen- rich outer layers are blasted off into space, generating a display akin to a true supernova. However, with the energy released the star can breathe a sigh of relief and settles back down to complete its evolution. Oxygen continues to burn quietly until it is exhausted and the star is forced down the well-trodden path to its doom with a collapsing core of iron. A few decades after the star let rip the first time, its core implodes. The most likely outcome is a black hole or perhaps, if the star is spinning quickly enough, or has shed enough mass, a highly magnetized neutron star called a magnetar. The supernovae shockwave races outwards and catches up with the earlier expanding shell of hydrogen-rich material. As they collide it generates a massive pulse of light which re-invigorates the supernova’s display (Fig. 2.4 ). Some supernovae, such as SN 2006gy and SN 2006tf, appear to fit the bill for these extended and very brilliant displays. Magnetars are odd neutron stars with a magnetic field strength up to one trillion times that of the Earth. Most of these spin slowly having magnetically braked their spin against the surrounding plasma in interstellar space. Interestingly, there is the suggestion that the braking process might just be fast enough to power a final display after the star has gone supernova. As the magnetar decelerate from a few hundred revolutions per second to perhaps the speed of an old gramophone record, the energy is released as two rapidly expanding and highly magnetized bubbles of plasma. These should 42 The Complex Lives of Star Clusters catch up with the outgoing supernova shockwave and re-energize it generating a second peak in the luminosity of the explosion. A few supernovae, such as SN 2005bf, appear to show this. Even more massive stars, born with more than 130 times the mass of the Sun, could, if they can retain their mass against the effects of strong stellar winds, or overly amorous neighbors, might simply blow themselves apart when oxygen ignites. The effect of the pair instability is so great that the shock wave unleashed within the star eliminates it in its entirety. Such an explosion would release many Sun’s worth of iron, cobalt and nickel. Although it is possible—perhaps even likely—that some stars can die this way, this route has to be trodden only on the rarest of days. When astronomers examine the oldest present residents of the galaxy, they simply do not see enough iron contaminating them. Therefore, these iron-rich decimators must have been uncommon, otherwise they would have left their chemical fingerprint for astronomers to observe today. Massive stars are a diverse bunch of objects. This is hardly surprising when we remember astronomers classify everything with a mass greater than 8–10 times the mass of the Sun as a massive star. This is the broadest possible grouping you can imagine. The lowest mass stars, meanwhile, form a tightly knit, but highly populated community of objects with less than half the Sun’s mass. The range could hardly be more different. Perhaps most importantly, it is the death of massive stars that produce most of the elements of intermediate mass: elements ranging from oxygen, through the light metals such as magnesium and sodium, and up to, but ironically not quite including iron. Although iron is produced in some abundance in these explosions, this production line is considerably less efficient than others. Therefore, it probably does not contribute the majority of this element to the universe. Core-collapse supernovae are also believed to be the site of formation of the heaviest elements— those known as r-process elements. These are the elements above lead (Pb-207) in the Periodic Table. Such elements can only be produced in a hurry, within an environment where there are copious amounts of neutrons. It is thought that in many, if not most cases, that this element-production is done within the firestorm above the developing neutron star. Adventures in Stellar Evolution 43

Thus we owe rather a lot to these explosions. If nothing else, the water we need to sustain life and the oxygen inhaled to sustain respiration were forged within the maelstrom of core-collapse supernovae.

Intermediate Mass Stars

Somewhere below ten solar masses, stars with ten times the mass of the Sun, the ingredients cooked up in the early phases of the star’s life never get hot enough to fuse all the way to iron. The term “massive star” ceases to apply. Within this domain are stars that can heat their carbon to the point of ignition and in some cases can still explode as supernovae. But the majority—those with masses from two solar masses up to seven solar masses—never quite get hot enough to burn anything more complex than helium. That is not to say they can’t carry out some interesting chemical transfor- mations, but their deaths are far less flamboyant than those of their more massive siblings. These stars have life times measured in anything from around a billion years for stars with about twice the mass of the Sun, down to 30 million years or so for those with the largest masses in this category. These stars are white or light blue in color and release most of their energy in the blue and ultraviolet portions of the spectrum. The amount of ultraviolet is far less than in the most massive stars and not enough to cause surrounding gases to fluoresce. Instead, if they retain cloaks of gas and dust for any length of time, or, like the Pleiades, bump into it in their travels through the galaxy, this material reflects their sky-blue glow. This produces reflection nebulae, rather than emission nebulae. When the core of these stars runs out of energy it will collapse and heat up, just as it does for more massive stars. After 50,000 years or so, the star has swelled to become a bright red giant. As the expansion ensues the outer layers begin to convect. A deep cir- culation extends from the outside of the star down to the core and helium is dredged upwards. These currents also have the effect of keeping the outer layers well mixed. As the core grows hotter, a shell of hydrogen burns furiously around it. After a few tens of thousands of years the core heats up to 100 million degrees and 44 The Complex Lives of Star Clusters helium fusion commences. It’s a fairly passive affair but the new energy supply stirs up the interior and causes the star to contract and heat up. The star now settles down as a yellow or orange giant star for anything between one and a few tens of millions of years. Some of these stars spend some of their time in a portion of the HR diagram called the instability strip. It is here that we find the stars that Shapley was seeking when he was measuring the distance to the globular clusters. Within the instability strip, stars are just hot enough that helium is found as a neutral gas at the surface, but further down it is an ionized gas called plasma. This is a state where the gases electrons are able to break away from the nuclei. In a narrow zone between the two, these two states of helium are able to swap around. The outermost layer, which is full of helium atoms, is quite transparent to radiation, but the ionized plasma is not. When there is a lot of ionized helium present, energy is trapped within the outer layers of the star and this causes it to expand. However, as it expands, it cools down and the helium nuclei grab hold of the free electrons that form a sea around them. With the electrons soaked up, radiation is able to escape freely again, and the star deflates. As it does so it heats up, once again, until it is hot enough for one of the electrons to break free. Radiation becomes trapped and the star is forced to repeat the process. In a star with 3–9 times the mass of the Sun, these cycles of repeated expansion and contraction last a few days at a time. Chapter 3 examines these stars in more detail. The brighter stars are the most massive and the largest. Consequently, they take the longest time to cycle from small and hot to larger and cooler. This correlation forms the cornerstone of the period-luminosity relationship that makes Cepheids such wonderful stars for measuring stellar distance. As long as you know the period of the pulsation, you can determine how bright the star truly is. Then you can compare its observed luminosity, or apparent magnitude, with its expected luminosity, or absolute magnitude. And, as stars get dimmer the further away they are, it can then be determined how far away the star is. Once the helium runs low, nuclear reactions wane and the remaining helium is combined with carbon to make oxygen. Bereft of an energy supply, the core begins to contract once more. Adventures in Stellar Evolution 45

The star loops briefly towards the blue end of the HR diagram and then begins its final ascent. With helium burning in a shell around the inert carbon- oxygen core, the core grows in mass and continues to collapse. This increases the amount of heat and the speed at which the helium is consumed in the surrounding shell. The star is expanding, shedding mass from its outermost layers and growing ever brighter. Its final death dance is about to play out. Once the helium is all gone, a layer of hydrogen above the core reaches temperatures high enough to fire up. Over a hundred thousand years or so, this burns furiously, adding fresh helium to the outside of the core. When the helium-rich layer is thick enough, it too begins to burn, adding fresh carbon and oxygen to the core. But when helium burns, the hydrogen burning layer is shoved upwards and outwards until it becomes too cool to continue its activities, so it shuts down. Helium fusion lasts a decade or so, until all of this fresh supply if fuel is exhausted. When gone, the star repeats its pirouette and begins fusing hydrogen to make more helium, and so on and so forth. On the HR diagram the star zigzags ever higher, while cooling down. The star is ascending what is known as the Asymptotic Giant Branch, or AGB phase. The name comes from the trajectory the star is on in the HR diagram. This path to giant-hood roughly parallels the earlier red giant phase, but at higher temperatures and greater luminosity. You might imagine that this process could go on indefinitely, but there is a problem. Not only is fuel being consumed from below, to make the core bigger, but the star is also blowing its remaining fuel into space in enormous stellar winds. As the core gets larger, it also gets hotter and this increases the amount of energy it creates by fusion. This drives more and more powerful winds from the star’s exterior and these blow away the remaining fuel. The star continues its routine for a few hundred thousand years, pulsing in and out; brighter and dimmer as hydrogen and helium alternate in powering its dying structure. Within a million years of the pulses beginning the fuel is spent and the star begins to power down: but what a beautiful way to go. The core of the star is revealed as the last of the outer layers are ejected. This planet-sized sphere of carbon and oxygen grows ever hotter as 46 The Complex Lives of Star Clusters what is left of it contracts to become a white dwarf. Temperatures rise from 5,000 to over 100,000 K. Above this temperature most of the energy is released as ultraviolet and X-rays. The dispersing envelope of the star fluoresces, glowing like an enormous ghostly butterfly. These planetary nebulae adopt a variety of shapes that depend on how the final winds from the nascent white dwarf are blowing. The details of this are still being worked out by astronomers, but they can vary from chaotic splats of luminosity, through elegant butterflies, to translucent, near perfect spheres. Although several hundred planetary nebulae have been observed, until relatively recently they were unknown in globular clusters. This had led to some suggesting that the metal-poor stars of the clusters were physically incapable of forming such structures. Alternatively, the relatively brutal environment within a globular cluster might mean that the relatively fragile red giants would lose their envelopes of hydrogen and helium before a planetary nebula could form. As with many ideas built on a dearth of positive evidence, rather than a wealth of negative data, the Hubble Space Telescope soon dispelled this notion. Lying towards the outside of the globular cluster, M15, Hubble confirmed that an object known as Keustner 648 was a planetary nebula. Although confirmation took nearly 70 years, this luminous, oval of light was first observed by F.G Pease in 1928 using the 100 in. Mount Wilson Observatory. Pease managed to take a spectrum of the object that was buried within the light of the hundreds of thousands of other stars in M15. This was no mean feat given the instrumentation of the day. Unfortunately, for Pease, the telescope’s optics were insufficient to resolve the object more fully. This meant that its nature remained obscured for some decades to come. Since the discovery by Pease, only a handful of other planetary nebulae have been discovered in globular clusters. In part this is down to the limited numbers of stars that can be observed, but there is also a discrepancy that appears to be intrinsic to globular clusters. We shall return to this in Chap. 6 . Quite aside from the beauty cast by their eventual demise, stars in the AGB phase are particularly important for us. During this phase carbon from the core is swept upwards by convection during the brief wave of helium fusion. This carbon is mostly in Adventures in Stellar Evolution 47 a form called carbon-12. This has six protons and six neutrons. Carbon-12 soot then pollutes the shell where hydrogen will later burn. The important bit happens when helium fusion stops. When the helium engine wanes and the hydrogen burning layer re-boots, hydrogen and carbon-12 combine to form carbon-13. Carbon-13 has one extra neutron, but is otherwise still carbon. A hundred thousand years later this carbon-13 ends up back in the layer where helium burns and here then the reaction between helium and carbon-13 produces oxygen-16 and a free neutron. This little neutron is the instrument of change for many other elements in the universe. Neutrons are electrically-neutral so can stick with relative east to other nuclei. Throw in a supply of heavier elements, like iron, then the neutrons can glue themselves onto each forming progressively heavier elements, all the way up to lead. The process of neutron assembly is called the s -process and happens with relative sluggishness. A few billion neutrons are available in every cubic centimeter of the helium-burning layer and these can work their magic on the other elements within the star. By contrast, the r-process assembles elements with extreme rapidity, blasting nuclei with trillions of neutrons, taking the assembly line past lead towards the summit of the Periodic Table. AGB stars are also the site of another assembly line called known as hot-bottom-burning. Although not important for stars with masses less than three times that of the Sun, stars that can engage in these reactions are thought to produce most of the universe’s lithium and fluorine, as well as elements like magnesium, aluminum and sodium. The formation of sodium through hot-bottom-burning leads us into another, somewhat mysterious problem. Without going into the details just yet, there is an odd relationship between oxygen and sodium in many globular cluster stars. The stars that are richest in sodium don’t seem to become AGB stars when they run out of fuel. Why, isn’t yet known, but in Chap. 4 we will look at this oddity further. Over the few hundred thousand years that the star is firing both its engines, billions of tons of s-process elements are created. These have a distinctive chemical signature, both predominantly heavier than iron but lighter than lead and rich in neutrons. The upper limit is determined by conflicting radioactive decay. 48 The Complex Lives of Star Clusters

Although elements between iron and lead can be radioactive, the process only serves to change neutrons into protons in the atomic nuclei. This is what, ultimately, produces the new element. However, above lead, radioactive decay is more brutal, ejecting helium-mass projectiles called alpha particles. The s-process just can’t keep pace with this nuclear carnage so the assembly line stops. Nonetheless, during the relatively brief time it operates, the s- process creates vast amounts of elements like rubidium, strontium and lead. These are then vented into space to help form the next generation of stars and planets. While all of this activity is going on the last of the hydrogen is burning away and the remnant star—a white dwarf—begins its 300 billion year long journey to cold, inky blackness. Bereft of any internal supply of fuel the white dwarf fades slowly. The rate of cooling is fairly predictable which allows these moribund objects to be used to date star clusters, or the galaxy as a whole. In general the coolest white dwarfs that we can observe have temperatures of around 3,000 K, making them an orangey-red color. These are between 10 and 12 billion years old, depending on their chemical make-up, with the more hydrogen-rich dwarfs taking the longest to cool.

Sun-Like Stars

The course of life in stars with masses that range from 60 % of the mass of the Sun up to just over twice its mass is broadly the same as for the intermediate mass stars. There are a few differences which make them particularly useful to astronomers who are looking at the ages and dynamics of globular clusters. In particular, the time it takes these stars to become red giants, once their central store of hydrogen is exhausted is measured in hundreds of millions of years, not tens of thousands. The difference comes down to the mass of their helium core. In stars with masses over two and a quarter times that of the Sun, the core is already large enough to fire up once its temperature becomes high enough to do so. These also are hotter to begin with so the time taken to ignite the fuel is less. However, a star like the Sun leaves the main sequence with only one tenth of its mass burnt to helium. This is only a fifth of Adventures in Stellar Evolution 49 that needed to set up a helium burning inferno. This small core immediately contracts and becomes degenerate. This is a state where the particles in the core are confined to particular energies and the electrons in this state strongly resist further contraction. The core, thus, has to grow by the addition of fresh helium until it can become massive and hot enough to begin helium fusion. The Sun will take the best part of a billion years to complete this transition. Somewhat more massive stars will take correspondingly smaller periods of time, but it will still be a very protracted process. Consequently, stars with masses less than roughly Sirius, will track slowly upwards, cooling, reddening and brightening. This forms a particularly prominent feature in the color-magnitude diagrams of large star clusters (Chap. 1 ), which is useful in determining the age of the cluster (Fig. 2.5 ). More massive stars spend

B-V 0.0 0.5 1.0 1.5 12

Asymptotic Red Giant Branch 14

Horizontal Branch Red Giant Branch

Blue Hook

18 Blue Stragglers Main Sequence Turn-off

20 Main Sequence

BA FGK M Spectral Class

FIG. 2.5 Re-capping the basic features of a colour-magnitude diagram of a globular cluster. The prominent band of stars extending from the main sequence to the tip of the red giant branch and beyond in these clusters forms a convenient tract showing how low mass stars evolve 50 The Complex Lives of Star Clusters such fleeting times evolving from main sequence to red giant that you really can’t track them as they dance across the HR diagram. Moreover, the proportion of massive and intermediate stars in a cluster is much lower than that of smaller, Sun-like stars. Both of these factors conspire to make it highly unlikely that you will spot a massive star will make the jump from main sequence to red giant. Along the red giant branch there is a little upwards bump, as if the star has tripped on its path to giant-hood. Evolving low mass stars abruptly move upwards at nearly constant temperatures mid- way along the red giant branch, but why? The origin of this little bump lies hundreds of millions of years earlier when the star was still burning hydrogen in its core. Some of this helium ended up stirred upwards around the core, making a broad area that was depleted in hydrogen but richer in helium. Above this layer, convection kept the composition roughly constant. Eons later the shell, in which hydrogen was burning, crept outwards as the core continued to grow in mass. Perhaps 500 million years after the star’s core ran out of fuel, the shell reached the base of the layer in which hydrogen and helium were still convecting. The increase in the amount of low mass fuel causes the star to briefly dim then brighten once more. As the star has to traverse the same part of the HR diagram three times, stars stack up causing a small clump in the red giant branch. This is the red bump. It was first predicted by H. C. Thomas in 1967 with further work by Icko Iben (University of Chicago) published in 1968. However, it wasn’t until 1985 that Christopher King (Yale University) managed to spot the bump in a large sample of red giant stars in the globular cluster 47 Tuc. In intermediate mass stars, convection currents can dip down all the way from the surface of the star to the outside of the hydrogen burning layer and drag helium upwards. In lower mass stars like the Sun, these two layers keep their distance and the stars don’t show as much change in the amount of helium in their spectra. In those stars that do show the change astronomers refer to it as the first dredge up . Once this little drama is over the star continues to brighten a bit more until the core has become hot enough to burn helium to carbon and oxygen. At this point these stars chart the same general course to their more massive cousins. These stars, pretty much regardless of their mass spend approximately 150 million years Adventures in Stellar Evolution 51 burning helium to carbon and oxygen. Like their slightly more massive cousins, these stars drop out of the red giant branch and, depending on a number of factors—some of which are unknown, populate a region extending along the HR diagram called the horizontal branch. Stars with plenty of metals form a clump of stars just below the red bump. More metal poor stars, or those that have lost some or most of their hydrogen envelope, spread out to higher temperatures. Once all of the helium is used up, these stars follow the same path away from the red giant branch and onto the region where white dwarfs cool and fade. For these stars, the transition lasts around 1 million years or so, and like the intermediate-mass stars is marked by a series of alternating pulses of hydrogen and helium fusion until the entire envelope has been shed. However, with less mass to play with, there are less pulses overall—perhaps less than three. During these burns, some s-process elements are produced, but with fewer burns of hydrogen and helium, the yield is lower. Figure 2.6 illustrates the fate of low mass stars. In globular clusters and some open clusters there is a “shortage” of red giants and something of a glut of hot helium-burning stars. The evolution of these stars differs from the more conventional Sun-like stars because these have lost their outer-most layers that are rich in hydrogen. These small subdwarf-B stars (or extreme horizontal branch stars) burn helium then turn directly into white dwarfs once their fuel is gone. Without a large opaque shell of hydrogen around the core, which would absorb energy and allow the stars to expand, these never re-emerge as AGB stars when their fuel is exhausted.

The Fate of the Smallest Stars in the Universe

Stars with less than 60 % of the mass of the Sun are red in color and known as red dwarfs. These stars are extremely numerous, comprising up to 75 % of the stars in the visible universe. Despite their abundance, these stars contribute the least to future generations of stars. The most massive (carrying more than half the mass of the Sun) may just get hot enough after their hydrogen is gone that the helium ash can be burnt: but most will not. In addition, only those red dwarfs born with more than a quarter the Sun’s mass will expand 52 The Complex Lives of Star Clusters

E D1

C2 D3

D2 E

FIG. 2.6 The evolution of low mass stars like the Sun can follow a conventional track (top half of figure) or in some cases, diverge and follow a radically different path (lower ). In the top branch the G or K class star expands into a red giant (A , B ) then ignites helium in its core (C1 ). After 100–150 million years the helium is exhausted and the contracting core triggers a re-expansion of the star (D1 ). This AGB star ejects the rest of its envelope and what’s left becomes a white dwarf (E ). In dense stars clusters (or in close binary systems) the red giant’s envelope is stripped off ( C2 ) leaving a sub-dwarf B (sdB) star. If it has enough mass, helium ignites producing an extreme horizontal branch star (D2 ); if not the sdB star becomes a helium white dwarf. This sdB star burns helium for a 100– 150 million years before its helium fuel is exhausted. The remnant then quietly shrivels as a white dwarf. A few sdB stars have enough hydrogen to briefly re-expand as AGB-Manqué stars (D3 ) before their final transfor- mation is complete. The size of each star is not drawn to scale to become red giants. Even so, these will soon thereafter shed this skin to expose a low mass helium-rich white dwarf. The smallest and most numerous red dwarfs, born with less than a fifth the mass of the Sun, grow neither red nor giant. Instead these stars flirt with expansion during their extended middle age, but as they age and run low on fuel these stars grow progressively hotter. These parsimonious stars effectively skip the red giant phase and move straight towards the domain of the white dwarfs. After anything up to 10 trillion years, these red dwarfs become white dwarfs and slowly fade from view. They are the only stars that will spend most of their visible lives burning hydrogen. The rest spend 90–99 % of their lives shriveling and fading away as white dwarfs. Adventures in Stellar Evolution 53 Corpses

As we’ve seen, when a star dies, in most instances it leaves a corpse behind: some sort of dense, inert ball of matter that slowly fades from view. We can quibble over the true nature of those remnants made by the most massive stars—the black holes. However, even these are effectively inert and do not directly contribute to the fate of neighboring stars directly. It is only when these objects find themselves in very close proximity to others that they can influence their future. This can be through two mechanisms. If the remnant finds itself close enough to another star to begin stealing its mass, it can profoundly affect the fate of the living star, and in some cases find itself becoming a rather impressive firework in its own right. Chapter 5 looks at some of these shenanigans in more detail. However, even when the interaction is more distant, the long arm of gravity can have profound effects on the star as a whole, accelerating and potentially ejecting it, not only from the star cluster, but potentially from the galaxy as a whole. Many such hypervelocity stars have been discovered that are speeding away from the galaxy at over 100,000 km/h (65–100 km/s). Very recently a hypervelocity cluster was spotted speeding its way out of the nearby massive galaxy M87, in the constellation of Virgo. Its origin probably lies with a close encounter with the super-massive black hole at this galaxy’s heart. The fate of these stars is to live out the rest of their days in splendid isolation. It is hard to determine just how many lone stars there are, drifting through the wilderness between the galaxies of the Local Group. However, in the Virgo cluster a rather large cohort of lone stars and clusters are known. Some of these will be hypervelocity stars, while others will have been ripped from their parental nest by violent collisions between the galaxies of the cluster. As the Virgo cluster is over 50 million light years away, only the brightest stars—predominantly red giants— have been discovered. You can expect that for every bright giant there are a hundred smaller and less evolved stars lurking in the shadows. 54 The Complex Lives of Star Clusters Cosmic Recycling

When stars die they shed much of their mass into interstellar space—or in the case of hypervelocity stars, intergalactic space. The precise amount varies from star to star and is dependent on three main factors. The first is the mass of the star. The more massive a star is the stronger its stellar winds blow—and these also shed a vast mass of material when they go supernova. A star with ten times the mass of the Sun will lose around nine solar masses of hydrogen and helium, enriched with elements the star has forged during its life. A 100 solar mass star will lose perhaps 90 solar masses in such a way—much of this before it has had time to leave the main sequence. Smaller stars like the Sun shed around half their mass, mostly towards the end of their lives. Aa star with one-tenth the mass of the Sun may shed less than one hundredth of its mass throughout its prolonged life. For a star with a mass above ten Suns, all of this mass loss happens in less than 15 million years and tends to accumulate in plumes that extend above the galactic plane. In earlier galactic epochs, the halo of the galaxy would have been pockmarked with hot, rapidly expanding bubbles of gas vented from these stars. Gas flow isn’t even and observations of star-forming regions like 30 Doradus show hot voids pock-marked by tendrils of denser, colder gas that are still, in some cases, condensing. Lower mass stars shed very little mass as they complete circuit after circuit of the galactic centre. Yet when they become red giants most of their mass is shed in pulses that last a few million years. With much less energy, this gas tends not to disperse as widely as that shed by massive stars. As the elements made by stars also varies with their mass you might expect strong regional variation in the chemical make-up of stars that form in subsequent variation. But interestingly, the picture is quite mixed. Almost all first generation stars in the galactic halo have the same composition, indicating they were born in an evenly stirred pot. Within that population, some stars show differences in the amount of elements that suggests there were pockets of gas with a composition more or less enriched in heavier elements. This can be explained in Adventures in Stellar Evolution 55 a number of different, yet related ways. For example, the generally mixed composition suggests that supernovae—particularly Type II, or core-collapse supernovae—spread their debris far and wide, and that these were further mixed by the movement and collision of gas clouds in the halo. So-called s-process elements are produced over longer periods of time and are spread less widely when these more sedate stars die as red giants. Pockets of gas, enriched in s-process elements, will tend to accumulate where less massive stars have been born. The end result depends critically on how many stars are formed during the time when clouds are still collapsing in the halo and, shortly thereafter, into the early galactic disc. Many models have been produced over the years such as those by Claudia Travaglio and colleagues at the Max Planck in Heidelberg. None of these entirely match observations. However, given the complexity of the processes involved, this is not particularly surprising. What is apparent, however, is that the processes that shaped the halo of the Milky Way appear to be fairly universal and have left the same chemical fingerprints elsewhere. Within the globular clusters an even more complex pattern emerges. Ten to twenty percent of the stars appear to be first generation stars with a chemistry matching the isolated halo stars. The rest appear to have formed a little later on and can have very different chemistries. How, this came about is still under fairly fierce debate.

Conclusions

Stars lead rather exotic lives, transforming from one type to another under a directive that is driven by their mass and composition. Although all born in an equivalent way, much like people, what they inherit at birth determines how they will live out their subsequent lives—and how they will ultimately fare. Like people, a star’s life has only two certainties: death and taxes. Death gets us all in the end: the only thing that varies is the manner in which death gets us. Some live precocious lives, producing an abundance of elements before dying violently. Stellar winds exact a heavy tax throughout their lives, stealing away 90 % of their mass. Others take a gentler path from cradle to grave, burning their fuel gently; driving forward to a glamorous, colorful end as planetary 56 The Complex Lives of Star Clusters nebulae. The smallest stars are miserly, simmering quietly away in great numbers in the shadows of the universe. While these embers eke out a living, only the barest whisper of gas escapes, leaving them effectively unchanged over periods of time that dwarf the present age of the universe. Over time the universe recovers not only much of its initial investment in the stars it has nurtured, but also gains an increasingly sophisticated repertoire of heavier elements that not only add to the mass of subsequent stellar generations, but also enriches them in elements that can alter the ways in which they evolve. 3. Variable Stars

Introduction

Variable stars and star cluster go together like cheese and wine. In Chap. 1 , we saw how Shapley used RR Lyrae variable stars to map out the general scale of the galaxy; and later how Hubble used Cepheids to show that M31, the Andromeda galaxy, lay distant to our galactic shores. Stellar variability is integral to our understanding of how stars work and this chapter will explore these idiosyncratic stars and what they tell us about the evolution of stars, as well as how star clusters are assembled and ultimately die. Globular clusters hold a rich variety of variable stars and the younger open clusters more still. In order to map these out, each type of variable star will have its journey traced from cradle to grave. For the majority, an exploration of their fates within globular clusters is the most useful path. For the few that are confined to young clusters, their evolution will be followed separately. In broad terms, an individual star may encounter all of the regions of instability and variability, or none at all. What happens to each star depends crucially on the star’s mass and to some extent on its chemical composition, or chemotype. We will look first at variable stars in a young, massive cluster; then move onto variable stars in globular clusters. However, it must be clear that the latter types of variable stars are found not only in globular clusters, but in open clusters (of the right age) and the galactic disc as well.

© Springer International Publishing Switzerland 2015 57 D. Stevenson, The Complex Lives of Star Clusters, Astronomers’ Universe, DOI 10.1007/978-3-319-14234-0_3 58 The Complex Lives of Star Clusters Low and Intermediate Mass Stars from Birth Through Middle Age and Death

Main sequence stars, with masses between half and twice the mass of the Sun, are found in class F (yellow white stars) and the coolest end of class A (white stars). These lie in a region on the HR diagram called the instability strip (Fig. 3.1 ). In this region temperatures rise to around 50,000 K a short distance below the star’s photosphere. At these temperatures, helium loses one of its two electrons to become (singly) ionized. This acts something like the lid on a pressure cooker. When the helium is ionized the liberated electron can take up energy from passing photons and effectively trap it. The extra energy causes the layer of plasma containing the ionized helium to inflate and the star expands and cools. As its temperature drops, the electrons lose sufficient energy that they are grabbed by the helium ions.

Instability driven by ionization of iron RV Tauri Classical Cepheid Stars Variables

β-Cephei Variables Mira Variables

Type II Cepheids: Semi-regular W Virginis (SR) Variables

RR Lyrae δ-Scuti Variables Variables Hydrogen-Burning Main Sequence ZZ Ceti Variables (White Dwarfs)

Instability Strip

Instability driven by Instability driven by ionization of helium ionization of hydrogen

FIG. 3.1 Variable stars and the HR Diagram. Stars in the instability strip pulsate because of ionisation of helium just under the surface of the stars. β-Cepheids pulsate because of radiation periodically trapped and released by a layer of partly ionised iron, some distance under the surface of the star Variable Stars 59

This makes the gas more transparent to radiation, which then can stream out of the star. The star then contracts; the layer heats up again; helium becomes ionized—and the process repeats. This constant see-saw effect causes regular and very predictable changes in luminosity for all stars in this region of the HR diagram. A band stretching from the white dwarfs through the main sequence to the supergiants is affected. This is the reason for the name “instability strip” in this region of the HR diagram (Fig . 3.1 ). As the stars become larger the band tilts over towards slightly cooler temperatures. This makes affected white dwarf and main sequence stars somewhat hotter than the equivalent giant and supergiant stars. Along the strip there are regions of classical variable stars. Instability strip white dwarfs are called ZZ Ceti stars; in the main sequence they are called delta-Scuti stars; in the lower mass region of the giants they are referred to as W Virginis, or Type II Cepheid, stars; while at the more luminous regions, the instability strip creates classical Cepheid stars. A region of the HR diagram which is rather unique to old star clusters also lies in the instability strip. These are called RR Lyrae variables (Chap. 1 ). All of these have the same “helium-valve” driving pulsations which we detect as variability. A similar mechanism operates at the coolest end of the giant and supergiant branches on the HR diagram. Here, it is the ionization of hydrogen rather than helium at much lower temperatures that drives the variability. In this region there are two regions of variable stars: the semi-regular (or SR) variables and the Mira variables, named after the prototype Mira (o-Ceti) (Table 3.1 ). On the hot side of the Mira variables, and probably falling within the classical instability strip, is another set of variable giant stars, called the RV Tauri stars. These are all highly evolved giant stars that are transforming into white dwarfs. Together, these variable stars form something of a sequence which we will explore through the metaphorical eyes of a star as it ages from pre-main sequence towards its death as a white dwarf. a

FIG. 3.2 Different routes to instability. (a ) The star has a thin layer, lying just under its surface, where hydrogen, helium or metals can be ionised. Ionization traps energy causing expansion and cooling. In turn, cooling causes the loss of ionisation and the release of trapped energy. This then causes contraction, trapping of energy and re-heating, and so the process continues. Hydrogen ionisation occurs in cool, M class giants and supergiants; helium ionisation forms the classical instability strip, while metal ionisation may drive pulsation in very hot main sequence (β-Cephei) stars. In b the generation or release of energy from the core is unstable, which causes the star to pulsate or erupt. This process may occur in massive stars that have left the main sequence. Pair instability is one type of nuclear instability that may occur in very massive stars. In c the transport of energy by convection is unstable and in very wide giant and supergiant stars, changes in the transport of energy by convection may become visible. The variability in some semi-regular variables, such as Betelgeuse, may be caused by convection Variable Stars 61

Table 3.1 Variable stars in clusters older than 100 million years old Type of variable Magnitude (single range Spectral stars) Period (visible) class Evolutionary state δ-Scuti 1–2 h 0.003–0.9 Mid F to Main sequence or late A early post main sequence W Virginis 1–4 days 0.3–1.7 Mid F to Transition to RGB, (Type II Early K helium burning Cepheid) or transition to AGB RV Tauri 30–1,000 Up to 4 F or G Low mass, days changing post-AGB depending cyclically on to K or M sub-type Pulsating K 2–4 days 0.005– K class Helium burning Giant (although 0.015 giant? one example maybe 26 h) Semi- 20–350 days 2 Late K to Low and regular (43–97 in early M intermediate (SR) globular mass RGB or Variable clusters) AGB stars Mira 200–1,000 1–4 Early to late AGB Variable days M RR Lyrae Several 0.6 A-late F Horizontal branch, hours to 1 burning helium day ZZ Ceti 100s–1,000s 0.01–0.03 Early B to Cooling white of seconds Early A dwarf Most of these variables are seen in both old open and globular clusters. However, RR Lyrae stars are more common in globular clusters and RV Tauri stars are only seen in globular clusters (Population II stars)

The Music of the Stars

Astronomers use two key terms when describing the pulsation of stars: fundamental mode and first overtone. For those of us with a largely musically-illiterate background, the fundamental mode (or sometimes frequency) is the lowest frequency that a chord or 62 The Complex Lives of Star Clusters column of air will vibrate in when excited. The first overtone is the next frequency above the fundamental mode that vibration will occur. Stars tend to vibrate in the fundamental mode or the first overtone, and only sometimes in both. The RR Lyrae stars come in two sub-classes, RR Lyrae ab and RR Lyrae c, which prefer one or the other mode of vibration, but never both. The nature of the vibration, or pulsation, reveals the inner working of the star and is of fundamental importance in astrophysics. Sound waves act like seismic waves on and in the Earth. Waves of different kinds take specific times to cross through the Earth’s interior and some will only pass through some layers but not others. Similarly, by observing how pulsations make the star as a whole change in brightness, astronomers can piece together a profile of the star’s interior by modeling how sound waves propagate through it. The RR Lyrae stars paint a very interesting portrait of stellar evolution and point to some interesting effects that the star’s chemistry has on its structure and evolution. There is much more on this later in this chapter. Before we get there we need to take a star through its life. When it reaches the RR Lyrae phase we will look again at how stellar music illuminates (or perhaps sonicates) the interior structure of these stars.

A Varying Journey Through Time

In fits and starts a star begins its journey through life. Initially shrouded within the cloud that gave birth to it, its nuclear fires are only barely lit. This star lurks inside a dark nebula much like the Taurus molecular cloud. Jets of matter periodically erupt from its magnetic poles and shoot several Earth masses of gas into the surrounding nebula. Driving these is the erratic accretion of gas onto the star’s equator through a declining disc of matter. As the hot, gaseous matter crashes onto the star’s photosphere, copious amounts of ultraviolet light flare outwards. The angular momentum of the in-falling material is transferred to the star and out into the jets, keeping the star spinning fast, but not so fast that it breaks up. Variable Stars 63

T Tauri Stars

As clouds drift to and fro around the star, the star brightens and fades from view. From Earth we see it, barely, as a T Tauri star. Astronomers know that within 10 million years, the nebula will have dispersed and the T Tauri will begin to settle down. With a little over 1.5 solar masses of gas, this star will become an F-class star, positioned squarely in the middle of the instability strip.

Delta Scuti Stars

F-class main sequence stars show only small variations in brightness as their pulsations are relatively small. These delta Scuti (δ-Scuti) stars are very abundant in the Milky Way, coming second only to white dwarf variable stars in number. Pulsations have a period of about 1–2 h, with a variation of 0.003–0.9 magnitudes. The extent of the variation changes with the period: those showing the largest brightness variations are the largest stars and take the longest to pulsate. This creates a period-luminosity relationship and is used to estimate distances to clusters or nearby galaxies, such as the Magellanic Clouds (Chap. 1 ). However, these stars are only bright enough to be visible over relatively short distances and are too dim to be visible over millions of light years. For these distances, much brighter stars are needed. Delta Scuti stars show two different kinds of pulsation. For one, the stars pulsate as a whole: these are called radial pulsations as they affect the star in every direction around its core. In addition, delta Scuti stars also show non-radial pulsations, as the star wobbles something like a jelly. These variations may be driven in part by convection in a very thin layer lying above the helium- valve. As the helium layer changes in depth and opacity (how transparent it is to radiation) it will affect the depth of convection, which is fairly meager to begin with, in these stars. Altering the depth of convection will affect how efficiently energy is trans- ported causing the star to vibrate. Within globular clusters, stars are hotter for any given mass than their counterparts in the galactic disc. This is because they are poorer in metals and so are less “puffed up” by radiation coming out from their cores (Chap. 1 ). These clusters, plus the 64 The Complex Lives of Star Clusters older parts of the Milky Way (and other galaxies), contain their own class of δ-Scuti-like stars, called SX Phoenicis variables. As they are slightly smaller and less massive than classical δ-Scuti stars, their periods of pulsation are shorter, on the order of 42–114 min with brightness variations of about 0.7 visual magnitudes. In globular clusters there are no remaining stars of the appropriate mass that formed with the cluster. Instead all SX Pho variables are blue straggler stars, stars that have grown more massive through one or more mergers with neighbors (Chap. 6 ). The extra mass positions them within the instability strip. Other than that they follow the same rules: their period and luminosity vary with one another. The brighter they are, the longer the period of pulsation. At the hot end of the instability strip are a number of rarer pulsating stars called rapidly oscillating or, ro-Ap, stars. The “A” in their name comes from their spectral class, while the “p” demarcates their pulsation. These stars have a very thin layer of convection immediately under their photospheres in which helium ionization periodically occurs. Any hotter than this and the star will have ionized helium throughout its mass: pulsations then cease. RoAp stars have miniscule magnitude variations that vary over a matter of minutes, making them tricky objects to characterize. The standard delta Scuti star spends about 1 billion years on the main sequence, quietly changing hydrogen into helium. As its core becomes richer in helium, it contracts, grows hotter and burns fuel more vigorously. The star steadily heats up and expands. To an observer, the star’s period of pulsation will grow steadily longer as its girth increases. As the two go hand-in-hand the period-luminosity relationship continues to hold. Once about 10 % of the star’s mass has been turned into helium the star exits the main sequence and briefly becomes a sub-giant. It may still be a delta-Scuti variable at this point, but as it steadily expands and cools, the region in which helium is at the Goldilocks temperature to trigger pulsations deepens. By the time the star has cooled to a temperature similar to the Sun, pulsations weaken and become effectively invisible to astronomers. Variable Stars 65

Exit Stage Left

Over the ensuing hundred million years, the star expands to become a red giant. The core of the giant is growing ever more massive as it is fed by a furiously burning shell of hydrogen fusion surrounding it. As the core grows hotter, the outer layers expand and cool, while the previously thin shell of convection bites deeper and deeper towards the core. After nearly 500 million years the star approaches the tip of the red giant branch. Its outermost layer has chilled to a cozy 4,000 K. Hydrogen now emulates the ionization process that helium performed half a billion years earlier. In a narrow band under the photosphere hydrogen can ionize and begins trapping radiation from the fiery core deep underneath. The star expands in a large pulse that lasts between 20 and 350 days. At its peak, the star is nearly two magnitudes brighter. As before, with expansion comes cooling, and the hydrogen nuclei recapture their wandering electrons. The opacity of the layer falls; the star contracts and heats up; and the process begins again. While en route to the tip of the red giant branch a few stars become unstable while still hotter, orange, K-class stars. These are relative newcomers to the scene of variable stars. Relatively few are known and exactly where they are in their evolution is unclear. Some may be older stars which have already exhausted their supply of helium, while many more may be stars on their first ascent to become red giants. Perhaps the most famous is one of the first identified: the bright giant star, Arcturus, in the constellation of Bootes. Peter Edmonds and Ronald Gilliland (both of the Space Telescope Science Institute) discovered a handful of these K-giants in 47 Tucanae in the mid-1990s and carried out extensive charac- terization of them in an attempt to better understand what sorts of stars these are. Once again, the similar age and chemical composition of the stars in the globular cluster made analysis of these stars easier. Pulsing K-giants have periods lasting between 2 and 4 days, with one perhaps as short as 26 h. Each variable of this type has very low amplitude: their magnitudes vary by far less than a tenth of one magnitude—typically between one hundredth and one thousandth of a magnitude. The combination of relatively long 66 The Complex Lives of Star Clusters period and miniscule variation in brightness explains their belated discovery in the mid-1980s. Edmonds and Gilliland concluded that these K-giants were a mixture of first and second ascent (AGB) red giant stars. Arcturus, itself, is thought to be a helium-burning giant, but because it is richer in metals than the 47 Tuc stars, it lies closer to the red giants than the horizontal branch stars found in globular clusters like 47 Tuc. The underlying cause of their instability isn’t known. It might be convection; surface features, such as star spots rotating into and out of view; or pulsations driven by something other than helium. Perhaps a combination of these explains their unusual properties. The best match is with some sort of pulsation in the first or second overtone, perhaps driven by the ionization of hydrogen, although this doesn’t match the one star identified in 47 Tuc that had a period of 26 h: this is simply too short. Pulsations driven by sound waves might do the trick here. These will be driven by convection within the star’s outermost layers. Whatever the cause, the variable orange giants provide a bridge between the larger, cooler red giants, with long periods, and the smaller, hotter main sequence and helium-burning phases. Continued observation is needed to better understand how these variable stars originate and how they fit into the broader scheme of stellar evolution.

Semi-regular Variable Stars

The coolest red giants pulse inwards and outwards with periods of hundreds of days. These pulsations are called radial pulsations as they affect the whole surface of the star roughly evenly. On top of this radial pulsation, these red giants also have erratic but powerful convection currents surging upwards from their cores. Karl Schwarzschild, best known for his work on black holes, proposed that such bubbling motion might account for some of the variability seen in these semi-regular variables. Indeed the theoretical timescale of this variability matches the observations for some of these stars: around 200 days for convection to cart energy from the core to the surface and back again. However, it seems clear that most semi-regular variables also pulsate and that these pulsations are driven by a hydrogen ionization-valve, much like the classical helium valve on the instability strip. Variable Stars 67

SR variables are sub-divided into a variety of smaller groups based on chemistry, spectral class and period. Within globular clusters the SRD sub-class is most abundant. M13 contains a dozen or so SR variables, with periods around 43 days up to 97 days. Those with longer periods are both more luminous and larger than those with short periods. Patricia Whitelock (South African Astronomical Observatory) carried out analysis of SR variables around 30 years ago. She identified a period luminosity relationship, but one which was dependent on the metal content of the stars. So within a cluster (or a group of similar clusters with similar metal content) SR variables followed a line. At a luminosity of around 3,000 Suns and a period of around 350 days the SR variables vanished. It wasn’t that they were dropping dead; rather they were becoming something else: Mira variables. The data was telling Whitelock that SR variables were of two types in some clusters, but of one type in others. In more metal-rich clusters and the galactic bulge SR variable stars could be found towards the tip of the red giant branch: i.e. stars that were yet to burn helium. However, in other star clusters, those with less metals, SR variables were only found on the second red giant branch, the asymptotic giant branch (AGB: see Chap. 2 ). SR variable stars were only found above a set luminosity and stars in metal-poor clusters simply never got this bright before they ignited their helium cores. In effect, at the tip of the red giant branch, the metal-poor stars were too hot but dim to reach the region where hydrogen ionization had its effect. These stars never became cool enough to pulse. Therefore, in many globular clusters the SR variable stars are all much older and more evolved than their metal-rich brethren elsewhere. We shall return to these later in the chapter, after a horizontal excursion…

The Horizontal Branch

At the tip of the red giant branch all suitably massive red giants fire up their helium cores. This happens at around 2,000–2,500 times the luminosity of the Sun (Chap. 2 ). This convulsion shatters the core and mixes hydrogen down into it, while helium is mixed up 68 The Complex Lives of Star Clusters into the envelope. The hydrogen burning shell that has kept the star going, is belched upwards until it is too cool to operate efficiently. Although firing on all cylinders below, the total amount of energy helium fusion creates falls short of that pumped out by hydrogen fusion. Bereft of a suitable energy source, the giant collapses inwards before stabilizing at 30–50 times the diameter of the Sun. These helium-burning giants have surface temperatures of around 4,000–5,000 Kelvin and luminosities around 100 times that of the Sun. If that was it for our little star, the story of instability would pause here for around 100 million years, while its helium was fusing to make carbon and oxygen. However, in many star clusters, particularly the globular clusters, things take a stranger turn—or more precisely have taken a turn by the time the helium is lit. In many globular clusters, and to a lesser extent in open clusters, the stars on the horizontal branch are far lighter than their immediate predecessors on the red giant branch. Typically, stars ascending the red giant branch in today’s globulars have masses approximately 80 % that of the Sun (0.8 solar masses). However, stretching across the HR diagram, there are hundreds of very hot, low mass helium-burning stars. These stars could only have come from red giants that were at least as massive as the red giants that come before them. Without going into the details of their catastrophe (which is outlined in Chaps. 4 and 6 ), these stars have had much of their outer layers removed by other stars. This is very unfortunate for the star, but great for astronomers. For not only does it reveal some of the inner workings of dense star clusters, but it converts rather dull, helium burning red giants into rather small, yet perfectly formed curiosities called RR Lyrae stars. About 90 % of all variable stars in globular clusters are of the RR Lyrae type. Each has a mass of around 50–65 % that of the Sun with a luminosity around 40–50 times that of the Sun. They are also about 40–50 times the Sun’s diameter—approximately half that of untainted, helium-burning red giants. These stars also lie in the instability strip and like their kin also pulsate, this time with a period of several hours to a day. RR Lyrae stars were discovered by Williamina Fleming at the Harvard Observatory in 1901. Fleming also famously characterized thousands of other stars which were used in both stellar classification and in the later synthesis of the Variable Stars 69

HR diagram. As Chap. 1 described, RR Lyrae variables were central to our understanding of the galactic distance scale. Shapley was to later use them to map the structure and scale of the Milky Way. As was briefly mentioned earlier RR Lyrae stars come in two sub-classes: RRab and RRc. The RRab sub-class stars are cooler and somewhat larger and pulsate in the fundamental mode, while the smaller, hotter and less massive siblings in sub-class RRc vibrate in the first overtone. The pulsations of the RRc sub-class also produce largely sinusoidal patterns, while the RRab sub-class has an asymmetrical light curve. This effect was predicted in stellar models produced by Robert Christy (Caltech) as early as 1966, clearly well ahead of observations that would later confirm it. There is no overlap between the RRab and RRc sub-classes: they always form distinct groups on color-magnitude diagrams, indicating that it is the depth and mass of the envelope that contributes to the difference in pulsations. The different sub-classes of RR Lyrae stars also allow astronomers to discriminate between different clusters. The Oosterhoff Dichotomy is a division of star clusters into two groups, Oost I and Oost II, based on the proportion of stars in each of the RRab and RRc sub-classes. Each cluster type has stars that have different mean periods as well as differing ratios of numbers of RRab to RRc sub-types. It turns out that the Oost I clusters were more metal-rich than those that fall into the Oost II group. Globular clusters such as M3 have red-only end horizontal branch stars so that the few RR Lyrae stars they do contain are all redder, larger and have longer periods (RRab class) than the equivalent RR Lyrae stars from M15. M3 is an Oost I class globular cluster, while the bluer M15 is a Oost II cluster and contains predominantly RRc class RR Lyrae stars. Thinking back to Chap. 2 , the presence of metals (elements heavier than helium) makes the star redder and cooler than the equivalent metal-poor star. Thus M3 with more heavy elements, has a redder, cooler horizontal branch than M15. RR Lyrae stars therefore prove their worth not only as standard candles but also as tracers of stellar chemistry, thus opening up further avenues of investigation into these glistening balls of light. Some RR Lyrae stars—the prototype RR Lyrae is a case in point—show an unusual and, as yet, unexplained pattern of variability called the Blazhko effect, named after the Russian astronomer 70 The Complex Lives of Star Clusters

Sergey Blazhko. Stars exhibiting this phenomenon show regular changes in the strength of their pulses; the phasing of the pulse cycle; or sometimes both the strength and phase of the cycle are altered. Thus the light curve changes in a predictable manner from cycle to cycle. For RR Lyrae the Blazhko period approximates 39.1 days. Given the structure of these stars, a number of possible causes exist. These could be alterations in the production of energy in the core or in the thin shell of hydrogen surrounding it, or the flow of energy through the envelope of the star. Whatever it is, the mechanism is very regular. In almost all globular clusters, astronomers can find RR Lyrae stars. In the Large Magellanic Cloud (LMC) the exception is the unusually blue cluster Hodge 11 (Chap. 7 ). Unusually blue, because the cluster is more than 11 billion years old—and old usually means red. The horizontal branch is more of a blob, and one located at much higher than normal temperatures (15,000– 40,000 K). Such helium-burning stars are simply too hot to fall into the strip so no RR Lyrae stars are found here. In most globular clusters there is a pronounced gap near the region RR Lyrae stars are found. This is called the RR Lyrae gap and it is somewhat peculiar—and a little bit misleading. In astronomical circles RR Lyrae stars were both informative and annoy- ing. Placing a star on the HR diagram that wouldn’t stay put was tricky. Therefore, many RR Lyrae stars were simply left out—leaving a bit of a hole on the horizontal branch. However, this is not the full story. Although some of this gap is down to astronomical preference, in many clusters there truly is a gap near where the RR Lyrae stars lie. Gaps in the HR diagram usually mean that stars evolve quickly across these regions. Whether this is the case, or whether it is down to some other reason, remains unclear. Aside from peculiar gaps, many other globular clusters have red ends but not blue and, as we’ve seen with Hodge 11, some had only blue horizontal branches. Thus the HR diagrams of many globular clusters have not so much a horizontal branch, but a twig and a floating leaf. The gap lies with stars with an absolute magnitude of about zero and a temperature of 7,000–9,000 K. It is within that gap that the dense RR Lyrae variables fuse helium to carbon in a core with a size similar to the planet Earth. Lying above this is a thin shell of hydrogen in which some nuclear Variable Stars 71 reactions are occurring, but overall this layer contributes very little to the energy budget of the star. As RR Lyrae stars age and consume more and more of their fuel, the star slowly expands and rises gently up above the horizontal branch. The period of pulsation increases with the increase in size and luminosity once again, maintaining the period-luminosity relationship for these stars. Most RR Lyrae stars will not return to the red giant branch: they simply lack sufficient hydrogen in their outer layers to trap radiation and inflate the star once its helium core is spent and contracts once more. Such stars dawdle around the HR diagram above the horizontal branch, before moving towards the top of the region where white dwarf stars are born. These fairly hot stars are called AGB-Manqué stars (Chap. 2 ). After a few thousand years, their denuded reserve of hydrogen is exhausted, their core filled with carbon and oxygen and their lives end. What remains of these former giants moves rapidly to the left of the HR diagram and onto the top of the white dwarf track at around 80,000 K with a luminosity a few hundred times that of the Sun. Bereft of fuel, these hot little stars cool rapidly and descend the white dwarf track towards more reasonable temperatures. It is unlikely many of the AGB Manqué stars will have enough of an envelope to form a planetary nebula, so their passing is a relatively unremarkable affair with the star simply cooling off after the last of its hydrogen is exhausted. Over the ensuing 5–10 billion years they cool down until they cross the instability strip once more. At temperatures of around 10,000 K these planet-sized ZZ Ceti stars pulsate with a period measured in minutes. They wander slowly across the strip until they are no longer hot enough to ionize what helium they contain. Life becomes a steady decline into inky blackness tens of billions of years hence. Yet, for the more massive horizontal branch stars the good times are not quite over. Within the depths of these stars the carbon oxygen core contracts and heats up. The remaining helium lies in a shell around it and now fires up. Over the next few million years the star expands and cools down. As it does so it passes through the instability strip, where the ionization of helium can drive pulsations. Indeed, it may have done this once before while en route to the horizontal branch. However, this step would have been very brief—perhaps a few thousand years at most and very few stars are ever caught here by observational surveys. 72 The Complex Lives of Star Clusters

More metal-rich stars in modern open clusters don’t move onto the horizontal branch at all, but rather form a clump mid- way along the red giant branch. These metal-rich stars may spend most, or all, of their time in the instability strip burning helium, but at lower temperatures and higher luminosities than the horizontal branch stars.

W Virginis Stars

As a dying giant passes into the instability strip waves of pulsation grip its outer most layers. Over a time period lasting anything between one and 50 days, the star pulses in and out. Although the period might lengthen somewhat as the star expands and grows brighter, it is unlikely to spend much of its life in the strip. Metal poor stars, with masses between 60 and 80 % of the mass of the Sun, are called Type II Cepheids. Those with pulsations lasting between 1 and 4 days are called BL Herculis stars, while those with periods up to 20 days are called W Virginis variables. These have spectral classes F6 to K2 and are clearly the brighter and probably the more massive, with larger diameters. The key difference may be the starting mass, or more likely how much mass these stars lose while they are on their first ascent of the red giant branch. Classical Cepheid stars have masses between three and nine times that of the Sun. These stars will cross the instability strip when they first leave the main sequence and make their very quick hop across to the red giant branch. Many more will return there when they are burning helium. Stars in the latter phase will have the longest period of time in the strip, but many of those will simply pop in and out of it when helium first begins to burn and then, later, when this fuel is exhausted. Both sets of stars, the Type I (classical) and Type II Cepheids obey period-luminosity relationships. However, they are not quite the same and it was this difference that led Shapley to mis-calculate the diameter of the galaxy. Shapley, if you recall, didn’t realize the distinction between each type of star. He took the dimmer Type II Cepheids as equivalent to the classical ones. Using intrinsically dimmer stars made him think that they were further away than they were (Chap. 1 ). Indeed, it wasn’t until 1942 that Walter Baade recognized the distinction between the two types when he analyzed Cepheid variable stars in the Andromeda galaxy (M31). Variable Stars 73

After a few million years, a dying giant star will have expanded and cooled further. The helium is all but exhausted in the shell around the inert carbon-oxygen core. Low metallicity stars at this stage will probably still be too hot to become semi-regular (SR) variables (see above). However, a more metal-rich star will have expanded and cooled enough to reach this region where hydrogen can become ionized some distance below the tip of the AGB. These stars will pulsate due to ionization but soon they will also pulsate due to changes that are underway deep inside their cores. By ana- lyzing the number of stars on the red giant branch and the number that were SR variables, it was possible to determine how long a star would remain in this stage. Patricia Whitelock showed that around 300,000 years was typical for these stars. They brighten and grow bigger as the core slowly grows in mass. After the luminosity has reached to around 3,000 times that of the Sun the SR phase gives way to the Mira phase. Mira variable stars are the most luminous stars on the Asymptotic Giant Branch (AGB). They are losing copious amounts of mass in stellar winds that remove up to a Sun’s worth of mass every hundred thousand years. Miras are the most luminous stars in globular clusters but are also found throughout the galaxy and the universe at large. Mira variables come in three different types, of which only one is found in globular clusters. Carbon-rich Mira stars are found in the galactic bulge, Large and Small Magellanic Clouds and elsewhere. They are relatively rare stars and often highly luminous. Oxygen-rich Miras are much more common and there are two types of these. Those rich in titanium oxide are the only ones found in globular clusters. The others, rich in silicates and zirconium oxides, are found throughout the galactic disc, including in some open clusters.

Mira Variables

Mira variables are formed from all stars with masses of at least 80 % that of the Sun up to around 7–8 times the Sun’s mass. The lower limit will be set by the amount of hydrogen a red giant holds onto on its first ascent of the red giant branch. Stars that lose too much mass end up as extreme horizontal branch stars. From here they either become AGB-Manqué stars or directly fall onto the 74 The Complex Lives of Star Clusters white dwarf region of the HR diagram. Miras pulsate with a period longer than 100 days and all show variations of at least 2.5 magnitudes in visible wavelengths and 1 magnitude in the infrared. Mira itself varies from naked eye invisibility up to a visual magnitude of +4. The variation in luminosity is slightly misleading: Miras expand and cool so that grains of dust form in their outermost layers. These grains both block visible radiation but re-emit this as heat (infrared). The overall amount of energy released doesn’t vary much at all, it’s simply that the amount of light emitted in the visible portion of the spectrum that varies. Mira variables also show emission lines of hydrogen. The origin of these may seem somewhat perplexing as emission requires a lot of energy and these cool stars might not be expected to generate such enthusiasm. However, Mira variables, much like other giant stars, pulsate with such vigor that they can generate shock waves deep within their extended atmospheres. As these shocks propagate outwards they excite hydrogen atoms to the point at which they emit light in the red portion of the visible spectrum. Mira variables have diameters around 250–350 times that of the Sun. This value varies both because the stars are pulsating every hundred days or so and because these stars undergo so- called thermal pulses (Chap. 2 ). During most of the star’s time as a Mira variable it is burning hydrogen into helium. Indeed, the thermal pulse phase begins this way once the core helium has been exhausted. After a few thousand years a layer of helium has been built up that is sufficiently dense to fire up. Helium then ignites explosively and burns for a few decades at most—less than the typical human lifespan. During helium fusion, the hydrogen burning shell is blown outwards to cooler regions and winks out. This means that the supply of helium is limited. Once exhausted, the star’s inner regions contract until the lowermost region of hydrogen is hot enough to begin burning once more. In principle each of these thermal pulses should cause the star to swell and contract with a period of a few tens to hundreds of thousands of years. Hydrogen fusion produces the most energy, making the star expand, while during the short period in which helium burns, the star will contract. This should lead to measur- able changes in the diameters of these star, but given the brevity of the helium burning phase—and because these stars are heavily Variable Stars 75 obscured in mire of their own creation—confirmation of this has not yet happened. A few Miras appear to change their pulsation period over time scales measured in decades to centuries, which indicates that the physical size of these stars is changing. It is just possible that these precious few stars are being caught in the midst of a thermal pulse, where the star is swapping its fuel source from hydrogen to helium and back again. As these stars continue to fuse hydrogen and helium in alternating waves, the star’s core grows slowly more massive and hotter and consequently makes the star expand and grow ever brighter. The rate of pulsation slows with increasing size, which could also be used to determine how evolved AGB stars were. What is clear, at the very top of the AGB strip, stars begin to physically shrink. Observations suggest that globular cluster AGB stars that lie at the tip of the asymptotic giant branch shrink from around 250 times the diameter of the Sun down to half this value. Their envelopes are now severely depleted in hydrogen and dust and the last phase of their evolution is fast approaching.

OH/IR Variables

At the very summit of the AGB, stars can morph into another two types of variable star. At this point in their lives they have lost so much material to their surroundings that they are often completely obscured and all but invisible except at infrared wavelengths. Energy from the star is caught up in a funeral wreath of dust and gas and re-emitted at longer, lower energy wavelengths. This infrared is still very intense and the star is still pumping out as much or marginally more energy as it was while a Mira variable. Indeed, at infrared wavelengths these are by far the most luminous stars in the sky. These stars are known as OH/IR variables. The “OH” part of their name comes from one of the most abundant chemicals in their shroud: the hydroxyl molecule—a compound made of one hydrogen and one oxygen atom. In Mira’s this can act to concentrate infrared radiation into masers: the microwave version of a laser, where microwaves of one or more specific wavelengths are amplified as they pass through this compound, generating coherent, broad beams of radiation. 76 The Complex Lives of Star Clusters

RV Tauri Stars

While the ascent of the AGB through the semi-regular variables takes a few million years, the passage through the thermally- pulsing Mira stage lasts perhaps a hundred thousand years at most. The final OH/IR variable stage lasts a few hundred years as the last of the star’s outer layers are expelled. After all but a couple of hun- dredths of a Sun’s worth of hydrogen and helium remains above the core, the outer layers become too thin and transparent to absorb the energy streaming out of the star’s core. What remains begins a century’s long period of contraction. As it does so, the surface temperature rises from 2,000–3,000 Kelvin up past 6,000 K. In this region a few stars become unstable and emerge briefly as RV Tauri stars. These are all low mass, metal-poor, Population II stars with 50–70 % the mass of the Sun. As they are Population II, they are abundant in the galactic halo, and far from uncommon in globular clusters. RV Tau stars are encased in a thick clumpy mass of debris shed in the previous red giant phase. RV Tau stars come in two, broad groups: RVa Tau stars, such as R Scuti, have no overall change in brightness from one maximum to the next. The other is RVb variety, of which RV Tauri is the prototype. These stars also show variation from one cycle to the next. All of these stars are highly luminous, each a few thousand times brighter than the Sun. The variability changes their brightness by up to four magnitudes, and some are greater than this. Over their cycles RV Tauri stars vary in color from yellow- white (spectral class F-G) at their brightest through to orange-red (K-M classes) at their coolest and dimmest. These changes take anywhere between 30 and 150 days to complete in RVa and RVb types. The RVb Tau stars have two overlapping cycles. The cycle from cool to hot—dim to bright—and back again lasts the standard 30–150 days. However on top of this cycle, RVb Tau stars also have a much longer cycle lasting 600–1,500 days. Over this cycle there is a super-imposed cyclical variation in the star’s brightness so that these stars have two maxima and two minima in their light curves. It is thought that the two types of RV Tau star are physically related. The RVa type have much thinner shells of dust around them than the RVb type. It is speculated that the RVb variety have generated dust close to the surface of the star and that this Variable Stars 77 dusty shell is then projected outwards, perhaps by a shockwave generated as the star pulsates. This shockwave eventually dissi- pates the layer of dust and the star morphs into the less murky RVa variety. Over decades this process repeats as more shells of dust are generated close to the star, only to be driven away in pulses. It is also thought that these Population II stars are mostly binary stars and it would seem likely that the binary partner also plays a role in shaping the flow of material away from these aging giant stars. Once again, the precise role remains unclear. All of this activity can’t last, particularly as these stars have very little material left to give back to the universe. These stars have only the breadth of a human lifetime, before they cast off their mortal coils as planetary nebulae. The RV Tau phase marks the final breaths of the star as pulsations drive away the leftovers of its outer layers. As the remnant star then continues to heat up, its surface temperature ascends to tens of thousands of degrees. Within 100 years the planetary nebula will take shape and the wan- ing nuclear fires will finally turn off. The pulsations of the final stages in the star’s productive life may be visible as shells within the planetary nebula. But soon, these will be erased as the hot, fast wind of the emerging white dwarf blows the funeral shroud away.

ZZ Ceti Stars

ZZ Ceti stars are white dwarfs with approximately half to three quarters the mass of the Sun. Their interiors are inert balls of carbon and oxygen, with a few made of pure helium, or a mixture of oxygen, neon and magnesium. They have a surface temperature between 7,000 and 10,000 K and lie at the lowest extreme of the helium-powered instability strip. These planet-sized stars come in a few sub-flavors depending on what kind of atmospheres they have. Those white dwarfs with atmospheres dominated by hydrogen are called DAV or ZZ Ceti stars; while those stars with a helium-rich atmosphere are called V777 Herculis stars. Those with a mixture of carbon, oxygen and helium in their atmospheres are called GW Virginis stars. All vibrate with periods of hundreds to thousands of seconds. The vibrations are concentrated within the hot atmospheres of these white dwarfs; the interiors of which behave essentially as solids. Variations in light output are small in most cases (a few percent 78 The Complex Lives of Star Clusters

Stable Upper Layer

Thunderstorm Boundary Atmospheric Gravity Waves Unstable Lower Layer

FIG. 3.3 An atmospheric gravity wave on Earth. This photograph was taken in the mid-west and shows a very dramatic billowing sheet of cloud. There is a layer of unstable air near the ground over which a mass of stable air has been set in motion, perhaps by a nearby thunderstorm or because it has been forced to rise over a mountain

differences) but a few show quite substantial changes with changes up to 30 % of their total output at visible wavelengths. ZZ Ceti stars are rather peculiar in the sense that they don’t pulsate in a standard sense. Instead of the surface of the star moving inwards and outwards, the surface wriggles with what are known as gravity waves. These waves are seen on Earth, often in the vicinity of strong thunderstorms. A strong updraft moving from one layer to another generates a wave in the upper layer, which spreads outwards. Atmospheric gravity waves make for some impressive skies over the mid-west of the US and on occasion in the UK when so- called Spanish Plumes arrive, bringing hot, thundery air (Fig. 3.3 ). In the white dwarf similar waves, perhaps caused by convection in the outer layers, or heat released from the core, drive waves within the thin, outer, gaseous part of the white dwarf. As all white dwarfs are born with temperatures above 80,000 K and must steadily cool into oblivion, all must pass through this region of the instability strip. Our little main sequence star is no exception. After more than a billion years, evolving from one form to another, and having made multiple passes through the instability strip, our star makes its final passage this strip and heads to a chilly grave 300 billion years hence (Table 3.2 ). Variable Stars 79

Table 3.2 Young and massive variable stars frequently found in star clusters with ages less than 100 million years Type of Visual magni- Spectral Evolutionary variable star Period tude range class state T Tauri Non- 0.5–1.0 Late M to Low mass Stars Periodic/ mid-F protostars Erratic class Herbig Ae/ Non- 0.5–1.0 Early B to Intermediate Be Stars Periodic/ early F mass Erratic class protostars Be Stars Minutes to Up to 1.5 Late O to High mass hours magnitudes Late A main with class sequence to longer blue term supergiant variability over months to years β-Cephei 2.4–13 h 0.01–0.3 Early to late Main sequence Stars magnitudes B class to early post-main sequence Rho Erratic with 0.5–1.5 visual F to G at Helium Cassiopeia monthly magnitudes brightest, burning stars to decadal but highly dropping supergiant drops in variable to K to M (or evolving visual at visually towards this output dimmest state) S Doradus Erratic low 0.1–0.3 Mid-late B Post-main amplitude magnitudes but down sequence variability but large to G after hypergiants, with large outbursts outbursts at varying outbursts are several evolutionary occurring magnitudes stages on millennial timescales (continued) 80 The Complex Lives of Star Clusters

Table 3.2 (continued) Type of Visual magni- Spectral Evolutionary variable star Period tude range class state Wolf Rayet Erratic Mostly less WN— Post main changes than 0.2 nitrogen, sequence, due to magnitudes hydrogen helium clouds of but with and burning or dust but rare helium; later with outbursts WO— larger, over several oxygen rarer magnitudes and outbursts helium known WC helium and carbon Type 1 0.5–41.4 0.5–1.0 Early F to Mostly helium (classical) days (RS magnitudes mid G fusing Cepheids Puppis) giants, post-helium fusing giant Technically the Be stars include the Herbig Be stars but are separated here for clarity as they are a very diverse group of objects. Many red supergiants are also semi-regular variables and these are included in Table 3.1

Young, Fickle and Massive

In the youngest clusters none of the previously described variable stars exist. There has simply been too little time for such stars to emerge. The δ-Scuti stars are the youngest of these and although these intermediate mass stars may emerge around the same time as a clusters most massive members, they will be effectively invisible in the glare of any massive siblings.

Herbig Ae/Be stars

A cluster less than 5 million years old will play host to a variety of massive stars, alongside a menagerie of intermediate mass stars and low mass protostars that are still condensing. Of the low and intermediate mass protostars some will appear as T Tauri stars and Herbig Ae/Be stars. The former are protostars with masses less than around twice that of the Sun. These will become stars of Variable Stars 81 classes M through K and G to F. The Herbig Ae/Be protostars have masses between twice and eight times that of the Sun. They are all less than 5 million years old—the time it takes a two solar mass star to condense from its natal cloud. As these stars are forming inflowing matter is fed towards them through a swirling accretion disc. As this crashes onto the surface of the star it generates copious amounts of X-rays which are detectable from Earth. These stars may also be magnetically active with further X-ray emission coming from flares. Herbig Ae/Be stars also are a common source of energetic jets which shoot out of the rotation poles of the star—undoubtedly guided by magnetic field lines generated either within the disc or the young protostar. These can stretch over 1 light year from the star and glow from radiation emitted by shockwaves moving through the jet, or from sites where the jets violently impact surrounding clouds of gas and dust. The T Tauri protostars, named after the prototype, T Tauri in the Taurus dark molecular cloud, lead longer lives than their more massive brethren, the Herbig Ae/Be stars. This is simply a function of mass. Smaller stars take longer to do anything in the universe than their more massive cousins. While the formation time of massive stars may be measured in tens of thousands of years, these smaller stars takes tens of millions to billions of years simply to get going. A star like the Sun took just over 10 million years to condense, but a much smaller star will take proportionately longer. As long as material still flows onto these stars, the same sorts of activity seen in Herbig Ae/Be stars will occur. All of these stars show strong emission of infrared radiation. This isn’t coming from the star, but from material in the accretion disc, which is heated internally by friction and by radiation coming from the central star. These sorts of emissions are strongest from the more massive Ae/Be protostars as both the supply of gravitational energy and the supply of gas and dust is greatest. Both Herbig Ae/Be stars and T Tauri stars may show variability for a number of reasons. The most prosaic is the intermittent obscuring of the central protostar by surrounding clouds of dusty gas—and this includes the proto-planetary disc that surrounds the star’s equator—if the disc is tilted towards our line of sight and precessing with the star. Changes in the position and density of 82 The Complex Lives of Star Clusters this material will obviously affect the amount of light that can reach us from the protostar. Changes in the supply of material to the central protostar will affect the amount of X-rays that are emitted as it collides with the surface of the star. Magnetic activity in the protostar will also affect X-ray emission, particularly from T Tauri protostars. The amount of star spots covering the surface of T Tauri stars, and the speed at which the protostar rotates, also affects how much bright surface is visible and for how long in each stellar “day”. Most T Tauri stars and Herbig Ae/Be stars rotate once per 24 h—considerably faster than our Sun does now. Finally, the amount of material in the jets, directed away from the protostar, affects the brightness of the jet and the amount of energy it emits at various wavelengths from X-ray to radio wave. All of these are very hard to predict and therefore these stars, unlike the other variable stars discussed earlier, have no fixed periods or matching patterns of variability. Each has its own particular fingerprint over the generic types of radiation these youthful protostars emit. Once accretion stops the stars settle down and their intrinsic variability comes to an end. The majority of intermediate and massive stars show limited variability while on the main sequence. There are, of course, exceptions. Hot B-class stars with masses between 7 and 20 times that of the Sun often show pulsations. Like their lower mass cousins, these are driven by a deep instability caused by ionization. In this case the element in charge is iron (or related metals, such as nickel, cobalt, copper and zinc). At roughly 200,000 K, iron atoms have lost most of their electrons, but the iron nucleus, with 26 positively charged protons holds more of an attraction for what remains than helium does for its lowly pair. A few of the remaining electrons can be exchanged in just the same way as hydrogen and helium do in low and intermediate mass stars. In this case, however, the zone of ionization lies much deeper within the star. Pulsations are therefore, relatively weak as much more mass needs shifting up and down in response to the iron-driven pulsations. Pulsations have periods of a few hours and only change the brightness by less than one third of a visual magnitude. However, given that these stars are intrinsically very bright to begin with, their effects are readily detectable. Most β-Cephei stars will evolve off the main sequence within 20 million years of birth—many a lot faster than this. Their fate Variable Stars 83 is to become red supergiants within a few tens of thousands of years of their central store of hydrogen becoming exhausted. The lower mass β-Cephei stars will cross the instability strip at least three times: once when the first become supergiants; the second time after helium is ignited and a third and possibly fourth time after helium is exhausted. In each crossing these stars will become classical Cepheid stars with periods of several days. On a distant cosmic shore an astronomer may be watching their pulsations and come to the realization that their home galaxy is but one of millions visible in a sprawling universe.

Be Stars

Once on the main sequence stars with masses from roughly twice to roughly 11 times the mass of the Sun may manifest variability associated with the presence of a disc of material excreted from their equators. These stars, known as Be stars (“B” denotes class, while “e” denotes spectral emission), show some emission lines of hydrogen as well as emission in the infrared portion of the spectrum. The first star identified as a Be star was Gamma Cassiopeia. This was observed by Angelo Secchi in 1866. Indeed, Gamma Cassiopeia was also the first star ever observed with emission lines. Both types of emission arise within the disc of material that orbits the star’s equator and is being fed from it. Infrared radiation comes from re-radiated heat from dust within the disc; while the emission lines come from fluorescence of hydrogen that is induced by ultraviolet radiation from the star. The discs are regarded as “Keplerian” in that they rotate around the star in broadly circular (strictly speaking, somewhat elliptical) paths. Be stars are relatively common, comprising roughly 20 % of all B-class stars. Well known examples include the main sequence stars Pleione and Achernar. However, the numbers vary according to location. For example within M54, the Pleiades, six of the seven brightest B-class stars show Be characteristics, Pleione included. Be stars show variability on periods from minutes to hours at some levels, but often with broader changes in luminosity over the course of months or years. Variability is associated with the movement of material within the disc and changes in its density (Fig. 3.4 ). 84 The Complex Lives of Star Clusters

FIG. 3.4 Shell stars in the Pleiades open cluster (M45). Each of the arrowed stars are 3–5 solar mass B-class stars that are rotating so rapidly (200– 300 km/s) that their shapes are distorted into discs. Be stars are further surrounded by extended discs of out-flowing matter. Studies by CoRoT also showed that their cores are around 20 % larger and more massive than they should be. The observed spectrum of these stars is influenced by the viewing angle (bottom of figure)

Be stars are quite a diverse bunch of objects and strictly speaking span a range of spectral types from O7 at the hot end (Oe stars) to A0 (Ae stars) at their coolest and displaying changes in luminosity up to 1.5 visual magnitudes. The hottest, O-class stars, comprising roughly 5 % of the total population, are often seen switching from Be-like spectra to more classical O or B-class spectra over relatively short timescales, suggesting that there are changes in the mass of the disc surrounding the star. The phenomenon is also seen in stars of different evolutionary stages. Herbig Be stars (above) are pre-main sequence, while some Be stars are post main sequence blue supergiants; the cores of very young planetary nebulae or symbiotic stars (see later in this chapter). The diversity of the phenotype (its characteristic) suggests that the disc can arise in many different ways. How can astronomers be confident of the high speed of rotation? For one, some stars are physically distorted. Seen edge on, Achernar is shaped more like a candy Minstrel than a soccer ball. Variable Stars 85

Other Be stars may be viewed at different angles and are hence less easy to ascribe a shape. However, when a star rotates its light is shifted towards or away from the red end of the spectrum, depending on which area is being viewed. Looking at a star as a whole, the rotation speed can be gauged by how the movement broadens spectral lines. This is known as Doppler Broadening . This technique can be used to determine the speed of rotation with a fairly high degree of precision and has been done for many stars for which there are clear spectra. The vast bulk of the material in a Be star’s disc is neither falling into or moving away from the star: it simply orbits it. Material is fed into the disc by stellar winds, or where the disc touches the surface of the star, from the stellar surface itself. Material leaves the disc, eventually from the combined effects of radiation from the star heating it and driving it off the disc, or from the directly erosive effects of the stellar wind. However, the origin of the disc must vary with the evolutionary state of the particular Be star. Herbig Be stars are still accreting material so the presence of a circumstellar disc is to be expected. The disc is feeding the star material from which the star is growing. In more mature stars the presence of the disc is linked to the rotation speed of the star: All Be stars rotate at a break-neck speed of 200–300 km/s. At this speed the star is very close to breaking up and is able to expel a disc of material. Yet this cannot be the full story. Many B-class stars show such mighty speeds of rotation but do not show the Be characteristics so something else must be helping the star expel material. How, then do some B-class stars manifest the Be behavior? In some stars there is evidence for a companion star—and given the B star’s moderately high mass, this is often a neutron star. The presence of the neutron star not only ensures that the Be star may have a high orbital velocity, but during the final stages in the life of the neutron star’s progenitor, it could have donated much of its material to the future Be star. Immediately before death, the progenitor of the neutron star would have been a red supergiant. As the B-star accreted material from its supergiant companion, it would have been spun up to high speeds. However, although this can explain many Be stars, many more are not part of binary systems. These could simply have been born with very high rotation speeds, or 86 The Complex Lives of Star Clusters perhaps are the products of mergers between two lower mass stars. Indeed, in the Small Magellanic Cloud cluster NGC 330 there is a significant population of Be stars—and what appear to be a population of relatively massive blue straggler stars that lie above the main sequence turn-off, the position where stars of a certain mass run out of fuel and leave the main sequence (Chap. 2 ). Other Be stars appear to possess relatively strong magnetic fields measured at a few hundred times that of the Earth. In these Be stars, the magnetic field might be responsible for cor- ralling part of the B-star’s strong wind and holding it within a torus around the star’s equator. This may be particularly true of those Be supergiants that have been shown not to possess companions. Joseph Cassinelli (University of Wisconsin) and colleagues suggested that Be stars generate a strong magnetic field in the plasma torus that surrounds the equatorial plane of the star. Richard Robinson (Catholic University of America & Computer Sciences Corporation) observed the 17M Be star γ-Cassiopeia, which is a copious producer of X-rays. Like other Be stars, this is rotating rapidly at 230–310 km/s. In cool stars the presence of X-ray emission is often a sign that they posses strong magnetic fields. However, since stars as hot as spectral class B cannot generate these fields by convection within their envelopes, where are they made? Cassinelli and his co-workers proposed that the disc forms not from material out-flowing along the equatorial plane, but rather that this is a stable disc generated in situ from material in the star’s wind. Plasma leaves the star in a strong stellar wind, but magnetic field lines channel the gas into a stable, orbiting disc, where rotation supports the contents against the gravitational pull of the underlying star. Eventually, plasma escapes the disc in the equatorial plane of the star where the field buckles outwards. The disc is then periodically replenished by matter flowing into it from the stellar surface. This scheme would explain the X-ray emission and the overall variability of the Be stars. Similarly, Robinson concluded that changes in X-ray output and overall variability from the Be star γ-Cassiopeia was also best explained by the presence of a magnetic field. However, the authors of this study did not state whether the field was likely generated in the Be star or in the disc that orbited it. Such a magnetic field is probably seeded in the star but is then amplified by shearing Variable Stars 87 motions occurring within the disc. At the inner edge of the disc (about two and a half stellar radii from the star) the magnetic field will tend to drag material back into the star. However, outside this radius, the field should impart enough energy to the disc to prevent it from spiraling inwards onto the star’s photosphere. Murugesapillai Maheswaran (University of Wisconsin) extended this theme and showed that massive stars are adept at developing relatively strong magnetic fields in such Keplerian discs. In B-type stars the field is of the order of 1–10 Gauss (between 2 and 20 times stronger than the Earth’s) but O-type stars may develop fields as strong as 500 Gauss through the same mechanism. These magnetic fields can clearly generate significant X-ray emission— and be a source of variability in emission across the electromagnetic spectrum. Still other Be stars exhibit pulsations akin to the β-Cephei stars (described shortly, in this chapter) and these might be responsible for driving material outwards. The latter are known as Lambda Eridani variables, after the prototype. Gamma Cassiopeia is an archetypal shell star , with a more pronounced disc. The classic shell stars do not show Doppler line-broadening despite high rotation speeds. This is because their disc is seen edge on and Doppler- shifted emission is hidden by the bulk of the disc. As a sub-group, even shell stars can be further sub-divided, based on their spectral classification.

Supergiant Variable Stars

Stars with masses above nine solar masses tend to skip the instability strip—at least in its formal guise. These stars are so massive that star zip across its summit towards the red supergiant phase so quickly that they don’t manifest any of its symptoms. That is not to say that they won’t experience its effects, rather that they will be so brief as not be apparent to observers. Once they reach their largest and coolest extents, they tend to pile up where hydrogen ionization becomes important. A number of red supergiants are cool enough for hydrogen ionization to drive pulsations. These stars, of which Betelgeuse is the best known, become semi-regular (SR) variables with periods measured in hundreds of 88 The Complex Lives of Star Clusters days (see earlier in this chapter). Like their lower mass cousins, variability is likely to be caused by two principle factors: convection within the outer envelope, which causes variability with a period of around 200 days; and physical pulsation of the entire envelope. On top of this, red supergiants are so physically dis- tended that gravity has a fairly weak hold on the stellar surface. This may cause it to distort in differing directions with convection, magnetic fields, the presence of companion stars (or planets) and the formation of dust as contributing factors. All red supergiants will eventually expire as core-collapse supernovae and there is just a hint that as the end approaches the most massive of these may undergo violent explosive outbursts which drive off parts of their outer layers. Eran Ofek (Weizmann Institute) investigated the progenitor of the supernova SN 2010mc and found that 40 days prior to the star’s decimation, it had ejected a tenth of a Sun’s worth of matter in a single blast. The eruption matched the predictions of two other astronomers, Eliot Quataert and Joshua Shiode, both of the University of California, Berkeley. Both astronomers had made models suggesting that the interior of a highly evolved and moribund supergiant may undergo violent oscillations as different nuclear fires burn up what fuels are available. And this star isn’t alone. SN 2006jp was preceded 2 years earlier by an outburst that jettisoned a helium-rich shell of material at thousands of kilometers per second. In 2009, SN 2009ip had a series of violent outbursts, each scattering matter at hundreds of kilometers per second. This was followed by a much meatier explosion, which was likely the supernova that ultimately claimed the life of the star. At the moment the interpretation of these events represents astronomers playing catch up with nature. Stars, like their more extravagant human counterparts appear to do odd and unpredictable things as they approach their end. At the moment for those closest to their ends, the source of all this trouble is unclear, but much further work is underway. Expect news soon. Somewhat more predictable, or at least understandable, is the temperamental behavior of stars with masses over 30 times that of the Sun. At the low end of the mass range are stars such as Rho Cassiopeia. This yellow hypergiant is in the midst of a transition from red to blue. Soon after it left the main sequence Variable Stars 89

Rho Cassiopeia expanded to become a red supergiant. Unlike the wimpier Betelgeuse, its core generated so much energy that it began to shove off its outer layers. As more and more material was expelled the underlying star grew hotter as deeper and deeper layers were exposed. In many ways, this parallels the demise of low mass stars at the end of the AGB phase. However, Rho Cassiopeia has much more fuel to spare and won’t die quietly. Once several Sun’s worth of material have been expelled Rho Cassiopeia should re-emerge as a hot blue supergiant, somewhat like the less massive Rigel (β-Orions). Perhaps a few hundred thousand years later, its fiery end will come as a supernova. The precise evolutionary stage of Rho Cassiopeia is unknown. It may be burning helium or just about to do so. It is thought that many of the more massive supergiants will execute what is known as a blue-loop when helium ignites. Other less massive stars undergo these loops when helium fires up or just after the main supply of this fuel is exhausted. A blue-loop is visible on the theoretical path for intermediate and massive stars. When helium ignites, the star trips backwards to the hotter, left of the main sequence—in a somewhat analogous fashion to the horizontal branch of low mass, Population II stars. The transition from red to blue may be quite dramatic for some stars. In 2008 Howard Bond (Space Telescope Science Institute) identified a transient source of radiation in the galaxy NGC 300. This source of energy, labeled NGC 300 OT, for o ptical transient), released nearly 1040 Joules of energy in an apparently non-lethal eruption. In the process, a thick shell of dusty material was driven outwards and the star increased in brightness to an absolute magnitude of −12: this is well in excess of a nova but shy of a standard supernova. The outburst was comparable to that of many Luminous Blue Variable stars in magnitude. Underlying the eruption was an object with 55,000 times the energy output of the Sun. This sort of output would be expected for a star with a mass 10–15 times that of the Sun and was eerily similar to another outburst observed in the same year SN 2008S. Although, bearing the label “SN” of a supernova, this, too, appeared to mark the expulsion of a large dusty shell of material from a fairly massive star, rather than the dramatic death of the star. 90 The Complex Lives of Star Clusters

Later observations of NGC 300 OT suggested that the star had expelled material in two dusty lobes, which were now moving away from the poles of the star. Although a supernova model had been suggested for SN 2008S, the best fit for both objects is a pair of stars that are executing blue loops after turning on helium fusion. If true, these chance observations are a very lucky catch for astronomers, as this evolutionary phase is not expected to last long in stars of this much mass. That it seems to occur with a dramatic burst of energy is perhaps all the more remarkable—but certainly useful when astronomers come to follow the brief lives of massive stars. Although NGC 300 OT’s central star and the survivor of “SN 2008S” may only briefly turn the skies blue, for the most massive stars the switch from red to blue may be permanent if enough of the outer layers are lost through strong stellar winds. Rho Cassiopeia’s expelled materials cool down and once their temperature falls below a few thousand degrees soot and other material begins to condense as dust. This opaque blanket contributes to the variability of Rho Cassiopeia stars. The underlying star emits a fairly constant gaze, but periodically, the dust obscures much of the star’s light. Over time this dusty layer is blown away only to be refreshed by further material blowing out of the star. Tens of thousands of years later, when the star turns blue, the dust will be broken down by fierce ultraviolet radiation from the star. In time the coming supernova shockwave will scatter the star’s former envelope into the neighboring universe.

Luminous Blue Variables and Wolf-Rayet Stars

Stars with masses above 30 times the mass of the Sun can morph into different forms. The very most massive O-class stars become luminous blue variables (Chap. 2 ) before (usually) shedding most of their mass in powerful stellar winds. What remains is a Wolf- Rayet star—the shriveled core of the former heavy-weight. Many of these stars were once the universe’s mightiest, with masses in Variable Stars 91 excess of 50 times that of the Sun. However, in these stars stellar winds prove to be a harsh tax and steadily remove up to (and in some cases) well over half the mass of the star, before it can explode. While they persist, the LBVs emit a steady slow wind of dusty material with speeds up to a couple of hundred kilometers per second. Periodically, these giants have fits and expel much greater quantities of material in single outbursts. P Cygni did so in 1601, shedding a tenth of a Sun’s worth of gas in a violent eruption. The prototype, S Doradus has had a long history of steady fizzing, interspersed with sporadic eruptions. When in eruption the spectral class drops from B to F as cool, dusty gas, ejected by the star condenses around it. Eta Carinae is considered by some to be a prototypical member of the class of S Doradus stars. However, some suspect that its enormous eruption in 1843 was caused not by some instability within the star, but by a collision between it and a second star (Chap. 6 ). Both LBVs and WR stars are shrouded in the murk of their own creation. As more and more mass is expelled from the massive star and it morphs into a Wolf-Rayet (WR star), the speed of its winds accelerates. Large clots of dusty gas punctuate the steady stream the star is emitting. Over the space of a year a WR star will lose up to a Jupiter-mass of material, or nearly 1,000 Earths. At this rate of expulsion it isn’t hard to see how something as massive as 100 solar masses can be reduced to a flickering candle in only a few million years. It is for this reason that WR stars and LBVs are only found in the very youngest stars clusters such as NGC 3606 in our galaxy, or 30 Doradus in the LMC. Nature is not kind to the most spendthrift creatures. Waste not want not; and for the universe’s mightiest stars, their carefree life leads to a premature death. By 10 million years all that will remain of these mighty stars is a speeding black hole or neutron star, kicked briskly from the cluster by the violence of its own demise.

Cataclysmic Variables

We now leave the solo stars behind to briefly investigate those stars that achieve astronomical notoriety through the illicit acquisition of material from their unwitting companions. The cataclysmic variables are a class of binary white dwarf system where 92 The Complex Lives of Star Clusters the white dwarf steals material from a main sequence or giant star companion: all are by definition low mass systems as only these will produce a white dwarf first from two main sequence stars. Cataclysmic variables comprise a sub-set of low mass X-ray binary systems. In these a white dwarf, neutron star or black hole orbits the common centre of gravity with another less evolved star, commonly a red giant or main sequence star. Many of these are strong sources of X-rays, hence their positioning within the X-ray binary class. It would take an entire book to cover these systems in their own right. Instead, this chapter will provide an overview of the white dwarf systems, while Chap. 6 takes a deeper look at those systems containing either neutron stars or black holes. Globular clusters contain a wealth of cataclysmic binary star systems by virtue of the close proximity stars find themselves in within these tight balls of stars. One might intuitively expect that there would be a similar enhancement in the number of cataclysmic variables as neutron star Low mass X-ray binaries (LMBXs, Chap. 6 ). However, this is not the case. There are roughly twice the number of cataclysmic variables per star as there are in the disc. There are more than ten times the number of neutron star low mass X-ray binaries. The difference is the gravitational pull of these two types of star. White dwarfs are less massive than neutron stars, so their pull on neighboring stars is correspondingly less. Consequently, white dwarf stars are less able to grab hold of passing stars and draw them into the sorts of relationships that will ultimately result in the release of X-rays from the white dwarf. The binary systems in star clusters can be either primordial , in that they formed with the cluster, or secondary , forming later when stars interact with one another, most likely when a pre- existing binary system encounters another star (or binary). Regardless of their origin, a white dwarf must lie sufficiently close to its partner to gravitationally tug upon it. The companion, which could be a red giant, sub-giant, main sequence star, helium star, or on occasion, another white dwarf, must fill its Roche Lobe (see Fig. 3.5 ) so that matter from it can be acquired by the white dwarf. These binary systems are known as semi-detached systems as the two stars are separate but in contact via a stream of matter from the companion to its white dwarf. Variable Stars 93

White Dwarf

Roche Lobes

FIG. 3.5 Detached versus semi detached systems. In a the two stars are far apart and both lie within a boundary called the Roche Lobe. Here, they have complete gravitational control over their mass. In b , the stars lie closer together and the larger, main sequence star has a smaller Roche Lobe because the white dwarf is able to exert its gravitational pull on the larger star. Material overflows the edge of the Roche Lobe (known as Roche Lobe Overflow ) of the main sequence star and moves towards the white dwarf. As the two stars are orbiting one another, the material forms a disc around the equator of the white dwarf (an accretion disc)

As the white dwarf is a small, Earth-sized star, it must lie close to its companion in order to feed from it. This means that the orbital period (the time the two stars take to orbit their mutual centre of gravity) must be short: typically a few weeks to months for a red giant companion; a day or less for main sequence companion; and down to a matter of minutes for a white dwarf companion. Cataclysmic variables can emit bursts of radiation from the high energy X-ray portion of the electromagnetic spectrum, through visible and infrared down to radio wavelengths. Emission comes from two main sources: the disc of material around the white dwarf and from the material raining onto the surface of the white dwarf. A few white dwarf stars harbor intense magnetic fields up to a couple of hundred million times the strength of the Earth’s. These systems, known as AM Hercules systems, have material streaming from the companion in a stream towards the gravitational juncture between the two stars, known as the L1 Lagrangian point. This is where the two Roche Lobes touch. From here material is directed in two streams towards the magnetic poles of the white dwarf. The most intensely magnetic example is AR Ursa Majoris, with a magnetic field measuring 230 million times that of 94 The Complex Lives of Star Clusters the Earth (230 Million Guass)—roughly equivalent to that found around rejuvenated millisecond pulsars (Chap. 6 ). Most, however, have fields measuring 10–80 million times that of the Earth: still substantial but nowhere near the strength of those fields around neutron stars. Here there is a jet of material spirals from the L1 point, part of the way towards the white dwarf, before funneling inwards across the last few thousand miles to each magnetic pole. If the magnetic field is weaker (1–10 million Gauss), or the two stars are too widely separated, the two stars don’t co-rotate. These systems are called DQ Hercules or “intermediate polar” systems. In these systems material streams from the L1 point towards the white dwarf, but in many cases is able to accumulate into a disc around the white dwarf. However, as it spirals inwards it soon encounters the white dwarf’s magnetic field. Here it is fun- neled onto the white dwarf as before. In a few cases the spin of the white dwarf can fire blobs of material back towards the companion stars, along the white dwarf’s magnetic field lines. In each of these cases the impact of the material on the white dwarf’s poles can create a situation analogous to pulsars where the X-ray emission pulses in time with the star’s rotation. In this case the pulses come from the hot, X-ray emitting magnetic poles of the white dwarf, not from beams of radiation generated by the magnetic field of the pulsar. Four possible configurations of cataclysmic variable are shown in Fig. 3.6 . Cataclysmic variables show both periodic and random variations in their output. Output may vary cyclically as the white dwarf and companion orbit one another and periodically block out the X-ray emission coming from the white dwarf. This will be particularly obvious where the companion is a large star, such as in Fig. 3.6a . There may also be variation in the detected output of the white dwarf caused by changes to the amount of material falling onto its surface. This is also true for changes in the amount of material flowing onto the accretion disc (Fig. 3.6b ). This can be caused by the two stars having elliptical orbits, or by the presence of magnetic fields in the companion star that affect the flow of plasma from its surface towards the white dwarf. A small number of cataclysmic systems emit low energy (soft) X-rays continuously. These super-soft X-ray systems have a partnership balanced on a knife-edge. In most cataclysmic systems Variable Stars 95

Red Giant

b Hot spot

FIG. 3.6 Four of the possible types of cataclysmic variable. In a , a red giant sheds material and a white dwarf captures much of the stars wind. These systems may appear as recurrent novae, with repeated outbursts separated by decades of relative calm. In b , the standard cataclysmic variable has material streaming from a small main sequence star towards the white dwarf. This forms a disc around the white dwarf. Where the inflowing stream from the companion impacts the disc there is a hot spot (or splodge) and this is where most X-ray emission originates. In c , the disc is disrupted by the white dwarf’s strong magnetic field. Material from the edge of the disc flows in streamers towards the magnetic poles of the white dwarf from the inner edge of the disc. Hot, X-ray emitting spots mark the position on the white dwarf’s surface where its magnetic poles lie. d represents an AM CVn System where two white dwarfs, one a lower density helium dwarf (or a helium-rich star) orbit one another closely. Helium flows from the lower density star to the most massive white dwarf causing periodic eruptions

the companion star is a small red or orange dwarf star with a very low mass. These stars are only able to emit a small amount of matter each year—usually amounting to less than a hundred millionth of a solar mass. These systems are generally powered by the flow of material onto the disc surrounding the white dwarf, rather than the impact of this material onto the surface of the white dwarf. However, in super-soft systems there is perhaps a thousand times 96 The Complex Lives of Star Clusters more material flowing onto the white dwarf; so much in fact that it continuously burns to make helium. X-rays are emitted predominantly from the hot surface of the white dwarf as nuclear reactions power away a short distance under its surface. What isn’t burnt is expelled in two broad outflows away from the poles of the white dwarf. What makes super-soft systems special is their apparent ability to burn hydrogen stably and thus build up the mass of the white dwarf. Should that mass reach the so-called Chandrasekhar limit (1.382 solar masses for a typical white dwarf) then the core of the white dwarf will implode and ignite, sending an explosive wave of nuclear fire through the white dwarf. This, then, quickly decimates the white dwarf as a Type Ia supernova. Astronomers rely on these supernova to chart distances to galaxies and probe the expansion of the universe. In 1998 two groups of astronomers, led by Saul Perlmutter (Lawrence Berkeley), and separately Brian P. Schmidt (High-z Cosmology Project) and Adam G. Riess (Harvard University) interpreted data from Type Ia supernova revealing that the expansion of the universe was accelerating, contrary to expectation. Tying down the expansion rate is thus fundamental to our understanding of our origins and places Type Ia supernovae at the heart of modern cosmology. It is, therefore, critical that astronomers understand how these supernovae occur. Super-soft systems perhaps provide one gateway to this understanding. The problem with most cataclysmic variables is that the puny rate of acquisition means that fuel cannot be burnt continuously. Instead hydrogen builds up over millennia until it is thick, dense and hot enough to ignite. What results is spectacular—although not quite a supernova. Instead, hydrogen begins to burn through the carbon-nitrogen cycle: nothing obviously problematic there. In the case of the white dwarf, nuclear reactions start quite cold and slow. However, as the star begins to assemble helium from its stock of hydrogen, nuclear reactions create nitrogen-13 and oxygen-15, neither of which is very stable. As these decay into carbon-13 and nitrogen-15 (respectively) they emit positrons, antimatter electrons. These annihilate with electrons and emit gamma rays which rapidly heat up the layer of hydrogen. As the temperatures rise, the rate of nuclear reactions accelerates and more of the unstable nuclei are made. Decaying, these Variable Stars 97 cause further heating and the whole layer catastrophically ignites. A maelstrom of energy engulfs the outer layers of the white dwarf, blasting them into space. In minutes the placid white dwarf has morphed into a planet-sized thermonuclear bomb, which balloons in size. Although the core of the white dwarf remains blissfully unaware of the torment castigating its surface, the wholesale expansion of the outermost layers of the star causes it to brighten up to 100,000 times. This radioactively-powered outburst blasts material away from the white dwarf, disembowels its accretion disc and broils the surface of its companion. This is not enough to disrupt the system and the process should repeat on te thousand year cycles. Through each outburst it is predicted that the white dwarf loses a little of its mass so that over time the white dwarf and companion should drift slowly further apart. Whether this attrition is sufficient to end the cataclysmic variable’s activities is unclear. Between the extremes of the super-soft systems and the novae systems are recurrent novae. A few are known and these systems contain a red giant and white dwarf. The two stars are usually relatively far apart, but close enough that the thick wind from the red giant is harvested by the white dwarf. In these systems the white dwarf collects between one ten millionth and one hundred thousandth of a solar mass of hydrogen each year. It’s just enough to keep the fires periodically burning on the surface of the white dwarf, but not enough to power the continuous blaze that is seen in the super-soft systems. Instead hydrogen burns in explosive waves that periodically blow material off into space. Systems, such as RS Ophiuchi, erupt on multi-decadal timescales, with eruptions observed in 1898, 1933, 1958, 1967, 1985, and 2006. Whether the white dwarf in these systems gains or loses mass is unclear. However, it remains possible that at least a few of these systems have white dwarfs that are growing in stature and could, 1 day, detonate as Type Ia supernovae. All cataclysmic variables lead relatively brief lives—less than 100 million years. The reason is simple: math. Even a system where there is a red dwarf companion, transferring one hundred millionth or less of its mass per year to its white dwarf companion will be reduced to planetary dimensions (mass, being the most important) in roughly 100 million years. Indeed, if a 0.1 solar mass 98 The Complex Lives of Star Clusters red dwarf were to transfer one hundred millionth of a solar mass per year it would be reduced to nothing in 10 million years… These systems by virtue of their nature self-destruct in a (relative) blink of an eye. Many systems have relatively helium-rich but low mass stars as the donor stars. These can only have arisen if they were once much more massive—perhaps akin in mass to the Sun—but now have lost most or all of their outer layers and some of their cores to their companion. What you see now is a husk of a star, withered to less than half its original mass. The fact that star clusters, and globular clusters in particular are rich hunting grounds for these systems implies that they are constantly created: for how else could a ball of stars, with an age measured in billions of years, play host to a cataclysmic variable with a lifetime measured in tens of millions? The interplay of stars and star systems (Chap. 6 ) within clusters is an integral part of their existence. It is through successive and continuous exchange of stars that new cataclysmic variables can be produced in sizable quantities. Given the approximate frequency of cataclysmic variables in the Milky Way as a whole (around 0.01 % or 10 million systems) you would expect around 100 cataclysmic variables in globular clusters. Instead, through interactions between stars, roughly twice this number are predicted—if not quite found, yet. Of these, around 25 % are expected to be survivors from the first population of binary stars that was formed with the cluster itself. The remainder will have been formed later through tidal capture of stars by white dwarfs, or through interactions between pre-existing binary systems and passing stars. These routes are explored more fully in Chaps. 6 and 7 .

Conclusions

Not only do star clusters allow astronomers to probe the basics of stellar evolution, they also provide a natural laboratory for the examination of stellar exotica by providing host to a wild and var- ied zoo of stellar exotica that is either rare or extinct elsewhere in the galaxy. Young star clusters, such as 30 Doradus, play host to some of the universe’s most massive stars and may provide a birthing Variable Stars 99 ground for intermediate mass black holes where ultra-massive stars can form through collisions (Chaps. 6 and 7 ). These clusters may thus be the living fossils of an era where galactic assembly, 13 billion years ago, produced the first super-massive black holes that now dominate the interiors of all large galaxies. Upon examination even the apparently more sedate, elderly globular clusters are a hot-bed of intrigue, with a diverse population of variable stars and binaries. Although much of this comes down to a myriad of binary partnerships (examined fully in Chap. 6 ), the large populations of stars naturally allows for large numbers of proportionately rare objects to exist. Examination of red giants allowed Patricia Whitelock and colleagues to probe aspects of the evolution of these stars from inception to their ultimate deaths. Likewise new types of uncommon variable star have been found within clusters. Like our cities, where millions congregate, there is much to see within the confines of star clusters that might be impossible to find elsewhere in the more rural neighborhood of our galaxy. Undoubtedly more secrets remain to be discovered, particularly as constantly improving telescopic resolution allows for greater dissection of the light coming from young, massive star clusters in the local universe. Expect much in the coming years. 4. Globular Cluster Formation

Introduction

Globular clusters are a unique system of stars in the universe. With the exception of the Ultra Compact Dwarfs (Chap. 1 ), no other stellar grouping contains so many stars compacted into such a small volume. A typical cluster will contain over 100,000 stars in a volume of 1,000,000 cubic light years. However, this is slightly misleading. Most of the stars within the globular are found within a sphere less than 50 light years across, so that the central core of the cluster will typically have stars separated by distances comparable to the Solar System. The globular cluster can be subdivided into urban and suburban regions. The urban core contains half the amount of stellar mass but is typically only a few parsecs (less than 10 light years) across. This is known as the half-mass radius . Surrounding this region, in a large sprawl over 100 light years (30 parsecs) across are the remaining stars. The stars in this region move with velocities as high as 10 km/s and are less tightly bound to the core. More recent discoveries—the “faint fuzzies” and ultra compact dwarfs (UCDs), described in Chap. 1 —form outlying populations that in many (but probably not all) cases are linked to the globular clusters, and may share a common origin with them. How do we make sense of a population of very massive star clusters with the differing range of properties that are observed? Do they all share the same origin?

The Stars of Globular Clusters

The stars that populate globular clusters have some unique features, as well as many that are common to all halo (population II) stars. In general the stars are poor in elements heavier than helium.

Typically, they have between one hundredth to one thousandth the amount of these elements compared to other star populations. Aside from hydrogen and helium, the most abundant elements are those which were formed in massive stars and released in explosions called Type II supernovae (Chap. 2 ). Globular cluster stars are typically poor in iron, which is primarily manufactured in explosions called Type Ia supernovae. These involve white dwarf stars, which take at least a few tens of millions of years to form. Such white dwarf stars are lower in mass and may not have been common in the universe’s infancy. It is generally assumed that most of the universe’s first population of stars were massive enough to die as Type II supernovae, rather than quietly as white dwarfs, and this is validated by the chemistry of the halo stars. Were there more of the sorts of deaths that manufacture iron, then this element would be more abundant in the universe. Within the galactic halo, the proportion of light elements up to oxygen is fairly constant, although more mature red giant stars often fiddle with the proportions of these element through the nuclear reactions occurring in their cores. Elements from silicon upwards are also fairly constant. There is, however, a swathe of elements between oxygen and silicon that show odd patterns of star-to-star variation, but only in the stars of the clusters, not the halo as a whole. These chemical fingerprints point to some interesting phases in the development of these stars clusters; evolutionary steps that only happened in the globulars but not the more loosely bound open clusters. Such processes were unknown until quite recently and paint a dynamic picture of star formation in the clusters which is akin to the formation of the smallest galaxies of today.

Cluster Formation: A Reprise

Recall that all stars are formed from collapsing clouds of gas and dust called molecular clouds (Chap. 2 ). The process takes anywhere between 100,000 and 2 billion years to transform a low density ball of frigid gas and dust into a dense ball of self-sustaining plasma. As most stars are formed from clouds of gas tens of light years across, it isn’t surprising that most stars form in groupings (Fig. 4.1 ). Globular Cluster Formation 103

FIG. 4.1 General monolithic model for the formation of the Milky way and its star clusters. In a the original gas from the Big Bang, which is polluted by small amounts of heavy elements from the first generation of stars, forms a roughly spherical ball. In here, the globular clusters condense. These shed heavier elements into the remaining gas and this settles further under gravity to form the thick disc (b ). Finally, star death in the disc and halo contributes the remaining gas which settles further to form the thin disc (c )

The globular clusters are clearly a world apart from most clusters because of their sheer size. These micro-universes imply, perhaps untruthfully, a special set of circumstances that are peculiarly efficient in the bulk manufacture of stars. So what, if anything, was peculiar about the circumstances in which they formed and could it happen again?

What Do Observations of Globular Clusters Tell Us About How They Formed?

Most of what astronomers know about the nature of globular clusters comes from observations in the last 30 years. Instruments such as the Hubble Space Telescope have allowed astronomers to 104 The Complex Lives of Star Clusters peer deeply into these balls of light and distinguish the individual contributions the different sub-populations of stars make to the overall metropolis. Before Hubble, astronomers could clearly discern the giants and the brightest main sequence stars. However, the dimmest stars, whether they were white dwarfs, red dwarfs or any other object with a luminosity much less than the Sun, weren’t clearly resolved from the overall peach fuzz glow of the cluster until Hubble came along. Within that fractured ball of light a few more interesting characters were apparent, even though they couldn’t be directly observed. X-ray emission from globular clusters implied a healthy population of exotic objects involving pairings of stellar corpses to still living stars. These zombies retained life after death, by stealing matter from neighboring stars in binary partnerships. Although they couldn’t be observed directly, as they accreted gas from their companion, the gas was heated up until it emitted copious amounts of X-rays. These streamed outwards to be intercepted by our detectors, betraying their presence. Globular clusters contain a wealth of these stellar corpses in binary systems, which is perhaps unsurprising given the close proximity of their stars (Chaps. 3 and 6 ). What is remarkable, when one compares a Hubble image of a cluster such as 47 Tuc with an older ground-based one, are the gaps between the stars (Fig. 1.1 ). Ground-based images simply showed a homogenous glow of light, which gradually fragmented into individual stars at its periphery. Hubble peered into these balls and not only fragmented the light into individual stars, but also revealed a wealth of dim stars, notably many formerly unseen white dwarf stars. White dwarfs are important as they can be used to date not only the cluster itself but other white dwarfs in the galactic disc and halo. These, in turn, also form a sign post in the formation history of the galaxy as a whole. With the enhanced resolution provided by Hubble, enough stars could then be counted and dissected using spectroscopy. This provided the detail that allowed astronomers to reveal the over- arching picture of how globular clusters formed, but also raised some interesting problems about the big picture itself. Remember, astronomers believed that star clusters, both globular and open, formed all of their stars at the same time through the monolithic Globular Cluster Formation 105 collapse of the nebula (Chap. 2 ). Would Hubble provide any new insights here? With an ability to discern individual stars, the chemistry of each was available for analysis, which allows insight into the evolutionary processes of individual cluster members. Astronomers could then begin to compare the evolution of these stars with their models. Once again, were there any differences from expectation? Hubble has provided some rather tantalizing clues to the inner workings of globular clusters and forced astronomers to go back to the drawing board. Much of what astronomers thought they knew has, and continues to be, re-written. This chapter, more than any other, examines these issues.

Secret Agents: Dwarf Galaxies Masquerading as Star Clusters

Omega Centauri and Its Kin

In 1999 Young-Wok Lee (Yonsei University, Seoul) published an analysis of the red giants of Omega (w) Centauri. The research demonstrated that there were multiple, most likely four, red giant branches, with the reddest, most metal-rich branch indicative of a population 2 billion years younger than the bluest. Although it seemed most likely that the majority of galactic globular clusters formed exactly as the textbooks suggest, from the collapse of a cloud of gas and dust, it was now likely that at least some of the large globular clusters were actually the remains of dwarf galaxies. This put the largest globular clusters, Omega Centauri included, on a par with the UCDs described in Chap. 1 . These weren’t so much clusters, but the nuclei of hapless galaxies that had wandered too close to the Milky Way and been torn asunder. Omega Centauri wasn’t alone. The Milky Way’s largest cluster, M54, is actually part of—and indeed forms the core of—the Sagittarius Dwarf galaxy. Discovered in 1994, this elongated spheroidal galaxy orbits the poles of the Milky Way. What is interesting about this galaxy is the presence within it of four globular clusters, M54 simply being the largest. A direct comparison of the Milky 106 The Complex Lives of Star Clusters

Way and M54 might place M54 as an analogous structure to the Milky Way’s galactic bulge within the Sagittarius Dwarf galaxy. For the last few hundred million years gravitational forces have been steadily stretching this galaxy into a long tendril of stars (and clusters) through a prolonged process. At some point over the ensuing billion years or so the Sagittarius Dwarf will become absorbed into our galaxy as a circling stream of stars. Was M54 always the heart of the Sagittarius Dwarf? Did the cluster assume the position after the Milky Way grabbed hold of the cluster, or was it always at the heart of the small galaxy? This is a rather important question as its answer leads us towards or away from the idea that most (or even all) globular clusters are actually the central portions of galaxies that have long since been ripped apart—and this clearly affects whether differences in the chemistries of the stars point to a complex history for star formation. Instead of M54 being the core of the Sagittarius Dwarf galaxy, could it instead be the case that gravitational forces caused it to fall inwards as the Milky Way began to tear the galaxy apart? It is nearly impossible to distinguish between these two scenarios. If M54 truly is the heart of the Sagittarius Dwarf galaxy, it’s possible that many other globular clusters are the cores of long extinguished companion galaxies to the Milky Way. At present the case is open. If this were true it would also help resolve the issue of chemistry. The different populations of globular cluster stars would have arisen in a much more massive structure which was able to retain gas and form more than one stellar population. The ultra-compact dwarf galaxies (UCDs for short—see Chap. 1 ) are a step up the mass-ladder from the globular clusters. These have luminosities up to 100 times that of typical globular clusters and they have half-light radii of more than 10 parsecs (approximately 33 light years). Globular clusters have half-light radii of 1–5 parsecs, for comparison. Their origin is also disputed as they have a range in masses from globular clusters right up to small elliptical galaxies. The largest UCDs in the Virgo cluster of galaxies have very red colors, in this case suggestive of relatively high amounts of metals in the component stars. The metallicity is comparable to the bulges of spiral galaxies or to elliptical galaxies. Their masses and radii are also clearly analogous to these galactic structures. Although they now look very like large globular Globular Cluster Formation 107

FIG. 4.2 The formation of large globular clusters and UCDs by tidal stripping. Some globular clusters and the largest UCDs may be the remnants of small galaxies that wandered too close to a larger neighbor (a ). Most of the small galaxy’s stars are stripped off by tidal forces (b ) as it orbits the larger galaxy leaving a small core embedded in a faint stream of stars (c ). The largest, reddest UCDs almost certainly all formed this way clusters, their broad diameters and similarity to elliptical galaxies or spiral galaxy bulges, is highly suggestive that these large, red UCDs are galaxy cores that have been stripped bare of outlying stars (Figs. 4.1 and 4.2 ). Smaller UCDs still have the relatively large half light radii of more than 10 parsecs, but otherwise their colors and diameters fall closer to the largest globular clusters, notably Omega Centauri (Chap. 1 ). Thus Omega Centauri forms a clear link between the UCDs and the globular clusters, with UCDs most likely representing the cores of larger galaxies than the largest globular clusters. A final interesting observation also concerns Omega Centauri. It’s entirely possible that within such merging star systems individual globular clusters could fall into one another, producing hybrid clusters. Although not likely to be the source of many multi-generational globular clusters, Raffaele Gratton 108 The Complex Lives of Star Clusters

(Osservatorio Astronomico di Padova) considers it at least possible that some hybrid globular clusters may have formed this way. A case in point was the discovery by Young W. Lee of a separate cluster of stars moving within the body of Omega Centauri. Like a Matryoshka Doll, Omega Centauri contains a small, metal-rich cluster of stars that move coherently. Observations by Francesco Ferraro (Osservatorio Astronomico di Bologna) and colleagues suggested that the cluster was accreted by OMega Centauri at some point before it too merged with the Milky way. The discovery of systems like Omega Centauri complicates the picture of globular cluster formation even further. We have mergers of one cluster with another; multiple generations of stars within the cluster; rounds of stellar pollution; and the eventual exodus of many stars to the halo. Globular clusters are far from the simple orbs of single generation stars that they were once thought to be.

The Problem

Omega Centauri had already forced a reconsideration of the origin of globular clusters. However, it and M54 were thought to be isolated examples of globular clusters that have multiple generations of stars—and their origin was thought to lie in small galaxies that had, or were, merging with the Milky Way. However, in 2007, a team led by Giampaolo Piotto of the University of Padua showed that the globular cluster NGC 2808 contained multiple populations of stars (Fig. 4.3 ). By plotting the luminosities of the main sequence stars against their colors it became apparent that there are not one, but three main sequence tracts. This implied three distinct populations of stars within the cluster. Now, the problem is that globular clusters were thought to be the products of monolithic collapse: the wholesale implosion of a cloud of gas and dust that created the cluster of stars in its entirety. However, if that was true then there would only be one main sequence tract. Moreover, later observations of many, if not all, globular clusters revealed that they too have multiple populations, each with a slightly different chemistry. The number varies from two populations in the 400,000-strong cluster NGC 6397, to three in the Piotto’s millionaire cluster, NGC 2808, to four in multi- Globular Cluster Formation 109

RHB

EHB

Sodium-rich; Sodium-poor; oxygen-poor oxygen-rich stars (He-rich?) stars

FIG. 4.3 NGC2808 has three converging main sequence tracts. The bluest (blue dots ) is the richest in sodium, but poorest in oxygen. It may also, controversially be richest in helium. ω Centauri has four main sequence tracts. The most helium-rich stars are expected to form the extreme (blue ) horizontal branch (EHB), while the least produce the red end (RHB). At the moment all that is known is that sodium-rich stars form the blue end in most globular clusters. The issue of how much helium is present in these stars is far from resolved millionaire cluster Omega Centauri. The nature of these different populations poses some unusual puzzles. The bluer main sequence is fainter and these stars are richer in sodium but poorer in oxygen. Indeed, there are no oxygen-rich and sodium-rich stars: all seem to be either one or the other. For the best part of four decades observations of globular cluster stars have confirmed this pattern which is known as the “sodium-oxygen anti-correlation”. In simple terms this means that stars which are rich in sodium are poor in oxygen and vice versa. There are no oxygen-poor, sodium-poor stars, or oxygen- rich, sodium-rich stars. An anti-correlation is also known as a negative correlation, which means if one unit goes up, then the other linked unit (or variable) goes down. This anti-correlation is a very big deal. Imagine travelling to a city center populated with only old men and young women, while in the suburbs there are only old women and young men and never 110 The Complex Lives of Star Clusters the twain shall meet. You’d certainly wonder how this distribution had come about. This anti-correlated pattern of chemistry is only seen in globular cluster stars: nothing like it is seen in any star in the galactic disc, nor in stars of equal antiquity within the galactic halo. Something odd is going on within the stars in a globular cluster that means that sodium and oxygen really can’t get along, but only in globular cluster stars. In stars there are other similar (anti-correlated) pairings of elements, such as aluminum and magnesium, but these are relatively straightforward to understand. These two elements are paired up in nuclear reactions, so that if a star consumes magnesium it will produce (amongst other things) aluminum: they are partners. Although not miles apart in terms of their structure and production line, sodium and oxygen do lie rather far apart in the Periodic Table, and there is a sufficiently large difference in the manner in which they are produced to question how they could be linked in an opposing pattern. Moreover, the circumstances in which sodium and oxygen are made and consumed couldn’t be more different. Other elements such as carbon and nitrogen also vary universally in very predictable ways; the same goes for heavy elements from silicon and above. It is only elements with masses between nitrogen and silicon that don’t seem to want to play ball. Any star born with reasonable masses—perhaps two or more times that of the Sun—burn most of their hydrogen through the CNO cycle (Chap. 2 ). Carbon serves as a catalyst, forming a core upon which helium is built by the sequential addition of hydrogen. When the fuel is running low the amount of available carbon goes down and nitrogen goes up as the reactions begin to falter before they can complete the synthesis of helium. Essentially, there is no longer enough hydrogen to go around, and helium production ceases. What hydrogen there is ends up locked onto the carbon core, transforming it into nitrogen. The next steps are blocked so no more helium is spat out and no more oxygen can be made. Stars at the end of the main sequence tend to show an increase in the amount of nitrogen and a decrease in carbon: oxygen shouldn’t really change. The temperatures at which this happens are anything from 20 to 70 million degrees, but the highest end of this range is really only realized in the most massive stars, or late on as the star is evolving into a red giant. At this point, the core is shrinking and Globular Cluster Formation 111 growing hotter, while hydrogen is burning in an increasingly hot shell around it. At temperatures above 20 million degrees sodium can be produced through hydrogen-burning reactions that consume any neon that is present. Hydrogen adds in threes to neon to make a succession of sodium isotopes until the stable isotope, sodium-23, is produced. Calculations suggest that the temperatures required should be achievable in any low mass star that is burning hydrogen in a shell after it has left the main sequence. Moreover, anything more massive than one and a half times that of the Sun could produce this element throughout their lives if there is enough neon present. You would expect, then, stars to get richer in sodium as they age—but only those stars with sufficient mass. The stars that are sodium-rich on the main sequence in today’s globular clusters are far too wimpy to make this element, so they must have either acquired it at birth or inherited it from a nearby, dying companion. By contrast, oxygen can be made by almost any star that gets hot enough to burn helium. Yet, the majority of it comes from supernovae—in particular those explosions that mark the ends of short-lived, massive stars in so-called Type II, or core-collapse events (Chap. 2 ). Countless observations confirm that the gas from which the globular clusters formed was greatly enriched in oxygen. In general, this oxygen-enriched gas was found throughout the territory in which the Milky Way and its entourage of globular clusters formed. Sodium appeared later, when lower mass stars began to die and spew out this element while they were red giants. As sodium is produced at a slower rate than oxygen, it can be used to trace the formation of second and third generations of stars in globular clusters. More peculiarly, stars with an amount of oxygen less than half that of the Sun seem to destroy (or at least hide) what oxygen they have as they transform into red giants. Sodium-rich stars further transform their chemistry to eliminate what little oxygen remains. The oxygen-annihilating process only kicks in once these stars have reached the red giant bump (Chap. 2 ), suggesting that something is happening within them once stage is reached. What’s weird about this is oxygen really shouldn’t get anywhere near hot enough to fuse into anything else while the star is still ascending the red giant branch. It could pick up stray hydrogen and be transformed 112 The Complex Lives of Star Clusters into carbon and helium. But if that was true you’d see more carbon or nitrogen appearing in the stellar spectra—and you don’t. Oxygen just seems to vanish as the star reddens and brightens. Is it something to do with processes that happen at the red bump? The bump marks the point in the star’s life where the zone in which hydrogen is burning reaches out to where the envelope used to convect. Here, the star briefly dims as the amount of helium-3 decreases, taking down the rate of nuclear reactions. The star then brightens again as the rate of nuclear reactions picks up again, compensating for the energy needs of the star and its growing core (Chap. 2 ). The sudden decrease accompanying the crossing of the bump would suggest that oxygen is being taken up by the core of the star—perhaps to be transformed into another element or perhaps just “dropped out” or “vacuumed out” into the still inert, but ever hotter helium core. Why the sodium-rich, oxygen- poor stars would do this and not the more oxygen-rich stars is a complete mystery. In all clusters examined the sodium-rich, but oxygen-poor, population predominates; with between 10 and 30 % of stars being sodium-poor and oxygen-rich. So, we have clusters that appear to show more than one population of stars which implies they formed separately. How, then, do we make a globular cluster with a mass of a few hundred thousand Suns that has more than one generation of stars? That is a fairly thorny and fundamental problem in astrophysics. Moreover, what is the connection with sodium?

A Question of Mass

If the obvious dilemma is how to make a globular cluster with two or more stellar populations, then the underlying predicament is just how massive were globular clusters in the past? The two questions are fundamentally linked. Why? Well, if you can beef up the primordial mass of globular clusters then you can start to play around with the process of star formation that occurred within them. However, if you find evidence that says they pretty much haven’t changed since they were born, then the chemistry of the stars comes back and bites you. We need massive clusters to produce enough gas to then synthesize the subsequent generations of stars that seem to be present within them. Globular Cluster Formation 113

The globular clusters populating the halo of the Milky Way seem massive by the standards of those that are forming today. With up to a few million times the mass of the Sun, they are comparable to the Universe’s smallest galaxies, the dwarf spheroidals. Yet, this present mass might well be misleading: an iceberg forms a reasonable analogy. A typical iceberg, floating away from Greenland or Antarctica, has roughly three quarters of its mass lurking under the surface of the water. What you can see on the surface grossly under-represents what is actually there. Now, it isn’t that globular clusters are hiding nine tenths of their mass; rather what you see now may well be but the flesh and bones of the former cluster, stripped down by forces acting inside and externally to the cluster. If we are to accommodate two or more stellar populations in the billions of years since these star cities formed, much of their mass must have been lost to the halo and disc of the galaxy. If you allow for this, you can produce different populations. For a typical cluster, with 70 % of stars sodium-rich and apparently being second or third generation, you had to have around ten times the current mass in stars to distribute enough gas to produce them. Although the clusters we observe today are too wimpy to generate and retain much if any gas, they might well have done when they first formed. Look at the problem that way and some pieces of the puzzle might well fall into place. So, let’s look again at the formation of the cluster, with this idea in mind. A nebula with several million times the mass of the Sun is shocked, perhaps by a nearby supernova or simply by the collec- tive radiation of millions of scattered Population III stars. Either way, gravity gets a hold of it and rapidly begins to crush it inwards. Random motions and changes in momentum and magnetic fields shred the star into millions of collapsing cloud cores. Millions of low mass stars begin to form, but are soon overtaken by a rash of massive stars—perhaps 10,000 or more in number. Within 100,000 years of the collapse beginning, the contracting nebula is lit up by a massive pulse of ultraviolet light as a cluster of massive stars fires up their engines. The question is what then happens to all the remaining gas once the massive stars fire up? If you view the globular cluster as essentially what we see today, the radiation and strong stellar winds accelerate the remaining gas to speeds in the region of 1–2,000 km/s. This is far in excess of the wimpy escape velocity of the cluster, 114 The Complex Lives of Star Clusters which may be anything from 3 to 80 km/s. Consequently, within a few million years—less time than it takes all of the hottest stars to die—the cluster is stripped bare, leaving the barest husk of hot tenuous gas. This is further pulverized and expelled by waves of supernova explosions as these stars expire. All that will remain after such carnage is the gas expelled by lesser stars, and this effectively diffuses out of the cluster over the ensuing billions of years: it is this slow moving gas that could, in principle fuel more star formation. Lower mass stars evolve through a red giant phase to become white dwarfs. These stars expel their outer shells of hydrogen- rich gas in winds with speeds of between 5 and 25 km/s. Although fast by our standards, this is a leisurely stroll for a star’s final breaths and more than easily retained by the star cluster as a whole, particularly if that cluster held a lot more stars when it was formed than are seen today. Such gas is only lost through the broiling action of the surrounding hot, diffuse gases within the galactic halo or through the pull of gravity from the galaxy as a whole. In the monolithic model, outlined in Chap. 2 (Fig. 4.1 ), star formation ceases, effectively the moment the first stars are formed. Initially, only one or two examples of multi-population star clusters were known and these were the true heavyweight clusters like Omega Centauri. Upon discovery it was assumed that these were not really globular clusters, but the stripped down cores of dwarf galaxies (see later in this chapter): problem solved. However, when lesser clusters also showed up with the same multi-generational populations, that solution seemed less ten- able. Applying the cutting edge of Occam’s razor, it appeared that today’s globular clusters were simply all remnants of larger objects that had spawned two, three or four stellar families. They didn’t have to be galactic nuclei, though this wasn’t ruled out.

A Helium Clue

The sodium-rich stars are bluer on the main sequence, which would imply one of two things. Either they come with less metals overall, or they have more helium than the sodium-poor stars. Globular Cluster Formation 115

Metals are quite liberal with the availability of their electrons, flashing them around like so much loose cash. These free electrons are able to soak up energy streaming out of the stellar core, before radiating it at longer wavelengths. This tends to make the more metal-rich stars redder (cooler) and somewhat larger than a more impoverished star with the same mass. Increasing the amount of helium, on the other hand, makes stars hotter for any given mass, because it’s an element that is uniquely reluctant to form chemical compounds. Parsimonious helium clings onto its two electrons with a religious fervor. This grip is hard to break except to all but the most energetic of ultraviolet photons—and this typically only happens at temperatures in excess of about 50,000 K. Helium’s tight grip means that heat and other radiation can escape the star easily, so any star rich in this element will be smaller and hotter than another that is poor in this element. Thus, when you look at a color-magnitude diagram for clusters like NGC6725 or NGC 2808 and you see two or more main sequences (Fig. 4.3 ) it’s tempting to think that some stars have more helium than others. Certainly, you can pretty much rule out that they have fewer metals than their redder brethren. Spectroscopic measurements can be made of these stars and, aside from sodium and oxygen, there are insufficient differences in these to explain the difference in stellar color. That leaves helium as the only viable cause of the difference in color. 1 The most-helium- rich stars form the hottest, bluest main sequence. Stars with less helium are redder for a given mass. However, helium’s recalcitrance in giving up its electrons makes it fiendishly difficult to observe in the sorts of cool stars that populate the main sequences in these clusters. To see helium clearly in these stars, you need to catch it in that act of giving up, or reclaiming, one or both of its electrons and this only happens at high temperatures. In cool stars high temperatures only transiently happen if the star emits a stellar flare. As these last seconds to minutes the odds of capturing these events in observing are low.

1 Differences in color and helium abundance might not be primordial. Instead fast rotation in some stars could dredge out more helium from deeper layers making the star appear bluer. Collisions or harassment between stars could alter spin. 116 The Complex Lives of Star Clusters

Therefore, astronomers can only infer the presence of helium from the effect this elusive element has on the star, making the inference feel a little tenuous, to say the least. A star could be hotter if it has recently lost mass, gained helium, lost metals, or was born with an odd admixture of both. Indeed, some researchers question whether it is there, in excess, at all—except in the outermost layers. In his thesis, Grant Newsham (University of Ohio) looked at whether it was conceivable that so much helium would be present during star formation and how such helium-rich stars would evolve. For one, helium is only produced in excess by two possible types of star: fast spinning, massive stars and AGB stars. It’s unlikely that there would be enough of the former to make helium in such copious amounts and then spill it out into the inter-cluster environment. AGB stars could do it, but again could they produce helium in the amounts needed to make the observed generation of blue stars? (Fig . 4.4 ). In Newsham’s models the type of helium-rich stars that were produced couldn’t match the observations, without making the stars unrealistically rich in helium. You have to remember that in the 13 billion years or so that stars have been manufacturing this element, the amount in the universe has hardly altered since the Big Bang. Helium tends to remain locked up in the interior of the star—or converted into other elements as the star evolves. Getting it out into the universe to make more stars is really rather tricky. This suggests that these blue stars weren’t so much wholly different objects, but simply regular cluster stars that had been grossly polluted with helium by neighboring giants. Think of stone- cladding: a 1980s construction fashion faux pas. Take a conventional house and make it look like a mansion by slapping lots of pieces of rock on its external walls. Beneath the pretence of its grand exterior it’s the same humble abode. Newsham’s model bor- rows this concept. Newsham isn’t alone, in 2012, Antonino Milone (Instituto de Astrofısica de Canarias) analyzed the main sequence stars of NGC 6397, which again show two, distinct populations. Although there was a suggestion of a very modest (5 %) enhancement in helium, it was nowhere near the 70–90 % increase suggested for clusters, such as Omega (ω) Centauri. Instead, their models suggest that Globular Cluster Formation 117

FIG. 4.4 The formation of sodium and helium-rich stars. In a the star is born enriched with helium (up to 40 % of its mass instead of 23–24 % in most galactic stars). This star is destined from birth to become an extreme horizontal branch star. Alternatively in b a conventional star accretes helium (and sodium)-rich gas from a companion (orange arrows ) and has its fate altered. At the end of its main sequence life, this accreted gas is mixed into the core and the star evolves into a hot, extreme horizontal branch star after a flirtation with the red giant branch the principle difference between the blue and red main sequence tracks are more likely to be due to differences in nitrogen, with the red sequence more nitrogen-rich. The fainter blue sequence was best explained if it was richer sodium and nitrogen (and modestly helium), while depleted in carbon and oxygen. However, just to complicate the story yet further, recent measurements by Andrea Dupree and Jay Strader (both of Harvard Smithsonian Institute for Astrophysics) of a red giant in Omega Centauri show an enhancement in the abundance of helium. Now, this can’t just be superficial as these stars have largely convective envelopes that should stir any helium, which was accreted earlier, deep into their interiors. However, it is just one star, and as Dupree and Strader point out, it may be that this particular star has a different origin and evolution to the other population of stars within the cluster. Superficial layering of helium-rich gas could 118 The Complex Lives of Star Clusters explain the difference in the composition of these main sequence stars, just as well as having them thoroughly mixed throughout. At the moment the jury remains out. To recap, the bluer (hotter) main sequence stars are apparently sodium-rich. It is inferred from their color, that they must also be richer in helium, as they appear to be hotter than they should be for their masses. In Newsham’s model, the bluer stars are only enriched in helium and sodium in their outer layers, not throughout their bulk. The extra helium has been acquired from neighboring, highly evolved stars. It’s possible that the sodium- rich stars had companions, or simply that the gas shed by the dying AGB stars loitered long enough within the cluster core for the remaining stars to acquire it passively. Under the skin, they are still oxygen-rich, sodium-poor stars. Because of the structure of these low mass stars, the helium doesn’t percolate down beyond the base of the outer convective zone—at least not during the time these stars are on the main sequence. At present, which model (if either) is correct is not clear. Part of the problem is clearly that helium does not betray its presence easily. However, more pressingly, the models astrophysicists use to create and evolve these alleged helium-rich stars are not fool-proof. You can really wave your hands in lots of different directions and get the same answer. The only fact that is absolute is that within most globular clusters there are multiple populations of stars, which are likely to be of different ages. Although you could envisage a nebula collapsing into a tight knit cluster of stars and preserv- ing different pockets of different compositions, this is unlikely. Why? Because as we’ve seen in Chap. 2 , stars follow rather chaotic and violent orbits within the cluster as it is condensing. A star will take at most 10 million years to cross back and forth through the core of the developing cluster. This would readily scramble their orbits and hence their differing chemistries. Instead, members of the cluster might have formed at the same time, but large numbers might have become polluted with helium-rich gas. Or second and third generations of stars might have formed from gas that was richer in helium (and sodium) at a later time. Once the overall architecture of the cluster has formed then it becomes somewhat easier to isolate pockets of gas and stars within it, particularly if one population forms more towards the Globular Cluster Formation 119 cluster core where there was less overall movement of individual stars to the cluster periphery. Indeed, observations of NGC 6725 by Christian I. Johnson (University of California) and Catherine A. Pilachowski (Indiana University) clearly show that the sodium- rich stars tend to lurk closest to the core of the cluster: this is where most of the slow moving gas from AGB stars would have tended to accumulate under the pull of gravity. Whether this gas simply polluted lower mass members of the same generation or whether it gave rise to subsequent generations of stars is open to question. There are certainly problems with both scenarios. The former produces stars with the correct characteristics on the main sequence, but can’t explain more evolved sodium-rich giant stars, while the latter requires (possibly) unrealistically large amounts of helium to be present to form the second and third generations. Given the distinct main sequences that are visible in color- magnitude diagrams, whatever the explanation is, it has to allow distinct waves of pollution or star formation. For if this was a continuous process all of the main sequences would merge into a broad band, rather than appear as the distinct bands that are observed.

Evidence from the Physical Distribution of Stars

When astronomers dissect globular clusters like Omega Centauri or 47 Tucanae, they observe that the centrally distributed, sodium- rich stars all tend to move rather slowly and in no particular direction. They simply aimlessly wander around the cluster core. The more diffusely distributed sodium-poor stars orbit around these idle stars. The sodium-rich stars also have more strongly dipping orbits than their oxygen-rich brethren that take them in and out of the central regions of the cluster. This tends to favor the idea that the sodium (and presumably helium)-rich stars formed later on, within the core of the older oxygen-rich star cluster (Figs. 4.5 and 4.6 ). In this scenario they were spawned from gas shed by AGB stars with 5–8 times the mass of the Sun. Indeed, despite some contradictory evidence this is by far the most widely accepted model to date. Further validating this model is that the central sodium- rich stars also show enrichment in s-process elements, the very 120 The Complex Lives of Star Clusters a b c

FIG. 4.5 The evolution of stellar orbits over the last ten and a half billion years of globular cluster history. In 47 Tucanae the first generation of stars (blue ) have largely circular orbits that randomly flow around the heart of the cluster (a ). 100–200 million years later a second generation of stars is formed from the debris cast off by the first (b ). These form towards the centre of the cluster and have more elliptical orbits, possibly tracing the original motion of gas towards the cluster core. Over the ensuing billions of years, the orbits of these second generation stars gradually elongate until they are radial, like comets orbiting the Sun. These orbits take them repeatedly into and out of the cluster core where they engage in multiple relationships with other stars and binaries—and when they age, cause them to lose their outer layers (c ) ones synthesized by AGB stars (Chap. 2 ). Therefore, enrichment in helium certainly works and there is good circumstantial evidence to support it. Perhaps, given that the present day globular clusters still have enough gravity to overcome the energy in AGB winds, the process of pollution could still be going on. Thus different populations of stars could emerge from the smog of successive generations of stars. Although these aren’t new stars, in the sense that they are born from the current ashes of stellar evolution, they are at least stone-clad to give them the veneer of modernity. The question is can you make these populations sufficiently distinct in terms of their chemical make-up. It’s certainly possible, but confirming it will require a lot more computer modeling. The entire model for the formation of these multiple generations of stars hangs on one principle: that the globular clusters we see today were considerably more massive in the past and thus able to hold onto sufficient gas to manufacture multiple generations of stars. But is this true? Globular Cluster Formation 121

a b cooling flow c

e f

FIG. 4.6 A possible solution to the sodium-oxygen problem in globular clusters? In many globular clusters there are two populations of stars. The youngest are sodium-poor and are formed from primordial gas. ( a ) These form directly from the nebula. Stellar winds and supernovae clear away the remaining gas (b ). Less massive stars generates low winds that fill the cluster with sodium-rich, oxygen-poor gas. This gas drains into the cluster core in a so-called cooling flow, large arrows in c triggering a second wave of star formation (pink stars in d ). Therese stars are concentrated towards the cluster core where the gas concentration was highest. Over time the number of first generation stars dwindles ( e ) as they are stripped away from the periphery of the cluster ( small arrows ), leaving predominantly sodium (and helium)-rich second generation stars seen today ( f )

Do Observations of Young Globular Clusters Back Up This Model?

The worst thing that can happen to a good model is awkward data. A lot of time has been put into producing a model of globular clusters that accounts for the chemical properties of different groups (or populations) of stars. However, this model hangs by a couple 122 The Complex Lives of Star Clusters of fingers on a rather loose ledge. The problem with it is due to observations of some galaxies suggesting that globular clusters were not a lot more massive in the past than they are now. In order for that to be possible the original cluster must have been massive enough to hold onto the gas shed from its stars— particularly its intermediate mass stars. It was those earlier intermediate mass stars that would have produced the necessary quantities of helium-rich gas, from which the subsequent population (or populations) of stars would condense. Both the quantities of helium-rich gas required—and the speed at which this gas would have been released from these stars—match the observations of the helium-rich stars. The model works only if the original globular clusters are much more massive—perhaps ten times as massive—as the ones seen today. If they weren’t on that scale, there wouldn’t have been enough stars to produce the observed pollution, nor would the gravity of the cluster have been strong enough (in most cases) to hold onto the gas from these AGB stars. To investigate this Nathan Bastian (University of Liverpool) and Jay Strader (then at Michigan State University) used Spitzer infrared observations to examine 12 LMC clusters and one in the SMC. Within these middle-aged massive clusters models suggested that up to 10 % of the cluster mass should/could still be in the form of gas. Despite extensive probing, no gaseous reservoir was found. Indeed, Bastian and Strader’s observations constrain the mass of gas to less than 1 % of the mass of the cluster. If this observation is extended too much older clusters, then it would seem unlikely that globulars could have held onto enough gas to form this later population of stars. What if these clusters just aren’t big enough and the earlier clusters have lost most of their stellar mass? OK, this can’t quite be ruled out completely, but in another study, involving Jay Strader (this time at the Harvard-Smithsonian), Søren Larson (Radboud University, Nijmegen) and colleagues examined a population of very metal poor globular clusters associated with the dwarf Fornax galaxy—one of the Milky Way’s satellites. The Fornax dwarf galaxy is a rather unusual star city in that a disproportionate number of its stars are found within the five globular clusters it contains. This means that if stars have been lost to from the globular clusters they should now be found within the galaxy as a whole. However, Globular Cluster Formation 123 by examining the chemistry of the present stars in the clusters and the bulk galaxy, the authors of the study conclude that the clusters couldn’t have lost more than 80 % of their mass. Now, that sounds like a lot, and I suppose losing more than half your stars would seem rather brutal. However, thinking back to the models suggested earlier, they require that globular clusters were up to ten times more massive in the past than they are now, if they were to hold onto enough gas to fuel a second generation of stars. The Fornax globular clusters indicate that they can’t have been massive enough in the past to account for the presence of the current helium and sodium-rich population of stars that dominate many globular clusters today. Oh dear. The Fornax dwarf galaxy also provides clues to the sort of population of clusters that may or may not have existed early in the universe’s history. We’ve already seen that clusters dissolve over time. Perhaps then, there were far more clusters associated with galaxies in the past than are found presently. However, once again Fornax suggests that the present number is likely to be very similar to that formed early in the early history of the universe. Models of globular clusters that have them both much more massive and able to retain gas, were produced to match a con- temporaneous observation, not the other way around and this is always a problem. However, perhaps the halo of the Milky Way and other large galaxies contained a far larger amount of gas and dust early on and this helped to corral the stellar winds blowing from within the cluster? In which case the mass of the clusters might not have been so great and the requirement for star loss low- ered. Gas that was shed from the AGB stars was naturally trapped within the cluster and then able to fall back in and synthesize another crop of stars. The scenario is a little ad hoc but perhaps not unreasonable. Thus, globular clusters appear to be telling us contradictory stories. There is much evidence to suggest that helium is enhanced in some main sequence stars and that this must have come from moderately massive AGB stars. However, this asser- tion rests on a number of assumptions and is refuted by observations of some globular clusters in the Fornax dwarf spheroidal galaxy. Alternatives, such as fiddling with the amount of nitrogen seems more reasonable but they may not account for clusters 124 The Complex Lives of Star Clusters where there seems to be a very large difference in the color of the main sequence bands. Therefore, assume nothing at the moment. This is a truly strange phenomenon which does not yet have a satisfactory explanation. Element 23, therefore, presents something of a problem for our interpretation of the history of globular clusters as a whole. But that’s not all. Sodium also seems to affect the fate of the individual stars.

Salt in the Diet

In humans, a diet rich in sodium salt leads to poor health. High sodium not only puts stress on the kidneys, which can, in extreme cases, cause their failure, but it also leads to elevated blood pressure. In turn, this can damage arterial linings, which can cause stroke or cardiac arrest, depending on the site of the damage. Either way, a sodium-rich diet is a predictable route to a premature death. With a certain irony it now appears that what is true on Earth is also true in Heaven. To give you an idea of the problem, Campbell (Monash University) found that 70 % of stars in NGC 6752 were sodium-rich—a rather typical figure for globular clusters—and none of these sodium-rich stars became AGB stars: all of them died prematurely as Extreme Horizontal Branch (EHB) stars (Chap. 1 ). The only stars to return to the red giant branch after burning helium were the less populous oxygen-rich, sodium-poor stars. Whatever the issue is, it is unique to globular clusters. The implication is that the presence of sodium alone determines what will happen to the star once it leaves the main sequence. No other element seems to have this predictive power in astronomy—except perhaps helium—and as we’ve already seen, that’s a little trickier to pin down. To sum up: astronomers have found a first generation of stars that are rich in oxygen but low in sodium and essentially behave themselves according to both the theory of stellar evolution, following the trajectory of isolated halo stars. These objects do all the things other stars in the galaxy do that have their bulk. After exhausting hydrogen in their cores, they first become red giants; Globular Cluster Formation 125

E1 D1 C1

A C2 D3

D2 E2

FIG. 4.7 A reprise of the evolutionary paths of stars in globular clusters. The sodium-poor, oxygen-rich stars that comprise around 30 % of most globular clusters follow path A through B to E1. The majority of stars are more sodium-rich and oxygen-poor and these follow the path through C2, D2 to E2. The AGB phase is missed in its entirety fire up their helium core; take a second run at being a red giant; before expelling their outer layers and retiring from the pitch as white dwarfs. Sodium-rich stars, on the other hand like to do things differently. These don’t play ball: they run through the façade of becoming red giants, but then they appear to shrink and become very hot, blue helium-burning EHB stars before expiring (Fig. 4.7 ). There is no second ascension; no AGB phase. These objects are not found lying loose in either the halo or the galactic disc. They are unique to the globular clusters—if not all of them, most of them. Remember these stars are the majority in the cluster, therefore, their behavior is alarming. Once again, many astronomers turn their attention to the chemistry. Perhaps the (largely inferred) helium-rich main sequence stars become these upstarts that evolve into hot helium- burning stars: certainly this would make sense, wouldn’t it?

Do Helium-Rich Main Sequence Stars Become Helium-Rich Geriatrics?

To address this question, you must first look at humans. If you look at a population of people some are clearly wealthier than others. Some are truly rich, and perhaps, the luckiest of these are 126 The Complex Lives of Star Clusters also young enough to really enjoy their wealth. However, there are probably as many or perhaps more old, wealthy people than young ones. We’ve already seen some clusters contain sodium-rich and oxygen-poor stars as well as oxygen-rich and sodium-poor ones. The question is just like wealthy humans do all young, or less evolved sodium-rich, oxygen-poor stars turn into old sodium-rich, oxygen poor ones? In a nutshell: do the sodium-rich stars we see on the main sequence become the sodium-rich hot extreme horizontal branch stars, or are did they reach their apparent wealth (in sodium) through different means? The problem with this question is the main sequence turn off and the red giant branch. If you look at the color-magnitude diagrams of clusters, such as NGC 2808, which have multiple main sequences, these converge at the main sequence turn off: the point at which stars turn away from the main sequence and begin to transform into red giants. Sodium-rich stars reappear at the hot end of the horizontal branch, while sodium-poor ones occupy the cooler, red end. The hot stars are called extreme horizontal branch stars. However, are the two sets of sodium-rich stars truly related, or did they get their wealth of sodium in different ways? Imagine criminals escaping the scene of a crime in a mini. They drive off and dart through the traffic, while weaving in and out of buildings. You lose sight of them, before another mini appears in the distance. Is it the same mini, with its entourage of crooks, or is it a completely separate vehicle? Unless you’ve actually followed the full journey the mini has taken, you cannot be sure that the mini you see heading off into the sunset is the one that left the alarm bells ringing in the bank. The problem for astronomers is similar: you can’t actually watch a star evolve. You must resort to population studies if you wish to turn a snapshot into a complete anthology. Let’s turn our attention to the Extreme Horizontal Branch (EHB) stars, as these are rather unusual objects. They are mostly very hot stars with radii about half that of the Sun down to around the diameter of Jupiter. They are, therefore, rather dense objects that are made up mostly of helium with a thin layer of hydrogen sitting on top. Less than one hundredth of the star’s mass is hydrogen. Temperatures rise swiftly from the stellar surface to the hot core underneath where helium is actively being converted into Globular Cluster Formation 127 carbon and oxygen. This nuclear reaction powers the star. Any hydrogen fusion that’s occurring atop the core provides very little energy to this dense little object. Astronomers can be confident that the hot extreme horizontal branch stars have, proportionately, a lot of helium (and now, little hydrogen) because their high temperatures ensure that these elements are visible in spectroscopes. We also know that most have low masses: not far over the limit required to fire up helium fusion—around roughly half the mass of the Sun. Once again astronomers can be confident of this figure as it is determined by measurements of radial velocity. In this technique—now gain- fully employed to spot planets—systems containing two stars are probed by a spectroscope. As each star pivots around the other the distance and velocity of the two stars can be measured. This allows their masses to be calculated. What is peculiar about these stars is that they must all have come from stars that had roughly the mass of the Sun. That means these EHB stars have lost nearly half their mass while they were ascending the red giant branch. Stars of this caliber really shouldn’t be able to lose so much mass through stellar winds. A star, like the Sun, will probably lose about one fifth of its mass as it expands to become a red giant—strengthening stellar winds whisking the Sun’s outer layers off into space. The more metal-poor stars in globular clusters are expected to blow much gentler winds and hence lose much less mass. So, if that’s the case, how can a star that began life much like our Sun end up a shriveled but fiercely hot EHB star? In EHB stars astronomers can detect the abundance of sodium with ease, for the same reason that they can measure the amount of helium. What we don’t know is did the sodium-rich star of yester- year become the sodium-rich EHB star of today? If it did then the high abundance of helium in the EHB stars must have been present early on as well. Well, maybe: the helium might have been made in the star’s core… And, come to think of it, so might the sodium. In EHB stars, the hydrogen-rich outer layer is almost gone: it has clearly been stripped off in some manner. It is the loss of the hydrogen-rich outer layer that accounts for the missing 50 % of mass that was described above. Of the original mass of the star a 128 The Complex Lives of Star Clusters paltry 20 Jupiter masses worth of hydrogen remains in a very thin, inactive layer. Most astronomers believe that the hydrogen-rich outer layer is removed through some form of violent interaction with another star. Indeed, observations made by Pierre Maxted, then at the University of Southampton in the UK, suggest that at least two thirds of EHB stars are members of close binary systems. In these systems, the EHB star is paired up with another low mass star—sometimes another EHB star. Of the remaining third of EHB stars there is a high probability that these are also binary systems—but ones with longer periods, where each star takes up to 100 days to complete an orbit. In all of the detected systems, the binary partner would have easily fitted within the outer hydrogen- rich layer of the star. In the binary scenario the two stars orbited one another and one expanded into a red giant, the less evolved, smaller star began to enter the red giant’s outer layers. As it did so, friction would have caused the star to spiral inwards and scatter the increasingly fragile, outer hydrogen layer into outer space. After most of the outer layer was ejected (perhaps 98 % of it) the thin residual shell of hydrogen was too transparent to hold back the flow of energy from the core and shrank back onto it. The end result is a hot, planet-sized star. In many cases the resulting star, sometimes called a sub-dwarf B star, or sdB star for short, is too puny to ignite its store of helium and the star dwindles away as a helium-rich white dwarf. In many other cases, the former red giant is able to assemble a big enough core of helium to fire it up before all the hydrogen is removed. The result is an EHB star that can burn its helium fuel for around 100 million years or so. As the binary case is pretty much done and dusted, how does the sodium and helium fit in? EHB stars might make sodium as they are becoming red giants. Helium is easy: the amount of helium should be higher in these little stars as the hydrogen-rich outer layers have been stripped off exposing the helium core. Alternatively, this could be a grand cosmic coincidence. As the sodium (and helium)-rich stars are concentrated towards the cluster core there is a far greater chance that the star will become paired up in a binary system (Chap. 6 ). As such, the sodium-rich stars might just be in the wrong place at the wrong time. Being trapped within the cluster core makes them far more likely to become paired up in Globular Cluster Formation 129 binary systems and that ultimately results in them having their hydrogen-rich layers stripped off. This would then leave the small, helium and sodium-rich star behind. The more aloof sodium-poor first generation stars orbit further afield and evade such carnage. Interestingly, the binary route to helium-rich horizontal branch stars makes the connection to helium-rich main sequence stars irrelevant. It really doesn’t matter much what they were made of while they were on the main sequence. The loss of the hydrogen-rich outer layer exposes the helium-rich layers underneath. And as sodium can be made in this region, as well, any link between the original star and its horizontal branch descendent is lost. But what of oxygen: to where is this element disappearing and how? At present there really isn’t a convincing explanation for the loss of oxygen. It must be draining into the core of the star, but quite how and why remains to be elucidated.

A (Somewhat Silly) Gedankenexperiment

Just for fun, let’s reject the idea that these sodium-rich stars just happen to be innocent bystanders in the dramas that execute within the cores of globular clusters. Instead is the extra wealth of helium in the majority of globular cluster stars converting oxygen to sodium through some hitherto unknown route, then direct- ing the star on a catastrophic path? Imagine a contagion spreading through the cluster. One star accretes helium from a neighbor and the extra helium sets it on a course where oxygen is converted into sodium as the star evolves into a red giant. This star also ejects its new helium-rich envelope before or as it begins to burn helium within its core. The star then becomes an extreme horizontal branch star—but the helium it ejects, contaminates and alters further, neighboring stars. The process spreads through the core of the globular, where stars are in closest proximity and the risk of contamination greatest. Soon, nearly all of the stars in the globular core are helium and sodium rich. This is a nice idea, which is charmingly simple—and undoubtedly wrong. Could you really contaminate enough stars with helium in this manner? Maybe, if accretion caused the stars to spin 130 The Complex Lives of Star Clusters up and mix more helium from their cores into their envelopes. But it’s something of a long shot. So, at the moment we can assume sodium-rich stars on the main sequence are the progenitors of sodium-rich stars on the EHB—which are also rich in helium, suggesting the sodium-rich main sequence stars are as well. Whatever the solution is, it must be something unique to globular clusters, where the stars are the most densely packed. It must also be highly efficient, for not one of the EHB stars observed by Simon Campbell was sodium-poor or oxygen-rich, or conversely, none of the AGB stars were sodium- rich and oxygen-poor. It might be interesting to see if the younger clusters seen in the Large Magellanic Cloud (LMC) also show the sodium-oxygen anti-correlation, and to observe if sodium-rich stars also fail to become AGB stars. If they do, then we could have a unique opportunity to study the earliest events in the formation of our surprisingly complex globulars.

A Summary: A Confusing Picture Painted with Salt

The story of sodium in globular clusters is clearly a very complex one with somewhat contradictory, yet very tightly interwoven (or perhaps tangled) strands. On the one hand the sodium story suggests that globular clusters sported more than one period of stars formation and this implies that they were more massive in the past. There is also the implication that the abundance of sodium is tied to that of helium, with many globular clusters having a large enrichment in this second element—which then profoundly influences the evolution of the cluster stars. These results suggest that after a few tens of millions of years globular clusters produced a second, third or perhaps fourth generation of stars. These were produced from the ashes of intermediate mass AGB stars. More profoundly, this implies that globular clusters must have been more massive in the past. In part this is to allow the cluster to hold onto sufficient gas to make stars in the first place, but more importantly if you are to make 100,000 Globular Cluster Formation 131 plus stars you need an awful lot of star forming material to have come from the previous generation. That some globular clusters were actually the cores of small, companion galaxies to the Milky Way, suggests that in at least some cases this requirement was accommodated by the cluster lying within (or forming the core of) another galaxy. However, this cannot be universally true. Some observations indicate that many globular clusters cannot have been substantially more massive than they currently are. Therefore, the origin of all this gas to fuel second and third generations of stars is unclear—unless our clusters are different to those in the Fornax dwarf—and this seems unlikely. Finally, if sodium predicts a different pattern of star formation, does it also provide clues to the fates of stars one they have left the main sequence? Again, the evidence is equivocal at best. Yes, sodium-rich (and helium-rich) stars end up as EHB stars rather than AGB ones. However, is this chemistry or geography? Is it the location of these stars that is important in determining their fate rather than their underlying chemistry?

Have Globular Clusters Been Consigned to the Dustbin of History?

Leaving aside the trauma of stellar chemistry, it is apparent that the ancient globular clusters found within the halo of the Milky Way formed in very gas-rich environments. These formed sufficient numbers of stars that the clusters were massive enough to retain much of the gas shed when stars died. Although today’s clusters are often too wimpy to continue this process, when they formed around 10–12 billion years ago, they had, perhaps, ten times the number of stars, and hence mass as we can see now. In a few places, giant molecular clouds of gas and dust are compressed to the point that they spawn tens of thousands of stars. In star burst galaxies, such as M82 or the Antennae, collisions or near misses between galaxies have allowed giant clouds to collide and form stars with high efficiency. Observations of these galaxies in the visible and infrared portions of the spectrum show dense conglomerations of stars with the sorts of dense profiles seemingly unique to the globular clusters. Overall their masses 132 The Complex Lives of Star Clusters are comparable to the small to medium sized globulars found in our galactic halo. Therein lies the problem. The majority of the massive, or rich, clusters that are being formed now have tens of thousands of stars like our current population of globulars. Therefore, once completed they will look like these star cities. However, remember that what we see in orbit around the galactic core now may well be a pale shadow of what was once there. In order to make the sodium-rich stellar population most astronomers envisage clusters that had started out with ten or more times the present number of stars—we also, would of course, lose a lot of massive stars early on through natural wastage. Leaving Jay Strader and colleagues’ observations of the Fornax dwarf galaxy aside, if the present population of clusters were not substantially more massive in the past, then the cluster would not have been able to hold onto its sodium-enriched gas. Its gravitational pull would have simply been too weak. As far as we can tell there are precious few millionaire clusters currently being formed2 , let alone clusters with tens of millions of members. This epoch of star formation seems long gone—there simply aren’t places in the universe where there is enough gas to form this number of stars at once. Interestingly, recent observations by Hakim Atek (École Polytechnique Fédérale de Lausanne, Switzerland) indicated that shortly after the universe was formed, dwarf galaxies were forming the majority of stars and did so for several billion years. These objects have masses only marginally higher than today’s globular clusters but they appear to be dou- bling the number of stars they contain every 150 million years. Similarly, Volker Bromm and Cathie J. Clarke, both from the Institute of Astronomy, University of Cambridge, suggested that the globular clusters formed during the construction of dwarf galaxies. It seems likely, then, that today’s globular clusters are the descendents of this wave of massive star formation that constructed the dwarf galaxies. These multi-millionaire star cities then merged to form the elliptical and spiral galaxies we see today.

2 MGG 9, in M82, has approximately 1.3 million stars while MGG 11 (also in M82) contains 650,000. Most others hold less than 500,000 stars. 30 Doradus in the LMC has approximately 450,000 stars, for comparison. Globular Cluster Formation 133

This idea brings us back to the concept that some or perhaps all of the globular clusters we see today are really galactic fragments— the other stars having been ripped away by different forces acting in the ensuing billions of years as the larger galaxy assembled. Today’s globulars are in many ways unique. They probably cannot be formed again, at least not in their original form. We can make star cities with populations as wealthy as the old clusters appear now, but not as they did originally. Therefore, if we assume that globular clusters have lost many (most?) of their stars to accommodate the apparent enrichment in sodium/helium then the clusters must have held around ten million stars originally. We don’t see clusters of this magnitude forming anywhere in the universe today. Did the star cities we see today form only during the universe’s formative years? Probably not. Imagine a gradual deceleration, rather than a dead halt. Although the majority of these clusters did form when the universe was forming its current population of large galaxies, waves of later cluster formation have occurred. Around particularly massive galaxies, such as M87, there exist over 13,000 globular clusters. Although many formed with the galaxy, many more have formed when this giant of the Virgo cluster consumed smaller, gas-rich galaxies. Not only did these cannibal- ized galaxies contribute their own population of globulars, but the violent collision of their gas-rich discs with the hot gas surrounding the Virgo galaxy cluster core, generated a further wave of massive star formation. During the mergers 2–5 billion years ago, the metal-rich gas from the discs was compressed and birthed several hundred further globular clusters. As a mirror reflects our image, the formation of the younger globular clusters reflects the formation of the youngest population of stars within the old clusters. Just as the sodium-rich stars are more concentrated towards the cores of the old globular clusters, the youngest and most metal-rich globular clusters in galaxies like NGC 7252 and M49 are found more towards the centre of the galaxy than the older clusters. The oldest globular clusters of these galaxies formed within its halo as the galaxy was condensing from the void. Later collisions shed copious quantities of gas in long tidal streams into the halo. As these ribbons of hydrogen and helium-rich material crashed back in towards the cores of the 134 The Complex Lives of Star Clusters

colliding galaxies, the gas was violently compressed against the hot gas swirling around the core of the galaxy. This led to a population of slow moving clusters deep within the halo of these galaxies. Just like the distribution of stars in globular clusters, the galaxy hosts fast-moving, metal-poor outer populations of clusters, with the metal-rich clusters slinking within the inner halo of the galaxy. In principle this could happen today, but the longer the universe waits, the smaller the clusters will be. Imagine making cakes. Early on, with many ingredients lots of cakes can be baked. As the supply of one ingredient after another declines, the chances of making a big batch of cakes decreases, as well as the chance of making any cakes at all. So, as observations suggest, collisions between gas-rich spirals, such as the Antennae, do produce clusters of stars with the pretence that they are globulars, but the supply of cake mix is such that the total batch of cakes the galaxy can produce will never approach those made when the galaxy, and the universe as a whole, was in its infancy.

Conclusions

Although the basic premise describing the formation of the initial globular clusters is fairly well understood, it’s apparent that most if not all of these clusters have led rather complex lives. The precise details of this are controversial, with differing pieces of information seemingly contradicting others. In one scenario—supported by chemistry—the globular clusters formed their stars over a fairly prolonged period with distinct bursts. After the initial formation and violence of its youth, most clusters engaged in one, two or three subsequent rounds of star formation. These later rounds of synthesis occurred inside the deep, dense cores of the original globular cluster. Largely hidden from view, these later generations of stars had peculiar chemical compositions. For reasons that are still largely unclear, these odd chemistries have led these younger stars to evolve rather differently to their predecessors. Over time, the loss of the older generations has left clusters dominated by the younger, chemically odd stars. Globular Cluster Formation 135

Despite the appeal of this scenario the physics doesn’t quite back it up. There is no direct evidence that globular clusters were ever much more massive than they are now, this despite a steady loss of stars through processes that we will discuss in Chap. 7 . If they were not substantially more massive in the past, the chemistry becomes unsustainable as the star clusters would never have produced enough gas to fuel the vigorous star formation that is implied by the plethora of sodium-rich stars. Moreover, the assumption that the bluer stars are more helium rich than their counterparts is not directly confirmed in most cases, and where it is there evidence is weak. If this is wrong, then there are further problems. Finally, the bridge between the (assumed) helium rich main sequence stars and the helium-rich horizontal branch stars is surely broken. The latter can easily exist without the former through processes that are currently well understood. What remains to be established is quite how the EHB stars come about. The implication is that these emerge violently through interactions with neighboring stars, deep within the cluster’s heart. However, this remains to be proven. The science of these clusters is evolving at a pace as increasingly high resolution observations become more commonplace. Lurking within globular clusters is a rich tapestry of exotic objects, produced within their dense interiors, over their multi-billion year histories. As the resolving power of telescopes increases ever further astronomers will be able to dissect the lives of the inhabitants and hence get a better handle on the processes that have sculpted the cluster as a whole. A city is more than the sum of its buildings; its architecture more than the beams and arches that support its constructs. Over the coming years a wealth of new data will illu- minate how the various parts of these star cities operate and have evolved in step over the last 12 billion years. 5. Open Clusters

Introduction

Open clusters have played as much of a central role in astronomer’s understanding of stellar evolution as globular clusters have. For instance, by examining the overall population of the stars in open clusters, astronomers have learned how stars of different masses go from their stable hydrogen-burning phase to red giants and beyond (Fig. 5.1 ). Star clusters are populated by sufficient stars that they carve paths across the HR diagram that graphically illustrate how they evolve. Until recently, it was assumed that globular clusters were simply ancient relics of some bygone era of star formation. Meanwhile, the open clusters that populated the galactic disc were assumed to be all that the universe could synthesize nowadays from the scraps of hydrogen and helium that were left over. Yet recent observations have begun redrafting this picture. So, what are the differences, if any, between open and globular clusters, other than the overall number of stars that they contain? Are there any differences in the way they form and eventually fall apart?

The Structure of Open Clusters

Astronomers know that there are no millionaire open clusters and no globular clusters populated by as few as 1,000 stars. Instead, there are a range of clusters, from the sparsest of open villages, through sprawling towns, to the giant star cities such as M54. With each change in the scale of the population, in general, the density increases in step. The concept “generally” is important, as there exist the odd massive open clusters, which are very low in density. To make an analogy, compare a typical British city, such as Nottingham, with a typical US city such as Boston. In a British city, space is used very efficiently with housing close together.

0.00

2.00 Colour - Magnitude Diagram for the Pleiades

4.00 Main Sequence Turn-Off

6.00

8.00

Visual Magnitude 10.00

12.00

-0.20 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 14.00 Colour Index

FIG. 5.1 Colour-magnitude diagram for the brightest stars in the Pleiades. There are sufficient stars to show the main sequence and the location where the most massive stars (currently about five times the mass of the Sun) are evolving away to become red giants . The data for this plot came from the following website: http://www.atnf.csiro.au/outreach/education/ senior/astrophysics/stellarevolution_pleiadesact.html

This is a function of the cost and availability of land. The typical suburban garden measures less than 5 m by 10 m. By contrast, a suburban US household will be embedded in an area up to half an hectare in size, with an abundance of space. Thus a typical US city will sprawl over a larger area, per head of population, than a British one, where land is more limited in availability. Similarly, the stars in an open cluster tend to sprawl—and although the total area covered is less than that filling a globular cluster, the space between the stars is “used less efficiently” with a larger star-to-star gap. That said, we can make some general statements about those clusters that are not typically referred to as “rich”, such as the central cluster of NGC 3603 in the Milky Way or NGC 1805 in the Large Magellanic Cloud. For example, in most cases it is assumed that open clusters are looser and proportionately smaller aggregations of stars than globular clusters and that these are generally young or in a few cases of “intermediate age”, meaning they have ages approximating a billion years or so. However, there are a number of problems with this definition that we have already encountered, notably the “rich clusters”: clusters that are young, yet as Open Clusters 139 massive as some present-day globular clusters. How do these fit into the scheme? Given that open clusters follow the same overall structure as globular clusters, this implies some common rules for their formation and development. An open cluster usually averages about 40 light years across. Approximately half of the cluster’s stars are found within a dense core that is approximately 3–4 light years across. The remainder of the cluster’s inhabitants occupies a more diffuse corona that extends the full 20 light years in radius around this core. To put that in perspective, a globular cluster may be 200 light years across with half its mass filling a core at most 10–20 light years wide at its heart. More mature clusters with ages in tens to hundreds of millions of years have a generally fairly smooth appearance with little evidence of internal structure. The smooth appearance is a product of internal processes more than anything else. That said, there is a lot of variation within open clusters that relates to the manner in which they formed. Typically, the more massive the cluster is, the greater the chance that there will be massive stars present. The Pleiades and Hyades, while too old to host massive stars now, almost certainly never had any to begin with. There is no evidence that either cluster hosts any neutron stars or black holes, though it is possible that any that they had have been ejected from the cluster when their parent star died in a supernova. Such violent deaths often give the nascent neutron star or black hole a kick that would easily overcome the weak gravitational bonds of the open cluster. Massive rich open clusters, such as that which forms the core of NGC 3603, may host tens or even a few hundred massive stars. However, in terms of sheer stellar numeracy many of these rich clusters resemble young globular clusters more than open clusters, like the Pleiades. Exactly why some nebulae produce mostly low mass clusters while others synthesize a wide range certainly owes something to the overall mass of the nebula that spawned them. However, many other factors are certainly at work to affect the overall efficiency of star formation (Chap. 2 ). A simple glance at NGC 604 shows a lot of stars, potentially hundreds of thousands, but their distribution is anything but globular. 140 The Complex Lives of Star Clusters Old Yet Open?

In Chap. 4 we saw that some galaxies appear to sport globular-like clusters that are young. Is the converse true: are there any open clusters that are old? There are some relatively old globular clusters with ages measured in billions of years. Astronomers know of many dozen clusters in the Milky Way including the 625 million year old Hyades, or the old, rich (and officially open) cluster M67. Indeed, 21 years ago Randy Phelps and colleagues at the Carnegie Institute carried out an extensive literature and CCD-based photometric survey of potential old open clusters. In this they identified 19 clusters older than 5 billion years—the official age of M67. Of these, one, Berkley 17, appeared to be older than 10 billion years of age. However, you must put this in perspective. The 19 clusters older than 5 billion years that have been identified are approximately 1 % of the 1,100-plus open clusters that exist in the Milky Way. Therefore, although we might say these old open clusters blur the edges of the open versus globular categories, it is only a slight overlap. Why then are these cluster types different? Why are there so few old open clusters? This question will be considered at length after a brief jaunt through the somewhat awkward classification system of these clusters. Once this is done, you will understand something of the diversity of open clusters and perhaps get more of a feel for the problems cluster formation presents members of the astronomical community.

Classification of Open Clusters

Open clusters, like globulars, have their own system of classification based on their physical appearance. The Sawyer-Hogg system is used to classify globular clusters based on the size and central density of these clusters. The Trumpler classification system aims to do the same for open clusters. This system is more somewhat more detailed, incorporating three separate characteristics for each cluster, each signified by a separate number of letter. The first number is a Roman numeral from I to IV indicating both the concentration of stars and how readily the cluster is distinguishable Open Clusters 141 from the neighboring field of stars in the galaxy. The second number is an Arabic 1–3. This describes the range in brightness of the stars within the cluster. Three indicates the largest range. Penultimately, there is a letter p , m or r , which describes the overall richness of the cluster ( p oor, m edium or r ich). Last, but not least, some clusters carry a final, fourth letter n if they have any associated nebulosity. The Pleiades, thus, carries the designation I3rnn, meaning that it is strongly concentrated towards its middle; has a broad range of stellar luminosity; is rich in stars; and it has some nebulosity present. The classification system is superficially easy to grasp. However, as Fig. 5.2 shows, putting it into practice is far more difficult. Trumpler was something of an ace at doing this.

IIIIIIIV

FIG. 5.2 Some examples of open clusters inserted into the Trumpler classification system. Unlike the globular clusters, which follow a fairly predictable pattern of luminosity and density, the Trumpler system is rather more difficult to visualise. From left to right clusters become more diffuse (using Roman numerals I through IV ). From top to bottom , clusters are assigned r (rich); m (medium); p (poor) depending on the nature of the stars the clusters contain. Given the vagaries of astro- photography the pinning down the differences in photographs is really rather tricky. The second character in the system (not shown) would make this square a cube and denotes the range of brightness of the stars 142 The Complex Lives of Star Clusters

The main obstacle for any amateur trying to duplicate his efforts is the resolution of the optical instrument he or she is using. Many of the open clusters lie fairly distantly to the Earth, in the galaxy’s spiral arms. Distance naturally diminished the luminosity of the component stars, limiting the numbers that are visible. Distance also adds an increasing mist of interstellar dust and gas that blurs the differences between component stars, as well as fading out those naturally dim stars. This dimming (and reddening) effect is known as interstellar extinction . Unless these factors are managed it becomes difficult to correctly assign the cluster. By contrast, most globular clusters lie far above the galactic plane and suffer considerably less interstellar extinction, making them far easier to understand. Nonetheless, given these complications, the Trumpler system is something of a triumph and clearly required extensive skill and patience. The system, with three (and occasionally a fourth) variable makes an interesting stellar cube, to lie alongside the Sawyer-Hogg square used in globular cluster classification.

How Birth Determines Life

As was alluded earlier, there is a great and largely unexplained diversity in open clusters. In the Local Group of galaxies the most obvious contrast lies between NGC 604, in the nearby Triangulum galaxy (M33), and R136, the central cluster in 30 Doradus. Both are only 3–5 million years old and each contains a few hundred thousand stars. NGC 604 hosts around 100,000 stars, while R136 has around half a million: the exact numbers cannot be known as the vast majority of the lowest mass stars are too faint to be seen in the glare of their more massive neighbors. The mass of each cluster is comparable, within an order of magnitude, registering around 450,000 Sun’s for R136 and 100,000 for NGC 604. However, that’s where the similarities end. NGC 604 spreads it mass over 1,500 light years, while R136 concentrates its mass and light within a sphere a miniscule 35 light years wide. As a result, NGC 604’s central cluster is more of a stellar association than even a fairly dispersed open cluster (Fig. 5.3 ). The central association of NGC 604 is nothing like the concentrated, globular-like appearance of R136—or even the Milky Way’s comparable cluster, NGC 3603. Open Clusters 143

FIG. 5.3 The internal structure of the core of NGC 604 in M33. This 1,500 light year wide stellar association comprises a number of visible structures, a few of which are indicated by shading, here. Blue ovals represent smaller sub-clusters of stars that are dominated by a few massive O and B class objects. Pink and lilac blobs represent wider associations of stars, again dominated by several massive O-class stars. Blue lines represent shells of massive stars. These clusters and walls of massive stars form crests on a turbulent sea of low and intermediate mass stars. The total mass of the region exceeds 100,000 Suns. Undoctored, original images credit: NASA/ESA

Similarly in M82, two massive young clusters, MGG-9 and MGG-11, have reasonably similar masses (1.5 million and 350,000 solar masses, respectively) but these, too, are very different beasts. The less massive MGG-11 is a lot more compact than its very close neighbor. The half-mass radius (the radius in which half the mass of the cluster sits) of MGG-9 is 2.6 light years, while MGG- 11 has a half-mass radius of 1.5 light years. Thus MGG-11 has a little less than one quarter the mass of MGG-9, but is far more centrally concentrated than MGG-9. Therefore, as both examples show, the density of a cluster is not simply a function of its mass: other factors must be at play. This makes it tremendously 144 The Complex Lives of Star Clusters hard to issue predictions about how clusters form and what will ultimately emerge once the gas clouds from which they are born disperse. Subtle differences in the size of cloud, the distribution of material and the manner in which the cloud was compressed and collapsed has serious consequences for the fate of the ensuing cluster of stars. Imagine a billion butterflies beating their wings in slightly different ways. Minor differences between the location and timing of each wing beat seem to make all the difference to what is eventually formed. Not only do the initial conditions within the cloud determine the size and distribution of the stars, this in turn affects how the cluster will mature.

How Open Clusters Come Apart

An open cluster may contain up to 10,000 stars, most of which are low mass. Again, remember that the figure 10,000 is very loose and that the distinction between open and globular clusters is very grey indeed. The cluster may have been born part of a larger association of stars called an association, which may itself play host to several open clusters, all of which formed at roughly the same time. Again, NGC 604 is a case in point. Even a cursory glance at its constitution reveals a plethora of small clusters, embedded within a fairly well structured, yet very loose, agglomeration of stars. Such associations are not gravitationally bound and clusters within them are largely free to move in isolation. So, how do open clusters and associations fall apart? Imagine two blobs of cold molasses held together on a string. This provides a useful analogy that explains what happens to this family of stars. When the association is young a considerable amount of mass is held in massive and intermediate mass stars. When these die, they shed most of their mass into the surrounding space where stellar winds, supernovae and friction against surrounding gas eventually disperses it. This scatters much of the mass of the association, further weakening the bonds between the stars. The stars that remain in such an afflicted open cluster suffer violent relaxation. The wholesale expulsion of mass alters the gravitational forces that act upon the remaining stars so that their motion Open Clusters 145 within the cluster is disturbed. This randomizes their orbits and sets in motion the later acts in the drama that will tease the stars away, one by one. About 150 light years from R136 lies the slightly older Hodge 301. At 25 million years old almost all of its massive stars have now exploded, leaving a 4,800 solar-mass open cluster embedded in a violently broiling sea of hot gas. Once the last star has exploded (most likely some blue stragglers—see next chapter) and this gas has dispersed, what remains of Hodge 301 will begin to settle down as a more sedate open cluster. Acting from outside the cluster is the force of gravity from surrounding clouds of gas and dust. An association and its clusters may be spread over a few hundred light years, differences in the pull of gravity across the cluster will do most of the damage. You can broadly break the death of the open cluster into two phases. The first phase lasts a few million years and occurs immediately after the cluster has formed. This primarily afflicts those massive clusters that have a reasonable population of massive stars. Stellar winds and supernovae cause much of the mass of the star cluster to be pulped and scattered into interstellar space. The remaining stars are now held together with much less gravitational force than they were before. As a result they tend to spread out. This weakens their bonds and causes them to redistribute their sluggish orbits around the cluster’s centre of mass. Many small stars on the periphery of the cluster may escape altogether at this point, while others, which were part of binary systems, find themselves thrown off at tangents when their more massive partner self-destructs. This process is called violent relaxation. This is a lovely term for a chaotic mess of stars that are slewing around in all directions. Most clusters survive this phase, battered and bruised, but still gloriously alive. Many smaller clusters such as the Pleiades and Hyades may never have encountered this phase as they never gave birth to sufficiently massive stars. Both the Hyades and Pleiades appear to have birthed stars with less than 7–8 times the mass of the Sun as there is no evidence for neutron stars or black holes in the presently observed clusters. Only those clusters with a retinue of massive stars will suffer the traumatic machinations of violent relaxation. 146 The Complex Lives of Star Clusters

However, over the next few tens and hundreds of millions of years the gradual loss of matter, as stars become red giants, also causes the cluster to spread out. It is from this point onwards that the galactic gravitational field acts . After the carnage wrought by violent relaxation there are other, longer-lasting forces at play which will be looked at in the next chapter. In essence this variety of forces tends to operate on timescales longer than most open clusters can keep together. Open clusters are most vulnerable to the effect of the galaxy’s pull as a whole. To think of how this happens think again of our two blobs of molasses held on a string. You want them to keep together, but various forces are causing you to lose cohesion, or stickiness. When the blobs of molasses are cold this is akin to the star cluster when it is first formed. The supernovae that happen soon after the cluster forms heat up this sticky analogy, causing it to become soft and malleable. At this point the odd blob can come away on its own. This is the process of violent relaxation, with the escaping blobs representing some of the cluster stars. Now imagine you’re not only heating the masses up but swinging them about your head. Forces acting on the string, cause soft blobs to slide along it, which allows pieces to fly off. An open cluster of stars is relatively diffuse with one side closer to the centre of the galaxy than the other. Moreover, within the plane of the galactic disc lie other clusters with their own gravitational pulls and giant molecular clouds of gas and dust. These also exert a gravitational pull on the cluster. Therefore, as our two blobs of molasses swing around one end is pulled more strongly than the other. This causes the stars nearest the mass that is tugging on them to move faster towards it than the stars that are furthest away. Within the galaxy, those stars closest to the centre of any neighboring mass (or the galactic centre) will be pulled slightly more than those further away. Over the time it takes the cluster to orbit the galactic centre, differences in gravitational forces will tease the cluster apart by changing the amount of acceleration each star experiences. This causes the distances between the stars within the cluster to increase, until the entire cluster is stretched out and loses its internal structure (Fig. 5.4 ). These tidal forces generally destroy open clusters within 1 billion years of their formation—barely a few orbits for most of them. Open Clusters 147

a b c

Giant Molecular Cloud

FIG. 5.4 The dissolution of open clusters in the galactic disc. As a weakly bound open cluster orbits the galaxy it will encounter other star clusters or massive clouds of gas and dust, known as giant molecular clouds (GMCs), with up to several million times the mass of the Sun. Stars closest to these large masses experience the strongest gravitational forces and accelerate more than those further away. Over time, this stretches the orbits of the stars within the cluster until they are able to escape its gravitational confines

There are exceptions, of course. We’ve already met them: the old open clusters. Berkley 17 (Be17) has an estimated age of 10.07 billion years, making it a contemporary of the globular clusters. Be17 is still recognizable as an open cluster today but its numbers have been stripped down to around 400 individual stars and binaries. How is it Be17 survived? The simplest reason is probably that it was born massive. Perhaps bequeathed with a few tens of thousands of stars, which were suitably close to one another, Be17 was teased and kneaded by the galaxy’s gravity but had enough internal strength not to be spaghettified out of existence. It’s not as if the globular clusters are immune to this effect. Two of the smallest and youngest globular clusters, Palomar 5 and 12 are already at their ends (Chap. 7 ). For example, high resolution imaging of Palomar 5 shows that the central cluster with its tens of thousands of stars is embedded in a long, diffuse train of stars that extends 13,000 light years from the cluster core. The whole mass moves coherently through the halo of the galaxy, but within it, those stars that are outside the cluster are now moving with enough energy to prevent them falling back into its core. 148 The Complex Lives of Star Clusters

The escape velocity—the speed at which an object will have to go to break the clutches of the cluster—is around 3–4 km/s (6,750–9,000 miles/h). Although this speed is quite fast by our standards, it’s an easy pace by stellar ones. To put the number in some sort of astronomical perspective, the Earth orbits the Sun at more than ten times this speed (67,208 miles/h). Place the Earth in Palomar 12 at its orbital speed and it would be long gone in a few hundred thousand years.

Location, Location, Location

The idea that mass determines how well a cluster will hold together works but it cannot be the whole story. For one we do not know how many stars Berkley-17 had to begin with. It is simply too difficult to trace back the orbits of any surviving stars that have escaped this cluster. Old clusters such as M67 have around 700 solar masses of stars, while the 12 billion year old Berkely-17 has been reduced to around 400–500 solar masses, depending on the number of very low mass stars that are present. Excluding that shed by stellar winds and early supernovae, a vast amount could still have been lost to the surrounding galaxy as stars were system- atically winnowed from it. Yet, within the galaxy there are profound differences in the strength of the forces acting on clusters. Those closest to the center of the galaxy experience the harshest tidal forces as they whizz around the central super-massive black hole, while those far out in the disc have a fairly leisurely ride around our galaxy. Could this then be a deciding factor in determining the fate of open clusters? Indeed, is it far more important than the initial mass of the clus ter? Simon Portegies-Zwart (working at MIT) and colleagues generated computer models of clusters that are found within 200 parsecs from the central super-massive black hole. They found that most clusters—even those larger ones with the characteristics of the present day Arches or Quintuplet clusters—would dissolve within 70 million years if they were located within a radius of 200 parsecs (around 700 light years) of the galactic center. On top of this, in less than 20 million years, forces acting on the cluster Open Clusters 149 stars would effectively dissipate the cluster to a point at which it was unobservable from the Earth. To be clear, the cluster would still exist in some form, but the density of stars would be so low as to make it unrecognizable as a cluster from our vantage point. Thus, even quite significant rich clusters, with large populations of stars, will be shredded if they are located in more “challenging” environments. Meanwhile out by the Sun, gravitational forces are somewhat gentler, thus explaining the longer lifespan of such stellar families here. Indeed the presence of the Arches and Quintuplet clusters is somewhat puzzling. Each contains around 10,000 solar masses of stars inside a dense, globular-like structure. In each, around 5,000 solar masses of stars is found within the central 3 light years of the cluster, a density readily comparable to that found in a modern day globular cluster or R136 in the Large Magellanic Cloud. At present it isn’t clear quite how these clusters formed. They certainly haven’t migrated inwards as they are too young to have done this. Instead, they probably formed within clouds of gas and dust dragged inwards from further away. Any remains of this original cloud would soon have been dissipated by a mixture of strong stellar winds from the stars in the cluster, and tides acting near to the galaxy’s heart. Further work by Portegies-Zwart, this time at the University of Amsterdam, in collaboration with Piet Hut (University of Tokyo), compared models of how open clusters come apart at greater distances. An interesting outcome was the timescale involved for those clusters far removed from the galactic core. Although most models had shown that the average lifetime of an open cluster orbiting far from the galactic centre was less than 300 million years, Portegies-Zwart’s simulations of larger open clusters indicated that these could survive for up to 4 billion years, and in some cases much longer. Within the first 300 million years or so the mass of each cluster decreases by about 20 %, or one fifth, as the most massive stars die and shed much of their mass into interstellar space. Although this does loosen the bonds between the stars, it isn’t enough to completely dissipate the cluster. Instead secondary forces acting within the cluster are reduced (next chapter) and as the remaining stars hold together more securely, gravitational tides from the galaxy are not enough to pull the stars apart for billions of years. 150 The Complex Lives of Star Clusters

Among the more important processes is something called the “relaxation time”. This is looked at more fully in the next chapter, when the fate of globular clusters is examined. In essence, it is the time it takes a star to change its direction by 90°. In a star cluster this is important because it means that any star that is “relaxed” has not only had the time needed to pass close to another star, but to gain enough energy to change its direction. This force tends to cause the stars to spread out. In Portegies-Zwart’s work, this time is longer than the original models suggested, at least for clusters born with more than 10,000 stars. In these clusters, stars drift around the centre of gravity at such slow paces that they rarely come into close enough contact with one another to affect one another. Consequently, the cluster retains much of its integrity and the stars are only affected by tidal forces over longer timescales. Thus open clusters born in the galaxy’s spiral arms should be identifiable for billions of years not tens or hundreds of millions as was once thought. Using this model, the continued presence of M67 or Berkley-17 becomes more understandable. What does matter, from an Earthly perspective, is whether the stars are identifiable as belonging to a cluster. Not only do the numbers of stars decrease with time, but those that go first are the brightest. Once those stars that were born with more than twice the mass of the Sun have died (approximately 800 million years) those that are left are relatively faint. If instead you view star clusters as a smooth distribution of objects of different masses, rather than specific categories, then one would expect a variety of outcomes. Those born with lower mass, or within regions with strong gravitational tides, will lead shorter lives than those born in more benign climes. Another distinction between globular and open clusters is thus blurred. That we see differences today is down to the circumstances in which these clusters formed as much as the number of stars they contain. Remember when the galaxy was forming not only was there a far greater abundance of fuel, but the gas was distributed over a greater volume of space. The motions and temperatures within the gas differed from those of today and the clusters, once formed, predominantly, led fairly passive lives drifting through the largely empty void of the galactic halo. Open clusters lead more dramatic lives with less mass to hold them together. Gravitational forces Open Clusters 151 are more extreme within the disc of the galaxy. Pass too close to the clutches of a million solar mass, giant molecular cloud and you could be history. An interesting analogy has been made between open clusters and radioactive decay. The latter is a process that afflicts many isotopes of chemical elements and it is an inherently unpredictable process. When physicists or chemists talk about the decay timescale they use the term “half-life”. This refers to the time it takes half the mass (or half the number of nuclei of atoms) to decay. This is convenient as you cannot say which nuclei will decay at which time. Similarly, open clusters (indeed all clusters) can be thought to have a half-life. You can’t say which cluster will fall apart and when, but looking at a population of clusters you can say with some confidence that a proportion will fall about in a given time. A half-life for clusters is thus a useful idea if you have a sufficient population with a particular set of characteristics. Cluster half- lives range then from 150 to 800 million years and this is broadly in agreement with predictions made for star clusters of different masses and in different parts of the galaxy. The fate of these clusters is to dissolve into co-moving agglomerations of stars called “moving-groups”. These stellar associations move through the galaxy as a coherent (or fairly coherent) mass but they are no longer bound together by gravity.

Planets in Globular Clusters

In the core of an open cluster there are approximately 1.5 stars per cubic light year, depending on the nature of the cluster. Compare this to the density near the Sun of less than 0.003 stars per cubic light year. However, were we to place the Solar System within the core of a cluster such as M67 (Hyades) there would be little chance of a collision between the Solar System and a neighboring star. Thus our planet would most likely continue to orbit the Sun safely for the time the Sun remained on the main sequence. Given that most stars, the Sun included, formed within open clusters, you would think it a fairly safe conclusion that there would be planets orbiting other stars in these stellar villages and towns. 152 The Complex Lives of Star Clusters

Despite this, many astronomers still doubted that planets would maintain stable orbits around the stars of an open cluster. Their doubts were finally allayed in 2007 when Epsilon Tauri b was found orbiting a red giant in the 625 million year old Hyades cluster. This effectively put an end to the argument. To rub in the point, a further three planets were found in the open cluster M67 and two (Kepler 66b and 67b) in the 1 billion year old open cluster NGC 6811. Although the number of such planets remains small, the detection of metal-rich gas in the atmospheres of many open cluster white dwarfs implies that planetary systems are commonplace. That said, the detection of dust in the atmospheres of these stars is not a direct detection of planets per se, rather it is the astrophysical equivalent of crematoria ash as a proxy for former life.

Conclusions

Open clusters are the jewels in the crowns of most star-forming galaxies. They form a continuum of masses and ages that extends back to the globular clusters that formed with the galaxy. A combination of deep observations and more accurate computer modeling has revealed a more hardy personality for open clusters. Despite their apparent frailty, it turns out they have a lot more back-bone than was once more thought. Although, in general more ephem- eral in nature than the (generally) more massive globular clusters, place these butterflies in a gentle breeze and they can fly just as well as the crows that swoop around them from on high. Remember that open clusters populate the galactic disc where harassment can happen multiple times in each orbit around the galactic central point. A globular cluster, by contrast, only experiences such trauma twice per orbit around galactic central point. This is when they penetrate the galactic disc. Despite their limited exposure to the brutalities of the Milky Way’s disc, it is clear that they, too experience post-traumatic stress disorder and show ample signs of distress on each passage (Chap. 7 ). No cluster is immune to the effects of wear and tear: decrepitude and eventual dissolution afflict them all. As Chap. 6 will illustrate, the impact of this stressful existence can be rather dramatic and Open Clusters 153

exciting. Yet, as Chap. 7 continues, too much excitement can be a bad thing: even the globular clusters will succumb. The forces that bring about the downfall of star clusters are universal. By the close of the final chapter, even the galaxies that our beautiful open and globular clusters inhabit will suffer the same affliction. Thus the stories the open and globular clusters tell are but a mere summary of the story that will affect every bound stellar system in the universe. Read on. 6. Stellar Soap Operas

Introduction

Within the heart of an open cluster stars are spaced perhaps 1 light year apart. Nine trillion kilometers is a long way between stars, particularly if those stars are only moving at a few kilometers per second. What are the chances that they would ever encounter one another given their apparent distance? You’d think an encounter would be unlikely, and that might be true for a single star drifting through the core of the cluster. However, if that star is part of a binary system—a partnership of two stars—then the effective distance between the binary and the neighbor is reduced because the binary system is much larger than its component stars. The binary, in effect, forms a much larger target for some sort of collision. Take the story to a globular cluster and things get even spic- ier. At the heart of these star cities, stars are separated on average by a diameter comparable to the Solar System. It is then very likely that each star will feel the pulling power of its neighbors. Put a binary system in the mix and the chances of an encounter become very likely indeed. But just how likely is “likely”, and what sorts of scandalous behaviors are possible when these stars come close to one another?

Binary Star Systems

When stars form it had been thought that the majority, approximately 60 %, were formed as pairs or higher associations such as triplets and quadruplets. This was the perceived wisdom for decades. However, when Charles Lada (Harvard-Smithsonian Center for Astrophysics) took into account the universe’s dimmest stars, the red dwarfs, that figure took something of a nose

© Springer International Publishing Switzerland 2015 155 D. Stevenson, The Complex Lives of Star Clusters, Astronomers’ Universe, DOI 10.1007/978-3-319-14234-0_6 156 The Complex Lives of Star Clusters dive. Instead of 67 % of stars being binaries, 67 % turned out to be singletons. Now, contrast the proportion of binaries in the open galaxy with those found in dense stellar clusters. In the fairly old open cluster NGC 188, 76 % of stars are found in binary systems. Similarly, within very young clusters, the majority of massive stars—around 70–80 %—are also found in binary systems. That there are similar numbers of very young massive binary systems to old binary systems might lead you to conclude that all of those found within the old cluster formed early on as the open cluster was condensing from its natal cloud. If the binaries formed later, you might expect an increase in the number of binary systems with age, and though this effect may be present its impact is probably not enough to raise the proportion by more than around 10 %. However, in star clusters simple counts of binaries are misleading. Although more common in the denser, globular clusters, it appears that binaries are constantly created and destroyed as these star clusters age. Calculations made by Simon Portegies- Zwart and co-workers suggest that approximately one binary system is destroyed roughly every 30 million years. Obviously, if old clusters have marginally more binary systems than young ones, this implies that they must be created at a similar, if slightly higher rate than they are destroyed. Therefore, the fact that an old cluster like NGC 811 has roughly the same numbers of binary systems as the younger systems does not mean that young binaries become old ones. Instead there is a rather dramatic interplay between the components of binary star systems, both with their birth partners and with other interloping stars within the cluster. It is these interactions that produce some of the most interesting parties in globular clusters. A loose Hollywood analogy is the number of marriages: there isn’t much of a change in number with increasing actor age. Yet, that simple observation belies the rapid exchange of partners that occurs as each actor (or at least many of the actors) ages. The abundance of binary systems also suggests that the environments that spawn these star clusters are also very efficient at forming binary stars. Although it isn’t entirely clear why this would be so, the most likely reason will come down to the distribution of momentum of the material that forms the stars themselves. Stellar Soap Operas 157

If the collapsing cloud had a lot of momentum, particularly a property of spinning matter known as angular momentum, then it would be likely that the stars that were forming would end up as binaries. Imagine once more the analogy of molasses on a string. This time take one blob and imagine it to be a protostar in the throes of its formation. This gloopy mass of sugar will tend to break up into smaller portions if it is flung around. Much like a spinning ice skater pulling his or her arms in, as the cloud of gas collapses, the conservation of momentum ensures that it spins ever faster as its diameter shrinks. Given enough spin, the matter will end up forming more of a hoop-like structure within the inner portion of the disc. This hoop then fragments into stars that are bound by gravity into overlapping orbits around their center of mass. These binaries are known as primordial because they formed with the cluster. These primordial binaries will then evolve in different ways, depending on their mass and the extent of their separation. Wide binaries are the most likely to experience interactions with neighboring stars as they form a large target for a collision. Imagine trying to shoot a grape with a bullet from 100 m, then, compare this with attempting to shoot a watermelon from the same distance. It’s a lot easier to hit the watermelon. Conversely, the stars within tight binaries tend to experience interactions with one another rather than with outsiders. This is particularly true as the component stars evolve away from the main sequence, gain size and begin to overlap one another’s field of influence (Fig. 6.1 ).

General Principles

Where stars are formed more than 5–10 A.U. (Earth-Sun distances) apart the two stars are unlikely to affect one another very much during their lives. Aside from the incessant pull of gravity as one star and then another become a red giant there is a sufficiently wide separation of the stars to prevent much movement of material from the giant to its neighbor. Perhaps once in a while a white dwarf might grab hold of its neighbor’s stellar wind to fire off a nova, but otherwise too little material will flow between the stars to do anything other than pollute one star with gas from its sibling. 158 The Complex Lives of Star Clusters

FIG. 6.1 Interactions between stars. In a , a lone star is passed by another star. The sizes of the two stars relative to the distance between them makes an encounter, let alone a collision highly unlikely, even within a dense star cluster. In b , a tight binary system present a larger target for a collision or interaction with a third, passing star. Yet, the chances of a collision are still relatively small, as the size of the orbits is still small (a few million kilometers) compared to the size of each star. However, these stars are close enough to one another to allow them to interact directly. In c , a wide binary presents a large target (perhaps billions of kilometers) for a passing star. Within a star cluster, the chances of a collision are far higher. In most cases one star in the binary (usually the lightest) is ejected and replaced by the interloping star

Yet, the evolution of a star can have other consequences for the partnership, and for other stars in the cluster. When these stars evolve into red giants, and beyond, they lose roughly half their mass which weakens the bonds between them. This can be enough for the two stars to become separated if another star ventures close enough to pull upon them. These wayward stars can then either escape the cluster altogether, or form new partnerships with other stars. Where the interloping star becomes bonded to the remaining white dwarf, more interesting fireworks can take place. Stellar Soap Operas 159

However, one of the key processes that takes place is a tightening, or hardening , of binary partnerships when the two stars encounter a third. Like a threatened couple, they can be driven closer together as they give up some of their energy to the passing star. In return the passing star, having now gained energy, accelerates away. This sort of interaction can help stave off the effects of gravity on the cluster as a whole, keeping it afloat, so to speak. The consequence is that some initially rather wide binaries get squeezed until each of the component stars are forced to interact or even collide. Consequently, many more stars become engaged in rather wild behaviors that make star clusters particularly interesting places to investigate. The numbers speak for themselves. Although all the globular clusters combined contain only one thousandth mass of the galaxy, these star cities contain around 20 % the number of so-called low mass X-ray binaries, partnerships between neutron stars and Sun-like stars. Within the Milky Way they also hold around half the total population of binary pulsars and over half the galaxy’s store of millisecond pulsars, spun-up, pulsating neutron stars. These partnerships have been born through encounters within the close confines of the cluster. Although less densely populated than globulars, open clusters contain a high fraction of binary stars, many of which are engaged in exotic behaviors. The nature of the stars in young and more mature clusters changes the kinds of interaction which are possible. Therefore, this chapter is broadly split down the lines of age.

Young Clusters

Besides a similarly high fraction of binary stars, there are some notable differences between the sorts of partnerships seen in open and globular clusters. On the whole open clusters are identifiable as younger than their globular cousins, at least within the Milky Way. Most importantly, the youngest clusters play host to the galaxy’s most massive stars. These stars lead such short lives that they are almost exclusively found inside a galaxy’s open star clusters and associations. They tend not to be born alone and certainly don’t live long enough to escape from any cluster that they 160 The Complex Lives of Star Clusters were born in. Therefore, it’s in the youngest clusters that the most massive stars will live and die. When astronomers look at young open clusters, they see frequent pairings between particularly massive O and B-class stars. Within 30 Doradus in the LMC, lies R144. This strong X-ray source is apparently the most massive binary system known. The masses of the individual stars in this binary aren’t clear, but each must weigh in at 100–150 times the mass of the Sun. They are a dazzling pair of stars, each as mighty as Eta Carinae. When it was discovered, R144 displaced a pairing of stars in the Milky Way’s cluster NGC 3603 as the local universe’s most massive couple. The pair in NGC 3603’s weighed in at only 212 times that of the Sun. When tight pairings of massive stars are still burning hydrogen, they can still provide astronomers with a dazzling display. Not only are they visually bright with luminosities up to a few million times that of the Sun, they also produce violent stellar winds that blow at velocities up to 2,000 km/s (45,000 miles/h). Where such behemoths lie close to one another, the orbital speed of the two stars can be added to the wind speed. This becomes a rather significant figure extending into the thousands of kilometers per second if the stars are massive and orbit one another closely. As a result, the already toasty winds shed by these stars slam into each other with significant energy and this generates copious amounts of X-rays. As the stars move backwards and forwards to one another while they orbit their center of gravity, the strength of the collision varies and with it the intensity of the X-ray emission. Such variation was used about a decade ago to pry apart the component stars in the Eta Carinae system. The primary star in the Eta Carinae system and the brighter of the two is a Luminous Blue Variable and has a mass of 100–120 times that of the Sun. It is shedding large amounts of cool, dusty gas into surrounding space; too cool in fact to generate X-rays at all. However, despite the wimpy nature of Eta carinae’s wind, the system is a strong source of X-rays. Moreover, every 5.52 years there is a sudden drop in the X-rays that are detected. This is accompanied by drops in the amount of radio waves picked up on Earth. These observations are best explained if the Eta Carinae system is a binary, with an unseen O- or early B-class companion that is generating a fast wind. This fast wind slams into the Stellar Soap Operas 161 slower wind of Eta Carinae. This violent collision is the source of all those X-rays. However, as Eta Carinae and its large dusty shells swing in front of the X-ray source, it eclipses it, preventing the X-rays and radio waves from reaching our shores. Eta Carinae, effectively, overshadows its still impressive companion. Obscured by Eta Carinae’s light and dust, the impressive 30 solar mass companion is all but invisible. It was only by measuring the periodic variation in the system’s X-ray emission that its presence came to light. Impressive though main sequence O-class stars are, it is when they evolve into other forms that their true power and influence comes to light. The most massive O-class stars become luminous blue variables (LBVs, mentioned above). These highly luminous stars would outshine several hundred thousand Suns, generating as much energy as the Sun does in a year in the space of a second. Although giant by the standards of our Sun, they are not the universe’s largest stars. Most are between 50 and 100 times the diameter of the Sun. The largest stars are the red supergiants with diameters that would stretch out to the orbit of Saturn and beyond, were they to be placed within our Solar System. An LBV may pack over 100 times the mass of the Sun into this volume, while a punier red supergiant, stretches its mass out into a sphere with more than 100 times the capacity. What LBVs and red supergiants share, however, is a propensity to lose weight and lose it fast. Both classes of stars will shed many solar masses of gas before death comes a-calling. Pair such a moribund star with a close companion and all of this gas becomes accessible to it. Not only is the companion now able to gain considerable weight of its own, but it and its companion’s fate are irrevocably altered. Figure 6.2 illustrates how this happens. In this chapter we will examine some of the more interesting outcomes of such interactions.

Lurid Marriages and Messy Divorces

In globular clusters, and to a lesser extent in open clusters, stars can encounter one another frequently. As we have seen where two stars form a binary system, the effective size of the system is much greater, making collisions much more likely. 162 The Complex Lives of Star Clusters

a1 a2

Roche Lobes Roche Lobes

b1 b2

Semi-Detached Binary System Detached Binary System

FIG. 6.2 Detached and semi-detached binary systems. In a1 and a2 the two stars are close enough to one another (within at most a few astronomical units) that they can interact when they leave the main sequence and become giant stars. Material from the most massive star, which evolves first, can flow to the companion, increasing its mass b1 . Where there is a large difference in mass the material overflows the companion and can form an envelope around both stars. In a2 and b2 the two stars are so far apart that only their stellar winds can reach one another and no material is directly transferred from one star or the other. A few binary stars orbit so close to one another that they physically touch. These contact binaries (not shown) are the surprisingly common in the Milky Way

What makes binary systems even more interesting is the amount of energy they contain. Depending on the mass of the two stars and their separation, binary stars may orbit one another at speeds up to 2,000 km/s. This is far greater than the typical speeds of stars in globular clusters of between 5 and 15 km/s. Stars in open clusters typically move around these with a speed of a handful of kilometers per second at most. Thus, when a single star encounters a binary system many interesting things can occur, not all of them pleasant for either party. The slower moving star will be tugged upon by the binary stars as they whizz around their center of gravity. This allows momentum to be transferred invis- ibly between the binary star and the interloping star. As a result, the interloping star may be accelerated and driven away from the binary, while the binary system loses energy and shrinks in diameter. If, however, the interloping star is more massive than one or both of the stars in the binary, it can bully its way into the Stellar Soap Operas 163

system and eject the most vulnerable, smallest star. Not only that, it can steal the ejected star’s partner and take up residence as a new binary partner. Over the billions of years that have elapsed since globular clusters formed, such interactions have led to an increase in the number of binary systems in the cluster’s heart and a loss of these cluster’s lowest mass stars. In some cases, the ejected dwarf stars trail behind globular clusters in long streams. However, the dense confines of globular clusters have made a wide variety of what are otherwise unusual partnerships. Before looking at these in more detail it’s worth examining the work that uncovered the propensity of stars to interact with one another; for it was only a couple of decades ago that most astronomers would have considered such relationships highly improbable. This chapter focuses on the work of two key groups: Jarrod R. Hurley and Michael M. Shara working at the American Museum of Natural History; and Simon Portegies-Zwart, working with Piet Hut in various Universities. Around a decade ago, Hurley and Shara used the GRAPE-6 super-computer system to model the activities of tens of thousands of stars in globular clusters, while Portegies- Zwart focused on the interactions of massive stars in young star clusters, in particular how very massive stars could form in these dense environments. What these groups found was illuminating to say the least. Hurley and Shara produced one of the best titled and utterly anthropomorphic research papers of all time: “The promiscuous nature of stars in clusters”. GRAPE-6 is an acronym for “GRAvity PipE 6”, a rather impressive computer in its day with 32 parallel processors that could carry out one trillion operations per second. This allowed it to model a star system with over 50,000 stars from birth to death, over 10 billion years in the space of 1 week. Each model particle, representing a star, had mass, a starting position and a velocity which gave it a realism unseen in previous computer simulations. The initial motivation for their work appeared to be improving understanding of the formation of so-called blue straggler stars. These hydrogen burning stars exist on the main sequence at temperatures and masses above the point all of the cluster’s other stars have already evolved into red giants. Somehow these stars have 164 The Complex Lives of Star Clusters retained youth well past their years. The two prevailing theories of the day were that these stars formed when two smaller, less massive stars merged, or when one star in a binary stole matter from a companion. Either was a viable explanation, but the merger theory suffered because it was generally thought that mergers of two stars would be highly unlikely. This contradicted the high numbers of such stars that could be observed in all globular and most open clusters. For example Hubble observations of the globular M30 showed a high number of blue straggler stars, but a curious dearth of red giants (see Chap. 4 ). In M80, 305 blue stragglers were found by Hubble in the cluster’s core. Along with blue stragglers, X-ray binary systems (see later) were present at 1,000 times their frequency elsewhere in the galaxy. This pattern repeats with star systems that include cataclysmic variables (Chap. 3 ). These have a white dwarf that is accreting material from a neighboring low mass star. As the material falls onto an accretion disc it can release bursts of energy from its outer edge—rather than from the point at which it contacts the white dwarf. Cataclysmic variable stars come in various guises, but all require a white dwarf that is accreting, absorbing, material from a lower mass star, often a red dwarf. Another type of star, virtually unique to globular clusters, are sub-dwarf B (sdB) stars. These stars have a diameter of between the Sun and the Earth. They are all low in mass, about half that of the Sun, and very hot. All appear to be formed when red giant stars lose their outer hydrogen-rich layers. Although red giants can do this when they approach the end of their lives, in globular clusters they appear to be forced down this route prematurely. The abundance of sdB stars clearly correlates with the lack of red giants. It is therefore likely that red giants become involved in untoward relationships with other stars, get stripped down to their cores, then are forced to live out the rest of their days as sdB stars. These hot little embers can follow one of two paths. If they have enough mass, somewhere between two fifths and half the mass of the Sun, they ignite their helium and become extreme horizontal branch stars (Chaps. 2 , 3 and 4 ). The remaining stars aren’t massive enough and quickly exhaust what hydrogen they have left. Once this is gone they contract further and begin cooling off as white dwarfs. Stellar Soap Operas 165

The GRAPE-6 simulations reveal a lot about how stars behave while “married” in binary systems. In one run, Hurley and Shara observed 20,000 stars over the equivalent of 5 billion years of their history. During this time, 500 interactions occurred, involving a total of 730 stars. Of these some stars were particularly promiscuous, flirting with and exchanging partners with 494 other stars. In this group 105 stars swapped partners twice; 48 swapped their partner three times; 27 made four exchanges; with 14 stars push- ing the boat out with five exchanges. One star had a remarkable 22 exchanges, where 10 “remarriages” occurred. Amongst these interactions, flirting resulted in temporary exchanges before the “star changed its mind”, returning to its original partner. For our Casanova star, that means despite 22 interactions, the majority were mere flirtations that resulted in a messy threesome, before the Casanova was rejected. A good thing, too, one might add. From the perspective of solving the blue straggler problem, 13 mergers occurred as a result of unstable pairings. In these, the transfer of mass from one star to the other caused the two stars to come together. Such mergers might be significant sources of energy in their own right (see later in this chapter). It is of note that from an observers perspective, it might well be worth tracking down likely star systems that are going to merge in star clusters, most likely globular clusters. Pick the right pair and some spectacular fireworks are likely. Although the odds of catching a merger in our lifetime are low, with the appropriate automated search system should be able to spot some merge bursts that are attributable to globular clusters.

Two Routes to Blue

As we’ve seen, blue stragglers are stars that are hotter and bluer than the most evolved main sequence stars that have reached the point where they are “heading north” to become red giants. These stars can only be present if they have somehow gained enough mass sometime after they were born. If they had been born massive they would have died young: these stars are old but have still managed to retain the veneer of youth. Although blue stragglers 166 The Complex Lives of Star Clusters are best known in older open and globular clusters, young open clusters, such as NGC 133, also sport a population of them. There are two possible ways that such youth could have been engineered into them. In the simplest, two stars collide and merge while they are still both on the main sequence. The collision will release enough energy to thoroughly shake up the star, sending helium-enriched gas upwards from the core into the envelope and driving fresh hydrogen down into the core. The extra helium in the star’s outer layers makes the star bluer (Chap. 4 ), while the extra mass makes the star hotter and more luminous overall. An interesting speculation might be that the bluer, second main sequence found in many globular clusters might actually comprise stars produced by mergers (or harassment from a stellar near miss), rather than a true second generation. Whatever their origin, such reinvigorated stars won’t live any longer than their precursors, as their extra mass makes the core hotter and burn their fuel faster. The extra helium in the outer layers allows more heat to escape making the star smaller and denser for its mass than a star built more extensively of hydrogen. Stars could divide in one of two ways. In the simplest the two stars simply encounter one another like two misdirected trains on one piece of track, plowing into one another. The collision will heat the material to very high temperatures and potentially drive much of the outer layers off. Assuming the stars do not collide with too much energy, most of the material will fall back together again and the blue straggler will commence its new existence. Perhaps, more interestingly, what material that escapes the maw of the newly-formed star will collapse into a disc around it. Within this swirling mass, it is possible that a new generation of planets could be formed. Whether these nascent bodies would remain in stable orbits is another matter, but there is the potential, at least, for these planets to find a hold. Despite the prospect that two stars might have some kind of head-on collision the most likely scenario involves a collision inside a binary or triple star system. Here, perhaps because of the effects of an interloping star, or because the orbits of three stars become unstable, the two innermost stars fall into one another. This kind of collision is not head on. Instead, the orbit of the two inner stars gradually decays. Sooner or later the two innermost Stellar Soap Operas 167 stars merge. The process takes a few months, during which the two stars merge first their outer layers then their cores. Much of the material from the stars may be driven outwards in strong winds directed away from the poles of these merging stars. Moreover, a disc of material may spiral outwards from the equator of the colliding objects. This outward moving disc is called an excretion disc . Once again, it is in the disc that planets might just form in the ensuing millions of years. Ultimately, the two cores will fall together and much of the remaining gas and dust will fall back onto the central, reinvigorated star. In the second scheme, blue stragglers form when one star in a binary steals material from its partner through an accretion disc. These stars are known as Algol-type binaries. In the prototype an aging, low mass, orange sub-giant star is losing mass from its outer layers. Much of this gas is trapped into a disc around a more massive B-class partner. In time the orange sub-giant will lost all of its outer layers. It probably won’t keep enough mass to allow it to ignite its helium ashes and will fade away as a helium-rich, white dwarf. The now more massive companion is now a blue straggler—a star more massive than any other cluster star, but one still fusing hydrogen into helium. In time this will then expand to become a red giant, lose its outer layers and either became a hot extreme horizontal branch star, or another helium white dwarf, if too much mass is lost. The question is, which of these two mechanisms, merger or acquisition, is the route to most blue stragglers? Aaron Miller (University of Wisconsin) and Robert Mathieu (Northwestern University) examined the population of blue straggler stars in the old open cluster NGC 188. Computer models predicted that the majority would be formed by collisions between stars, yet observations showed that most of these systems were binaries where the companion was almost always about 0.5 solar masses, half the mass of the Sun. Although feasible that such a preponderance of cloned systems arose naturally and then suffered a collision, it is highly unlikely. Instead, it would be far simpler if the blue straggler and the low mass helium-rich star arose simultaneously: the blue straggler was originally the less massive star. As its companion left the main sequence and expanded, the lower mass companion acquired much of its bloated companion’s outer layers. 168 The Complex Lives of Star Clusters

This boosted the mass of the companion, turning it into a blue straggler, while reducing the companion to a sub-dwarf B star. At half a solar mass these remnant stars should just about be able to fire up their helium cores and become EHB stars. In time the blue straggler will run out of fuel and expand to become a red giant. It will transfer back some of its acquisition to its hot, dense partner. Just to show that different roads lead to Rome, Chengyuan Li (Kavli Institute for Astronomy and Astrophysics) and colleagues demonstrated that two separate mechanisms were at work in the LMC globular cluster Hodge 11. This star cluster is the best part of 12 billion years old. Blue stragglers are found throughout the cluster, but they appear to have different origins, depending on where they are found. Li and co-workers examined the colors and brightness of 162 blue straggler stars. The population was clearly split into two camps, with differing colors. Those in the cluster core were, on the whole bluer than those further out. This suggested that, within the cluster core, blue straggler stars were formed through the direct collision of two stars—most likely within a binary system. Beyond the cluster core, and in the majority of its volume, blue stragglers appear to have been formed, mostly, through the process of acquisition: one binary star accreting material from its companion star. Moreover, given the ages of the stars in the system, it appeared that most of the blue stragglers were formed in some form of cluster crisis 4–5 billion years ago. In the next chapter the likely nature of this crisis will be examined. The accretion route appears to be central to the formation of blue stragglers in open clusters. The distances between the stars of these grouping are such that mergers are unlikely. Although binary routes, either mergers or acquisitions, are the most likely route for formation of blue stragglers, a few have been identified which show a more complex pattern of merging. In the old and fairly massive open cluster M67, one straggler was so massive that it must have formed through repeated collisions, or acquisitions. Such massive blue stragglers are called super-blue stragglers. Although rare, it is clear that given the right circumstances multiple mergers are possible. This is important, for it not only suggests that there exists a very dynamic and potentially Stellar Soap Operas 169

violent environment inside star clusters, but that the scenario evoked to explain the formation of super-massive stars is perfectly reasonable. Some of the consequences of such cluster violence, particularly events that may unfold in very young massive clusters, is explored later in this chapter. As a coda to this section, Lee Clark (San Diego State University) and colleagues also identify what they call “yellow stragglers”, while others have suggested so-called “red stragglers” in the low mass globular cluster Palomar 13. Yellow stragglers are defined as sub-giant, F or G-class stars that have completed hydrogen burning and are now expanding to become red giants. Yellow stragglers, much like their blue siblings, are located above the main sequence turn-off for most of the cluster’s stars. These objects are either the descendants of blue stragglers, stars which have now exhausted their rejuvenated store of hydrogen; or they are the products of mergers between stars that have merged soon after one or both left the main sequence. Red stragglers are even more evolved stars that are now ascending the red giant branch. Like the blue and yellow stragglers, they lie above the location of other red giants at a given temperature indicating that they have anomalously high masses that are most likely the descendants of earlier collisions between main sequence stars. They could not have been formed through the collision of red giant stars, as an impact at this stage in the star’s life would have dispersed the outer layers of both, leaving one or two sub dwarf B stars. Interestingly, Clark and co-workers study shows an interesting distribution of blue stragglers and other stars within Palomar 13. As would be expected, blue stragglers are highly concentrated towards the core of the globular cluster. They form a tight ker- nel within a larger mass of red giant stars. These evolved giants are also generally confined to an outer core of stars. Beyond this lies predominantly low mass, main sequence stars. This paints an interesting picture of how this small globular cluster is evolving in response to the stresses put upon it by the Milky Way (Chap. 7 ). More massive stars evolve the fastest, so the red giants must be the descendents of the clusters most massive stars. These have clearly segregated very efficiently towards the core of the cluster, leaving the lower mass main sequence stars drifting further out. However, within the very centre of the cluster blue stragglers 170 The Complex Lives of Star Clusters dominate. These main sequence stars must be actively produced and falling towards the cluster core. As Chap. 7 will explore in more detail, Palomar 13 is in the midst of destruction by its host, the Milky Way. The cluster appears to be very puffed up and low in density. This is the result of a fatal and repeated process called disc shocking (Chap. 7 ). This is allowing stars to redistribute themselves within the cluster. In the case of Palomar 13, it is hastening the collapse of the core of the cluster, which has resulted in the formation of many blue straggler stars. The red giants have also migrated towards the core of the cluster as they are the most massive evolved stars. However, it is here, that most will be disrupted through repeated encounters with other stars. The lowest mass stars are most affected by disc shocking and have been driven towards the cluster periphery, where soon, galactic tides will rip them away from the remainder of the cluster.

Beyond the Blue: The Twisted Fates of Cluster Stars

Hurley and Shara continued to explore the fates of the blue stragglers as the cluster continued to age. Amongst their “lurid examples” was an object that began its life as an F-class main sequence star with 1.33 solar masses of material. After 4.1 billion years, not only had this object survived on the main sequence, but it had grown sixfold in mass to 7.7 solar masses. Its life included phases as a blue straggler, a singleton super-blue straggler, a super blue straggler in a binary and a blue straggler member of a triple star system. This star got around. In another example, an initially wide binary system, consisting of two stars, one a red dwarf with a little over half the Sun’s mass and another with 99 % the mass of the Sun, became involved in an altercation with another binary system after 3 billion years. The most massive of the stars from each binary then paired up, ejecting the lightest of the three stars. This new system consisted of the 0.99 solar mass star and another, with a mass a little greater than the Sun. Orbiting further out was the lightest star. The star that was ejected had so much energy that it left the cluster altogether (Fig. 6.3 ). Stellar Soap Operas 171

The innermost two stars merge forming a blue straggler star with 2.1 solar masses, while the third star is ejected Two binary systems merge 3 The lightest of after 3 billion four stars (the red years dwarf) is then At 4 billion years ejected the merger product marries another similar BS star, both roughly 2.1 M in mass …while the other two These two stars merge merge to make 4.3 M star meets one 4.3 M at 4.5 another binary GYa One star is ejected with two sun-like stars at 4.6 Gya

x x

The product star dies at 4.7 Gya

FIG. 6.3 The twisted life of a Sun-like star that began paired with a red dwarf. After 4.7 billion years, and multiple fleeting relationships, the Sun-like star had morphed into a mid-B class object with over five times the mass of the Sun. Not long, thereafter, death overtook it and it ended its days as a cooling ember of an oxygen-neon white dwarf

Soon thereafter, changes in the orbit of the remaining three stars drove the central pair to merge and the outermost to fly off, alone, into the cluster. All the while the system was sinking towards the core of the star cluster. The merged pair become a blue straggler with just over twice the mass of the Sun. Now relatively massive, the 2.11 solar mass star fell towards the cluster core, arriving there after a total time of 4 billion years. Here the blue straggler met another of its kind forming yet another, albeit wide, binary. Vulnerable to harassment from neighboring stars, this pair was forced violently together after only a few tens of millions of years, when their orbits became unstable. Now a single 4.3 solar mass super-blue straggler, this went on its way for another 100 million years before it met up with yet another binary star. The more massive of the two stars, with a little over 50 % more mass than the Sun paired up with it, and its former bride was divorced. Not content with that, this binary then collided 172 The Complex Lives of Star Clusters with another! As the orbits of these four stars rearranged two of the stars collided forming a single star with more than five times the mass of the Sun, while the other three stars either went their way or remained as a binary partner for this vastly engorged star. Having led a rather impressive life, the 5.78 solar mass star left the main sequence to become a super-AGB star—a star that is burning carbon in its core, but is too lightweight to explode as a supernova. Death followed, soon thereafter. The point Hurley and Shara make is that although the process of mergers is rather random, the outcome tends to be the same. Collisions drive the engorged star that is produced further towards the cluster core. Here, in the dense central confines of the cluster, further encounters are almost inevitable. The results also show that in many cases, the lightest star in a binary is the most vulnerable to becoming ejected, in some cases from the cluster altogether. Over time this should result in a very skewed population of stars in the cluster. Hopefully, you can see the link with the sodium-oxygen anti- correlation problem (Chap. 4 ). To re-cap, in all globular clusters stars are either rich in sodium or oxygen, but never both. Sodium- rich stars appear to lie mostly towards the cluster core and are made from material that was likely processed through a large generation of intermediate mass stars. These, now dead, stars threw out helium and sodium-rich material that then came together to make the majority of the stars currently found in globular clusters. The remaining long lived sodium-poor stars form a more diffuse skeleton that predominate towards the outside of the cluster. Sodium-rich stars do not follow the rules and fail to re-expand to become red giants when their stock of helium runs out. They inhabit a region on the HR diagram called the Extreme Horizontal Branch and are much reduced in mass. This has created a problem in astronomy. Why do sodium and helium-rich stars fail to become red giants? The focus was on the chemistry: the extra helium. As it was explained in Chap. 4 , it was thought that the extra helium denied them access to the second red giant branch: the AGB. However, some researchers queried how much helium would be needed to cause this. The helium idea didn’t jibe with observations of EHB stars. These were too low in mass—only just able to ignite helium at all—and this could only happen if Stellar Soap Operas 173 these had lost a heck of a lot of mass while on the first red giant branch (Chaps. 2 and 4 ). Moreover, observations suggested that the vast majority, and possibly all, of the EHB stars were found in binaries. This leads to the inescapable conclusion that these EHB stars end up the way they are because of their geographical location, not through birth. These stars would become the hot, low mass objects we see today even if they had no more helium at birth than their siblings. Its geography, rather than birth right, that determines fate. These stars really are in the wrong place at the wrong time. If you want a fitting human analogy, perhaps a study of twins might work. A pair of twins is born in the US. One moves to Japan, while the other stays in, let’s say, California. After 50 years the Californian twin develops Type II diabetes and heart disease, while the Japanese emigrant with the better diet stays healthy. While their underlying genetics is the same, the very different diet and lifestyle led by each has led to a different fate. In the case of the stars, the distribution of oxygen and sodium-rich stars may be the clue to the likely stellar fate, rather than the chemistry per se which determines it. Unfortunately, given that the second generation of stars formed towards the core of the cluster, it then become inevitable that these stars, more than any other, which would become wrapped up in the lurid relationships discussed here. Returning to Hurley and Shara’s illuminating work, they show that, at least in simulation, it is possible to make blue straggler stars through collisions, even in open clusters. In one example a low mass red dwarf moves slowly out of the cluster core at less than 3 km/s. After just over a billion years it hits a Sun-like star. The collision produces a star with 1.3 solar masses of material. As this is now quite heavy, it sinks back towards the core of the cluster. The single-most important factor in determining the likelihood of a collision is the diameter, or semi-major axis, of the binary. This sets the system up for encounters. If this does not directly lead to a collision, it does allow for instability in the binary which then leads to its collapse and the merger of the two stars. In another simulation, Hurley and Shara produced a tight partnership of two helium white dwarfs. The loss of gravitational 174 The Complex Lives of Star Clusters radiation drove these two stars together to form an object that was still low in mass but now able to ignite its helium. Most likely, this object would appear as a sub-dwarf B (sdB) or EHB star, suggesting that at least some of these stars are produced through direct collisions, rather than the direct stripping of fledgling red giants. A final, rather interesting example brought a neutron star into intimate contact with a helium star—an object burning helium in its heart. Thorne-Żytkow Objects (or TZOs) are a type of stellar exotica that we will come across again later in this chapter. They are named after Kip Thorne and Anna Żytokow (then both at Caltech) who investigated whether a star could possess a neutron star in its heart and still remain stable. Although only barely, and still controversially, recognizable as observed objects, TZOs there are a few candidate systems that include a giant or supergiant with an unusual surface chemistry. These include the Small Magellanic Cloud’s red supergiant HV2112, and two yellow supergiant, variable stars U Aquarii and VZ Sagittarii; with the unusual black hole binary system, GRO J1655-40, also proposed as a descendent of a TZO. TZOs can be divided into two broad theoretical groups. In the lower mass versions a neutron star is absorbed into, or collides and merges with a low mass red giant star. The neutron star settles into the core of the red giant and begins to work its magic. Matter falling onto the surface of the neutron star begins to fuse into helium and then carbon at very high temperatures. These same high temperatures cause very vigorous convection within the envelope of the red giant, which continually fuels the fires burning below. In low mass red giants such reactions provide some of the energy, but the release of gravitational energy as the star slowly falls onto the neutron star provides the bulk. In more massive supergiants, nuclear reactions provide the bulk of the energy needed to hold the star up against gravity. As long as the supply of hydrogen is maintained, and the helium removed hydrogen can also fuse with heavier elements through a series of rapid reactions called the rp -process. Here, hydrogen nuclei (protons) fuse in long chains onto heavier elements forming proton-rich isotopes of elements such as rubidium, strontium, and zirconium. Normally, such reactions occur only on the surfaces of neutron stars that are stealing hydrogen from Stellar Soap Operas 175 companions, where temperatures can reach up towards a billion Kelvin. They don’t happen in red giant or supergiant stars where temperatures never exceed a 300 or 400 million degrees in the most massive AGB stars. Therefore, when these isotopes are spotted, astronomers immediately suspect that something odd is going on within the core of the red giant, something that appears to necessitate the presence of a very hot neutron star. TZOs are also predicted to produce the fragile, but hugely economically important element lithium. This element is the core of modern mobile existence for humans. Lithium batteries power our mobile communication networks, from laptops to phones and tablets, because the element packs a lot of chemical energy into a low mass. Despite lithium’s claim as one of only three elements fawned upon by the Big Bang, it is very hard to produce in stars, yet incredibly easy to destroy. Lithium fuses with hydrogen at low temperatures to make helium. Such puny temperatures are achievable, not only in every star in the universe, but also in a large fraction of brown dwarfs that can’t even fuse hydrogen directly. Thus lithium’s frailty makes it an especially interesting element and one which is difficult to manufacture in today’s universe. Yet a few stars do appear to be rich in this element. Some of these are well-known characters: AGB stars. Lithium is made by the amusingly named hot bottom burning (HBB) process. Hydrogen is brought down by convection from the envelope where it can fuse to make helium at around 40 million Kelvin. In the right circumstances helium-3 and 4 combine to make beryllium-7. Normally this is destroyed quickly in the star’s interior, combining with more hydrogen to make beryllium-8 which is unstable and breaks down to make two helium nuclei. However, if beryllium-7 can escape the star’s nuclear fires in time, it can be swept upwards by convection to cooler climes, where it morphs into lithium-7 through radioactive decay. Although both TZOs and AGB stars can make lithium, only the TZO can make lithium and rp -elements such as rubidium, strontium and zirconium. In the case of the three candidates mentioned above, they were initially classified as different objects. The red supergiant HV2112, was thought to be an asymptotic giant branch (AGB) star (Chap. 2 ) and these never attain temperatures high enough to 176 The Complex Lives of Star Clusters fuse hydrogen into the observed proton-rich elements. The yellow supergiant variable stars U Aquarii and VZ Sagittarii were thought to be another type of rare variable star called an R Corona Borealis (RCrB) star. These carbon-rich objects are probably formed when two white dwarf stars collide. Because RCrB stars are born with very little hydrogen they are readily distinguishable from TZOs which should be hydrogen-rich: they are, after all conventional red giants, at least on the outside. TZOs, formed from low mass red giants, die when the supply of hydrogen is exhausted. As long as the red giant didn’t add more than a Sun’s worth of matter to the neutron star, this hot neutron- rich object should survive. The red giant dies pretty much as it would do if it was changing into a white dwarf. The outer layers, now very rich in helium and rp -elements, are exhausted by a combination of strong stellar winds and nuclear reactions occurring on the surface of the neutron star. Meanwhile, the now the bulked up neutron star settles back down to its original form, surrounded by a dispersing planetary nebula. If the neutron star fell into a much more massive supergiant, things get more interesting. After a hundred million years or so, feasting on the inside of this supergiant, the neutron star crosses the divide between stability and implosion. Once its mass exceeds 2.5–3.0 times the mass of the Sun, gravity takes over and implodes it, forming a black hole. At this point the eating can accelerate and the innards of the star are consumed. Within the doomed supergiant, an accretion disc swallows vast chunks of star-stuff. Magnetic fields form, twist and buckle within the disc, launching jets of material outwards along the rotation poles of the doomed star. Within seconds these blast out of the star shredding it and potentially generating a Gamma Ray Burst and accompanying supernova. However, there is an alternative track. GRO J1655-40 is an unusual star system, one apparently so unlikely to have formed the way it is that some suspect that it is the child of a TZO. It consists of a low mass sub-giant star orbiting a black hole. The partnership generates highly energetic outbursts, driven by the accretion of material from the sub-giant onto the black hole. A pair of opposing, very high energy jets is launched by the system, which earn it the distinction of being classed as a micro-quasar. Stellar Soap Operas 177

These jets flicker in output in a manner that is reminiscent of quasars in the distant universe. What is unusual about this system is the underlying pairing. In most systems with black holes the partner star is also massive: pairings of low and high mass stars—or in this case the remnant of a very high mass star and its low mass companion. A few other such pairs are known but it is quite difficult to form such systems, given the scarcity of star systems that could form them. Astronomers then propose that these systems formed when the former TZO imploded. When the black hole formed inside the supergiant, it consumed the material rather quietly. What remained of the supergiant was left in a disc around the newly formed black hole. It was from this disc that the low mass star then formed. If this scenario is true then the companion would be expected to be rich in the same rp -elements as the former supergiant. Lithium, unfortunately, would be long gone by the time the companion became a sub-giant at the end of its main sequence life. So does the sub- giant hold a repository of these elements? In Hurley and Shara’s work the TZO story begins with a binary pairing of a 10.8 solar mass star and 5.3 solar mass star. When the more massive star evolves off the main sequence it begins to transfer matter to the less massive partner. Because both stars have very different masses the transfer is unstable and soon both stars are swallowed up in a common envelope of gas. Within this, the core of the giant and the main sequence star spiral towards one another. After a few tens of thousands of years the two remaining stars are a 2.42 solar mass helium core, now fusing helium to carbon, in orbit around the 5.31 solar mass main sequence star. When the helium core was exhausted the star expanded once more. This time the two stars were more evenly matched and the more massive main sequence star acquired much of the helium giant’s outflow, leaving it with a little over six times the mass of the Sun. After a few hundred thousand more years the helium giant was transformed into an oxygen-neon white dwarf in orbit around its blue straggler companion. Fifty-five million years later the blue straggler reached the end of its main sequence life and it too began to expand. After another common envelope phase the white dwarf was left in a tight orbit around a naked helium star. 178 The Complex Lives of Star Clusters

When the helium star ran out of fuel it expanded and the white dwarf companion grabbed hold of most of the material. In time it became so massive that it imploded forming a neutron star. It is generally thought that this kind of white dwarf will simply implode, probably accompanied by a weak Type Ib supernova— one rich in helium but where hydrogen is absent. The simulation of this system concluded that after a brief pause the helium giant continued to expand until the neutron star was absorbed into its mass. The neutron star spiraled into the core of the giant and the TZO was produced. This sort of TZO would be very different to the one described earlier. In that example the neutron star became the core of a hydrogen-rich red giant or supergiant. In this case the neutron star lies only a short distance below the surface of a dense helium star. Quite how this object would evolve is anyone’s guess. One would assume helium would be burnt at very high temperatures to make first carbon then iron. However, without hydrogen, these stars would not show much in the way of the rp-elements thought to characterize TZOs. Its fate: with sufficient mass, the neutron star might still implode to form a black hole, depending on how massive the helium star was once the neutron star had taken up residence. What is evident is that there are almost as many possible outcomes for stellar collisions as human imagination will allow. This is a truly fascinating area of astronomy which is over-looked at our peril. For many astronomers might rush to conclude that some wacky, unusual supernova that they have identified has a particular cause. Given the rarity of some of these supernovae, it might be the case that some of these were formed through the chance interaction of a number of stars over billions of years, potentially a scenario that is only acted out once or twice in the observable universe. It would be very interesting to see, where possible, which supernovae are associated with star clusters, either young or old, rather than simply identifying the type of galaxy in which they occur. A galaxy bereft of massive stars isn’t necessarily going to be bereft of the types of supernovae massive stars create. Hurley and Shara conclude that the GRAPE-6 system throws out so many oddball events that it would be worth having some sort of system to store this information. Given time, wouldn’t it then be nice to compare the rarest, most peculiar stars, mergebursts and Stellar Soap Operas 179

supernovae with this library? To reiterate a point made earlier, it is of significance that the frequency of binary star systems consisting of a pair of white dwarfs is far higher in clusters than in the rest of the galaxy. This would imply that a disproportionate number of Type Ia supernovae should happen in star clusters. Is this true? Surely, observations are needed.

W Ursa Majoris Stars

W Ursa Majoris (or W UMa) stars are contact binaries comprising two stars with masses comparable to the Sun, either somewhat more or somewhat less, and comparable in most cases to one another. As the name implies, a contact binary is one where the two stars are physically in contact with one another and share at least some of their outer envelope. Initially, two types of W UMa systems were identified on the basis of their spectral type. In the A sub-class both stars are more massive than the Sun and have spectral classes A or F with orbital periods of 0.4–0.8 days. In these systems both stars have a broad area of contact between them with extensive exchange of matter between each. There is a relatively small difference in the mass of each star. The W sub-class have components in spectral classes G or K. Since these are smaller stars, their orbital period must also be shorter if they are to be in contact. Thus these binaries have periods of 0.22–0.4 days. These stars are, generally, in less physical contact with one another and at least in some cases the W-class systems might evolve into the A-class, however, given that most W-systems are of lower mass than A-systems, this cannot be generally true. In both A and W systems, both stars have surface temperatures that lie within a few hundred Kelvin of one another, at most. A third sub-class, B, was introduced in 1978. These contain two stars with a greater difference in mass and hence surface temperature. Further investigation by Szilard Csizmadia (Konkoly Observatory, Hungarian Academy of Sciences) and Péter Klagivik (Eötvös Loránd University) revealed a fourth sub-class, H, with even greater differences in the mass of stars in these systems. 180 The Complex Lives of Star Clusters

One star may be several times more massive than its partner which leads to some interesting exchanges in momentum between the two contacting stars as they rush past one another. The greater difference in the mass of the two components also restricts the effectiveness of the transport of energy between the two stars and thus explains the greater difference in their surface temperatures. In most cases the two stars are both on the main sequence. However some W UMa systems have one component that has just left the main sequence and is expanding into a sub-giant star, en route towards the red giant branch. As the two stars are physically in contact, the two stars take a lot less than a day to orbit one another. Their orbital periods may be apparent from periodic changes in the detectable light from the star if the eclipses are aligned with the Earth. The light will dip each time the two stars are pointing towards (or away from) the Earth. That the stars share their outer envelope allows each to share energy as well. The more massive component of the binary is somewhat under-luminous while the less massive star is far more luminous than it should be as a result of this energy transfer. Lifang Li (Chinese Academy of Sciences) found that most W UMa systems were probably formed from detached binaries. In these systems the two stars orbit one another with periods less than approximately 2.24 days, but otherwise do not physically interact. If these sorts of systems can lose angular momentum through different mechanisms (see below) they can then spiral slowly together and come into contact with one another over a period of about 3.23 billion years. There is a suggestion in the scientific literature that the hotter A-type systems evolve into the cooler W-type systems. Mehmet Yildiz (Ege University, Turkey) investigated the properties of A- and W-sub-type systems and concluded that the A-systems were on average 200 million years younger than the W-systems. Both had ages comparable to the Sun (4.4 and 4.6 billion years, respectively). On the face of it this conclusion might be true, but this really depends on the nature of the hotter stars in A-type systems. More thorough literature searches do not really bear this out with what is a very mixed set of messages about each sub-type of W UMa star. A-type systems have components with A and F-class components which normally have lifetimes that are much shorter than Stellar Soap Operas 181 this. If these stars are truly more massive than the Sun, an age of over 4 billion years is impossible. The stars might be hotter than expected if the shenanigans in the system has caused one or both stars to heat up, making it appear to be more massive than it really is. Lifang Li concluded that those W UMa systems with the lowest masses also have lowest overall difference in mass between the two stars and the lowest angular momentum. If you recall, angular momentum is a property given to masses that are rotating, which includes orbiting another mass. A system with a high level of angular momentum would have (proportionately) the highest orbital speeds and stellar rotation speeds for any given mass. The pattern Li observes suggests that as the two stars in the system move closer, they shed angular momentum (and probably significant mass) leaving lower mass systems with slower orbital speeds. How could they accomplish this? Two routes are available. In the simplest magnetic fields work together with strong stellar winds, to cart off the momentum of the two stars. The magnetic field of the stars reaches out to and brakes against the surrounding magnetic field of the galaxy, while stellar winds carry away mass, which also carries away momentum as the two properties go hand in hand. In a more brutal mechanism, as the two stars approach one another material is flowing from the brighter component to the smaller, denser component of the W UMa system. Typically the star with the largest mass will be losing mass to its neighbor as this star is physically wider, less dense and more readily fills its Roche Lobe. However, where there is a large difference in the mass of the two stars, material flowing from one to the other often overspills the lower mass star, which is unable to hold onto all of the gas flowing onto it. This gas simply has too much angular momentum. The excess gas (which can be a very significant proportion of the total material that is being transferred), skips across the smaller star and continues outwards in an outwards spiraling stream. Eventually, this is lost from the star system in its entirety. As this spirals outwards it carries away the excess angular momentum of both stars and allows them to move ever closer. The outcome of the process depends on the ability of the smaller star to “cope” with all of the attention bequeathed upon it by its larger, but diminishing companion. In some W UMa systems the smaller secondary star is given so much material and energy by its companion that it too begins 182 The Complex Lives of Star Clusters to puff up. If this happens the smaller star can begin to give back some of its mass to the larger, more massive star. The interplay then forces the two stars to separate and the W UMa system is destroyed. If this is true then there should be a so-called period minimum in W UMa star systems: an orbital period where the orbit lasts the briefest time. This is when the two stars are closest. In other systems, the smaller secondary star can keep taking on more mass at a rate that keeps pace with the shrinking orbit of the two stars. If enough angular momentum is lost from the system in stellar winds or outflows, then the two stars will merge into a single, rapidly rotating star—a blue straggler. Eric Sandquist (San Diego State University) and Matthew Shetrone (University of Texas) examined the four known W UMa stars in the relatively aged open cluster M67. On top of the changes in brightness caused by the orbit of the two stars, there was further, more extended variability over timescales of days to months. The pattern was most likely caused by changes to the surface of the stars. Remember that each of the stars in a W UMa system has a mass similar to the Sun, it should exhibit star spots, flares and other active phenomena on its surface caused by convolution of the stellar magnetic fields. That the two stars also rotate around one another, whole their surfaces remain in contact, should only serve to enhance these magnetic fields. This should lead to the stars dimming when large star spots swing into view, or brightening when they are obscured. One star, AH Cancrii, showed changes in brightness that the two authors suspected were caused by prominences erupting from the surface of the stars. In Sandquist and Shetrone’s work they also identify one star they dub a red straggler: a red giant that is clearly too bright to have recently left the main sequence. Such a star is most likely the descendent of a blue straggler formed by the merger of two lighter stars.

How Binary Stars Can Affect One Another: SN 1993J

SN 1993J was found in the nearby spiral galaxy M81. This supernova was one of the brightest detected in its day and began as a so-called Type II event. This meant that astronomers could detect Stellar Soap Operas 183 hydrogen in its spectrum. Over the ensuing weeks the amount of hydrogen faded until helium came to dominate the spectrum of the light from the explosion. Although it wasn’t clear until a decade later, it was suspected that much of the outer hydrogen-rich layer of the progenitor star had been transferred to a close companion. In 2004, Justyn Maund and colleagues from the Universities of Cambridge, Hawaii and Oxford published the confirmation of this hypothesis. A relatively bright, massive B-class companion star had shown up in Hubble images of the region in which the star had detonated. This was the first companion star detected in supernova searches. The system that gave rise to SN 1993J began its life with two stars, the largest, primary having around 15 times the mass of the Sun. Its companion was slightly less massive: 14 times the Sun’s mass. Over a period of a few thousand years, around two thirds of the mass of the primary star was either transferred to the companion, boosting its mass to 22 solar masses, or was lost to surrounding space. The stripped down primary star then blew up, leaving the engorged companion behind. It isn’t yet clear, whether the companion has been set adrift by the explosion of its partner or whether they will remain bound. If the latter is true, then in a few years the system will appear as a so-called high mass X-ray binary (HMXB). These are discussed in more detail below. A few years before SN 1993J detonated perhaps the most famous supernova of all lit up the Southern skies. SN 1987A was a peculiar beast. Although this supernova was most definitely the Type II explosion favored by theorists, the explosion was spawned by a compact blue supergiant star. More unusually, the supernova was flanked by three, unusual, hydrogen-rich rings. Again, the influence of a companion was sought to explain the properties of both the supernova and its surroundings. The current favorite theory is that around 20,000 years before the explosion the progenitor of SN 1987A was an expanding red supergiant. As it grew in diameter mass was transferred to a close, massive companion star, much like the scenario envisaged for SN 1993J. However, in this case the transfer of mass became unstable. In essence, too much material was catapulted onto the companion star, far more than it could gobble up (see W UMa systems, above). As the two stars whizzed around one another, matter from 184 The Complex Lives of Star Clusters the red supergiant completely consumed the smaller companion in what is known as a common envelope. This is a dense cocoon of gas and it has the effect of robbing kinetic energy from each star through friction. The two stars then spiraled in towards one another. Before all of the gas could be driven away, the two stars collided and merged. The collision drove off shells of gas forming the intricate ring structures that are seen today. Around 20,000 years later what was left of the merged star system detonated as an odd Type II supernova. Why were these two explosions so different from one another? In the case of SN 1993J, the two stars were of similar mass. This allowed the smaller companion to effectively hold onto the material shed by its companion. However, in the case of SN 1987A, the companion was too small, so most of the gas shed by the larger star spewed into the space around both of them, forming the envelope. The companion was only able to steal some of the larger, primary star’s mass before it collided with it. In young open star clusters, with up to 70 % of massive stars having close companions, we should expect the majority of supernovae to occur in binary systems. Of fundamental importance is the timing of the transfer of hydrogen (and in some cases helium) to companion stars. Not only does this reduce the mass of the companion, but both stars should orbit one another with some considerable speed—up to a few thousand kilometers per second. It appears that a rapid rotation is essential for the formation of two violent stellar phenomena: magnetars and long gamma ray bursts. We look, first, at magnetars.

How Westerlund-1 Solved the Puzzle of Magnetars

It is generally assumed that if a star begins its life with more than 30–40 times the mass of the Sun, its inevitable fate is to implode to form a black hole. This is simply because calculations suggest that the amount of material the star can get rid of before it dies is insufficient to lower its core mass to the point that only a less massive neutron star will form. In general, it is thought that a star with more than 30–40 times the mass of the Sun will leave a core Stellar Soap Operas 185 with more than 2.8 times the mass of the Sun. This core is not stable: gravity will inevitably crush it “out of existence”. (This depends on what you think a black hole really is, but we can leave that for now.) Smaller stars will produce a core with a mass somewhere between 1.4 and 2.8 times the mass of the Sun, and it must be said the upper safe limit may be less than this. This material is compressed until the majority of it becomes neutrons, overlain with a thin crust of iron-rich material. Such neutrons stars are around 20 km across and are bequeathed magnetic fields up to one trillion times that of a typical fridge magnet. The theory is fairly sound, but as always observations like to come along that punch holes in astronomers best laid plans. One such headache was the discovery of the galaxy’s rarest inhabit- ant, a magnetar, in the massive Milky Way cluster Westerlund-1. Magnetars are the show-offs of the neutron star fraternity. Not content with a magnetic field one trillion times that of the Sun, these hellish objects pack a magnetic punch 100 times stronger. Other than that, these are still neutron stars, with the same mass. Westerlund-1 holds at least 100,000 stars, probably more, depending on the masses of the stars contained and despite its location behind a large cloud of gas and dust, forms a unique counterpart to the halo’s old globular clusters. The problem was that there were stars with more than 40 times the mass of the Sun still burning hydrogen in the cluster, yet there was an object, the magnetar, that was too light to have been formed through the death of one of these stars, let alone anything more massive. The stars that were currently most evolved should have become black holes, not lightweight neutron stars. So, what was going on? The solution to this problem was the partnership in which the magnetars was found. The exquisitely titled CXOU J164710.2-455216 must have been born with roughly 40 times the mass of the Sun. As all the cluster stars formed effectively at once CXOU J164710.2-455216 cannot be any older then these. The magnetar is also very young so it formed from stars only marginally heavier than the cluster’s most massive, current members. Simon Clark and colleagues solution was to pair an initially very massive star with a slightly lower weight one, much like SN 1993J. In this case the magnetar’s parent had an initially more massive companion which was running 186 The Complex Lives of Star Clusters out of fuel. As it did so it began to expand into a supergiant. The smaller, less evolved parent of the magnetar then accreted much of the gas and was spun up in the process. Now the fast spinning parent of the magnetar lost mass through strong winds, becoming a Wolf-Rayet. Some of this shed material was re-captured by its now smaller partner, changing its chemical composition. Soon, thereafter, the fast spinning star detonated as a supernova. But the combination of the parent’s fast spin and low mass meant that a neutron star was the product. Moreover, the fast spin allowed the neutron star to generate an unusually powerful magnetic field and the magnetar was born. It may sound a bit ad hoc, but there is evidence that this process happened. Simon Clark found the smoking gun from this crime scene. If the magnetar had a partner in crime, it should still be present. Moreover, having had its partner violently explode and lose a lot of mass, it would have been likely that the star would have received a bit of a kick and be moving quickly from the crime scene. Indeed, the star C1* Westerlund 1W5 was found using the VLT in Chile. This star is heading away from the magnetar at high velocity. This is precisely what would be expected if the magnetar had just destroyed its parental binary system. Moreover, the stellar divorcee has more helium and nitrogen than would be expected given that it is also rich in carbon. Nitrogen appears towards the end of the main sequence, along with helium as the original supply of carbon is turned into nitrogen through the CN-cycle (Chap. 2 ). A few tens of thousands of years later, helium ignites and begins rebuilding the star’s inven- tory of carbon. Given the difference in the stages that produce and consume each of these elements, a star can’t be simultaneously rich in both of them: it’s one or the other. C1* Westerlund 1W5 has both nitrogen and carbon in abundance so it must have picked up one of these elements from its now extinct partner (Fig . 6.4 ). Thus the frequent pairings of stars within clusters have two effects on stats fate. Most importantly, the pair of stars can dramatically affect one another’s destiny; and number two, as a result of interactions, stars can be rapidly ejected from the cluster as a whole. Many high mass, primordial, binary systems are not only disrupted, but the divorcee thrown out of the cluster as a whole. This is violent relaxation in the extreme. The ejected stars Stellar Soap Operas 187

Roche Lobes b

d c or

GRB Magnetar

FIG. 6.4 The path to a magnetar or a GRB. A pair of massive stars orbit one another closely ( a ). A few million years later, the more massive star (the primary) leaves the main sequence, expands and begins shedding material to its companion ( blue arrow ) ( b ). Before long the primary has lost all of its outer layers and has become a WR star (c ). Soon, its companion leaves the main sequence and begins delivering material to the WR star (blue arrow ), spinning it up. Little material is kept, though because the WR star has a very small mass compared to its companion. Before the WR star can get larger it blows up. The fast spinning core collapses into a black hole (left ) and generates a gamma ray burst (GRB, d ). If the core is less massive the collapsing core forms a magnetar. Dotted lines around each star represent regions called the Roche Lobes. These are the edges of the regions the star can directly control by its gravity. Outside this region gas can escape. Thick arrows on the companion stars (d ) indicate that the star may be ejected in the explosion spend what is left of their lives hurtling through the halo. In some instances, if the star is low mass and hence long-lived, it can not only escape the cluster but the galaxy as a whole. Such intergalactic stars are commonplace in galaxy clusters, but are otherwise hard to detect. Those that are spotted are the brightest red giants and a considerable number populate the spaces between the galaxies of the Virgo Cluster. 188 The Complex Lives of Star Clusters

Magnetars may be glamorous in their own right, but it is to the universe’s most secretive dark stars that many astronomers turn for sheer decadence. The most numerous black holes in the universe have the lowest mass. These so-called stellar mass black holes are believed to form in one of two ways. The most massive stars die, leaving black holes. These are born with 10–20 times the mass of the Sun, most likely, in explosions called Long Gamma Ray Bursts, or long GRBs, for short. Somewhat smaller black holes, born with a few times the mass of the Sun, are likely produced in explosions called short GRBs. These accompany the violent merger of two neutron stars, or perhaps a massive white dwarf and a neutron star. Short GRBs last a fraction of a second, the official cut-off being less than 0.2 s long. Long GRBs can last several minutes, with an unusual subset, known prosaically as ultra-long GRBs, lasting an hour or more. Long GRBs, like magnetars, are believed to be produced by fast rotating stars that have lost their hydrogen and helium-rich envelopes. All of those characterized so-far, excluding the very rare ultra-long GRBs, are accompanied by Type Ic supernovae. These explosions have spectra that lack hydrogen and helium. The problem with the basic scheme is that these carbon-rich stars have to shed their outer layers, leaving only the inner core of the star. If this process was driven by simple stellar winds, the star should slow its spin as it shed angular momentum in the wind. However, as with the magnetars, it is thought that the star must be spinning quickly in order to generate a strong magnetic field near its core. This field then helps drive two, opposing jets of material out of the star’s poles. It is within these jets that the gamma ray burst is generated through violent collisions between particles accelerating in their stream. The other problem with the basic WR scheme is how to get the star to lose its helium core and expose the inner carbon core. This requires a lot of huff and puffing, and it simply isn’t clear that stars with 40 or so times the mass of the Sun will be able to do this before their cores implode. The obvious solution is to pair up the progenitor of the GRB with another star. The companion can steal the others hydrogen and helium, leaving the exposed inner carbon core. To do this, Stellar Soap Operas 189 both stars must be orbiting one another very closely, less than the Sun-Mercury distance. After all, the left-over WR star is a rather small object with a radius not much greater than the Sun. As with the formation of the magnetar, the key to solving the mystery of GRBs will be to find the smoking gun: a companion star with an unusual chemistry, which has either been kicked out of the star system, or remains bound, but in a very elliptical orbit. Although it isn’t impossible that such evidence will be found 1 day, the problem with long GRBs is their distance. Most are hundreds of millions or billions of light years away. Although teloscopy is good these days, it isn’t that good. At present, it is impossible to resolve individual stars at the distances that most GRB occur, so finding their progenitors, or indeed finding their leftovers, will be hard indeed.

Pair Instability: The Unfolding Stories of SN 2006gy and SN 2007bi

Although astronomers know much about long GRBs come about, there remains a lot of speculation about the universe’s most gen- erously proportioned stars. Stars born with masses greater than about 95 times the Sun, could, in principle, follow a different path to their deaths. If these stars could keep hold of at least half their mass at death, the core could become unstable through a process known as pair instability (Chap. 2 ). Pair instability involves a battle between highly energetic gamma rays and the core of super-heavy stars. In all massive stars the temperature rises to about 800 million Kelvin when carbon is ignited. At the point of oxygen ignition, a few years later, the temperature escalates beyond 1.2 billion Kelvin. At these temperatures most of the radiation is so energetic that the gamma rays can turn, reversibly, into particle-antiparticle pairs, following Einstein’s E = mc2 . This equation, simply put, states that matter and energy can be converted one to the other, because they are, in essence, the same thing. In most massive stars this to-ing and fro- ing of matter and energy is not a big deal. In these stars, the core is forced to generate energy at a faster rate to keep it held up against gravity. For these stars, they simply consume fuel even faster, but otherwise come out of the battle unscathed. 190 The Complex Lives of Star Clusters

For the universe’s most massive children this is not the case. In these stars the outer layers are so massive that as soon as pair instability sets in, gravity takes charge and crushes the core inwards at a catastrophic rate. Implosion generates more heat as the star’s gravitational energy is converted to kinetic energy then heat energy. The temperatures rise to over 3.5 billion Kelvin and carbon and oxygen fuse to form iron. Such violent reactions generate prodigious amounts of heat that reverse the in-fall and drive material back outwards in an enormous shockwave. In stars with masses somewhere between 95 and 130 times the mass of the Sun, this shockwave powers through the star with as much energy as a small supernova. The outermost hydrogen- rich layers are blasted off into space generating a supernova-imposter, an explosion with the power of the supernova, but one which leaves the star intact. Much like taking the lid off an overheating pan, the star then expands, cools and then relaxes back. However, if there is still enough mass leftover, the process can repeat several times until the star has been whittled down to around 47 times the mass of the Sun. At this point the core will remain cool enough to burn its remaining oxygen quietly, allowing the star to complete its evolution as a Wolf-Rayet star and leave behind a neutron star or black hole. As this process causes expansion and contraction of the star, potentially several times, these explosions are called pulsational pair instability events. Pair instability in stars with masses between 130 and 260 times the mass of the Sun generates enough energy to destroy the star in its entirety. Strictly speaking, it is these stars that are true pair instability supernovae. Even more massive stars collapse promptly into intermediate mass black holes when their cores become pair-unstable. That’s the theory at least. But do such stars even exist, or have they ever existed? The problem with pair instability is that it should produce a vast amount of iron, through the radioactive decay of nickel-56. If the universe had spawned many such supernovae then the stars currently living in globular clusters or loose in the halo would be rich in iron. As Chap. 4 described, they most definitely are not. Halo stars are rich in oxygen which is produced in abundance by regular core-collapse supernovae. Thus, if pair instability supernovae occur at all, they must be very rare. Stellar Soap Operas 191

However, there is a smattering of stellar explosions that seem a bit too powerful to be of the regular variety. Of these two seem to fit the bill for the pulsational pair instability fraternity. In 2006 two supernovae were observed in fairly distant galaxies. Of these SN 2006gy, held the title of the most luminous supernovae recorded until a somewhat earlier event, SN 2005ap, took back the title. SN 2006gy was first observed by Robert Quimby and Peter Mondol near the centre of the spiral galaxy NGC 1260 on September 18th, 2006. From early on the explosion seemed unusual. For one it was too bright to be either a core-collapse, Type II explosion or a supernova of Type Ia. The supernova displayed narrow emission lines indicative of hydrogen lying around the supernova, plus evidence that the supernova shockwave was powering through shells of hydrogen and helium-rich matter that surrounded the star. When this debris was examined in more detail it was found to add up to around 20–25 solar masses of gas and dust: the vast majority of this was concentrated into a shell at approximately three times the distance between the Sun and Neptune. Further analysis provided clues to the explosion and its immediate precursor. The explosion had generated a shockwave moving at 4,000 km/s and this had slammed into the shell, which must have been ejected a few years earlier. The collision converted the vast amount of kinetic energy in the shock into the visible radiation that continued to power the supernova for months afterwards. Alternative scenarios involving powering the explosion through radioactivity seemed untenable. For one, to power SN 2006gy you would need 22 solar masses of nickel-56, which is well outside the range that can be generated in any known supernova. SN 2006tf was very similar. Although a dimmer explosion, it was sill the third brightest supernova in its day, trailing SN 2006gy and another oddity, SN 2005ap. Around this supernova, the shell of material also lay at a similar distance to that surrounding SN 2006gy and was, itself, embedded within a relatively slow moving, clumpy outflow of gas, similar to that seen surrounding LBV stars. The speed of these layers suggested that the clumpy outflow was generated in the decades to centuries prior to 2001. In that year the progenitor of SN 2006tf, still packing over 110 solar masses of material, blew off its 20–25 solar mass shell at a thousand kilometers per second. Five years later the core completed 192 The Complex Lives of Star Clusters its evolution and presumably collapsed into a black hole. The supernova that accompanied its demise then blasted the outgoing shell generating the brilliant display. Interestingly, the pattern of events matches the theoretical predictions for a pulsational pair instability event on a star of its mass. Stan Woosley (University of California at Santa Cruz) ran a number of simulations for stars of various masses that were undergoing pulsational pair instability. Although these had been done with SN 2006gy in mind, the match to SN 2006tf was uncanny. Certainly, not proof in itself, but an intriguing match made between theory and observation. The problem with these two supernovae, and some others like them, is that the stars that generate them have to be very massive. Whether pulsational pair instability is the driver or not is irrelevant. The problem is these stars are just massive; far more massive than would be expected if they had formed with the mass they had at death. SN 2005gl, observed by Avishay Gal-Yam (Weizmann Institute), was another problem. Although not especially bright, indications were that this was also the death of a star with perhaps 150 times the mass of the Sun. In pre-explosion images, taken by Hubble, the progenitor star could be spotted lurking in a cluster of bright and presumably massive stars. The presence of profuse quantities of hydrogen-rich gas around it also suggested that the star was an LBV. The issue with LBVs is that they should mark a stage in the life of a massive star, just after it has left the main sequence. In the typical scheme of things, the LBV ejects most of its outer hydrogen-rich layers leaving a small star dominated by helium, carbon and oxygen, a Wolf-Rayet. Indeed, on HR diagrams there is a very obvious desert of stars in the top right corner. This region is bounded on its lower side at a luminosity of around 500,000 Suns; and on its blue side at a color similar to that of the Sun, but slop- ing upwards to higher temperatures as the luminosity is increased. This is the Humphrey-Davison limit. Stars can cross this divide as they evolve towards higher luminosities, but they don’t stick around for long. For instance, Eta Carinae, lies just on the cool side of the Humphrey-Davison limit. It has a luminosity of around 1,000,000 Suns. However, it is very unstable and clearly trying to shed much of its matter into space. If allowed to continue, within Stellar Soap Operas 193

1,000 R

100 R Any stars here are blue stragglers produced by collisions and may die through pair 10 R instability/pulsational pair instability 3 R WN-7 1 R 1 million η−Carinae Rho WC-7 Cassiopeia

WO 100,000 Luminosity (times solar)

Stars, here, die by conventional core collapse He-ZAMS H-ZAMS

Temperature (K) 100,000 32,000 10,000 3,200

FIG. 6.5 The location of the Humphrey-Davison Limit on the HR Dia- gram and relative position of Eta Carinae and some Wolf-Rayet stars. Stars that die as a result of pair instability must lie above this limit or in the WN category. Most very massive stars die as WR stars. Key: WO — oxygen-rich WR stars; WC —carbon-rich WR stars; WN —nitrogen-rich stars; R refers to multiples of the solar radius; ZAMS —Helium (He) or hydrogen (H) zero age main sequence for comparison a few thousand years it will evolve into a hotter, but less luminous blue star: a Wolf-Rayet (Fig. 6.5 ). Of somewhat lower mass, Rho Cassiopeia is a yellow hypergiant with a mass of around 30 solar masses and a luminosity a few hundred thousand times that of the Sun. It is actively losing its outer layers and growing hotter (Fig. 6.5 ). Again, within a few thousand years it will shed its hydrogen-rich outer layers and morph into a blue supergiant or WR star. Observations of massive stars suggest that these lose about a ten millionth of a solar mass of material in stellar winds each year that they are burning hydrogen. Mass loss steps up to 100 times the value it was on the main sequence, as these stars expand into supergiants and hypergiants. Stellar winds can then rip away a ten thousandth of a solar mass per year potentially stripping a hundred solar mass star down to half this value in half a million years. 194 The Complex Lives of Star Clusters

Super-imposed on top of this spendthrift lifestyle are more violent outbursts by Luminous Blue Variables (LBVs). These can have eruptions that dispose of up to a few Sun’s worth of mass in a matter of months. These eruptions lend the name S Doradus, or eruptive, variables to this class of highly luminous star. All of this waste exacts a heavy toll and after a short time a star with well over a hundred solar masses will be whittled down to less than 30. A 200 solar mass star such as R136a in the Large Magellanic Cloud’s 30 Doradus nebula can be pared down to less than 50 solar masses in the 5 million years that it will take the star to run itself into its early grave. In most simulations, even the most massive star ends up reduced to a shriveled wreck long before its core can approach the stage at which pair instability could develop. It’s even marginal for the most massive stars to reach pulsational pair instability. Yet, astronomers have gathered some observational evidence that a few of the most luminous supernovae (SN 2006gy, SN 2006tf and SN 2008am) have pair instability in a pulsational form at their heart. Others, such as SN 2007bi are at least suggested, albeit controversially, to be driven directly by pair instability. Although true pair instability events remain unproven, it seems difficult to reconcile the simultaneous presence of very massive, fast moving shells of gas around supernovae such as SN 2006tf, if they are not driven by some sort of internal instability that precedes the supernova. Pulsational pair instability seems a logical choice and it is consistent with the very large masses of these stars. Regardless of the mechanism of these explosions, astronomers are faced with a conundrum. Just how do some stars retain so much mass right up to the point the Grim Reaper strikes, when mass is being shed like it’s out of fashion throughout their lives? All of these massive stars should be slimmed down by the time death comes knocking, yet they aren’t. The only viable option is that the universe manufactures a small, hardy army of ultra-massive stars that are constantly rebuilding themselves as stellar winds carve chunks out of them. This is where massive star clusters come in. The most viable mechanism is a process of runaway collisions in the heart of these stars clusters. These stars then appear as blue stragglers, located above the point where the remaining main sequence stars evolve to become Stellar Soap Operas 195 giants or supergiants. Calculations by Simon Portegies- Zwart, then at the University of Amsterdam, showed that if a massive cluster of stars was sufficiently dense, runaway collisions would occur between the clusters most massive stars. The outcome would be a star with up to 800 times the mass of the Sun. Such a beast would encounter pair instability and implode forming an intermediate mass black hole with 500 times the mass of the Sun. The key to this outcome was the density of the stars. Make it too low and the stars evolve into WR stars and die long before they can collide. Just right, and the stars can collide frequently and merge forming true monsters. In related work, Simon Portegies-Zwart and colleagues examined other kinds of interactions in young massive start clusters that resulted in lesser beasts. Aside from the possibility of creating truly monstrous multi-hundred solar mass stars, runaway collisions produce a variety of interesting objects, paralleling the work by Hurley and Shara that was described earlier in the chapter. Simulations were run using a variety of believable stars, a proportion of which were in binary systems. One such binary consisted of stars with 57 and 26 times the mass of the Sun that are separated by over 100 billion kilometers. The more massive of the pair grows through successive collisions with passing stars until it weighs in at over 130 times the mass of the Sun. Within 5 million years it is dead: its core collapsed into a black hole. It is the initially large size of the binary partnership that sets the 57 solar mass star up for its eventual doom. One million years after its formation, the 57 solar mass star and partner have encountered another binary. As was clear earlier, the large size of the binary system guarantees the stars will interact with passersby. The passing binary system is torn apart by the gravitational forces exerted on it by the massive stars in the original binary system. The disrupted partnership doesn’t contribute directly to the original system as the two interlopers are violently ejected from the cluster. However, in transferring orbital energy to the interlopers, the original binary system is forced to shrink in size. At this point the two stars present a smaller target but interactions with passing stars will directly impact upon the binary system. Collisions now occur between the most massive star and three other less massive but still substantial stars over the course 196 The Complex Lives of Star Clusters of the ensuing 2 million years. The original 27 solar mass partner is ejected during the scramble of these encounters. All the while the stars are falling towards the core of the cluster where there is a much higher density of stars. This increases the number of encounters that are possible. In the case of our two protagonists, the original 27 and 57 solar mass stars, the pair have exchanged partners, now, a few times, but are now destined for a further, fateful encounter. Before this happens encounters between the 27 solar mass star and neighbors sends it briefly careering towards the periphery of the cluster. However, it has insufficient energy to escape and then falls back inwards. The pair of binaries, containing our evolving stars, now approaches one another again and engages in a dramatic collision. The 27 solar mass star and its new partner collides with the binary containing its former partner and its new companion. The two former partners exchange stars with one another and the 27 solar mass star returns to its former companion. Shortly thereafter the most massive of the pair implodes forming a black hole. Less than a million years later, or 5 million years after the pair formed, the 27 solar mass star explodes and the neutron star that is formed receives a big enough kick from the explosion to be ejected from the cluster altogether. This neutron star appears as a high-velocity star destined for inter-galactic space. The partnership has finally been broken. These titanic events are summarized in Fig. 6.6 . In Portegies-Zwart’s models many of the “binaries” are actually parts of much larger groupings of gravitationally bound stars. In one case the former 57 solar mass star is part of a group of seven stars all orbiting a common centre of gravity. This messy septuplet structure is highly unstable, with each star exchanging momentum with the others. This results in a constant realignment of their orbits and primes them for collisions, much like the far simpler V 838 Monocerotis - discussed shortly - or Eta Carinae systems. In the model the combined mass of the seven stars is 151 times that of the Sun: a lot of mass primed for catastrophe. Most collisions soften, or widen, binaries. This means that energy has been transferred to the binary from the incoming star. The incoming star must slow down and become deposited in the core of the cluster. If repeated more and more stars would end up in the cluster core, priming the system for more frequent collisions, Stellar Soap Operas 197

a f a b g e b a f c b d e a b f

a f b

FIG. 6.6 The twisted fate of massive stars in a cluster like R1367. A pair of 57 and 26 solar mass stars (a and b ) form towards the cluster core and encounter three stars ( c , d and e ). Stars c and d merge with a and b , while e is ejected. The engorged stars a and b encounter another binary and continue to grow larger. One star, f , is ejected from the partnership but soon returns and is finally recaptured, while the central pair a and b , merge, forming a 133 solar mass star. This explodes and collapses to form a black hole. Shortly thereafter, at 6 million years, also blows up and the remnant neutron star is ejected from the cluster. Other stars are involved but are left out for the sake of simplicity… with the outcome of any collision depending on the mass and velocity of all of the stars. As was mentioned before, where a low mass star swings passed a massive binary, the low mass star is more likely to pick up energy and be accelerated outwards. This leaves the two massive stars in a tighter, harder, orbit. Finally, as more stars collect in the core of the cluster, naturally more pair up and form more binary systems; and it is collisions between these that are most likely to result in collisions or exchanges of stars as they present the largest targets to one another. One would then expect the candidates for pulsational pair instability supernovae to emerge in dense, young clusters of 198 The Complex Lives of Star Clusters

massive stars, much like Westerlund-1 or R136 in 30 Doradus. Run the clock forward 3 million years and it would be interesting to observe whether the (roughly) 250 solar mass R136a has met this fate. Follow up observations on supernovae such as SN 2006gy are also needed. Although this star was clearly massive, its host galaxy was not a prolific manufacturer of young, massive stars. Was the progenitor of this supernova born in a dense cluster of stars, experiencing runaway growth before it died? Or are models of the rate at which massive stars lose mass out of kilter? Either explanation is viable, but neither is clear. Thus, whether stars such as R136a can hold enough mass to die in anything other than a conventional core-collapse supernova is anyone’s guess, but it can’t be ruled out. In slightly more evolved clusters such as NGC 330 in the Small Magellanic Cloud, the majority of stars that are now leaving the main sequence have masses around 11 times that of the Sun. Indeed, you can track sufficient stars from the main sequence across to the red (super)giant branch to confirm that this is true. However, above the main sequence turn-off there exists a small but noticeable number of more massive stars with luminosities compatible with masses as high as 20 times that of the Sun. Although this cluster does not seem massive enough to have sufficient collisions to manufacture stars of this mass, other types of less dramatic interaction are possible. The majority of these blue straggler stars will most likely have been formed through binary interactions, where one star has donated most of its mass (or had it stolen) while it was a red supergiant. If both stars are of similar mass then the bulk of the donor star mass that is lost will be accreted by the companion (see previously in the discussion of SN 1987A and SN 1993J). Thus, for some considerable time (several million years) after the last primordial massive star blows up there will be a retinue of other stars that have been beefed up through transfers of mass from more evolved companion stars. Perhaps more interesting than the specter of a pair-instability explosion is the collision that leads to it. Take 10 million, million, million, million, million kilos of very hot star and smash it into another body of equivalent mass at several tens of kilometers per second and you have rather a lot of energy available to power a Stellar Soap Operas 199 display. A standard supernova packs around 1044 Joules (1 followed by 44 zeros) in energy, of which 1042 Joules is liberated in visible radiation. Explosions, such as SN 2006gy, released up to 100 times this amount of visible radiation, through the efficient conversion of their abundant kinetic energy during the collision of the shockwave with the surrounding shell of gas. If you collide two stars, each with 100 solar masses of material into one another at 10 km/s you can access 10 40 Joules of energy, or one hundredth the power of a typical supernova. Although the explosion may not be as meaty as a true supernova, nor pack enough energy to destroy the resulting star, it is still a rather impressive affair that would be visible across intergalactic distances. Such mergebursts, as they are known are still largely theoretical beasts, but they must occur. Astronomers have observed hundreds of blue stragglers that appear to have been formed through the direct merger of lower mass stars. In 2002 V838 Monceratus erupted inside a cluster of young massive stars, with an energy equivalent to an LBV eruption. For some time, thereafter, it was the most luminous star in our galaxy. As most explanations involving stellar explosions were ruled out, astronomers turned to the system in which the eruption occurred. The dynamics of the two remaining stars was at least suggestive that something catastrophic had happened to another star within the system. Other such events have been recorded, such as V1309 Scorpii in 2008. Here a low mass star was engulfed and merged with an aging giant star. The resulting fireworks display bore striking similarities to V838 Mon, albeit with lower energy. Why less energy? The stars involved in this event were roughly the same mass as the Sun, so the available kinetic energy was an order of magnitude less than in V838 Monceratus. Take a step back and look at Eta Carinae’s great eruption in 1843, and the possibility that this was a mergeburst seems a distinct possibility: Philipp Podsiadlowski certainly thought so. The mysterious explosion drove over ten solar masses of gas into space with an energy equivalent to a supernova. The star survived but took on a very different persona. Could a mergeburst drive such a powerful event? Easily. The mass of the stars involved is enormous—more than ten times that in the V838 Mon system. Moreover, if the two stars were in orbit around one 200 The Complex Lives of Star Clusters another their velocities would have been around 1–2,000 km/s. Kinetic energy varies with the square of the velocity of the object. Therefore, if two stars, each with a mass approaching 50–70 times that of the Sun collided with this velocity, their kinetic energy would easily approach that of a conventional supernova. We certainly can’t be sure what the cause of the explosion in 1843 was, but suggestions such as pair instability don’t quite work: the star should have died by now and it is very, very much alive. Other sorts of internal instability, such as changes in the way energy is delivered from the core to the outside universe, also can’t quite duplicate the energies involved. Stellar collisions, in star clusters are therefore a very promising and intriguing solution to some of the universe’s most enigmatic explosions. There are two broad classes of collision, extrinsic and intrinsic. Extrinsic collisions are between a star and one that has come from some distance and collided directly with it. In intrinsic collisions, two or more stars have their orbits scrambled, perhaps through interactions with external stars, or simply through the exchange of momentum between them. This results in two of the stars spiraling in on one another until they collide. The latter sort of collision appears to account for V838 Monceratus, V1309 Scorpii and perhaps Eta Carinae. The other type of collision has never been observed but surely must occur where stars and binary systems are closely packed.

The Likely Tale of a Massive Straggler

In the Small Magellanic Cloud a Be supergiant (Chap. 3 ), R4, provides another potential example of a stellar merger. If you recall, a “Be” star is a hot B-class star that is surrounded a strong outflow of material from its equator. This material forms a disc around the star. In all Be stars, the outflow is driven by the fast rotation of the star. Here, the blue, B-class supergiant is paired up with an apparently more evolved, but less massive A-class supergiant. The system is also flanked by a cloverleaf structure similar to the hourglass that surrounds SN 1987A. The suggestion is that this odd partnership originally consisted of three stars. Two of them were less massive B-class main sequence stars, with a third more Stellar Soap Operas 201 massive partner. The more massive star left the main sequence to become the A-class supergiant. Meanwhile, one of the two lighter B-class stars followed suit and expanded. As it did so it swallowed up its remaining B-class partner, leaving a much more massive, but less evolved B-class supergiant in orbit around its now less massive, but more evolved A-class partner. The fast rotation of the Be stars is naturally explained by the merger of two stars that were rapidly orbiting one another. Furthermore, the cloverleaf nebula surrounding the system was ejected in a bipolar outflow similar to that observed around V1309 Scorpii. This outflow would have originated in the cloud of material ejected as the two stars collided. Clusters of stars provide a natural environment where binary stars can merge. Clusters are both plentiful source of such systems, but interactions between cluster stars can also drive the component stars of the binary partners to merge. In Chap. 1 we saw how passing stars can “pump up” the orbit of a star in a binary system until one of its component stars reaches escape velocity and is ejected. However, this is a two-way street. It is more common for mature binary systems to act as a reservoir of energy. A passing star interacts with the binary and steals some of the kinetic energy of the orbiting stars. This accelerates the interloping star causing it to move away quickly. The two binary stars, now low on kinetic energy, move closer together. Over time further interactions can cause the stars to merge completely. In a triple star system, the three stars can interact in the same way. The out-lying star steals energy from the other two causing it to move outwards, while the inner pair moves closer together. In Eta Carinae, R4 and V838 Monceratus there is at least good evidence that this scenario played out. The orbits of the remaining stars, in each system, are compatible with this idea: three stars; one outer star robs energy from the other two, causing them to violently collide. In each collision, a large quantity of star-stuff is thrown out forming a bipolar nebula around the remaining star system. In the massive open cluster, NGC 3603, the LBV star Sher 25 is flanked by an hourglass nebula identical to that around both SN 1987A and R4. Did this LBV form through the collision of two smaller, but still very massive stars? 202 The Complex Lives of Star Clusters X-Ray Binary Systems

Star clusters play host to the largest proportion of star systems called X-ray binaries. These systems host a main sequence or giant star paired up with a white dwarf, neutron star or black hole. X-ray binaries come in two general flavors, with a third suggested. If we exclude the white dwarf partnerships (cataclysmic variables) that were discussed in Chap. 3 , the first are partnerships of a neutron star or black hole with a high mass star (more than ten times the mass of the Sun). This partner can be a main sequence star, a supergiant or a stripped down helium-rich star, such as a Wolf-Rayet. High mass X-ray binaries (HMXBs) release their budget of X-rays when the stellar wind from the partner is captured by the compact companion. This escaping gas then forms an accretion disc around the neutron star or black hole. These systems tend to be fairly wide, as gas can be accreted from the relatively dense stellar wind of the companion over considerable distances. The X-rays are also generally harder. They are of higher energy than those emitted by low mass systems (below) and are (in most cases) associated with streams of hot gas impacting the magnetized poles of the neutron star. X-ray emission is also strongly pulsed. Pulses occur as the hot gas is focused onto two points, one at each magnetic pole. As the neutron star rotates, the impact sites swing into and out of view, giving the appearance of pulses (Fig. 6.7 ). The pulsed emission also implies that the magnetic fields in these neutron stars are still relatively strong and able to

a b c

FIG. 6.7 Three different kinds of X-ray binary. In each the compact companion that generates the X-rays is a black hole or neutron star. The mass of the companion varies. In a the companion has a mass greater than 10 times that of the Sun, making it a High Mass X-Ray Binary (HMBX). In b the star has a mass of between 10 and 1.5 times the mass of the Sun—an Intermediate Mass X-Ray Binary (IMXB) and in c , the mass of the companion is low, making it a Low Mass X-Ray Binary (LMXB) Stellar Soap Operas 203

concentrate the flow of hot plasma streaming from the companion star. Given that these systems include high mass stars, they must be young, and this in turn allows for a fairly youthful and still strongly magnetic neutron star to take part in the partnership. The other variety of X-ray binary is unsurprisingly called a Low Mass X-Ray Binary (LMBX). In these systems, a star, with a mass equal or less than one and a half times the mass of the Sun has the dark companion. Material is accreted directly from the visible star’s surface onto the neutron star or black hole via an accretion disc through “Roche Lobe Overflow”. Here, the star has completely filled the invisible field around itself that it has sole gravitational control over. Material escapes the star’s Roche Lobe and pours towards its unseen companion through a constant stream. X-rays come, primarily, from the accretion disc when hot plasma from the companion hits the surface of the neutron star. Pulses are rare in these systems, as the neutron star has had time to cool down and lose much of its magnetism before its partner was drawn into the dance (Fig. 6.8 ). Low mass X-ray binary systems can, however, flare unpre- dictably when hydrogen or helium that has been acquired by the neutron star ignites, releasing a torrent of soft gamma rays and X-rays. These X-ray bursts only happen to low mass systems. Between the two extremes are Intermediate Mass X-ray binaries, and guess what: they are black holes or neutron stars paired to a star with a mass 1.5–10 times the mass of the Sun! In these systems mass is lost from the visible star through Roche Lobe Overflow, but at a very high rate (Fig. 6.9 ). In systems with a neutron star, X-rays are emitted from the accretion disc and the surface of the neutron star. Where a black hole is present, there is no surface and X-ray emission comes exclusively from the accretion disc or inflowing material. Determining whether the invisible component is a neutron star or black hole takes some work. You obviously can’t see the black hole and though neutron stars could release a lot of energy, they are very small. Therefore, spotting them in the glare of the stellar partner and any accretion disc is nigh on impossible. Luckily, astronomers have a few tricks up their sleeves. One way is to measure the brightness of the visible star and then measure the other parameters of the system as a whole. In particular, by 204 The Complex Lives of Star Clusters

FIG. 6.8 An illustration of the binary power of globular clusters. The blue dots , hovering like bees around the disc in the super-imposed of the visible image of the Sombrero galaxy (left ) are all copious emitters of X-rays and are all coincidental with globular clusters. X-ray binary systems lurking deep within each provide the display. Image credit: NASA/ Chandra/Hubble/Spitzer measuring how quickly the two components orbited one another (their radial velocity) you could determine the mass of each. If the mass was likely to be greater than 2.5–3 times that of the Sun, the invisible partner was likely to be a black hole. Another is to examine the X-ray emission from the system. By assuming that the X-rays were all coming from the invisible component, astronomers could determine how likely it was this object was a neutron star or black hole. For example, material falling onto a solid neutron star should release harder (more energetic) X-rays than material plunging over the edge of the abyss into a black hole. Very subtle measurements of X-ray released from X-ray binary systems confirm this: those with black hole candidates, such as GRO J1655-40, show very soft emission when the systems are in their relatively quiet state. Moreover, the total X-ray output Stellar Soap Operas 205

FIG. 6.9 X-ray binaries are found alternating between high, soft states (a ) and hard, quiescent states (c ). In a , a disc of material feeds onto the neutron star or black hole and emits a lot of soft (low energy) X-rays. When either the supply of fuel runs low, or the disc releases so much energy that it disrupts the flow of material (b ), the disc is disrupted and only lower levels of hard X-rays are released from the system ( c ). Many other X-ray binaries also release jets of material that are shot out of the disc at right angles ( blue triangles ). These are often driven by magnetic fields that are generated within the spinning accretion disc of plasma. These jets turn off when the systems enter their hard state of the system can be used to infer what’s eating all the gas. Arthur Eddington coined a term called the Eddington Limit. A black hole can effectively eat more material than a neutron star because the black hole’s gravitational pull at its “surface” is greater than that of a neutron star. As the material spirals into the black hole or neutron star, the heat generated tends to push against the pull of gravity. Raise the temperature high enough and the material can’t be captured at all. Instead the push of the radiation will drive the gas back outwards. The boundary between suction and dispersal is called the Eddington Limit and all hot objects have this. It defines how luminous a supergiant star can be at the Humphrey-Davison Limit and how much food a black hole can absorb. Black holes have higher limits than neutron stars as the latter have lower gravity and can generate extremely high temperatures at their surfaces. 206 The Complex Lives of Star Clusters

This explains the predominantly hard X-ray emission from neutron stars but not black holes. However, as recent measurements have shown, there are some objects that defy Eddington’s rules. For one, a neutron star in a high mass binary in M82 clearly accretes far more matter than it should and consequently appears as an ultraluminous X-ray source. The data Matteo Bachetti (University of Toulouse) and colleagues acquired clearly showed that despite the high output, this source was not a black hole. Thus, it appears unwise to make assumptions about the sources of X-rays in these systems. X-ray binaries systems, containing black holes often alternate between two states: a harder, high energy state, in which there is relatively little, overall, emission. This is known as the quiescent state. In this state a very hot corona of thin plasma exists around the accretion disc. Soft X-rays coming from the disc interact with electrons in this wispy corona and pick up energy through a process called Compton scattering. The electrons lose energy, giving it to the passing X-ray photons. The second state is a softer, lower energy state, in which there is profuse emission of X-rays. This is known as the “high state”. In the soft state, the black hole or neutron star is actively accreting hydrogen and helium and the disc of material flowing onto it releases large quantities of soft X-rays. In the hard state, the accretion disc is disrupted and only a small amount of material makes its way towards the black hole or neutrons star (Fig. 6.9 ). Disruption happens because the rate at which material flows towards the compact remnant star is uneven. If the compact star is given more material than it can accrete it builds up around the object. This increases the amount of X-rays and other radiation released by the object, or in the case of a black hole, by the innermost part of the disc. This surge in radiation eventually exceeds the Eddington Limit and radiation pressure forces the disc away from the compact star. The innermost portion of the disc disperses leaving a thinner sea of very hot plasma that releases the hard X-rays, but at a much lower rate than before. After the energy has been released, much like a pressure cooker releasing pressure, the system returns to its original form. Gas can flow all the way towards the compact star, once more and the process repeats. Stellar Soap Operas 207

The transfer of matter can vary if the orbit of the two stars is elliptical, or if magnetic fields are involved in channeling gas from the visible star to its companion. These can vary in strength with the star’s internal cycles or a wobble in the tilt of its axis, which in turn varies the amount of gas the star can channel. An elliptical orbit will bring the star that is donating material closer to and further away from its dark companion. When it gets closer the star’s Roche Lobe will shrink and more material will be pulled towards its companion than if it is further away. There is a seemingly endless variety of partnerships in these systems. Where there is a very high rate of mass transfer from the visible donor to the compact star it is more likely that emission will be predominantly soft. This is in part to the high rate of transfer allowing persistent and effective accretion of gas, but also because gas in the system will tend to absorb hard X-rays and re- release the energy at softer wavelengths. That said, clearly there is a lot of potential for variation, and it can be quite a struggle to pin down the characteristics of all the systems astronomers observe. High Mass Binaries are only found in young star systems and clusters that are more than 3–5 million years old. This time allows at least the most massive star in the partnership to evolve into something exotic like a black hole or neutron star. A very large number of low mass systems are known and these can have two origins. The primordial systems form with the cluster when a massive star in a binary dies. Over time these systems can be disrupted by close encounters between stars. The cluster’s stock of binary systems is constantly replenished by close encounters between lone stars and pre-existing binaries. In many close encounters, a binary with two unevenly matched stars will be disrupted. The lightest of the original pairing is ejected, sometimes from the cluster as a whole, while the interloping star becomes bound to the more massive binary partner. This is particularly true of neutron stars, which are themselves usually a lot more massive than the stars they replace. Thus a binary consisting of a neutron star and a normal star is established: a new low mass X-ray binary. In the 12.7 billion year old cluster M4, there is a particularly interesting partnership of a white dwarf, millisecond pulsar and giant planet. This system, known as PSR B1620-26, was undoubtedly created when the millisecond pulsar invaded the original 208 The Complex Lives of Star Clusters partnership of planet and white dwarf. In order to make that system stable, a third, now long-gone star may have been ejected, although simulations by Eric B. Foralso (MIT) suggest that the system’s rather eccentric orbits may have pre-existed the arrival of the neutron star. However, that the planet survived in its present orbit is fortuitous, to say the least, given the radical change in the gravity of the system that would have accompanied the arrival of the neutron star. In 2002 Raymond White (University of Alabama) suggested that all Low Mass X-ray Binaries might have been formed in globular clusters, these dense star clusters forming the ideal environment to bring neutron stars into close partnership with other stars. Subsequently, some are ejected through tidal interactions with other cluster stars and become somewhat more dispersed throughout the galaxy. Indeed, until fairly recently, this was a very viable suggestion that was based upon observations made by the Chandra X-ray telescope. However, higher resolution studies were carried out by Rasmus Voss (Max Planck Institute for Extraterrestrial Studies). Rasmus examined the X-ray emission produced by objects in the Milky Way, M31 and the nearby elliptical galaxy Centaurus A, indicate that while globular clusters play host to the local universe’s brightest LMBXs, there is a clear distribution of fainter objects scattered elsewhere throughout these galaxies. The bulge of M31 hosts a significant number of these. Precisely why bright LMBXs appear to predominate in globular clusters is unclear. One reason might be that they are constantly regenerated in these environments through interactions between the neutron star and new partnerships. Although LMBXs lead interesting lives, they cannot live long by the standard of the Sun. When one partnership emerges from the death of a massive star in a primordial binary or through a fortuitous pairing of a star with a pre-existing, but single neutron star, the stage is set for the demise of the system. Death comes in two guises: neither is particularly positive for the neutron star’s unwitting companion. LMBXs can die through the direct effect of the X-radiation from the neutron star or its accretion disc, or through the effects of strongly pulsed but beamed radiation coming from the neutron star. The former was proposed by Icko Iben Junior (University Stellar Soap Operas 209 of Illinois), Alexander Tutukov and Alexandra Federova (both at the Russian Academy of Sciences). Research work showed that as the neutron star continues to accrete material from its unwitting companion, X-rays blast the companion’s surface. This heats it strongly, essentially evaporating even more material that the neutron star can acquire. However, this is somewhat like hunting game with a machine gun. You kill an awful lot more than you could ever hope to eat. Of the mass boiled off the companion, only about 10 % is captured and consumed by the neutron star. The rest is evaporated into the surrounding space. Such prodigious loss of mass tends to widen the orbit of the two stars, and given enough time, the companion can be whittled down to something more akin to a giant planet than a star. If this process doesn’t terminate the companion, another, more insidious one will. Over time, the accretion of material from the companion spins up the neutron star. Even though the neutron stars in these LMBX systems are old and somewhat worn out, they can acquire the façade of youth through their malicious interaction with their companion. As material is accreted it adds angular momentum to the spinning neutron star. After several thousand years or more, the neutron star is accelerated until it spins at a few hundred times per second. As the neutron star accelerates the strength of wind generated near its poles also increases. Given the right tilt, this wind is directed towards its companion. Much like the X-ray glare described above, this intense wind rips into the companion star and boils it away on a very short timescale. Such Black Widow pulsars are observed: millisecond pulsars fatally irra- diating the shriveled remains of their companion. In a few instances, the companion star is a red giant and the transfer of material can exceed what the companion can absorb. Here, the neutron star or black hole is swallowed into a common envelope—a cocoon of gas and dust in which both the core of the giant and the neutron star or black hole orbit. In most instances this dense cocoon is dispersed by the churning action and stellar winds generated by both stars. Perhaps in a precious few the neutron star makes it all the way to the core of the other star before the structure is destroyed. In this case a Thorne-Zytkow object is formed (see previously, in this chapter). 210 The Complex Lives of Star Clusters

For those neutron stars that remain bound to their partners, and for whom their partners can lead long enough lives, LMXBs can become something more exotic. As the neutron star accretes material it is spun up, as we have seen. This intensifies the wind it generates until it is able to blow away its companion. Now, if this scenario is true then astronomers should be able to find stars that are part-way to completing the transition from standard neutron star to millisecond pulsar. A few such rarities have indeed been located. When a neutron star is young, it releases radiation in beams that are detected as pulses as these beams sweep past the Earth. Such enthusiasm can’t last long though and after a few million years at most, the neutron star’s spin is slowed down as its magnetic field tangles and brakes against the field in the surrounding galaxy. Therefore, after a few million years the rotation rate slows to a few per second and the pulses stop. The magnetic field will have dropped from perhaps a trillion times that of a fridge magnet down to a paltry 100 million times or so the value that an Earthly magnet can hold. Unable to pulse, such neutron stars continue to cool into invisibility over the ensuing billions of years. When given a kick from in-falling matter, these aging neutron stars are resurrected. Initially, accreting matter may be channeled onto the poles of the neutron star, where the diminished but still powerful magnetic field is concentrated. This generates hot-spots at each magnetic pole. As the neutron star rotates, these hot spots swing in and out of view producing pulses of X-rays that are detectable on Earth. This also allows the spin of the neutron star to be measured. Such quiet neutron stars punctuate their slow but steady accretion of material with violent outbursts. Hydrogen gas pulled onto the surface of the neutron star is rapidly compressed and heats up to over a billion degrees, just as it will do in the heart of a Thorne- Zytkow Object. Where the neutron star is exposed to space, the hydrogen fuel ignites violently when its density reaches a suitable value. This sends a flame rapidly across the surface of the neutron star in a manner akin to a nova on a white dwarf, but the similarity ends there. Hydrogen can burn all the way through helium to carbon, and on occasion enough carbon is present to send it burning into a sea of iron. In a nova, the energy released by nuclear Stellar Soap Operas 211 reactions and radioactive decay, drives much or all of the burning mass off into space: this expanding cloud is what is observed as the nova. A neutron star is a tad feistier. All of the burning material, regardless of its composition, is trapped by the neutron star’s intense gravitational field and the fire is eventually subdued. In high mass systems fuel is captured and burnt almost continually. However, in low mass systems the supply of fuel is more miserly and fuel burns in stops and starts called X-ray bursts. During these eruptions copious amounts of X-rays are released along with some gamma rays. Bursts fall into two classes: Type I and Type II. Type I bursts are associated exclusively with low mass X-ray systems and are essentially thermonuclear explosions. They are sub-divided into standard bursts, lasting a few seconds, and super-bursts lasting several hours. The more common bursts occur when hydrogen is taken from a companion. When sufficient fuel has built up the hydrogen ignites, just like it does in a white dwarf, nova. After a prolonged period of bursts, there will be sufficient helium to power another, longer super-burst explosion. As with the standard burst, the intense gravity of the neutron star traps the flame and only X-rays and gamma rays can escape to space. After several super-bursts it is possible for the neutron star to build up a sufficiently thick layer of carbon that it too ignites. These will also be recorded as super-bursts but with an even longer flame as still greater masses of fuel are needed to power the eruption. At least six suspect carbon bomb super-bursts have been detected, the first by Tod Strohmayer (NASA’s Goddard Space Flight Center). All are low mass X-ray binaries. Type II bursts are somewhat more enigmatic and produce more erratic outbursts. Both black hole and neutron star systems can produce these and these are believed to be caused by large variations in the amount of material flowing into the accretion disc and then onto (or into) the central dark object. Fuel may be supplied unevenly by the companion star, particularly if its orbit is very elliptical. When furthest from its companion, the flow will be least; and if the star has a strong magnetic field of its own the two fields could line up and deliver additional fuel to the neutron star or black hole companion. 212 The Complex Lives of Star Clusters

With each capture and burn the neutron star will have an accelerated spin, and as the neutron star spins up, it is born again as a radio pulsar. Very few systems have been found that appear to show a neutron star in the midst of transitioning from battered decrepitude to rejuvenated millisecond pulsar. The first that was identified comes with the flattering name SAX J1808.4-3658. This system was investigated by Sergio Campana and co-workers from Italy’s INAF—Osservatorio Astronomico di Brera. This system consists of a low mass star in orbit around a young millisecond pulsar. The pulsar spins at over 400 revolutions per second, fast by young, run-of-the-mill pulsar standards, but not by millisecond pulsar ones. This system is clearly still actively transferring mass and spin to the neutron star and this object should continue to accelerate. In time accretion will slow down, in part because the transfer of material is uneven and some gets lost from the star system. This causes the two stars to move apart. Accretion will also decrease, because at some point the enhanced magnetic field of the pulsar is likely to drive a wind that will begin to boil the companion away. As the companion shrivels in mass it will reach the point at which it is unable to keep its internal fires lit. If it is a main sequence star it will shrink past the lowest mass at which hydrogen fusion is maintained and it will change into a brown dwarf star. If the companion is a red giant, the core will become exposed leaving an inert white dwarf star. Neither of these stars holds gas in a form we would recognize. Instead the gas is degenerate and when these objects form, they will shrink back inside their Roche Lobes, ter- minating the transfer of mass to the neutron star. The neutron star can only grab hold its diminished companion once more, if gravitational waves, the roiling motion of space-time generated by massive moving objects, can rob the system of energy. In this eventuality, the two degenerate stars will spiral in towards one another and the process can begin again, at least until the companion is reduced to either dust or, perhaps, into a planetary system. Binary systems containing neutron stars are abundant in globular and other clusters. They are a very diverse set of systems that may contain other neutron stars, white dwarfs, red giants, main sequence stars or planets. Their abundance is due in the largest part to the crowded conditions found in such clusters. Stellar Soap Operas 213

Globular clusters, in particular, form an ideal breeding ground for these systems where stars are frequently brought into close contact with one another. The planetary system PSR B1620-26 is a case in point. The millisecond pulsar almost certainly invaded the system after the planet had formed. It may have been the agent that transformed a former red giant into the present white dwarf, and could have been spun up through this encounter. More likely, the pulsar was already a fast spinner, having been promoted to the ranks of millisecond star through a fatal interaction with a previous partner. Either way, the neutron star must be billions of years old and has clearly had sufficient time since it formed with the cluster to engage in all manner of activities. Entire classes of X-ray binary systems involve white dwarf stars. There are called cataclysmic variables. These are also extremely common in star clusters, particularly globular clusters, for the same reason that other X-ray binaries are. The cluster’s environment brings stars close together and results in a profusion of tight binary systems where two stars can interact. White dwarf systems are an order of magnitude more common than neutron star systems simply because white dwarfs are more common than their heavier brethren.

The Universe’s Loneliest Supernovae

SN 2005E was discovered on January 13th 2005 by the KAIT team, led by Alexi Filippenko, on the outskirts of the galaxy NGC 1032. The presence of helium in the ejecta led to it being classified as a Type Ib supernova, which was unfortunate as most of these explosions are attributable to massive stars, something that led to more than a little controversy in the scientific community. However, its presence, far from any site of star formation, or even a star cluster led most to prefer a model involving a low mass star—most likely a white dwarf. The problem with this supernova and several others which have been discovered subsequently is its odd chemistry. By far the most abundant element in the ejecta was calcium (nearly 50 % of it) which is, quite frankly, odd. Titanium-44 appeared to be the principle radioactive power source, rather than nickel-56, which 214 The Complex Lives of Star Clusters powers most supernovae (Chap. 2 ). Making any supernova produce mostly calcium is hard to do at best and making it dim is another problem. On top of this is the location of these supernovae. SN 2005E was close enough in images to NGC 1032 to suggest that this was its home, albeit a fairly distant one. Sibling supernovae of this type that have been found, subsequently, don’t appear to fall into any galaxies. How do you get a star to explode so far from home? One suggestion was that these supernovae were found in globular clusters. Although some galaxy clusters play host to a large cohort of free-floating globular clusters (Chap. 7 ), deep imaging of the regions in which the supernovae occurred seem to rule out the presence of an underlying cluster. Instead Joseph Lyman (University of Warwick, UK) proposed that these supernovae truly are orphaned events, occurring in the inky void between galaxies, or at least well into their halos. The authors also suggested that all of the original models for these supernovae are wrong, except for the involvement of a white dwarf. In their model two stars, one with low mass and one with a much higher mass are paired up. When the more massive star dies the resulting birth kick on the neutron star sends the partnership flying off into intergalactic space. Over the ensuing billions of years the lower mass star evolves into a white dwarf and it is now that the fun begins. Over time, gravitational waves bring the two stars ever closer together. Eventually, the white dwarf is violently dismembered and its viscera spilled over the surface of the neutron star. It is here that the neutron star cooks most of the white dwarf’s helium and carbon into calcium. Now, the interesting bit of this story is what happens next (at least in some cases). Where the neutron star receives sufficient mass it will implode to form a black hole. In the ensuing chaos the small mass of the white dwarf that remains in orbit is sucked down into the hole generating a short burst of gamma rays. The authors do not suggest that all short gamma ray bursts (those lasting less than 0.2 s) are associated with the merger of white dwarf and neutron stars, but they do suggest that a proportion of these explosions might be associated with such mergers and that these kinds of events might be the source of much of the calcium in the universe. There is certainly a correlation between the distribution of calcium-rich supernovae and short gamma Stellar Soap Operas 215 ray bursts, the latter having been tied to the distant outskirts of galaxies by the SWIFT X-ray telescope. The vast majority of short gamma ray bursts probably originate when two neutron stars collide. However, there should be a greater number of white dwarf- neutron star pairings than neutron star-neutron star binaries, given the higher frequency of stars that form white dwarfs. This must be particularly true in globular (and perhaps other star) clusters where white dwarf-neutron star binaries should form frequently. It must be said that no calcium-rich supernovae have yet been convincingly tied to globular (or other) clusters. However, given the preponderance of white dwarf-neutron star binaries in globular clusters, is it just a matter of time before astronomers get to test this model for the generation of these calcium-rich explosions—and perhaps link some of them to short gamma ray bursts?

Type Ia Supernovae and Beyond

How many Type Ia supernovae are known to have occurred in globular clusters? Remember when considering your answer that there is good observational evidence that 80 %, or more, Type Ia supernovae occur when white dwarf stars collide and the mass of the product comes at least close to the so-called Chandrasekhar limit of 1.382 solar masses. Recall also that within globular clusters stars are packed tightly so that white dwarf binaries, the types of systems that should produce collisions and detonations, should be common. The answer (with a 90 % confidence limit (meaning that a betting man or woman would be likely to agree with this answer 9 out of every ten times) is 0 %. None. Nothing: not one. OK, statistically there was a 10 % probability that one or two might have been missed, but when Rasmus Voss and Gijs Nelemans (both from Radbound University) made deep searches of supernovae sites, not one explosion appeared to arise within a globular cluster. Is this a surprise? When Michael Shara and Jarrod Hurley were “playing” with the GRAPE-6 supercomputer, simulations suggested that globular and open clusters would be likely hosts of Type Ia supernovae, indeed they could well contribute the majority of these explosions. Very recent observations led by “SNfactory” member 216 The Complex Lives of Star Clusters

Richard Scalzo of the Australian National University go further, suggesting that quite a wide range of white dwarf partnerships might power Type Ia supernovae of varying luminosities. If this is so, surely it would mean that star clusters would be even likelier to host such explosions? Certainly, recent observations suggest that the most luminous Type Ia events, plus an odd variant called Type Iax, occur in young star systems but that doesn’t necessarily mean young star clusters. Type Iax supernovae are faint Type Ia events typified by SN 2020cx. They were characterised by Ryan Foley. There were, at the time of writing, 25 identified explosions of this type, each with low velocities, abundant iron and evidence of a very hot fireball. It is suspected that these explosions do not always destroy the white dwarf, but transform it into an iron star. To look at this a little more deeply, we need to think about the sorts of white dwarfs that are likely to explode. They must have (or acquire) enough mass to push them up to or over the so- called Chandrasekhar mass limit. This is the mass at which the interior of the white dwarf can no longer be supported against gravity by degeneracy pressure. This is the force exerted by electrons on one another as they are forced to stack up in various energy levels under very dense conditions. Very close to this limit where the white dwarf is relatively hot inside, carbon nuclei are forced together and a thermonuclear fireball ensues. Where two white dwarfs come together, in most cases the intense gravity of each white dwarf shreds the pair and the combined mass collapses inwards, igniting in the process. Studies and theoretical predictions show that most Type Ia supernovae happen in star systems that are relatively young: less than 1 billion years old. The rate of these types of supernovae declines quite rapidly thereafter. Perhaps then the old-guard of globular clusters are simply too old to host these supernovae? Yet, if globular clusters are constantly forming new binary partnerships between white dwarfs why can’t they maintain the rate of type Ia supernovae? Work by Michael Shara and Jarrod Hurley that was published in 2002 seemed to suggest that such stellar partnerships would form readily. It does seem intuitive, but there is a problem with intuition. It neglects other factors that will act against the formation of massive white dwarf binaries. King amongst those is the very process you might expect Stellar Soap Operas 217 to help produce Type Ia supernovae progenitors: binary formation. In dense globular clusters stars tend to interact readily. If you are an aging star, evolving to become a red giant, then you need to retain at least half your mass if you are to ignite your helium core. We’ve already seen in this chapter that in the heart of modern day globular clusters, hanging onto your proverbial hat is rather tricky when so many stars are plowing through your neighborhood. Red giants have a bit of a hard time, with many losing their envelopes and ending up as sdB stars. Instead, those red giants that do maintain enough mass to burn helium and become carbon-oxygen rich white dwarfs are low in mass: roughly half that of the Sun. Any binaries formed from these will not have enough mass to trigger Type Ia supernovae when they collide. Instead they will most likely become either an R Corona Borealis star, or some other kind of (as yet) unseen giant that is burning helium in its core. It will not, ever, explode. Type Ia supernovae, thus become the preserve of (relative) youth; the descendents of much more massive stars that are now only seen in much younger clusters. So what about open clusters? Do these show more Type Ia supernovae than the background population of the Milky Way and other galaxy discs and bulges? That’s a very tricky question to answer. Aside from our galaxy and the very nearest neighbors, spotting the precise home of Type Ia explosions is very difficult to do with current technology. You can say whether the explosion happens in a region of star formation, based on the color of the background region of the galaxy. However, as most supernovae are very distant, it is nearly impossible to say whether such a region is an open star cluster or simply a star in a more diffuse region of young stars. Moreover, most Type Ia supernovae occur in systems that are more than 50 million years old, with a peak somewhere between 100 million and 1 billion years. Recall, from Chap. 5 , that most open clusters dissolve into the background field of stars in this time. Thus it becomes highly unlikely that you will be able to pin an aging but primed star system to an open cluster. The lack of such evidence probably explains the dearth of articles on Type Ia supernovae in open clusters in the scientific literature. When one looks for Type II or core-collapse supernovae, you have 218 The Complex Lives of Star Clusters no such problem picking up articles linking them to clusters of young stars. Thus, it is clear that although clusters should pay host to an excess of Type Ia supernovae, pinning them to open clusters is highly improbable; and pinning them to much older globular clusters has come up empty-handed. The take-home message is a model is, at the end of the day, simply a model. GRAPE-6 simulations suggest one thing, but the reality of linking Type Ia supernovae to star clusters has proved to be frustrating at best.

Conclusions

Star clusters provide very interesting natural laboratories, where extreme stellar physics can manifest. Most importantly, such environments produce some of the rarest stellar phenomena from blue stragglers through various X-ray binaries to magnetars. Clusters also provide and important test-bed for physical modeling of star systems. Shara and Hurley’s work with computer modeling throws up many unusual and unexpected phenomena produced by stellar collisions and near-misses. However, not everything that emerges from these calculations has yet been substantiated. Most notably, number crunching had suggested a far higher frequency of Type Ia supernovae in star clusters than has actually been observed. This absence of evidence appears to be down to two factors. One: most star clusters are relatively low in mass and fall apart before many of the interesting stellar encounters can take place. On the other hand, where star clusters are suitably massive to survive for hundreds of millions to billions of years, stellar harassment within the cluster core may actually prevent useful collisions taking place. When those that do happen finally occur, the stars that are left over to merge are too puny to power Type Ia (or other) supernovae. That is not to say such events could never happen, but rather where stellar masses have been reduced to a fraction of that of the Sun, three or more collisions will be needed before the white dwarfs concerned ever acquire enough bulk to catastrophically implode and ignite. This is inherently unlikely, particularly as such low mass white dwarf stars are more likely to be ejected from the cluster through Stellar Soap Operas 219

interactions with neighbors than end up in the core of the cluster where they could meet and merge with neighbors. It is to this topic that the next chapter turns its attention: what happens to globular star clusters as they age? In terms of their resident citizens and the stellar city as a whole, how do stellar collisions and near-misses eventually lead to the downfall of the cluster and its eventual dissolution? 7. The Complex Lives of Globular Clusters

Introduction

When a star cluster forms, all of its constituent stars are moving relative to one another. The interstellar froth from which they formed was in constant motion and this continued as individual cloud cores (Chap. 2 ) continued to collapse into protostars then stars. For clusters with masses in the range of 100,000 to 1 million stars, the speeds are on the order of 5–15 km/s, or a few tens of thousands of miles per hour. Within this melée of moving objects, there will be a high fraction of binary stars that are moving around their common center of gravity at several hundred kilometers per second. All this mass moving around imbues each star with a lot of momentum and kinetic energy which is available to the cluster as a whole. This energy and momentum keeps the cluster alive for billions of years but ultimately helps to destroy it in the end.

Speed, Distance and Crossing Time

A cluster’s constituent stars lead lives that are determined by two principle factors: their mass and whether they are in a close binary, one where the stars are separated by a few tens of stellar diameters at most. In addition, over time stars will be subject to the sorts of shenanigans described in Chap. 6 . All the while, they will be drifting in and out of the cluster core. To understand how a cluster matures and eventually dies, you need to understand a few processes that take hold of it, as well as the sorts of timescales that are involved. The most important timescale is referred to as the crossing time : this affects the rest of the star’s life and is essentially the time it takes a star to

All the stars move with the same speed and direction. They are also moving in the same direction as the cluster: there is no velocity dispersion.

Each star moves with a different speed and direction. But they are also moving in the same direction as the cluster. The difference in speed and direction of the individual stars, within the cluster means that there is a velocity dispersion.

FIG. 7.1 Stars within an association may all move coherently as a group but not move relative to each other (a ). In a star cluster (b ) stars move coherently as a group, but they also move relative to one another. The difference in speed and direction is called a velocity dispersion and can be used to probe the mass and structure of a star cluster cross from one side of the cluster to another, under the influence of gravity. Imagine a pendulum swinging backwards and forwards. A star that begins its life on one side accelerates towards the middle of the cluster then decelerates as it swings out towards the opposite periphery. This holds for stars in open and globular clusters. You can imagine the process as essentially frictionless, so as long as the star isn’t affected by any other on its travels it should simply oscillate back and forth across the cluster indefinitely. Of course, that is not the case. But hold the thought for now. Given the typical speeds that stars have inside clusters, a few basic timescales can be deduced. The trickiest bit is working out the velocity of the stars. This is something called the velocity dispersion (Fig. 7.1 ). Essentially this is the random velocity of each star an astronomer observes in a cluster subtracted from the overall motion of the cluster through the galaxy. Take the cluster’s bulk motion away from the observed motion of every star, relative to Earth, and it is fairly easy to calculate how fast each The Complex Lives of Globular Clusters 223 star is travelling within the cluster. If you need an analogy, just think of an observer at a train station measuring the speed of the passengers on a passing train as it speeds its way past the plat- form. The velocity dispersion of the passengers represents their individual velocities (speed and direction) superimposed upon the bulk motion of the train as it passes. Given the typical speed of a few kilometers per second it will take a star approximately 300,000-400,000 years to cross through the densest part of a cluster and a few million years to traverse the full width of an open cluster. For comparison a star would take 20–30 million years to cross the diameter of a globular cluster. In terms of the likely fate of a journeying star, it is the time the star takes to cross the densest part of the cluster that matters the most, for it is here that collisions are most likely to occur. Astronomers typically divide open and globular clusters into two chunks. Take a typical globular cluster with a mass of 100,000 Suns. This may be 100 light years across. Although that is quite large, half of the mass of the cluster will fit inside a much smaller box, called the cluster core. This may be only 10–20 light years across. The radius of this inner box is called the half-mass radius ; and since there is a fairly straightforward relationship between the light a cluster emits and its mass, this box can also be called the half-light radius . We’ll stick with mass. A dense cluster (Chap. 1 ) with a Sawyer-Hogg classification of I–III has a dense core and the crossing time will be roughly the same as described above: a few hundred thousand years. Less dense clusters, with a higher classification number will have a broader core and the crossing time will run into the millions of years. The distinction is very important and also holds for open clusters. The crossing time sets the period in which stars are most likely to interact with one another. Think of it this way, you carry out most of your daily interactions while young or middle aged. In your later years, the number of social contacts tends to be less— and quite obviously, barring an afterlife, you have no social contacts after death. For a star the most important part of its life is also its longest: its time on the main sequence. For collisions that take place on the main sequence, will make the star larger and more massive and more prone to further interactions. If the transit time (the crossing time) is longer than the main sequence lifetime, 224 The Complex Lives of Star Clusters then the star will only suffer loss. If a red giant or supergiant collides with another star, it loses its envelope and much of its mass. Collisions that occur when it is a main sequence star tend to add their component masses together, with minimal loss. If the crossing time for the cluster core is on the order of a few hundred thousand years, the star will be highly likely to interact and possibly collide with other stars. A direct collision will produce an even more massive star and it is here that truly monstrous stars, such as R136a, or the progenitors of ultra-luminous supernovae SN 2006gy and SN 2006tf may come about. In star clusters, particularly those with many massive, short- lived stars, a second process becomes important after a few million years. Massive stars steadily increase their stellar winds as they age, culminating in their explosive decimation. Most young globular-like clusters in the local universe have a dense core containing tens of thousands of massive stars fringed by a gap about as wide as the cluster and then a flanking shell or halo of other massive stars. Embedded throughout this structure are lower mass stars that are still coalescing from the interstellar flotsam and jetsam. In our galaxy, the super star cluster NGC 3603 has a very dense stellar core with a pronounced “gap” before you reach the flanking wall of the giant molecular cloud. Within this brightly illuminated wall of gas and dust thousands of other stars are likely forming. The word gap is something of a misnomer however. Imagery taken with the infrared imager, ISAAC, on the Very Large Telescope in Chile shows tens of thousands of low and intermediate mass stars in various stages of formation. These stars are largely invisible in the visible-wavelength images taken by Hubble as they are simply outshone by their massive brethren. The nebula that surrounds the central cluster in NGC 3603 forms the largest HII region in the Milky Way. This fluorescing region of gas and dust forms an interesting analogue of R136 in 30 Doradus in the Large Magellanic Cloud. The key difference between these clusters is the lack of a shell of massive stars in NGC 3603 around the central core. This is very evident in 30 Doradus and indicates a fairly complex structure in the molecular cloud from which R136 was formed. R136 is a million or so years older than NGC 3603 and it remains possible that the latter cluster will produce a similar shell of stars from within the surrounding nebula in the next million years or so. The Complex Lives of Globular Clusters 225

At least 2,000 solar masses of star stuff is taken up in the central cluster of O and B class stars in NGC 3603. Many of the local universe’s most massive stars reside here. Of these 62 had been discerned and characterized by Nicholas Melena (Lowell Observatory) in 2008. The precise number of massive stars is only an estimate that is based upon the amount of light the region generates: the stars are too close together and too luminous to separate clearly with current technology. The exact number of massive stars could be in excess of 100, comparable if slightly less than the estimate of 120 O and upper B-class stars for the Arches cluster. While the Arches cluster is hidden behind an obscuring nebula, towards the galactic center, NGC 3603 is clearly visible in one of the Milky Way’s spiral arms. Estimates of the cluster’s age are based on the main sequence turn off position, the point of highest luminosity and hence mass, where stars leave the main sequence to become (in this case) hypergiant stars. The central cluster in NGC3603 is aged at around 1–2 million years of age, based on the presence of stars with masses in excess of 100 solar masses. But spectra reveal the presence of some slightly evolved, but still very massive, Wolf-Rayet stars that have clearly just left the main sequence. These have an age of around 3–4 million years. On the cluster’s outskirts, the massive blue supergiant, Sher 25, appears to be around 4 million years of age. Thus, there appears to be a spread of ages within the star cluster, with Sher 25 being somewhat more mature than the stars in the central cluster. Similarly, the age of the Arches cluster is not well defined, but the fact that some of its residents have already died indicates that it must be at least 4 million years old.

Violent Relaxation

Ian Stevens and Joanna Hartwell (University of Birmingham, UK) used Chandra observations to analyze the X-ray emission from the Arches cluster, NGC 3603, as well as several other, nearby, super star clusters. Spectra indicate the presence of very fierce stellar winds that are shedding a few ten thousandths of a solar mass of gas each year at a whopping 2–3,000 km/s. These winds are travelling at more than 50 times the escape velocity of the cluster and 226 The Complex Lives of Star Clusters are rapidly removing tens of solar masses of gas. This should cause a slow expansion of the clusters over the ensuing millions of years; a process that will culminate in 5–15 million years as each of the currently observed massive stars completes it life and explodes. In disagreement with the evidence from visible radiation, the x-ray data suggests thatthe Arches cluster has in fact got 5-6 times the number of massive stars than is implied in NGC 3603’s central cluster. This places the Arches cluster on a par with R136. It’s just a pity that the cluster is only clearly visible in the infrared and X-ray portions of the spectrum. It would make a very impressive sight, even from the distant vantage of the Earth, were it not obscured by intervening gas and dust. When the massive stars in these clusters die, most of their mass is ejected at several thousand kilometers per second from the entire cluster. This wholesale loss and redistribution of mass rearranges the orbits of the remaining stars. The overall effect is to aggressively erase the earlier underlying structure of the cluster. The process takes a few million years to complete, beginning 3 million years after the cluster formed, and ending roughly 15 million years later, when the last truly massive star explodes. The tail end of the era of massive stars is a little drawn out, thanks to binary systems. In some of these cases one star might acquire enough mass to allow it to explode as a core-collapse supernova (Chaps. 2 and 6 ). Since binary stars are abundant in star clusters, there must be a contribution from these systems for a few tens of millions of years, overlapping the demise of the last single massive star. The wholesale expulsion of much of the cluster’s mass takes 30 million years at most. This may be less time than it will take a star to travel once across the entire width of the cluster. As matter is violently redistributed, the entire cluster vibrates. Stars rapidly change their stores of gravitational potential and kinetic energy as their links to their neighbors are shattered, reformed and shattered once more. There is a wholesale redistribution of stellar orbits around and through the cluster core in the process astronomers call violent relaxation. It’s the stellar equivalent of nuking a city and letting the survivors find their way back to their battered homes: very messy and (relatively) very quick. What is left is a much smoother distribution of stars and protostars. The Complex Lives of Globular Clusters 227

While all this carnage is going on, it’s important to remember that the majority of cluster stars have a mass equal to or less than that of the Sun. These objects may be far from birth while all this chaos is occurring around them. By the time they are born, violent relaxation will have carved its name into the cluster and what is left will be a much calmer and more orderly place.

The Two-Body Relaxation Time

As the last of the newborns emerges, two other processes begin to act upon them. The first of these is called two-body relaxation time. Again, this sounds more complicated than it is. This process describes the time it takes a star to radically alter its trajectory through the cluster (Fig. 7.2 ). Precisely, it is the time it takes a star

A Y

A Z

FIG. 7.2 The two-body relaxation time. This is the time it takes a star to have its direction of motion changed by 90 o (a right angle). This takes many encounters between the star (A ) and others (or binaries) along its path (stars W , X , Y and Z ). The process of two-body relaxation leads to lighter stars being moved towards the outside of the cluster, while more massive stars move towards the cluster core. This reflects the transfer of momentum through gravity between the stars involved. It is easier to make a small star move faster then it is a more massive one, so the transfer always slows the more massive star down causing it to sink to the cluster core, while the lighter star is accelerated outwards 228 The Complex Lives of Star Clusters to change its velocity by 90°. This effectively means the time it takes it to become bound in a binary system, or simply the time to be driven onto a completely different trajectory through the action of gravity. As this process is gravity driven, and in particular driven by the gravity of individual cluster stars, it will happen most quickly to stars in the core of the cluster, where the density of stars is greatest. Typical times for globular clusters are on the order of 100 million years for a star near the cluster core, and a billion years for stars at the half-mass radius. Bump that figure up to 10 billion years for stars at the periphery of a globular cluster. Less dense open clusters will have longer times in general, but this strongly depends on the density and the size of the cluster, which is hard to generalize. As an interesting aside, elliptical galaxies can be thought of as giant star clusters (Chap. 9 ), with a similar structure centered on a vast stellar core. If you’re wondering, the two-body relaxation time for a large elliptical galaxy is on the order of 100,000 trillion years, rather longer than the current age of the universe. This number is important as it determines how long it will take a galaxy to begin falling apart in the same way as a star cluster today. In the final chapter, we will look at this again, but without too much pre-emption, our galaxy will survive in some state for at least 1,000 times longer than the two-body relaxation time. In a typical star cluster the two-body relaxation time sets the clock on how quickly we can expect one star or binary to encounter another. It is the time over which flirtations, interactions and collisions are likely to occur between stars. Within the core of a globular cluster you could expect a collision every hundred million years or so. This is not too different from the timescales identified by Michael Shara and Jarrod Hurley that were described in Chap. 6 . This time ensures that a star like the Sun, located within the core of a star cluster, would meet and interact with other stars at least 100 times from its birth until the end of the main sequence. Clearly, it’s hardly surprising that few red giants survive such torment in the cluster core and end up as stripped-down Extreme Horizontal Branch (EHB) stars (Chaps. 3 and 6 ). A key word associated with the two-body relaxation time is virialization. This term effectively describes the scrambling of the velocities of the stars in a cluster. If you want to think about The Complex Lives of Globular Clusters 229

virialization in terms of energy (and some light math), when a system is virialized the total kinetic energy (K.E) of the stars is equal to half their gravitational potential energy (PE), which has a negative value, so that 2KE + PE = 0. Virialized systems are thus balanced with the potential and kinetic energies of all of the stars distributed so that the total energy in the system in zero. Stars lose kinetic energy as they move outwards from the cluster core, but this loss is balanced by a gain in potential energy and vice versa. Young clusters are not virialized, while most of the Milky Way’s globular clusters are virialized throughout most or all of their volume. When a cluster is virialized it has a smooth appearance, with the density of stars steadily decreasing from the center towards the outskirts. Virialization is a product of two-body interactions. In a very young star cluster the stars are broadly distributed in a fashion that depends on how and where they were formed. Many star clusters, such is evident in R136, are very clumpy with rings, cores and shells of stars of similar but discrete ages. As gravity sets to work on these, the stars plough past one another and experience a mixture of strong (close) and weak (distant) encounters. Although the majority of encounters are weak and distant these are enough to redistribute the velocities of all of the stars in the cluster that are interacting. As a result of exchanges in momentum and kinetic energy the distribution of stars smoothes out. Mark Gieles (European Southern Observatory) and others have analyzed various stars clusters and shown that young clusters are not virialized as they have not had time for two-body interactions to produce it. Meanwhile, globular clusters and the cores of elliptical galaxies are virialized. The outskirts of elliptical galaxies are typically not virialized for two reasons: one they are still often still merging with neighbors which adds fresh layers of structure. Secondly, the stars within these more dispersed regions have not experienced many encounters and are not relaxed: the universe is simply not old enough for this to have happened. Simon Porteqies-Zwart and colleagues show that in very young and very massive clusters, such as R136, the two-body time will be on the order of tens of thousands to hundreds of thousands of years. This is a timescale amenable to the collision of multiple massive stars: it is hardly surprising those collisions are frequent and that monsters such as the 256 solar mass R136a emerge. 230 The Complex Lives of Star Clusters acb

FIG. 7.3 The evolution of stellar orbits over the last ten and a half billion years of globular cluster history. In 47 Tucanae the first generation of stars (blue ) have largely circular orbits that randomly flow around the heart of the cluster (a ). 100–200 million years later a second generation of stars is formed from the debris cast off by the first. These form towards the centre of the cluster and have more elliptical orbits, possibly tracing the original motion of gas towards the cluster core (b ). Over the ensuing billions of years, the orbits of these second generation stars gradually elongate until they are radial, like comets orbiting the Sun (c ). These orbits take them repeatedly into and out of the cluster core where they engage in multiple relationships with other stars and binaries—and when they age, cause them to lose their outer layers

Beyond the cluster itself, other processes set to work on timescales that run into the billions of years. It is these that seal the fate of the cluster as a whole. As we saw in Chap. 4 , most globular clusters, or at least the larger ones, have got more than one population of stars. Although still regarded as controversial (Chap. 4 ) this second (and sometimes third and fourth) generation formed within 200 million years of the first. If we start the clock on these 100 million years after the first generation forms, the cluster will have already undergone violent relaxation and several crossing times will have elapsed. This gives the cluster quite a lot of time for stars to have undergone pairings and rearrangements in their orbits, before the next generation of stars is born. Into the middle of the expanded but already partly evolved cluster, throw a large quantity of gas enriched in helium, sodium and other elements. From here, form anything between 100,000 and 200,000 new stars. What is the likely effect of these new stars? The Complex Lives of Globular Clusters 231

Most early studies neglect these. Harvey Richer (University of British Columbia) explored the motion of stars in the large globular cluster 47 Tucanae. His group used very deep images taken by the Hubble Space Telescope Advanced Camera for Surveys. This allowed the motion of individual stars to be easily resolved. Richer and colleagues mapped out the orbits of a whopping 30,000 different stars in the cluster. The second generation stars had strongly dipping orbits, akin to comets, that took them deep into the core of the cluster before sweeping back outwards towards its fringes. The less metal-rich first generation stars had more random and less inclined orbits that largely avoided the cluster core (Fig . 7.3 ).

Disc Shocking

Star clusters that orbit the core of the Milky Way and Magellanic Clouds will periodically dive through their massive discs. As the globular cluster approaches the disc the front end of the cluster is pulled more strongly towards it than the rear. This is because gravity pulls with a strength that is proportional to the square of the distance between two objects. The front end of the cluster is closer to the mass that is pulling on it than the rear, so experiences more force and hence acceleration. Imagine a globular cluster approaching the disc at high speed (tens to hundreds of kilometers per second). After a few tens of millions of years plunging forwards, the front of the cluster punches through the disc. Now that the disc and cluster are together, the cluster rapidly decelerates while the rear still is accelerating forwards. This tends to squash the cluster in a direction 90° to its direction of motion (Fig. 7.4 ). The cluster still has momentum and continues forward, all the while decelerating. As it leaves the disc and heads back towards the halo, the rear of the cluster is pulled more strongly than the front and the cluster is stretched again, this time in the opposite direction. This stretching and squeezing has much the same effect as stretching and squeezing a tennis ball. The whole cluster heats up. Now, the individual stars don’t get any hotter, rather the orbits of the stars pick up energy which means that they are moving faster. Faster moving stars develop lager orbits inside the cluster, 232 The Complex Lives of Star Clusters

B C

FIG. 7.4 The effect of tidal (disc) shocking and tidal forces on globular clusters. In a the cluster is accelerating towards the galactic disc. Stars at the cluster front accelerate the most. In b the cluster passes through the disc. The cluster decelerates, with stars experiencing a pull towards the mid-plane of the disc. At c the cluster is now moving back into the halo but continues to decelerate. Stars furthest from the disc experience the weakest pull back towards the disc. As the stars alternate acceleration and deceleration the stars pick up and lose energy which causes their orbits to elongate. The cluster expands as a whole ( red arrows ) but also stretches along its orbit which makes the cluster, as a whole, expand. This effect, which causes the puffing up of the cluster, is called disc shocking. Over time this leads to the spaghettification of the cluster a whole as stars move further and further away from the center of the cluster. Open clusters, moving through the disc experience a similar effect (Chap. 5 ). Their orbits carry them around the galactic core at different speeds to other star clusters and giant nebulae, called giant molecular clouds. The side of the cluster nearest the attracting mass accelerates more towards it as they approach one another than the side furthest away. After they pass the closest side experiences the strongest deceleration. The rapid change in gravitational pull then puffs these clusters up. After several hundred million years, small clusters will become scrambled beyond recognition, but most globular clusters, born with millions of stars, will survive for much longer than the current age of the galaxy. The Complex Lives of Globular Clusters 233 Core Collapse

An interesting effect that hits most large star clusters, after several billion years, is core collapse. You might imagine that having hundreds of thousands of slow moving stars, gravity would have a field day, rapidly squashing the swarming mass down to a rather impressive black hole. The fact that it doesn’t is down to the momentum of all of the individual stars. They survive quite happily orbiting their common center of gravity for billions of years. But gravity can play an interesting trick on the cluster. As stars move around within the cluster’s mass, interactions between low and high mass stars tend to fling the low mass stars towards the outside of the cluster. This process is called mass segregation and is visible to differing extents within old globular clusters. A few dense young clusters also show the effect, but with these youngsters, it isn’t clear whether this is a characteristic they were born with, rather than something that developed later. Imagine that each star is a little atom in a cloud. Each star carries kinetic energy with it: the energy in its moving mass. Stars that encounter binary systems tend to steal some of the energy of the binary stars, causing the two binary stars to move closer together. Astronomers refer to this effect as binary hardening. The interloping star now has more energy and can expand its orbit, taking it further from the cluster core. In general, it is lighter stars that tend to pick up speed and change their orbits in this way. When a large number of the low mass stars have been ejected to the cluster’s fringes, much of the cluster’s kinetic energy will have been taken with them. This leaves the core of the cluster low on energy and liable to collapse under the pull of gravity. Now, there isn’t so little energy that the core implodes completely, rather as the stars fall towards one another, interactions either form new binary systems, or interactions between lone stars and binary stars give the core enough energy to halt the collapse. Astronomers like to make an analogy between the process of binary hardening to the burning of fuel in a star. Each time a binary encounters another star and loses energy, the cluster as a whole can use this energy to prevent it collapsing, much a like a star burning hydrogen in its core. If enough energy is given to stars in the core, the core can 234 The Complex Lives of Star Clusters a b

100-1,000 million 3-5 million years old years old

20 million years old

1-10 billion years old e

More than 10 billion years old

FIG. 7.5 The approximate timescale for the various events that take place in globular clusters over tens of billions of years. (a ) The newly formed cluster has lots of structure and abundant gas. (b ) By 20 million years all of the massive stars have exploded as supernovae and expelled much of the mass of the cluster. The cluster responds by relaxing violently… The cluster then expands as mass is lost from it. (c ) The mature cluster at a few hundred million to a billion years begins to undergo core collapse ( small black arrows ) as many low mass stars have been expelled to the cluster fringes through interactions with binary stars in the core. Meanwhile, repeated passages through the disc are causing tidal (or disc) shocking (large blue arrows ). Disc shocking distorts the shape of the cluster and expands it. (d ) Small stars are evaporating from the cluster ( small black arrows ) as more and more of the remaining cluster is involved in core collapse. (e ) Core collapse is reaches its largest extent (10–20 billion years) but escaping stars continue to whittle down its mass (evaporation). In time further waves of collapse will develop as binary systems in the cluster core are destroyed and stars scattered outwards (inner, shaded sphere ) bounce back outwards, though not to its original size. Then, much like a bouncy ball on a string, it will repeatedly collapse, re-expand and collapse over and over again (Fig. 7.5 ). Each time some smaller stars will receive so much energy that they escape the cluster altogether. This cyclical process, where the cluster expands and contracts repeatedly, is given the rather grand title of gravothermal oscillation. The Complex Lives of Globular Clusters 235

Over time, the area affected by core collapse, advances slowly outwards towards the periphery of the cluster. Each time a collapse occurs some binary systems are destroyed—the two stars forced together and their orbital energy carried outwards through the cluster by the ejected star. Each time binaries are destroyed, stars in the core move closer together and more binary systems are produced, much like a star switching fuels from hydrogen, through helium to carbon. With each change in fuel hydrogen fusion moves progressively further out into the envelope of the star. The same is true for core collapse: the production of binary star systems advances outwards, much like a slowly propagating wave. As the wave moves, more and more stars interact with one another and many receive enough energy to leave the cluster altogether. Thus, over time the mass of the cluster is driven downwards by core collapse. The cluster, evaporating all the while, stays afloat against the incessant pull of gravity. An important point to remember, once again, is the presence of two or more populations of stars in these clusters. Core collapse begins within the innermost, densest portion of the cluster, where stars are most tightly bound together. It is the second-generation stars, those that are most concentrated towards the core of the cluster, that are hit hardest and earliest by core collapse. The outermost and most metal poor stars may remain untouched by its effects even up until a good 10–12 billion years after the globular cluster formed. John Fregeau (MIT) and colleagues carried out simulations of clusters of stars and included the effects of a small number of primordial binary stars: those formed with the cluster, rather than those that formed later in the cluster’s life. They found that in many (but not all) cases core collapse was prevented altogether and clusters simply evaporated into the halo of their galaxies, long before core collapse could occur. The outcome of this model ties with observations that many globular clusters show little evidence for core collapse having occurred, even after nearly 13 billion years. Although core collapse is evident within the Milky Way’s largest clusters, many smaller ones won’t experience it at all until many billions of years hence, because the density of stars in their hearts is too low. In these, there is more than enough energy in the individual stars to stave off the attraction of gravity. In many of 236 The Complex Lives of Star Clusters these the pull of the rest of the Milky Way will simply shred the cluster, long before gravity has any chance of imploding the core of the cluster. Other models suggest that star clusters with collapsed cores are close to death. These have achieved the maximum number of binary systems and will simply cycle between binary collapse, as their stars are forced together, and the formation of new binaries (Fig. 7.5e ). It appears that at present around 80 % of the Milky Way’s globular clusters are in the throes of core collapse, while 20 % have collapsed already and are currently consuming binary systems (so-called “binary burning”). Observations by Piet Hut, John Fergeau and others show, as expected, that those clusters which have collapsed cores have an over-abundance of binary systems compared to those clusters that are not yet collapsed or are in the midst of collapsing. In the end, many models show that all of the lightest stars are ejected, leaving a cluster filled with red giants, neon-oxygen or carbon-oxygen rich white dwarfs and the products of collisions between these stars. With no main sequence stars remaining the cluster is doomed to a quick decline. An interesting outcome from many simulations suggests that when a neutron star encounters a LMBX system, the outcome is a fast moving pair of neutron stars, with the less massive donor star abandoned. The incoming neutron star has so much momentum that the resulting binary has enough energy to leave the globular cluster altogether. The two dead stars fly off into the halo, leaving the former donor star to live in peace. These eloping pairs of neutron stars face an unpleasant future, as their orbits collapse through the emission of gravitational waves. These, in turn cause the binary neutron stars to spiral ever closer to one another. A few hundred million years later the inevitable outcome will be the formation of a black hole, accompanied by a short gamma ray burst. In this regard it is of interest that many of these explosions have been shown to occur in the halos of galaxies, exactly where you might expect these runaways to expire. Similar marriages between massive white dwarfs and neutron stars might end the same way, or death may come in the form of a calcium-rich supernova (Chap. 6 ), if the white dwarf is of the right composition. The Complex Lives of Globular Clusters 237

How certain can astronomers be that these processes take place over the timescales that are envisaged? Obviously you cannot observe these processes happening in an individual cluster. However, you can take evolutionary snapshots if you know where to look. One such place is the Large Magellanic Cloud. Not only is this conveniently close, but it seems to have churned out globular clusters (or at least very globular-like clusters) over the last 12 billion years. We can take R136 as the youngest, but a variety of others chart intervals extending back several billion years. One thing that is evident is that the size of the collapsed core changes with time. Young clusters have cores that haven’t yet collapsed, but that the size of the collapsed region increases with time. This should be expected because collapse happens first in the densest inner regions where stars will encounter one another most frequently. Over time, binary stars within the core are forced together, while new binary systems form. But given several billions or tens of billions of years (longer than the current age of the Universe) all globular clusters should develop a dense core of stars. Many rich open clusters may do the same. Clusters that have undergone core collapse tend to have the most interesting populations. After all, by forcing stars together many interesting and unlikely partnerships will evolve. As we saw in Chap. 6 , globular clusters contain the highest proportion of millisecond pulsars, X-ray binaries and cataclysmic variable stars, compared to the size of their overall population of stars. Bring enough stars close together and interesting things begin to happen.

The Large Magellanic Cloud: Snapshots of Creation

The Large and Small Magellanic Clouds are a fantastic and conveniently close natural laboratory for looking at the evolution of globular clusters. For reasons that are unclear, but probably have to do with the interaction between the Milky way and both satellite galaxies, these two small galaxies have churned out dense star systems for billions of years. As a result, as Table 7.1 shows there are globular-like star clusters spanning the entire age of the galaxy. 238 The Complex Lives of Star Clusters

Table 7.1 Properties of various massive star clusters in the Large and Small Magellanic Clouds Mass (multiples Estimated Age of a solar Current apparent (millions mass) apparent magnitude Cluster Location of years) (⊙) magnitude at 15 Gya R136 LMC 2–4 450,000 −10.9 −7.2 (NGC 2070) NGC 330 SMC 10 40,000 −9.9 −6.2 NGC 1850 LMC 50 610,000 −10.5 −7.5 NGC 1866 LMC 100 110,000 −9.5 −7.0 NGC 1783 LMC 1,600 310,000 −8.4 −6.9 NGC 1978 LMC 2,000 310,000 −8.5 −7.2 NGC 1835 LMC 12,000 1,100,000 −8.7 −8.6 The ﬁ nal two columns compare the current luminosity of the cluster (column 5) with an estimate of what the clusters would each have looked like if they had the same age of 15 billion years. This adjusts for the effects of stellar evolution and associated decrepitude and allows valid comparisons to be made between the globular cluster NGC 1835 and the others

The LMC and SMC are still relatively gas-rich with up to 25 % of their mass still found as gas and dust. The Milky Way, by comparison has used up over 90 % of its available star-forming material to produce its current population of stars. Thus there is a good supply of fuel available for star formation. The clusters within the LMC and SMC are found in a variety of locations from the predominantly old stellar bar of the LMC, through to the bridge that extends between the interacting pair of dwarf irregular galaxies. The youngest LMC clusters are predominantly found within the relatively flat disc of neutral hydrogen gas in which the rest of the LMC is located. Within this region, there are very obvious foci of star formation in which the youngest clusters are located. Meanwhile the old clusters, those with ages approximating 12 billion year old, are fairly evenly distributed within and around the central bar of the LMC. Of the 15 that are know, seven of them, such as NGC 1835, 1898 and 2005, are located within or are projected onto the bar, while eight others, including Reticulum, NGC 1466 and Hodge 11, occupy the space around it. The SMC has only one cluster of comparable age to this LMC old guard: NGC 121. The Complex Lives of Globular Clusters 239

Furthermore, the LMC has only one identified cluster with an age between 3 and 12 billion years. With an estimated age of 9 billion years, ESO 121-SC03 is a distinct loner in the LMC. Although ESO 121-SC03 (and perhaps a number of other recently discovered rich stars clusters) now partly bridges the age divide, there is clearly a two-way split with clusters that are either very old or relatively young. The SMC, by contrast, has clusters with a more evenly distributed spread of ages. The precise reason for this dichotomy is unclear, but it may relate to the way in which the LMC and SMC have interacted with one another, either stimulat- ing or (in the case of the LMC) largely suppressing star formation for billions of years. The differences in the ways in which the LMC and SMC have formed stars are also evident in the chemical make-up of them. The younger population of LMC clusters (those with ages less than 3 billion years) are all fairly rich in heavy elements. Understandably, the 12 billion year old clusters are all metal-poor. There are no clusters with an abundance of heavy elements between the two extremes. By contrast, those clusters in the SMC show a broad range of metal contents, which sprawl from very low abundances to values similar to the modern day population of young stars. Working with others, Gerald Kron (Lick Observatory) had determined the spatial orientation of the youngest clusters in the 1950s and 1960s and these measurements suggested that most were associated with a spiral-like distribution around the central bar—but with fewer in the bar of the SMC, itself. The globular clusters also tended to avoid the bar, more so than the open clusters. Given that the bar of the SMC holds the majority of the galaxy’s oldest stars, this is perhaps surprising. This might have indicated that the globular clusters formed before the bar of the SMC. However, more recent work shows that while the high-mass SMC clusters show no preferred location within the small galaxy as a whole, there is an obvious distribution of the youngest clusters around a broadly three-dimensional spherical structure called a super-shell. Katharina Glatt (University of Basel) and co-workers analyzed this patterning and found that it was also true of many of the clusters within the LMC. Super-shells form around giant clusters that are rich in O and B-class stars. Strong winds and shockwaves from supernovae push 240 The Complex Lives of Star Clusters vast quantities of hydrogen and helium away from the core of the cluster, forming a large void, fringed by a denser shell of gas and dust. The enhanced density of gas within the shell then triggers further waves of star formation which have formed the distribution seen today. Within the SMC, two waves of enhanced star formation occurred at 160 and 630 million years ago; while in the LMC corresponding waves occurred at 125 and 800 million years ago. The first of these may have been initiated by a close encounter between the LMC and SMC. We then have a scenario where the close encounter between both dwarf galaxies compresses gas and dust causing a core of massive stars to form. As they raced through their brief lives towards the inevitable violent death, vast shells of gas and dust were launched out of the plane of both galaxies. As these swept up material, further rounds of star formation were set in motion. The shell-like distribution of clusters forms the astronomical fossil that astronomers are interpreting today. Further analysis by Andrés Piatti (Observatoria Astronómico, Argentina) and colleagues of several SMC clusters showed that within the broad range of ages, there was evidence for two further distinct bursts of cluster formation at 3 and 6 billion years ago. Thus, like the LMC, the SMC has had waves of star formation, albeit less focused or as dramatic as those in the LMC.

Ring Clusters

Studies of star clusters within the LMC and SMC reveal a small, yet significant population of clusters unlike any seen in the Milky Way: so-called ring clusters. In all 78 such clusters were found by Felicia Werchan and Dennis Zaritsky (University of Arizona) in analysis of clusters in LMC and SMC. This represented 7 % of the population of all clusters indentified in their work. The origin of these clusters remained unclear, but may relate to the very evident halo or shells of stars seen around central clusters such as R136. Perhaps, these rings of stars are the aged remnants of older, more centrally-concentrated clusters of stars, where the central cluster of massive stars have all expired, leaving the ring of lesser stars around it. Werchman and Zaritsky found no evidence in a bias in their distribution within the LMC and SMC, other than they seem to be unique to these two dwarf galaxies (Fig. 7.6 ). The Complex Lives of Globular Clusters 241

Central cluster of massive stars Within the dense shell a new A bright ring of young stars is left generates strong winds that blow population of stars forms, while where the former ring of dense gas outwards into the surrounding the central cluster of massive lay, while the central cluster is reduced dense gas.This generates a shell stars self-destructs in multiple to a population of lowluminosity and around the cluster. supernovae. low mass stars which are largely invisible (not shown).

FIG. 7.6 A scenario for the formation of ring clusters? In one scenario a cluster of massive stars forms within a dense nebula, much like 30 Doradus. Radiation pressure, stellar winds and, eventually, supernovae generate a dense shell around the cluster in which further intense star formation occurs. This is seen around the central star cluster in 30 Dora- dus. When the massive stars of the central cluster die, the outer ring or halo is left with the highest concentration of bright stars. Although this may contain many low mass stars the concentration of them is highest here making the ring appear brighter than the ring’s centre

Hodge 11

One of the oldest clusters in the LMC is Hodge 11. With an estimated age of 11.7 billion years it should look rather red, having lost most of its hot, massive and blue stars early on in its history. However, Hodge 11 retains a very youthful, blue aspect. How has it achieved this? In some regards Hodge 11 has given itself a bit of a facelift. It’s not that it has spawned a fresh, younger cohort of stars. Rather its stars have interacted with one another to produce a very blue population. Once again, it is the force of gravity that has driven this change. The cluster core has collapsed and many rounds of stellar collision and interaction have occurred. Within the core, interactions between stars have been so frequent that the entire population of mature red giants has been eliminated. A red giant is a vulnerable thing. Large but low on mass, its gravitational hold on its envelope is slight. Even without teasing, about one fifth of a red giant’s envelope will be lost as it expands. Throw the possibility of close encounters into the mix and the situation for the red giant is much 242 The Complex Lives of Star Clusters worse (Chap. 6 ). With each close encounter, some of the red giant’s vulnerable envelope is either teased away by a close passage of another star or bulldozed into the vacuum of space by a direct hit. In their place sub-dwarf B (sdB) stars fill the niche the giants would have occupied. Most occupy the hot, blue extreme horizontal branch with some so severely depleted in hydrogen that they will never build a massive enough core to ignite their store of helium. The horizontal branch of Hodge 11 is so hot that all of the stars lie to the left, blue side of the instability strip. Consequently, there are no RR Lyrae stars, making Hodge 11 something of an oddity. Mixed in with these stars are other blue main sequence stars: blue stragglers. These also owe their existence to the power of stellar collisions (Chap. 6 ). Most are F and A-class stars (yellow-white to white); but the distinct lack of bright yellow or red stars ensures the cluster appears blue to an outside observer. What remains of the lowest mass stars is so dismal that it contributes very little light to the cluster as a whole. One can be pretty certain that the cluster formed with a very large population of stars with masses less than half the mass of the Sun. Yet, most of these are now gone. As Chap. 6 illustrated, it tends to be the smallest stars that suffer most the slings and arrows of outrageous fortune. These diminutive, and formerly very abundant, cluster members have long since been banished to the outside universe, driven off in untold gravitational encounters with neighboring, more massive stars. The best these stars can hope for is to top and tail the cluster as it swoops around the center of the Large Magellanic Cloud.

M33: A Brief Tale of Two Clusters

M33 is a small spiral galaxy situated 2.6 million light years from the Milky Way. It has a mass roughly equal to both Magellanic Clouds combined, or one tenth that of the Milky Way. The small galaxy orbits M31 along an elongated path that takes it periodically within 100,000 light years of Andromeda. Like the Magellanic Clouds, M33 is experiencing prolific star formation and it plays host to an area of star formation that rivals the Tarantula Nebula in the LMC. NGC 604 is roughly 1,500 light years across and over 6,000 times more luminous that the Orion The Complex Lives of Globular Clusters 243 nebula. Within its heart lies a very young stellar association, with more than 100,000 members; of which more than 200 are class O or WR (Wolf-Rayet stars). This 3–4 million year old stellar menagerie clearly parallels 30 Doradus in terms of mass and age (Chap. 5 ). However, its stars are far less centrally concentrated and more closely resemble an OB association (Chap. 1 ) than R136 in 30 Doradus. Quite why there is such a strong distinction is unclear but it clearly illustrates that collapsing a ball of gas can produce a very different outcome each time. Regardless of the status of NGC 604, M33 does contain a variety of star clusters of different ages. Rupali Chandra and colleagues, from John Hopkins University, identified a relatively blue globular cluster within M33 they tagged as M33-C38. Based on its color, the cluster appeared to be 5–8 billion years younger than the remaining population identified in the galaxy, with a mass of 50,000–90,000 times that of the Sun. Although relatively lightweight by the standards of most Milky Way globular clusters, it is compatible with the Palomar clusters. The authors ruled out the possibility that the blue color was due to the presence of horizontal branch stars.” and the next sentence change to “None were identifiable in the Mount Hopkins MMT spectra that would imbibe the cluster with an unduly blue hue. Quite where the cluster lay in the galaxy was unclear, but its mass would imply that it should have been tidally disrupted given the time it has persisted in M33. Thus it appears unlikely that it lies within the galactic disc, though it is not ruled out. More information is clearly needed, but it will be interesting to see if M33-C38 is not alone in M33. As such M33-C38 joins a small but growing list of clusters within the Local Group that play host to stars more youthful than those typically found in globular clusters.

Core Collapse in M33

M33 is a little bit larger than a typical dwarf galaxy, but unlike the vast majority of other sizable galaxies, M33 does not appear to have a super-massive black hole in its nucleus. Instead, this small spiral galaxy, lying 3 million light years away, has a core of relatively young stars. Lacking a super-massive black hole to organize 244 The Complex Lives of Star Clusters them, the stars appear to have followed the same organizational rules as considerably smaller globular clusters. Lars Hernquist (Lick Observatory), Piet Hut (Institute for Advances Study at Princeton) and John Kormendy (University of Hawaii) examined the motion of the stars making up the core of the galaxy. They found that, much like the cores of globulars, the stars were relatively slow moving with a low overall dispersion in their velocities. The nucleus of the galaxy appears to have undergone core-collapse over a time-scale of a few tens of millions of years. One of the outcomes of this relatively recent collapse is the presence of large numbers of X-ray sources: the core of the galaxy is dominated by diffuse X-ray emission that appears to come from a relatively large population of low mass X-ray binaries. Unlike the core of the Milky Way and Andromeda, which are populated by primarily young massive stars (and a super-massive black hole), M33 has a core populated with X-ray sources in an abundance comparable with a typical globular cluster. The core also has a very blue color, which in part may be due to large numbers of blue stragglers. However, the core’s youth also suggests that a fairly large number of intermediate mass stars are present. Why is the nucleus of M33 so different to M31 and that of the Milky Way? The underlying reason is undoubtedly the mass of the galaxy—which is only a bit larger than the combined mass of the Magellanic Clouds. The growth of M33 appears to have partly aborted and during its formative stages there was insufficient mass to form a broad central bulge like that seen in its two larger neighbors. Rather than a very large population of stars forming promptly early in the galaxy’s history, a slower drizzle of fuel has driven a more recent burst of star formation that has formed the low density core. Much like an enlarged globular cluster, the stars within this core have been too dispersed to collide and form a black hole when they were young. In simple terms the crossing time was so low that few collisions occurred and no very massive stars were able to form. Instead, the stars have formed multiple pairings of stars that have held the core up against a more profound implosion. Over the ensuing hundreds of millions to billions of years these binary systems will drive waves of expansion and contraction The Complex Lives of Globular Clusters 245 within the core as they are formed, destroyed and reassembled. The core of M33 thus resembles a globular, but despite this, the position deep within M33 will drive a longer term evolution that avoids evaporation. A constant supply of stars and a prolonged supply of star-forming material will allow the core to grow larger even as it staves off complete collapse. Over the ensuing billions of years, its fate will undoubtedly lie with that of Andromeda (M31) and the Milky Way to which it is gravitationally ensnared (Chap. 9 ).

The Young Clusters of M82: MGG 9 and 11, Density and Fate

A star cluster’s future depends not only on its environment but on the nature of its birth. To make one further anthropomorphic analogy, both the nature and the nurture of the child determines its eventual fate. What might now be a classic tale of how the quality of the cluster itself determines its fate was relayed some years ago by Simon Portegies-Zwart (University of Amsterdam). Much of this was mentioned earlier in Chap. 6 in relation to the formation of super-massive stars through successive collisions between massive stars. The important point to consider in this context is how two clusters, each with very similar masses, could have turned out so profoundly differently. In 2003 the University of California’s Nate McCrady and colleagues published the analysis of the light from two star clusters in the starburst galaxy M82: MGG 9 and MGG 11. The light of both is dominated by red supergiant stars and both contain a large retinue of very massive stars, typical of a dense cluster that is less than 10 million years old. Recall that you can determine the approximate mass and structure of a cluster by examining the cluster’s light profile and its so-called velocity dispersion. To recap, in early pre-Hubble images star clusters appeared as largely featureless blobs of light, surrounded by an increasingly diffuse halo of individual stars. In the 1960s Ivan King developed numerous ways of interpreting the distribution of stars, the cluster mass and density of stars in 246 The Complex Lives of Star Clusters globular clusters. These King models are routinely used today for clusters that lie far from the Milky Way and where individual stars are nigh on impossible to discern directly. King’s models use both the distribution and intensity of light to map the mass of the cluster: something that is relatively straightforward as one solar mass has roughly one Sun’s worth of light if you assume that most of the mass is in visible stars. Velocity dispersion can be determined by the spread in the width of some absorption or emission lines in a spectrum caused by the movement of the individual stars within the overall blob of light being examined. This clever technique can be used to determine the motion of individual stars within a cluster, even when these stars cannot be discerned from the overall ball of luminosity. In the case of MGG-9, the profile suggested that the cluster had a velocity dispersion of 15.9 km/s giving it a mass of approximately 1.5 million solar masses. The half-mass radius (the radius in which half the mass of the cluster sits) was 2.6 light years. MGG-11 had a velocity dispersion of 11.4 km/s and a half-mass radius of 1.5 light years and a mass of 350,000 Suns. Thus MGG-11 has a little less than one quarter the mass of MGG 9. You might then, quite reasonably, assume that MGG-9 might have the more interesting fireworks, as it clearly has more stars to play around with. However, what matters here is the half-mass radius. That of MGG-11 is smaller and that means it takes several hundred thousand years less for a typical star to cross this distance under the influence of gravity. That is a small, yet significant detail. As we saw earlier, a small half-mass radius and a correspondingly small crossing time will allow a far greater opportunity for stars to meet one another. In the case of MGG-11 sufficient opportunity was then afforded its stars to violently collide within 4 million years. This time limit is important: it marks the stage at which massive stars will leave the main sequence. If stars can collide in this interval after their birth, they can circumvent the effects of their powerful stellar winds and grow to quite preposter- ous dimensions. Portegies-Zwart’s models show that the process of stellar growth is slowed by strong stellar winds, which are constantly bat- tling against the gain in mass. However, if the star cluster is sufficiently dense, this cannot compete with the effects of mergers. The Complex Lives of Globular Clusters 247

In the case of MGG 9, the cluster, although much more massive than MGG-11, it is too broad and the crossing time to long. Stars can merge, albeit at a slower rate, but it isn’t enough to keep up with the effects of the result star’s strong stellar winds. In MGG- 11, the merger rate outstrips the effects of stellar winds and the stars grow rapidly in mass. The resulting blue straggler has an extended lifespan, perhaps 5 million years instead of 3–4, but in the end its fate is sealed as its core encounters pair instability. Unlike the cases we discussed in previous chapters, this monster cannot stave off the effects of gravity. When pair instability strikes the core immediately implodes to form a black hole, a rather large one with around 800 times the mass of the Sun. It’s unclear what remains of the outer parts of these stars. If the object is spinning rapidly the black hole might launch powerful beams of radiation from its poles and the star’s demise may appear as a rather impres- sively long gamma ray burst (Chap. 2 ). This may or may not be accompanied by a visible supernova. Alternatively, much of the star may simply flow down the plug hole and the star may die quietly in the dark (Fig. 7.7 ). Dheeraj R. Pasham (University of Maryland) and colleagues analysed the X-ray emission from M82-X1 and concluded that the simplest explanation for the variable output of the source was a black hole with a mass roughly 400 times that of the Sun. Although lighter (by half) the estimation of Portegies-Zwart’s models, it easily fits the bill of an intermediate mass black hole (IMBH). The lighter mass may be due to a greater effect of stellar winds on the merger product, or perhaps to a difference between the predicted and actual density of stars in MGG 11. But that’s not all. If you look at the figure below, alongside MGG-11’s IMBH there is another highly variable X-ray source called X42.3+59. It has been suggested that this source also har- bors an IMBH with a mass measured in the tens of thousands of solar masses. X-ray output from the vicinity of this IMBH appears far more variable than that of MGG-11, making it easily visible in some Chandra images, but not others. Perhaps it has a supergiant star moving around it in a very eccentric orbit. Only at certain times are the two objects close enough for the black hole to capture some of the supergiant’s material and then emit X-rays. Alternatively the burst of X-rays was associated with the shredding 248 The Complex Lives of Star Clusters

MGG 11; host of X42.3+59 M82-X1

MGG 9

FIG. 7.7 The location of the clusters MGG-9 and MGG-11 at the heart of M82. The position of MGG 11 is coincident with a variable X-ray source, ULX M82-X1, while the larger, but more diffuse cluster, MGG 9, has no equivalent X-ray source. The next brightest object in the image, X42.3+59, may also be an intermediate mass black hole. For perspective, the disc of M82 lies roughly perpendicular to the SW-NE aligned broad red X-ray emitting region on the larger image. The green cross marks the dynamic heart of M82. Original X-ray image credit: NASA/CXC/SAO of a star that happened to wander too close to the hole. However, the presence of two ultra-luminous X-ray sources close together in the starburst implies that given the right circumstances it is fairly easy to make a population of sizable black holes that could later merge to form the sorts of super-massive beasts found at the cores of all large galaxies today. These observations suggest that the key to forming these sorts of super-massive black holes is the formation of multiple massive star clusters. Most importantly, these clusters have to have the right density of stars to ensure that enough violent collisions happen between the most massive members of the cluster before they have time to die. A short crossing time measured in hundreds of thousands of years will allow intermediate mass black holes to form. The Complex Lives of Globular Clusters 249

Early in the history of the universe, this would have been common when the density and overall availability of gas was high. In today’s more rarefied environment, with most gas now tied up in stars, the only places that can recreate (or at least approach) the conditions in the early universe are starbursts. Here, a galaxy can recreate a population of intermediate mass black holes that will work their way towards the core of the galaxy through two-body interactions. Imagine a starburst: multiple large black holes form within a sub-set of the massive star clusters formed in the burst. Belying their presence through a curious scattering of stars that get in their way, the black holes eventually emerge from their clusters and migrate under their mutual gravity towards the center of the galaxy. Along their path the odd hapless star falls victim to its gravity and a titanic burst of X-rays betrays the brief and messy carnage that results. Once the meal is consumed, the slightly enlarged black hole continues inwards. Fully 99 % of the stars the hole encounters are scattered outwards through two-body interactions. The star picks up energy and may be ejected from the galaxy altogether, while the black hole loses some of its orbital energy and falls inwards a little bit more. M82 is not alone in hosting some rather interesting stellar clusters. In 2011 Amy Reines reported that the nearby dwarf irregular galaxy, Henize 2-10, also appeared to host an actively accreting black hole. The analogy with M82 was clear. Henize 2-10 is also undergoing a very active starburst and the galaxy contains many massive clusters of stars. One such cluster, positioned near the galactic center, hosts an object that produces an unusual combination of hard X-ray and strong radio wave emission. The radio luminosity is 7.4 × 10 28 Watts at 4.9 GHz and the X-ray luminosity is 2.7 × 1032 Watts in the 2–10 keV soft X-ray region. Although this is not a stunning luminosity, it is the combination of radio and X-ray emission that is suggestive. Not many objects can produce both these radiations at once on this scale: a black hole is the most likely culprit. Given these values the mass of the black hole can be calculated. Once again, Arthur Eddington provides the clue. There is a maximum rate at which an object can emit radiation: the Eddington Limit (Chap. 6 ). Above this value and the radiation will begin to tear the object apart. Thus by looking at the amount of 250 The Complex Lives of Star Clusters radiation coming from the source, astronomers can calculate how massive it must be to hold itself together. In the case of an accreting black hole, the radiation must be coming from the disc around the hole, as the latter has no surface to emit radiation from. The disc is held together by the gravitational pull of the central black hole. Plugging the numbers in to their calculations, astronomers derive a mass of 1 million times that of the Sun for the central black hole. This is a rather impressive black hole for such a small galaxy. It might suggest that many clusters within Henize 2-10 have formed intermediate mass black holes and that these have merged to produce the super-massive black hole the dwarf galaxy now hosts. Henize 2-10 may be a dwarf galaxy, but it has a central black hole to rival that of the Milky Way.

Giant Elliptical Galaxies

In the centers of many large cluster of galaxies lie one or more giant elliptical galaxies, called cD galaxies. These play host to vast congregations of stars often measured in the trillions. Within these peta-cities thousands of globular clusters exist (Fig. 7.8 ). M87 is the central galaxy of the Virgo cluster and at a distance of 50 million light years is sufficiently close that its globular clusters can be resolved as fuzzy points of light. M87 hosts an astounding 14,000 globular clusters which were divided up nearly two decades ago into two, broad groups based on their chemical composition. Rebecca Elson and Basilio Santiago (University of Cambridge) showed that the brightest clusters were also the bluest and contained the least metals. These were also the rarest, with the majority of clusters appearing redder in Hubble images. The red clusters have no equivalent in the halo of our galaxy and appear to have been formed in the relatively recent past during merger events. The blue clusters, by contrast, are analogous to those orbiting the Milky Way and are almost certainly the inhabitants of M87 or the galaxies with which it has merged. The outer population is mostly of the blue, metal-poor family, while the inner population of clusters is skewed towards the metal-rich, redder variety. The more metal-rich inner cloud of clusters are clearly younger than the outer, metal-poor cloud and their origin may The Complex Lives of Globular Clusters 251

FIG. 7.8 Hubble image of M87. Most of the galaxy appears s a largely featureless ball of yellowish stars . Two features are apparent. One is the pale blue jet projected at the 2 o’clock position from the core of the galaxy. This beam of high energy particles is powered by material falling into the sss black hole at the galaxy’s heart. The other remarkable feature is the thousands of star-like points of light that are scattered around the galaxy nucleus. These are many of the galaxy’s 14,000 globular clusters that orbit its core. Credit: NASA, ESA well lie with the violent history of M87, rather than the formation of the original elliptical galaxy. On top of these general findings, work by Aaron J. Romanowsky (University of California, Santa Cruz) and others indicates that M87 is likely still accreting globular clusters from neighboring galaxies that wander too close to its enormous mass. Largely obscured within the bright yet diffuse halo of the galaxy are streams of stars and globular clusters. At a distance of approximately 45 kiloparsecs (a little over 140,000 light years) there was evidence for a shell of globular clusters that was consistent with these having been acquired from a now disrupted gas poor elliptical galaxy. This chevron-like structure appears to contain anywhere between 500 and 1,100 globular clusters, several times the total globular cluster population of the Milky Way. Further out at over 137 kiloparsecs there was a smaller, colder stream of stars and approximately 15 globular clusters that had been identified previously (“Stream A”). Between the two structures (at least from our vantage point) there are a variety of 252 The Complex Lives of Star Clusters

gas-poor, low mass elliptical galaxies (NGC 4476, NGC 4478, NGC 4486A, NGC 4486B and IC 3443). Given their apparent proximity to the giant M87, it seems likely that at least some of these galaxies might well be the stripped remnants of larger, gas-rich galaxies that have wandered too close to M87. One of these galaxies, NGC 4486B has a disproportionately large super massive black hole. Several years ago it was shown that there is a clear relationship between the size of a galactic bulge (that is effectively equivalent to an elliptical galaxy) and the central, super-massive black hole. There is a very clear bulge mass to black hole mass that shows that the central, super massive black hole has roughly one thousandth the mass of the galactic bulge (or elliptical galaxy as a whole). NGC 486B’s black hole is simply too massive for its galaxy, implying that much of the galaxy has been stripped away. NGC 4476, 4478, 4486A and B also appear to be free of any globular clusters which is highly unusual. They should carry a few hundred each. It would appear that on one or more passages around M87, the giant central galaxy has stripped away many of the stars (perhaps 90 %) and all of the globular clusters of these, now denuded, elliptical galaxies. Similar results were obtained a few years earlier, by Eric Peng (Peking University, Beijing), confirming that M87 is a thief par excellence; a galaxy optimized for stealing globular clusters from any neighbor foolish enough to wander too close to its domain. One of the most interesting globular clusters in the local universe which is associated with M87 is HVGC-1: a hypervelocity globular cluster. Surrounding many galaxies are a population of fleeing low and intermediate mass stars, known as hypervelocity stars. These have attained velocities of several hundred kilometers per second, comparable with orbital velocities. The implication is that these stars were once paired with another, now extinct massive star. When this blew up, the partner was released from its gravitational clutches and driven off into intergalactic space. However, a high velocity globular cluster, weighing in at several hundred thousand times the mass of the Sun, is another matter. The only way to drive this off at a velocity determined at 2,100– 2,300 km/s is to have it messily encounter the central, super- massive black hole of M87. With a mass of around three billion times the mass of the Sun the progenitor of HVGC-1 must have The Complex Lives of Globular Clusters 253 had a terrifying journey. Perhaps through some close encounter with a passing galaxy, HVGC-1 was sent plunging towards the core of M87. Undoubtedly, scattering both neighboring stars and many of its own, after several tens of millions of years, it entered the core of M87 on a white-knuckle ride. Accelerating all the while, it soon was moving at over 3,000 km/s. Clearly within the sights of the super massive black hole, it ploughed onwards until tidal forces between itself and the black hole began stripping off the cluster’s outer layers. With stars streaming off its outer halo, and the remaining cluster gaining momentum all the while, the stripped remnant was propelled back out towards the halo of the galaxy. Looking more like a comet, HVGC-1 was now set on a dramatic escape route. Having shed stars to the central super massive black hole, the cluster would now have gained a considerable amount of momentum at the expense of those lost stars. The central super massive black hole appears to lie slightly off-center within the galaxy, something that is surprising considering its massive bulk. The most likely scenario is that a fairly massive companion galaxy has fallen into M87 in the relatively recent past (last 1 billion years) and distorted the shape of the galaxy around its dark, central hole. This would also explain the distorted distribution of globular clusters around the inner region of M87. An in-falling galaxy will scatter stars and change the overall distribution of globular clusters as it plows in towards its eventual fate. Perhaps HVGC-1 was sent on its journey by this merger event? HVGC-1 has gained so much momentum that it is not only able to escape M87’s clutches, but will eventually divorce the massive Virgo cluster of galaxies in its entirety. At some distant point in time, perhaps 10 billion years from now, the stellar remains of HVGC-1 will dissolve into the dark emptiness of a nearby intergalactic void. A diminishing stream of aging stars will mark its passage into the universe’s history.

Adrift in a Sea of Galaxies

Several years ago, the Hubble Space Telescope turned its attention to the Abel 1689 galaxy cluster. Abel 1689 lies a little over 2 billion light years away and is dominated by a large collection of aged, 254 The Complex Lives of Star Clusters massive, yellow, elliptical galaxies. Perhaps best known for the gravitationally lensed arcs of light (from even more distant galaxies) that surround it, John Blakeslee (NRC Herzberg Astrophysics Program, Dominion Astrophysical Observatory), and Karla Alamo- Martinez (National Autonomous University of Mexico) detected 10,000 faint, point like sources of light buried alongside the diffuse glow of the surrounding elliptical galaxies. From their spectra they concluded that each of these was a globular cluster that was cast adrift within the Abel 1689 cluster. Extrapolating the observed population to the total volume of the cluster as a whole they calculated that there must be over 160,000 globular clusters drifting through the void between the galaxies. Close examination of their distribution reveals that although many tens of thousands might well be associated with the halos of the overlapping galaxies, at least 80,000 appear to be truly orphaned. This would represent the largest population of globular clusters known in the observable universe, and not one of them tied to a particular parental galaxy. The total volume examined was a little over 1.3 million light years across (400 kiloparsecs) spanning part of the central portion of the cluster (Fig. 7.9 ). Interestingly, the distribution of globular clusters matched the distribution of dark matter within the cluster. Dark matter, though by definition not visible to the eye, can be mapped through its effects on the light of neighboring or more distant galaxies. All matter bends the space in which it sits, much like a water melon deposited on a thin rubber sheet. Light is forced to follow the con- tours of the underlying space, which affects the route it takes from its source to the observer. By observing the gross and subtle distor- tions in the shape and brightness of distant galaxies, astronomers have mapped out the distribution of dark matter within the cluster. What is, perhaps, surprising, the dark matter maps closely to the globular clusters, far more so than the surrounding galaxies. How did so many globular clusters end up orphaned within the Abel 1689 cluster, tagged to the cluster’s dark matter? There are two possible routes. In the first, the bulk of the globular clusters formed with the galaxies, then were simply stripped off their hosts when the galaxies collided. In the second, many of the globular clusters condensed from the gas that was shocked and compressed during the galaxy collisions. In the former model the globular The Complex Lives of Globular Clusters 255

FIG. 7.9 Globular clusters adrift in galaxy cluster Abel 1689. In the left image the bulk of the cluster is visible as yellow elliptical galaxies. The faint blue arcs are more distant, gravitationally-lensed galaxies. In the close up a sea of small points of light bathe the larger elliptical galaxies. Teach point is a globular cluster cast off into the cluster void. Image credits: NASA/ESA

clusters should have a narrow range of ages and metal contents as they would all date to a similar epoch in the universe’s history. In the second there would be a spread of ages and metal contents: the most recent clusters having the highest metal contents. What about the distribution of globular clusters within Abel 1689? When initially gas-rich spirals and gas poor elliptical galaxies merged, the collision of gas clouds produced many countless globular clusters in a starburst, much like M82 or the Antennae galaxies. Those stars that were formed deeper within the galactic structures continued to move with their parent galaxies. Stars that formed further out from the galactic nuclei, along with any clusters, would be vulnerable to being ripped away from each parent galaxy by strong tidal forces acting between each. During these events, gas is compressed and heated; what doesn’t ignite as stars 256 The Complex Lives of Star Clusters is boiled off into the hot, intergalactic medium that surrounds the galaxies. In many galactic collisions gas can be efficiently converted into star clusters that form long tidal tails projecting outwards from their home galaxy. Such a structure is visible as a bridge of gas, stars and star clusters between the LMC and SMC. Many similar structures are visible at the scenes of other galactic collisions and, as fossil shells and tendrils within the otherwise featureless balls of light that comprise elliptical galaxies like M87. As the collision continues to unfold the tails are eventually warped and broken leaving a scattering of star clusters (and individual stars) that continue to drift outwards into the intergalactic void. The gas within the cluster is kept hot by a combination of forces. Within the starburst galaxies countless supernovae are instigated during the starbursts. Shock waves from these become coupled to the effect of powerful jets launched by the galactic super massive black holes. The motion of each multi-billion to trillion solar mass galaxy, through the cluster, also shocks and heats the gas and keeps it broiling at millions of degrees. The visible matter is more “sticky” because it can interact with itself easily. This tends to slow the visible matter down more effectively, allowing the gas to collide and condense into stars. The dark matter, by contrast—whatever its true nature—doesn’t interact with itself except through gravity, so when the dark matter dominated galaxies collide, the dark matter tends to keep moving, while the galaxies slow down. The globular clusters that have been scattered outwards then pool within the regions of densest dark matter, where the pull of gravity is strongest, rather than with the visible galaxies, which are still milling around. And Abel 1689 is not alone in hosting tens of thousands of globular clusters. Both the giant Coma cluster and Abel 3558 also contain tens of thousands of globular clusters. You must remember that the mass of the population of globular clusters means that these contain up to 80 % the stellar mass of the entire Milky Way. By contrast our puny population of globular clusters only contributes a few percent of the Milky Way’s mass. Hundreds of millions of years on, the Abel cluster consists of several massive, gas-poor but still interacting elliptical galaxies. These are surrounded by a seething cloud of hydrogen and helium The Complex Lives of Globular Clusters 257 gas, in which countless loose stars and globular clusters continue to orbit. Over time some will fall back into galaxies, perhaps swapping homes repeatedly. Others may be sent further into the voids between the galactic clusters following further, violent interactions with galaxies: either way, what a ride it is.

Cluster Evaporation

The fate of all star clusters is to fall apart. Death comes from within and without. Internally, stars obviously grow old and die. However, if that was all that mattered the cluster would survive for trillions of years. It is obvious, even without the gift of pre- monition, that no star cluster, whether they be open or globular, can survive indefinitely. The forces that hold them together ultimately tear them apart, while the galaxy as a whole, around which they orbit, exerts a terrible influence upon them. Open clusters, orbiting within the disc are repeatedly exposed to shearing tidal forces as they rotate around the galactic center. Stars within the cluster experience differences in the strength of gravity across the cluster’s width. While passage close to neighboring giant molecular clouds causes changes to the gravitational strength that all test the integrity of the cluster. Although the majority of the orbit of a globular cluster takes it through the sparsely populated halo, each globular has to cross the disc twice in every orbit. That’s around 40 passages over the life of most globular clusters. Each transit of the disc puffs the cluster up and allows stars more chance to escape. Over time globular clusters develop tails of (predominantly) low mass stars. This is particularly evident in the dissolving cluster Palomar 5. The cluster is embedded in a long stream of stars more than 20 times the length (or diameter) of the cluster. The stars in this tail are not bound to the cluster at all, but simply follow it like lost sheep around the cluster’s orbit (Fig. 7.10 ). Red dwarfs are preferentially lost from these clusters as they are the stars most readily ejected from binary systems, or ejected following close passages to binaries. The binary system gives up some of its star’s kinetic energy, causing the star to accelerate. With their low masses, red dwarfs will be accelerated most 258 The Complex Lives of Star Clusters

0 Declination [deg. J2000]

−2

235 230 225 Right Ascension [deg, J2000]

FIG. 7.10 A postcard from the edge. This Sloan Digital Sky Survey image, shows Palomar 5 tracking along part of its rather convoluted orbit around the Milky Way’s heart. The cluster is led and tailed by a long stream of stars that are being steadily drawn away from the cluster core. At present there are only around 35 % of Palomar 5’s stars left in the cluster: the remainder having been drawn out into the leading and trailing tidal tails. Image credit: SDSS strongly in such encounters and thus get flung furthest from the heart of the cluster. Low mass clusters such as Palomar 13 are expected to be far into the process of evaporation after more than 10 billion years orbiting our growing galaxy. These clusters have low escape velocities (less than 4 km/s) and it isn’t difficult for stars to get away from the core. Indeed, Palomar 5, 12 and 13 do appear to be evaporating, leaving long trails of stars along their orbits (Fig. 7.10 ). However, it appears that some globular clusters have a little bit more to them than is apparent on first glance. Palomar 13 was investigated by Patrick Cote, of Rutgers University, and colleagues. What they expected to find was that most stars within this small cluster moved with pretty much the same speed: in other words they had a low velocity dispersion (Fig. 7.1 ). To recap, this means that the stars move with roughly the same speed and direction, essentially moving as one. Instead, the stars appeared The Complex Lives of Globular Clusters 259 to still be orbiting their center of mass. And although the velocity dispersion was low, it wasn’t negligible. Stars moved relative to one another at over 2 km/s. This implied that there was an awful lot more mass in the cluster than was apparent by simply counting the stars. Cote and colleagues suggest two possible explanations for the odd result. In one there are a lot of stars that are moving with the cluster but are now very spread out around the cluster in a large, diffuse halo. In the other scenario the cluster is bound to a large halo of dark matter. The dark matter idea would be plausible if Palomar 13 was perhaps part of a long gone dwarf galaxy that the Milky way had swallowed up; or perhaps if the halo of the Milky Way contains clumps of dark matter that Palomar 13 had become entangled with. The idea of a dark matter halo seems less compelling than the alternative: Palomar 13 recently crossed the disc of the Milky Way. That this small cluster could have been violently shocked and heated by its passage seems more plausible. What this means is that the visible cluster is just part of a much more extended, but poorly discernible grouping of stars that are now effectively free to wander the halo of our galaxy, alone. Palomar 13’s days are now numbered. Having lost control of a large fraction of its stars, its gravitational pull will be considerably lower. The next passage of the disc in a few hundred million years may be its last, with its stars being dredged out of the cluster and into the halo. A billion years hence, a faint stream of stars will be all that marks its passage. Ana Bonaca (Yale University) and colleagues recently discovered the shredded remains of a globular cluster in the outskirts of the Milky Way. 85,000 light years away in the direction of M33, lies a tidal stream of stars, measuring 18,500 light years long and 244 light wide. Bonaca and those on her team found the stream in a trawl of Sloan Digital Sky Survey data. The shape, distribution and color of the stars pointed clearly to the stream being the remains of a 12 billion year old globular cluster that had experienced one passage too many through the galactic disc. Other, similar, streams of stars are known and it points to a rather messy and destructive history whereby clusters are shredded by a combination of tidal forces and evaporation within the halo of our galaxy. 260 The Complex Lives of Star Clusters

This must be true of other galaxies, where many globular clusters will have been destroyed since the universe was formed. Based on calculations on the rate of evaporation, approximately five globular clusters are believed to die every billion years. This figure depends on the nature of the clusters, and it remains possible that there were many more clusters than this, but these were fluffier than those that dominate the population today. When the globular clusters of the Milky Way, M31 and the giant elliptical M87 are examined there is a bias towards brighter globular clusters at smaller distances from the center of each galaxy. The implication is, perhaps, that close into the cores of each of these galaxies, most of the smaller clusters have already succumbed to galactic tidal forces and evaporated into the void. Only the more substantial clusters, with correspondingly stronger gravitational fields, have managed to hold themselves together. Death comes mostly from the two-body effect, rather than through exchanges with binary systems. Most stars receive gentle nudges and evaporate rather slowly. Observations show that the majority of stars leaving clusters do so at approximately the cluster’s escape velocity, rather than anything speedier that would resemble an orbital velocity in a binary. Thus the two-body time is the key time for a star cluster. For dense clusters, two-body relaxation times are shorter than for more diffuse clusters. This is because the two-body time sets the interval for stars to segregate within the cluster. Interactions between the cluster’s stars drive the lower mass stars to the cluster periphery, while the most massive fall to the cluster core and set up binary systems. This sets in motion the loss of the lightest stars into the halo as they are both the least well bound, being further from the cluster core; and they have the highest velocities relative to one another of all of the stars within the cluster. The two-body times are on the order of 100 million years for stars in the core of the cluster, but a billion years at the half-mass radius. Although this time extends the further out into the cluster you go, there has been ample time for every globular cluster in and around the Milky Way to relax. Not all of those in the Small and Large Magellanic Clouds have so they may have considerable time left to orbit our shores. Thus it is now, some 10 billion years and more since the majority of globular clusters formed, that they The Complex Lives of Globular Clusters 261 can be seen disintegrating. While loosely bound open clusters have been forming and falling apart over the last 12 billion years, the globular clusters have remained virtually intact throughout this time. During this extended period internal rearrangements in the organization of their stars have set in motion the machinery that will destroy them. The smallest, Palomar clusters born with perhaps 50,000–100,000stars, are already lining their own death beds with stars hemorrhaging from their peripheries as they swing through the galactic halo. Many of the remaining, larger clusters show evidence of advanced core collapse with dense inner cores of stars, rich in binary systems. This implies that these clusters have either ejected many of their component stars or have driven them out to the cluster periphery, where they are now vulnerable to galactic tides. There is an obvious irony in this chapter’s tale: gravity is driving the eventual destruction of the cluster, but it isn’t pulling the cluster’s stars together in some frightening stellar union. Rather, it is pulling them apart and scattering them like so many broken toys in a child’s playroom. Gravity acts to make a relatively small percentage of stellar unions; then uses these to accelerate the majority of remaining stars, until they have sufficient energy to escape the cluster. Thus under gravity the cluster falls apart rather than implodes.

Multiple Populations of Stars: An Afterthought

Chapter 4 examined the now widely accepted idea that globular clusters host multiple generations of stars and that subsequent generations are more helium-rich (bluer) than their first generations. These generations are apparent as broadly parallel main sequence tracts that extend form the main sequence turn-off to the base of the detected main sequence. In most cases the detection of extra helium is ambiguous, but would fit the colors of the stars. These helium-rich stars are also more sodium-rich. The problem is how to get up to 180 % more helium into these stars than their older siblings. This is a lot of helium. 262 The Complex Lives of Star Clusters

Ideas that solve these come in two flavors. The simplest was contamination of some stars by helium-rich gas from older, giants that lay nearby. The alternative was that these helium-rich stars were younger and formed from helium-enriched gas shed by the cluster’s first generation. The latter idea works only if the cluster was much more massive in the past than it is now and that most of the older, sodium and helium-poor stars have been lost from the cluster. More mass is required to provide enough stars to manufacture the observed helium and sodium and to hold onto the gas once it has been expelled from these first generation stars. However, Jay Strader and co-workers’ observations of globular clusters in the Fornax dwarf galaxy seem to rule out the possibility that so many stars could have been lost from globular cluster by now (earlier in this chapter). Where does that leave us and are there alternatives? Figure 7.11 summarizes the options. If we assume that the observations of the Fornax dwarf globulars rule out multiple generations, then is there another option that would be in keeping with other observations of globular clusters? Well, perhaps. We know that stars in the heart of globular clusters are frequently embroiled in interactions with neighbors. The so-called very blue extreme horizontal branch stars largely, if not exclusively appear to be the result of interactions between red giants and binary partners or interloping stars. These close, but near misses strip off the red giant’s outer layers (Chap. 6 ) leaving a much hotter star behind. As red giants are the source of interstellar helium, could this stripped material be added to other stars inflating their quota of this element? Red giants also manufacture sodium in the hydrogen-burning shell around their helium core (Chaps. 2 and 4 ). Could this be stripped off and accreted by neighbors? Finally, a second route to sodium-and helium-rich stars might come from additional mixing inside stars. Although low mass main sequence stars can’t manufacture sodium through hydrogen- burning (they are too cool inside), extra helium might get stirred up into the envelope from the stellar core if these stars are spun up. This could work if the stars are members of close binaries and accrete material from a neighbor (above) or simply spin faster after they are captured by, and exchange partners with, a pre-existing binary. A combination of extra spin and accretion from neighbors The Complex Lives of Globular Clusters 263

cooling flow

a b

FIG. 7.11 Various scenarios for the formation of sodium (and helium?) rich stars. In a , either passively from the cluster medium, or actively in binary systems, a star accretes helium-rich gas from red giants . Verdict: plausible as interactions between stars in cluster cores are frequent, as shown by the number of EHB stars and dearth of red giants . In b , helium and sodium rich gas shed by the first generation is reacquired by the cluster forming a second generation. Plausible and favoured by many, but possibly at odds with some observations, which indicate today’s clusters haven’t lost much mass. In c gas shed by the first generation of stars is trapped in a denser surrounding shell of gas and can cool and collapse forming a second generation. Unlikely, as supernovae would almost certainly sweep away such a protective shell. In d , close binary stars spin one another up increasing mixing of helium from the core to the envelope. Possible, but wouldn’t explain sodium, which requires high temperatures for its formation

might do the trick. Certainly, as this chapter shows, it is the stars in the cluster core that are most prone to these sorts of interactions and it is these very stars that appear to have the oddest chemistry. Once again, is it nurture rather than nature that has determined their chemistry and their fate? Rather than being born with an odd chemistry, did they come by it through interactions with neighbors like the EHB stars? 264 The Complex Lives of Star Clusters Conclusions

Sometime, between 10 and 30 billion years from now, all of the Milky Way’s original population of globular clusters will dissolve into the halo of our galaxy. Even the lofty Omega Centauri will stretch and bloat into a long stream of stars. The only thing tying them to their past will be a common streak of stars carved through the halo of the Milky Way—or as the final chapter shows, the off- spring galaxy that the Milky Way will morph into. Globular clusters are fantastically complex stellar organisms with a very rich history. Yet what we see today is a shadow of their former selves. For when we look at the young globular clusters in the Magellanic Clouds we can see a far more glamorous population of stars (Chaps. 6 and 7 ) that have largely passed from our retinue of stellar megacities. Over the eons, stars have come and gone from their hearts. A constant stream of binary systems have formed and died as the clusters have sought to maintain balance against the ever-present pull of gravity, both from inside and from the unrelenting pull of the Milky Way. Many of the smallest stars have been ejected. Moreover, with much of the original mass gone, with the passing of the cluster’s high and intermediate mass stars, the present day clusters, although still spectacular to us, represent a severely depleted population of objects. And what of the young upstarts: the clusters forged now in starbursts elsewhere? Although these start their clocks ticking now, give them 10–20 billion years and these two will come apart in the same way. The majority of these have masses comparable with our present population of star cities. Given that many stars have gone from our present population, it implies that ours were born more massive still. Will the current population of upstarts be massive enough to be able to spawn more than one generation of stars, as many of ours appear to have done? Time can only tell. For all of the universe’s clusters gravity will win, ironically dismembering the cluster, rather than causing it to implode. Over the eons to come, many more unusual stellar partnerships will be forged, only to dissolve with the passing of further stars, en route to the inky blackness of outer space. One by one the cluster’s remaining residents will either merge or face permanent exclusion. This is a portrait of the future of the Milky Way and it is one that we The Complex Lives of Globular Clusters 265 will turn to in the final chapter. Globular clusters are effervescent structures, and despite their greater size, face the same ultimate fate as the smaller open clusters which dot the face of the Milky Way’s disc. In the end the combined wrath of the galaxy’s tides and the internal machinations of the cluster’s stars will doom the cluster to evaporation. If we could look at Milkomeda (to be discussed Chap. 9 ) 100 billion years hence, we would see a grimly bland ball of stars, perhaps strewn here and there with translucent streams of stars that mark the graveyard of the Milky Way and Andromeda’s former globulars. 8. From Science Fiction to the Reality of Planets in Star Clusters

Introduction

Many science fiction writers, including Isaac Asimov, have often envisaged what it would be like to live on a planet that orbited a stellar resident (or pair of residents) of a globular cluster. Visions of the sky vary from a relatively bleak dark sky with a smattering of bright stars, through a sky filled with a sea of stars as bright as Venus, to a sky as bright as a day on Earth burgeoning with the orange orbs of a dozen red giants. Are any of these visions accurate—and even if the most extreme versions of this dream are true, would such circumstances allow the survival of a habitable world?

Living Worlds

What does it take to turn an inert lump of rock into a habitable citadel of life? The answer can only be given in the loosest terms as we only have the Earth as our example, a situation that will hopefully improve in the near future. On Earth we see life pretty much everywhere, from hot, acidic springs, deep undersea volcanic vents temperate oceans and forests to tiny inclusions in rock or ice. Life has filled every niche that is available on our planet. Life has proved itself to be extremely tenacious and a remarkably adept institution, filling every niche that is available to it. This suggests that life, as a process, is fairly robust and should arise with similar ease elsewhere. Although this might seem a fairly unsteady foun- dation upon which to build our house, it is not unreasonable. The only problem you might have with this is the possibility that the

Earth was endowed with a set of chemistry that is so unique to it that the processes through which life arises can only happen here. So, is this an implausible proposition? Probably not. The reason for this rather sweeping claim is that life on Earth has built itself from ingredients that astrochemists can observe in many parts of the nearby and distant universe. As such these must be available on nearly every planet, at least within our galaxy and those nearby. Thus, with a universally, or at least regionally, common chemistry set and a universal set of rules about the use of energy, it would seem that life must be a truly universal phenomenon, or at least a set of linked phenomena.

Along Came a Spider: What Life (Appears) to Need to Arise

On the Earth life needs energy and a couple of key chemicals: carbon and hydrogen oxide, a simple molecule made of two hydrogen atoms and one oxygen atom. We like to call the latter water and it is almost certainly the most abundant compound in the universe (boring old hydrogen with its two, identical atoms is an allotrope rather than a compound). Carbon appears to be essential for life as it can form long, stable chains of atoms both with itself and with a variety of other atoms and ions. These include oxygen, nitrogen, sulfur, chlorine, hydrogen, calcium, magnesium iron and copper. However, it is not impossible that carbon might form some structures in living organisms with its near-neighbor on the Periodic Table, silicon. The problem with silicon is that the bonds it makes are weaker than those that link carbon atoms. This makes its compounds a little more unstable and instability is not a feature a living organism wants lurking within it. Silicon is most adept at forming long chains with oxygen and we call these silicates, the building blocks of planets. Although life on Earth does use silicon, aside from our love of computer chips, this is mostly confined to hard shells that surround some aquatic animals or as components lurking within the structure of plant cell walls. It is not a cornerstone of life on Earth, despite its ubiquity on Earth and in space. From Science Fiction to the Reality… 269

So, what we are looking for is a planet that has a reasonable abundance of water and carbon compounds (in addition to carbon dioxide). Given the ubiquity of these components, life should (or could) rear its head fairly often. What we want to know is how often and where it does emerge, and how likely it is able to become complex. To consider this further, we need to think about the like- liest scenarios in which life can evolve, the timescales involved and the pre-requisites for the development of complexity. Once we have these, we can relate them to important timescales in star clusters, notably the crossing time and the two-body relaxation time. On the Earth there is at least superficial, chemical evidence for life having emerged within 50 million years of the end of a catastrophic period called the Late Heavy Bombardment (LHB). The LHB was a period of intense asteroid and cometary bombardment that occurred throughout the inner Solar System from 4.2 to roughly 3.9 billion years ago. Despite the ferocity of the bombardment, some very rare rocks that are preserved in Greenland from 3.85 billion years ago show some distinctly biological chemical fingerprints. Skews in the ratio of different isotopes of carbon suggest living organisms of some type were working with carbon compounds. The evidence of what was going on has been thoroughly contaminated by subsequent geological activity, but non- biological explanations for the chemical signal are certainly less probable. Assuming the chemistry does point to the actions of living organisms, what does that really mean? At face value it says the Earth went from non-living to living (in a biological sense) in the space of 50 million years. However, the LHB was sufficiently brutal that the geological record of the time before it was largely erased. It is possible life arose much earlier but that its record has been wiped from geological history. This is important because what we see as a 50 million year incubation period might have been much longer—perhaps 500 million years. Looking at it in its simplest terms the apparently short gestation time implies life can arise easily, given the correct ingredients and opportunities to do so. As we know no better we’ll stick with that assumption. Presumably a planet on which life can arise has carbon, water and at least a few hundred million years to develop its biological 270 The Complex Lives of Star Clusters cargo, as that was the minimum time for life to arise on Earth. The next, and perhaps biggest requirement, is a means of harvesting and manipulating energy. On Earth there are three broad schemes employed by organisms to collect and utilize energy from the environment. Plants and many bacteria employ photosynthesis to harvest their star’s radiant energy and, subsequently, build up the useful carbon-rich molecules needed to sustain life. Other bacteria harvest chemical energy directly from our planet’s interior. These live in hot springs, often many kilometers under the ocean surface or even deep within the crust of the Earth. Minerals, rich in iron and sulfur are used alongside water to chemically reduce (add electrons and remove oxygen from) carbon dioxide. This converts this relatively inert gas into carbohydrates and other useful chemicals that living things can utilize. Both the photosynthesizing organisms and the chemical broth munchers are known as autotrophs, as they make their own food from inert, non-biological materials. The final set of living organisms includes us. These are animals, or to use a more biologically correct term, heterotrophs, which are consumers. Heterotrophs are any organism that acquires its nutrients by eating other organisms. On any planet, if there are sufficient raw materials, such as carbohydrates or amino acids that have been delivered by comets, heterotrophs could evolve first. Once these materials begin to run low, there would be selective pressures for organisms to evolve the ability to make their own ingredients, otherwise they would die out. In this scenario the true autotrophs evolve later. It would seem more likely than not life would begin simply with the most basic living organisms, most likely something more like a virus than a cellular organism. Such a life-form would have one task: reproduction. For as long as there was some sort of nutri- ent that it could use, which was in a form that it could use, there would be no selective pressure to do anything more complex. Perhaps for as long as the LHB was continuing and nutrients were being delivered, life stayed in this rut: simple, primitive, with only one operation: reproduction. However, as nutrients begin to run low with each loss natural selection will drive the development of some mechanism to compensate. As long as this innova- tion can be transmitted from generation to generation, life will evolve and develop to match its ever changing environment. From Science Fiction to the Reality… 271

The problem with this idea is that at the moment what Earth’s conditions were like early in its evolution is wholly guess- work. There is no record of this stage in our planet’s history, nor is there a record anywhere else, so any logical hypothesis is fair game. What evolutionary biologists need is knowledge of the set of chemical and biological tools which were available and this may never be forthcoming. Instead, we can run thought experiments and hope for the best. What would biologists want from astrochemistry? For one an accurate and comprehensive list of the molecules that would have been available on the early Earth. This is no tall order. For example, in 1953 Stanley Miller and Harold Urey ran the now famous and ingenious experiment to try and create biomolecules from what was then thought to be the first chemistry set of Earth. They used an atmosphere that was oxygen-free but contained copious ammonia, hydrogen, methane and methane. With some UV and electrical discharges the pair created over 20 amino acids, though many remained undetected in their original analysis. In more recent work, involving volcanic-like conditions and a composition of gases more likely true of the early Earth (nitrogen, carbon dioxide as well as hydrogen sulfide, hydrogen and water), a similar broth of amino acids has been produced. Interestingly, in some such experiments it is necessary to include carbonates (compounds of carbon and oxygen that are found in antacids) to ensure that amino acids are produced. Carbonates are abundant in rocks and would be present on the floors of the oceans of the early Earth where copious volcanic activity would have delivered such gases. Far from being Charles Darwin’s famous “little warm pond”, such a broth would have been a seething, hot stew of chemicals. Such a cauldron matches the conditions implied by the existing evidence.

Limited Clues from an Earthly Tree

If you were born living off an increasingly scarce set of natural resources, the ability to take resources from other living things is tremendously powerful. Unlike your earlier food supply that was dwindling in abundance, this new living resource can reproduce and keep its numbers up, thus sustaining you as well. The organisms 272 The Complex Lives of Star Clusters on Earth can be grouped into a “Tree of Life” that shows their relatedness to one another and provides very convincing clues to how they each emerged. The tree of life was originally constructed by simply looking at the behavior and the physical characteristics of each organism. Although this worked well, it was intrinsically unreliable: after all most animals have eyes, but not all eyes were created in the same way or at the same time. To firm up the science, biologists have increasingly turned to DNA sequencing. In this process the chemical instructions that tell each organism what to be and how to behave are broken up and read. The end result is a four-letter library of sequences which can then be compared between organisms. Although it took the best part of a decade to produce the first human DNA sequence, increasing automation of the process allows biologists to determine the DNA sequence of an organism in a matter of months. Such productivity has meant that very large numbers of organisms can be compared quickly. From such endeavors, biologists can re-examine the relationships between each living thing and determine its evolutionary links. The tree of Earthly life is centered on bacteria that enjoy living in hellishly hot conditions. From these, the tree of life fans out towards the sorts of life forms we come to recognize on a daily basis. This implies that life on Earth almost certainly began hot, perhaps in abundant hot springs or in a volcanic vent, deep within the oceans. This may, or may not be true of life that forms elsewhere in the galaxy.

Energy, Entropy and Evolution

Energy can be described as the ability to do “work”. As well as meaning gainful employment in the case of humans, this is the ability to move an object, its particles, by force over a distance. If an organism is to survive it must do work, otherwise its environment will get the better of it. Chemists and physicists may equate this propensity for increasing disorder as an increase in entropy . However, technically, this is not quite true—popular though the notion is. Before examining this in a little more detail the basic trend in the universe as a whole is to increase entropy. This notion From Science Fiction to the Reality… 273 is engrained in all systems, whether they be living things, the earth, stars or galaxies as a whole. In reality, entropy is a little more complex than simply a drive towards disorder, but only a little bit more. Entropy isn’t so much a measure of disorder but a measure of probability. Imagine a building made of bricks and cement. The building is only a building if the bricks, cement, roofing tiles, windows and doors are constructed in a particular way. A building cannot have its doors on the roof, windows in the interior walls (well, not usually) and roof in the cellar. That doesn’t work. More fundamentally, bricks cannot be inserted in the windows without causing them some “distress”. However, if you disassemble the building so that all the components are in a pile, then the pile can be organised in any number of ways; not an infinite number, but certainly rather a lot. Windows can lie on top of cement, bricks on top of roofing tiles, etc. Because the house has only a limited number of useful, or functional, ways that it can be constructed, it has low entropy. On the other hand, the pile of bricks and other debris has high entropy because it can be in many different states. The second law of thermodynamics states that entropy is always increasing: things are becoming arranged in such a way that there are many different possible states. Given time the entropy of the building (or its parts) increases as the building falls apart. Bricks become scattered, particles of cement of concrete dissipated and the gravitational potential energy in each of the bricks released as heat and sound when they crashed to the ground. The heat and sound that was released when each brick tumbled then stirs up particles in the air, increasing their disorder. It is highly unlikely, but again not impossible, that the tumbling debris will spontaneously reassemble into a functional building. It’s obviously more probable that a neglected building will continue to decay. If things have a propensity to fall apart how can life advance in the face of such chaos? Some creationists are even convinced that evolution runs contrary to the idea of entropy: if entropy was a driving force in the universe then life wouldn’t exist—it would fall apart and evolution would not occur. Indeed, this argument sounds perfectly plausible at first glance. If organisms became more sophisticated over time, surely their entropy must be 274 The Complex Lives of Star Clusters lower than something more primitive? Afterall, a highly evolved organism—and this does not mean a tree or a human versus a “primitive bacterium”—has lower entropy than all of its components were they to be messily disassembled. By “highly evolved” biologists mean supremely adapted to their environment. Thus all bacteria are highly evolved in the sense that they have arisen over billions of years, through the processes of evolution via natural selection. They may not have arms, write books or sip coffee in the local coffee shop while contemplating the state of the econ- omy, but they are perfectly in tune with their environment and perfectly suited to it. So living organisms have much lower entropy than the components that they are built from. However, the key to understanding entropy and life is to consider what life does, as much as what life is. All living things utilize energy to maintain or develop their structure and grow. These reactions are inherently inefficient and much energy is lost to the surroundings during them. This has the effect of stirring up molecules so that they have increasing entropy. Therefore, although the entropy of the living organism is lower than all the little molecules that it was constructed from, in the process of the construction of itself that living organism made a right old mess of all the molecules around it by shedding heat and by creating waste materials. Indeed an estimated 99.85 % of all the consumables we survive on throughout our lives is shed as waste heat or matter. That’s rather a lot. All of this waste energy and heat cannot be reassembled easily into its original form—its entropy has increased. Thus, an organism that has evolved from a set of simple rep- licating molecules to a sophisticated café-cultured being (and that includes the trillions of bacteria that cover the tables and chairs in such establishments) is a far more efficient creator of disorder, and of increasing entropy, than what came before it. Living things are entropy manufacturing machines! Therefore, there is absolutely no contradiction between evolution and the second law of thermodynamics. The fact that living things assemble is perfectly in tune with thermodynamics: by manipulating and disturbing their surroundings life always increase entropy. From Science Fiction to the Reality… 275 Capturing Energy

Now that we have considered how inept and inefficient living things appear to be in terms of entropy, let’s appreciate how amaz- ingly sophisticated and adept they are at surviving and utilizing at least a small chunk of the energy they ultimately receive from the Sun or other sources. But how did organisms come to capture energy and use it to their advantage in the first place? There are clues hidden within the chemical pathways that operate inside cells. These systems are clearly organized in a hierarchical way around the capture and use of energy. Respiration, the term given to the production of chemical energy in the form of a molecule called ATP, initially runs through a system that is inefficient but works in the absence of oxygen. This is known as glycolysis.

H2O

ATP Light Stator + O and H+ H H+ 2 Rotor H+ e-

e- Light-Harvesting Chloroplast Complex Electrons are passed Inner Lipid (Fat) to carbon dioxide to H+ membrane manufacture glucose

FIG. 8.1 The machinery of photosynthesis in cyanobacteria (“Blue-Green Algae”) and plants. Light is harvested by a complex of pigments and proteins that includes chlorophyll (left ). This releases electrons which funnel along a system of carrier proteins ( brown columns , middle ) to carbon dioxide. The carbon dioxide is then converted into carbohydrates. Meanwhile, the chlorophyll molecules steal back electrons from water, replacing those that have been lost to carbon dioxide. This splits water, releasing oxygen gas and protons (H+ ions). These ions build up on one side of a membrane in the chloroplast, before funnelling through a rotating protein engine which manufactures an energy molecule for the plant, known as ATP ( right ). This ingenious molecular machine works in an analogous manner to an electric dynamo 276 The Complex Lives of Star Clusters

The chemical reactions of glycolysis not only produce energy, but are central to the production of amino acids, the building blocks of proteins. Glycolysis then feeds into one cyclical system that probably served initially in the production of amino acids. This is called the Krebs cycle. For instance, the Krebs cycle manufactures the precursor of the amino acid, glutamate. For those of us partial to Chinese food, the ubiquitous additive monosodium glutamate (MSG), is constructed here. The layout of the chemical pathways is therefore very revealing as it illustrates how evolution took one molecule and built it into others, using pre-existing steps. The generation of novel reactions was relatively rare: most are derived from ones that primitive life already used. In life, evolution always finds the easiest pathways to develop. Why make something from scratch when you can mould something that is already present? The Krebs cycle generates waste in the form of hydrogen, which is bound to other molecules, two B-vitamins. On its own the Krebs cycle also makes a small amount of energy but if you wire it into the final stage of respiration around 15 times as much becomes available, and these require oxygen. During these processes, hydrogen is fed along a chain of specialized proteins and compounds to oxygen where the two combine to produce water. In the process large quantities of ATP are made. The original carbon compound that was used as fuel is reduced to a husk that is left is disposed of as carbon dioxide gas. The organization of this process suggests that the anaerobic stages (glycolysis and the Krebs) were evolved first to produce both energy and amino acids. Later the oxygen-utilizing stages were strapped on, increasing the overall efficiency of the process by many times. This would be expected, as on the Earth oxygen was only available in sufficient quantities after cyanobacteria evolved to produce it from water through photosynthesis. Evolution took simple systems and elaborated upon them as the environment changed, rather than inventing completely new systems from scratch. In essence this is entropy in action once again. Given multiple possibilities, different chemical networks will spontaneously arise as nature attempts to increase the possibilities that are available to it. “Spontaneously” may be a little over-enthusiastic a term, as these things take time and in the case of living organisms, mutations. Mutations generate diversity, and From Science Fiction to the Reality… 277 by increasing the possibilities in an organism’s genetic code are in effect also increasing entropy. The seemingly odd thing about living organisms on Earth is the production and use of oxygen. The production of oxygen in photosynthesis is very similar to its use in respiration. The underlying processes are identical, even though the machinery is not. The problem with these machines is that they are utterly unlike the machines used in glycolysis and the Krebs cycle. Whereas glycolysis and the Krebs use mere enzymes (protein macines that selectively carry out chemical reactions), oxygen production involves a chain of proteins and pigments that capture light and pump ions (mostly hydrogen, stripped of its outer electron) across membranes. Many of these proteins include iron and sulfur in their structures and these help carry electrons from water to carbon dioxide. The central molecule, chlorophyll, contains magnesium. In photosynthesis water splits to leave oxygen gas, while the carbon dioxide is harvested and added to other compounds to form glucose. Similarly, in respiration, where oxygen is used, the same sorts of pathways are used again, but this time in reverse. In respiration, electrons are stripped from glucose in glycolysis and the Krebs cycle, then delivered through a selection of metal-containing proteins, to oxygen, producing water. By reducing to a carbon husk, carbon dioxide, entropy is increased. The tricky bit in Earthly metabolism is how life came to develop the pathways that deliver electrons to and from water. The ubiquitous presence of metal ions and compounds suggests metals were integral to energy generation. The ability of life to produce oxygen production obviously came before its use and the process dates back over 3.2 billion years. Somehow life coupled energy production to the movement of electrons and once that was done it never looked back. Quite possibly living organisms that were attached to rocks used the rocks themselves as a means of transporting energy in the form of electrons. This non-living link was then incorporated within the cells themselves. In reality bacteria are miniature electrical engineers. Many bacteria, including some which cause serious harm to humans, have appendages that transfer electrons from internal molecules to those in their surroundings. In many cases they use this system to generate additional usable energy when their internal stores of fuel run low. 278 The Complex Lives of Star Clusters

In other cases, particularly in some human pathogens, it really isn’t clear what the bacteria are up to, but their ability to cause disease depends on this electrical system. These exoelectrogenic bacteria, as they are known, are widespread in nature. Investigations of some bacteria, such as Shewanella oneidensis , or the ubiquitous photosynthesizing bacterium cyanobacteria, reveal that many bacteria produce electrically-conducting structures in times of stress. Perhaps the most dramatic example was discovered relatively recently. Scientists discovered colonies of different bacteria that work together to form centimeter long tubes which conduct electricity from oxygen-poor to oxygen-rich areas. Quite what they are doing with the electricity is not entirely clear, but the bacteria at either ends of these tubes (effectively living wires) cooperate with one another to produce and use energy. That living organisms can conduct electricity suggests that these might have evolved first and then been incorporated into internal energy generating systems that were more efficient or adaptable than the pre-existing energy pathways. It was when these external systems became paired up with the older anaerobic systems that life really learnt to master its environment (Fig. 8.2 ). So, how long did it take microbes to master the splitting of water or the use of oxygen? On Earth this took about 1.3 billion years from the formation of the planet, or 300–500 million years from the first appearance of the chemical evidence for life for oxygen production to take commence. The production and use of oxygen is important for complex life, if not life in general. Oxygen reacts with itself in the presence of sunlight to produce ozone (O3 ), a highly unstable, noxious gas that just happens to be an excellent mop for ultraviolet light. Although life can probably live on pretty much any planet with an energy source, minerals, water and carbon dioxide, it is unlikely to get very far without oxygen. For one, the presence of ozone means that organisms can crawl out of the mud and out of the oceans and move onto land. A lack of ozone means a fairly rapid death by mutation or simple photo-bleaching when exposed to solar radiation. While there are organisms on Earth that can survive in the presence of abundant ionizing radiation, they are few and far between and don’t tend to thrive. From Science Fiction to the Reality… 279

Carbohydrates Fats

Acetyl CoA O2 Citric Acid

Citric Acid Compounds Amino (Krebs) Acids e- e- e- of iron, Cycle nitrogen Anaerobic respiration and and carbon external electron transport

FIG. 8.2 The central machinery of life. Glycolysis and the Krebs cycle are ancient chemical pathways, which make small amounts of energy as well as essential nutrients such as amino acids. Bacteria also move electrons between compounds to make energy either inside or outside of the cell make but appear to have evolved separately to glycolysis and the Krebs cycle. The electron transport pathway to oxygen is used by all complex life but almost certainly evolved from the extracellular (outside cell) pathways found in bacteria

Oxygen is also a potent reactive chemical that allows respiration to produce an awful lot more energy for every gram of fuel that goes in. Imagine trying to get your car going with the air intake valves blocked. The engine is rather reluctant to fire up, and it’s the same for cells. If you can use oxygen you get an awful lot more bang for your bucks: in biological currency (ATP) about 17 times as much. On the Earth many types of bacteria can use oxygen, but most are equally happy without it. Almost all higher organisms— those which possess complex cells that are needed to build the chicken, carrots, potatoes, and broccoli we have for dinner— require oxygen to power them. Oxygen is used by structures called mitochondria, the descendents of types of photosynthetic bacteria 280 The Complex Lives of Star Clusters

Organism A absorbs chemicals from its surroundings and uses these, inefficiently, to make usable energy (ATP) A new, hybrid organism that makes energy from chemicals taken in from its surroundings and then uses A internalised, electrical circuitry to maximise energy (ATP) production

Gene Swapping C

Organism B uses biological electrical circuitry on appendages on its outer surface to produce energy (ATP) from the chemicals that surround it

FIG. 8.3 Is this how bacteria came to produce, then use, oxygen? Many bacteria are surprisingly good at generating electricity. Most use this property to transfer electrons from food molecules to oxygen (respiration); or from water or hydrogen sulfide to carbon dioxide. Imagine a bacterium that was doing this outside its cell membrane, the outer surface of its cell. Another type of cell was simply chewing up food molecules to make energy. Bacteria are remarkably adept at swapping genes, something we’ve taken on board to make genetically-modified organisms (GMOs). If the two types of cell can come together and swap DNA, then the resulting cell can take up nutrients and make energy efficiently by both methods. The ancestor of the modern cell is born. Photosynthesis and aerobic (oxygen-using) respiration are one step away called purple, non-sulfur bacteria. The purple non-sulfur bacteria lie in cloudy water but are a highly versatile group of organisms that can live with or without oxygen, carry out photosynthesis in oxygen’s absence or switch to using oxygen when it is present. Again, evolution saw a rather large niche when oxygen began to pervade the atmosphere. A highly versatile organism like the purple-non-sulfur bacterium could photosynthesize if it needed to, or it could turn the system on its head and use the then-toxic pollutant that was filling the oceans and soon, thereafter, the air (Fig. 8.3 ). Sometime between 2 and 3 billion years after the Earth formed some purple non-sulfur bacteria ended up inside an ancestor of one of our cells and together they forged a wonderful partnership that ultimately led to all of the complex life we see on Earth today. The partnership allowed our cells to use oxygen and From Science Fiction to the Reality… 281 with this dramatic increase in available energy, evolution had a new handle on the world. Now we have a biosphere with both producers and consumers and the stage is set for an evolutionary arms race between the two. Animals seek food at minimal cost to themselves, while plants act to defend themselves, or perhaps even use some animals to their advantage to reproduce. The layout of the cell’s machinery suggests that life quickly bifurcated into producer and consumer systems: the former manufacturing food and the latter consuming it from others. Given that life probably arose from a common set of raw materials that came with or were delivered to the planet, it probably began with the consumption of what materials were available. When these ran low there would have been pressure on living things to begin manufacturing the missing or denuded ingredient from other materials. In such a scheme you rapidly generate entire biological pathways in which chemicals that are needed are built up from simpler components. Evolution through natural selection, and driven by increasing entropy, provides the means through which this can occur and indeed must occur, for without it we would not be here now. The engine of life would have stalled long ago otherwise. Over time organisms that relied on pre-existing natural resources would have either gone extinct or evolved to steal resources from those auto- trophic life forms that made their own food. Modern life would have emerged from the scavengers that first populated the planet. Why is all of this important? Remember the only clue we have to the ubiquity of life elsewhere in the universe, or the ease at which it can arise, is life’s manifestations on Earth. Terrestrial life is (at the time of writing) the only example of life in the universe that we can be confident about.

Is Life on Earth a Reasonable Model for Life Elsewhere?

What about life elsewhere? Did it (or can it) follow the same paths? This is not an idle question as the answer to it dictates our understanding of how likely it is that life will evolve elsewhere, perhaps on planets orbiting stars deep within star clusters. If the ballpark figure of 1 billion years is reasonable for the formation of cellular 282 The Complex Lives of Star Clusters life, then stars that must host life obviously have to exist longer than this: moreover, they have to be hydrogen-burning stars as these are the most stable and long-lasting stars in the cosmos. So you are looking at stars with at most 75 % more mass than the Sun. Although it might be rather glamorous to orbit (distantly) a brilliant, blue O-class star, a planet wouldn’t have time to assemble, never mind host life, before its blazing Sun departed in dramatic fashion (Chap. 2 ). That life can emerge in star clusters is an almost universal certainty; life is robust. However, how far it gets depends critically on the nature of the star the host planet orbits, and more subtly on the nature of the star cluster in which the star resides. The presence of water and in what form and volume is important. Ocean planets might well form life along hot springs on the ocean floor: life is abundant on these surfaces on the Earth. However, life might not evolve into many different forms if the entire surface of the planet is covered in a deep, salty ocean. In particular at depths above 90 km liquid water compresses and turns into a type of ice (Ice V). In its solid form, water is a lot less conducive to the formation of living things as the types of chemistry it can take part in are reduced. Moreover, a very deep ocean would be utterly dark in its depths and should life begin deep down, it might not develop photosynthesis which clearly requires visible light or infrared radiation. Neither of these travels far in deep water. Therefore, although life is not precluded, it is unlikely to progress very far. More subtly, the presence of land allows a planet to regulate its temperature more easily. Silicate rocks that are exposed to the air undergo various chemical and physical processes that we call weathering. As rocks are being attacked by the elements, carbon dioxide gas is absorbed and silicate rocks become carbonates. This lowers the concentration of the gas in the air. Land surfaces with active volcanoes also release carbon dioxide directly into the atmosphere, whereas in most underwater eruptions the carbon dioxide that is released by them is absorbed directly into the ocean. The opposing actions of carbon dioxide release and uptake keeps the level of this gas in relative balance, ensuring that there is neither too much, causing over-heating, nor too little, causing both rapid cooling and an end to any photosynthesis. On the Earth the balance of the two processes is known as the carbonate-silicate From Science Fiction to the Reality… 283 cycle and has kept the Earth clement and potentially habitable for at least 4 billion years. On a planet with a global ocean the level of carbon dioxide will progressively fall as it dissolves in the ocean. Such an ocean-world is at risk of freezing over if carbon dioxide levels are not maintained at a sufficient level to keep the planet’s temperature up. Land is almost certainly also needed if life is to become complex and intelligent. Land allows life to proliferate in direct contact with the atmosphere, and the atmosphere can hold an awful lot more oxygen than water. More oxygen means more vigorous internal chemistry in any organism and this allows sophisticated energy-thirsty structures like brains to develop. You don’t see many fish operating heavy machinery on the Earth, even though they have filled pretty much every aquatic niche that is available. The most intelligent life forms in the oceans are (with the possible exception of octopus) mammals. Each of these species evolved on land as air-breathing, warm-blooded organisms that later migrated back to the oceans or rivers. Indeed, it is possible that intelligent life needs to be warm-blooded if it is to run an efficient central nervous system with a costly, large brain on top. Intelligence, otherwise, might require what might loosely be called parallel processing: a hive or colony of less complex organisms that work together to act intelligently. Again, these seemingly academic points are important. On the Earth, multicellular life emerged after nearly 3 billion years, while intelligent life took well over 4 billion years. Of course, life on earth might be a slow developer, while other planets might race from organic slime to marketing executive in a matter of a few hundred million years, maybe less. Who knows? The underlying context for this whole chapter is something we came across in Chap. 7 : the two-body relaxation time. If you recall, this is the time it takes a star to encounter sufficient other stars to change its direction by 90°. Where this is short, stars have frequent encounters with one another and there is clearly going to be a greater chance that in the process its planets will either be shed from their orbits, or have their orbits so distorted that the climate is severely disturbed to the point that the conditions become intol- erable for life. In most open clusters the two-body relaxation time is measured in hundreds of millions to billions of years. In fact it 284 The Complex Lives of Star Clusters is so long that most open clusters dissolve before most of the stars have undergone relaxation. Open clusters, found within the disc of most spiral galaxies, are teased apart by gravitational tides over the course of a few hundred million years (Chaps. 5 and 7 ). As stars have very little time to encounter one another before the cluster is torn asunder there would have been very little prospect of planetary orbits being scrambled or even slightly distorted. We know that the orbits of the planets in our system were heavily modified by gravitational interactions between the member planets. The general course of this rearrangement was worked out by the Nice Group, a collaboration of several scientists based in the French city of Nice. In general, between 4.2 and 3.9 billion years ago the orbits of the giant planets migrated outwards, sending a wave of icy planetesimals crashing inwards towards the Sun. Although a rather messy time for the inner planets, it hardly prevented life from starting. Indeed, as was described earlier, the late heavy bombardment might have been an essential chapter in the development of life; for along with hundreds of giga-tons of water came a raft of carbon-compounds that might just have seeded life. Therefore, we shouldn’t be too bothered about some orbital “tweaking”, but the wholesale removal of a planet from its star might be a problem, though even where a planet is scattered from its star, life could still arise at or near hot springs that are powered by geothermal energy. Such life would be sustainable for billions of years, sheltered from the ravages of outer space by a thick carapace of ice. Life might even develop some interesting characteristics, hidden from the outside universe by its icy shell. However, microbial life buried under kilometers of life won’t build rockets or send Earth messages. If we want something a little more sophisticated then the planet must remain close enough to its host star to maintain temperatures suitable for liquid water. Contrast the situation of a planet orbiting a star in an open cluster with one orbiting another in a globular cluster. Here the two-body time is measured in tens of millions of years to a few hundred million years. This means that stars encounter rather a lot of stars and have their trajectories altered significantly. Such an environment is highly likely to disturb planetary orbits to the point at which the chances of life becoming complex are unlikely. Long before life can evolve far through its slime phase, the planet From Science Fiction to the Reality… 285 is more than likely to have its orbit so heavily stretched that the planet’s climate is altered beyond repair. Within a few billion years of formation, the planet will most likely be ejected from its orbit altogether. For such a planet the outcome will be a very rapid fall in temperature: first water, then carbon dioxide then finally nitrogen, oxygen and other lighter gases will be frozen out onto the surface. Had any complex life arisen it will be quickly snuffed out. The only survivors would be hardy microbial life found lurking deep within the planet’s crust or at the bottom of its, now frozen, oceans.

The Galaxy’s Oldest Planet?

Glorious though the night (and day) sky on a planet orbiting its globular cluster star would be, prior to 1993, no planets had been found. Prior to that, a wealth of observations had been carried out on the millisecond pulsar PSR B1620-26 by Steinn Sigurdsson and colleagues since 1988. In 1992 Aleksander Wolszczan and Dale Frail’s announced (to a very skeptical scientific community) the discovery of three planets orbiting the millisecond pulsar PSR 1257+12; the first planets to be found outside the Solar System. Sigurdsson’s observations clearly also showed evidence for an unseen partner (Fig. 8.4 ). Five years of observational data on PSR 1257+12 provided a wealth of highly accurate data on the period and slow-down rate of the pulsar. Millisecond pulsars are some of the universe’s most accurate clocks, only very slowly decelerating as the pulsar’s magnetic field brakes against the surrounding and very weak interstellar magnetism. Where another object is orbiting the pulsar, its gravitational field will periodically accelerate and then decelerate the spin of the pulsar as the two or more objects orbit one another. By 1993 it was clear that the pulsar’s spin was being affected by (then) unseen mass. Most of the unseen material was likely in a star. However, the periodic changes in the radio signal were better matched by two objects: one with the mass of a low mass star and the rest by a much smaller, planet-mass companion. Indeed, measurements suggested that this object was most likely only 2.5 times the mass of Jupiter, placing it squarely in the realm 286 The Complex Lives of Star Clusters

FIG. 8.4 The location of M4’s pulsar planet. Leftmost picture —Hubble image of the central two-thirds of M4. Middle picture zooms in on a smaller portion close to the SW of the cluster core and reveals individual stars. Lowermost image is zoomed in even more and reveals ( arrowed ) the dim white dwarf, around which the planet and pulsar orbit. Top image : Kitt Peak National Observatory 0.9-m telescope, National Opti- cal Astronomy Observatories; courtesy M. Bolte (University of Califor- nia, Santa Cruz). Central and lower images : Harvey Richer (University of British Columbia, Vancouver, Canada) and NASA (HST) of giant planets. So faint was the system, it took Hubble another 10 years to identify the low mass white dwarf. Using models of white dwarf colour and luminosity Brad Hansen (UCLA) produced an accurate mass for the white dwarf: 34 % that of the Sun. With the mass of this helium white dwarf pinned down, and the orbit of it tightly constrained, in 2003, Hansen, Harvey Richer, Steinn Sigurdsson and colleagues could say with confidence that the unseen low mass companion to the pulsar-white dwarf binary was a 2.5 Jupiter mass planet in with an orbital period of 100 years. The orbit suggests that the neutron star pulsar captured the planetary system while its parent star was either on the main sequence or evolving into a red giant. The neutron star grabbed hold of the From Science Fiction to the Reality… 287 red giant and began stripping it down. After a few tens of millions of years the red giant was reduced to a white dwarf, while the neutron star was spun up until it was reborn as a millisecond pulsar (a process explained in Chap. 6 ). In the process, the planetary orbit was stretched and moulded into a new form. Rather luckily for the planet, the capture of (or by, depending on your perspective) of the neutron star did not result in the planet’s ejection from the system. Instead, this ancient world bathes in a torrent of high energy radiation from its new millisecond pulsar star. With current technology it is highly unlikely the planet will ever be spotted, given the distance to M4. The only hope of a sight- ing will be if the planet orbits the star in such a fashion as to periodically block some of the white dwarf’s feeble light reaching the Earth. Such a transit is highly unlikely given the planet-sized orb of the white dwarf, the distance to the planetary system, and the apparent distance to the giant world that is somewhat greater than that to Uranus from the Sun (23 A. U.). Given that all of the stars in M4 date to around the same age of 11.7 billion years, it suggests that the planet is approximately this old. There is a certain irony in the age of this system. By the time the properties of the pulsar planet’s system had been confirmed in 2003, it had become almost orthodox to consider planets would only form in metal-rich environments such as the galactic disc. There were even books published on the subject to this effect. However, as Carl Sagan once observed, and has been so frequently repeated, absence of evidence does not equate to evidence of absence. The technology used in planet-hunting was still in its infancy with very low resolution in 1993, when it was difficult to separate the effect of a planet on its star from other conflicting influences. PSR B1620-26 was found using the Doppler method, whereby the orbit of the planet around the star (in this case) altered the pulsation period of the pulsar. Most other planets were found similarly, in this case by altering the position of certain bands in the spectrum of the parent star. These differences are very slight and prior to 2000, no planets had been found with masses less than Neptune. The lack of planetary detection in a globular cluster was, therefore, something of a self-fulfilling prophecy. Moreover, with most globular cluster stars being poor in metals, the strength 288 The Complex Lives of Star Clusters of spectral bands is often limited, which makes it all the more difficult to detect planetary signals on stars that are thousands of light years away. M4, for example is 6,500 light years from Earth. During the late 1990s, Hubble stared intently at the core of 47 Tucanae, searching for hot-Jupiters (Jupiter mass worlds orbiting close to their parent stars) that were transiting their host star and utterly failed to find anything, but the hunt did not stop. It is only relatively recently that technology has permitted the detection of planets around nearby stars with masses as small as the Earth. Although the Kepler craft was phenomenally success- ful at identifying planets through the transit method, it was aimed at a spot in the sky deep within the galactic disc, not at a cluster, never mind a globular cluster high above the galactic plane. To achieve similar results with globular clusters will take a similarly precise endeavor with a custom-built craft. At present this sort of mission is not even on the drawing board, so the Doppler method within its current limitations will have to suffice. That said, no such project is operational. If we really wish to discover whether globular clusters host planets a dedicated mission will have to be established. With current budgetary limitations in the US and in Europe implementing this idea is a distant prospect.

A Planet Pair for Kapteyn’s Star

Could a planet within a globular cluster remain habitable? The two-body time (above) would suggest not. However, the majority of stars born to any cluster are red dwarfs, even within globular clusters. Such stars are dim and cool and this means that their habitable worlds must keep close to their flame in order to maintain temperatures suitable for liquid water. Where such diminutive stars encounter others within the cluster, the outcome is usually that the puny red dwarf gets kicked about the most; and more often than not gets ejected from the globular cluster (Chap. 6 ). This may sound rather harsh, but the result is that should a red dwarf host a habitable (or potentially habitable) planet, if the stars and its system of worlds encounter another star, the result might be the fortuitous ejection of the star and its planets from the cluster. This would then limit the chances of a subsequent catastrophic encounter From Science Fiction to the Reality… 289 with neighboring stars. The Palomar clusters are in the process of melting away, leaving long trails of stars ahead of and behind the core of the cluster. Maybe within these evaporating trails of stars, there lurks a star with a habitable world awaiting discovery. As shown in Chap. 7 , over time these evaporating clusters are reduced to long trails of stars that weave their way through the halo of the galaxy like so many silky threads. In one such thread of stars lies Kapteyn’s star. Kapteyn’s star, or Gliese 191 as it is also known, is an ancient red dwarf that lies a little over 13 light years from Earth. Despite its current proximity to Earth, Kapteyn’s star is an interstellar hobo, making a brief pass of our system of worlds on its long circuit around the halo of the Milky Way. In the summer of 2014 a team of astronomers, led by Guillem Anglada- Escude from Queen Mary’s in Northern Ireland, announced the detection of two planets orbiting this star. The team had used a radial velocity method based at the HARPS instrument in Chile and HiRES at Keck to uncover these small, hidden worlds. As each planet orbited its host, the gravitational nudges produced characteristic movement of dark absorption bands in the star’s spectrum belaying their presence. The outer planet, Kapteyn c, has a mass roughly half that of Neptune, but is certainly too cold for life. The innermost planet, Kapteyn b, has 4.8 times the mass of the Earth and orbits squarely within the habitable zone around the star. In this region temperatures are high enough to allow the presence of liquid water on the planet’s surface. Although such a massive world is certainly a poor prospect for habitability, it does open a door to the possibility that there are millions of potentially habitable worlds orbiting, tightly, their red dwarf Suns as they weave their way through the disc of the Milky Way. Were any living organisms detected on Kapteyn b, it would make Kapteyn b the oldest known habitable world in the galaxy, and likely one of the oldest in the universe. On its own, the discovery does not seem especially remarkable, aside from the fact that the planet must be roughly as old as the Jupiter-like world discovered in M4. This further refutes the idea that planets are the preserve of more metal-rich environments. As a super-terran, a planet only slightly more massive than the Earth, Kapteyn b opens up the probability that globular clusters could support stars with rocky planets, not simply gas 290 The Complex Lives of Star Clusters

FIG. 8.5 the Kapteyn Star (Gliese 191) planetary system. The potentially habitable Kapteyn b has an eccentric orbit inside the star’s habitable zone ( beige hoop ). While, Kapteyn c orbits further out beyond the region water can remain liquid on a planetary surface. The whole system would fit rather snugly inside the orbit of Mercury. Super-terran, Kapteyn b has a mass approximately five times that of the Earth, while Kapteyn c has half the mass of Neptune. Both could be ocean-planets with thick atmospheres; or Kapteyn b might have a rocky surface much like the Earth. Inner pink ring represents the region where a runaway greenhouse is possible and the outer blue ring where carbon dioxide clouds cause runaway freezing giants. Indeed, Kapteyn’s star has a little more up its sleeve, and something that brings us back to globular clusters (Fig. 8.5 ). That Kapteyn’s star is an interloper from a long gone globular would be interesting in itself, but the star and its worlds tells an even more fascinating story. If you follow the orbit back in space and time it eventually converges with the large globular cluster, Omega Centauri. Indeed the entire stream of stars, known as the Kapteyn moving group, links its origin to this large globular cluster. That isn’t to say that Kapteyn’s star was born and bred here, but given the similar ages and chemical composition, it does, at least, imply that Kapteyn’s star was formed in the same neighborhood as the main body of stars that now comprises Omega Centauri. If it was once part of Omega Centauri it would suggest that interactions between Kapteyn’s star and more massive neighbors From Science Fiction to the Reality… 291 within the cluster core eventually drove the star outwards into the halo of the galaxy as a loner (Chap. 7 ). Were this true, it would not only graphically illustrate how globular clusters fall apart, but also demonstrate that though stars may be thrown this way and that during frequent encounters within globular clusters, some still could hold onto their worlds. Indeed, it is the red dwarfs, with their planets circling close to their flames, which are the surest candidates for such a strong bond. Sun-like stars cast their worlds further afield, making them more prone to tidal stripping by close neighbors. As we look out into the halo and to these ancient star cities, it seems likely that their dimmest residents will be the best places to examine for the presence of habitable worlds. With globular clusters routinely expelling red dwarfs from their midst, astronomers could do worse than cast their eyes on the tidal tails that extend beyond the cluster walls when searching for planetary systems.

Visions of Heaven: The Artistic and Visionary View from the Surface of a Cluster World

Long before there was any evidence for or against the existence of planets within star clusters, most astronomers would have concluded that they simply didn’t exist, at least not bound to stars. After all, the constant shuffling within the miasma of stars, which in globular clusters lie so close to one another as to be almost touching, would disrupt planetary orbits and scatter the stars and their planets. It was the domain of the science fiction writers and not astronomers to imagine such worlds. Isaac Asimov had the planet Lagash orbiting six stars in the story Nightfall , in addition to the somewhat more mathematically realistic planet Arkon orbiting one in the Perry Rhodan series of stories deep within M13. Indeed, M13 pops up in other stories, even in a 1968 episode of the popular sci-fi series Doctor Who . In most of these stories writers have tended to err on the side of light-polluted worlds, bathed in the radiation of a hundred nearby giant stars. The somewhat contrary sentiment that dominated within the astronomy community was built on many years of negative 292 The Complex Lives of Star Clusters results searching for planets in clusters. Globular clusters were searched with the often crude instruments of the day with no success. Taking a different approach, in 1974, the Arecibo dish in Mexico was used to send a message to M13. This contained details of our biological heritage as well as the position of the Earth in the Milky Way. Slightly embarrassingly, the message was sent to the currently observed location of M13 in the night sky. Of course, travelling at the speed of light it would take 25,000 years to get there. In that time M13 will have moved on in its orbit, so even if intelligent aliens exist within its starry depths, they will miss the message’s arrival by several billion kilometers. More realistic views of the night sky as it would be seen from the surface of a hypothetical planet have been longer in coming. However, several can now be seen on the internet sites such as that of Bob King’s Astrobob. This hosts a superimposed image of the view inside M4 from the surface of a hypothetical planet created with Stellarium software. Astronomy went one step further, in July 2014, with a series of images created by William Harris and Jeremy Webb (both of McMaster University). These tracked the progress and change in perspective of a planet orbiting a star on a steeply dipping orbit into and out of 47 Tucanae. In both Bob King’s website and Harris and Webb’s article, the surface of the planet experiences something more akin to dusk. The sky is black, but punctuated with a large number of bright stars, which shed enough light to give a perspective something akin to night-time, suburban streets. You could probably read a newspaper by the light but not safely drive a car without headlights. In the work of Harris and Webb, the planet is surrounded by over 100 stars per cubic light year, while it is in the cluster core. The night-sky has over 10,000 first magnitude (or brighter) stars: that’s over 300 times the number visible from Earth. Of these, 1,000 are brighter than Sirius (apparent magnitude −2) while a sizable handful of stars will have an apparent visual magnitude brighter than −9: at least 100 times the brightness of Venus. With 130,000 stars brighter than our naked eye limit, the night sky would be rather impressive, even with the combined light pollution from the heavenly star city. Moreover, of those stars approaching and exceeding the brightness of Venus, on a cloud-free day these would be visible during the daytime as distant Suns. From Science Fiction to the Reality… 293 Planets in the Open Cluster, M67

Despite the earlier discovery of a planet in the dense globular M4, there were still doubts regarding the presence of these objects in clusters in general. M4’s planet might be a fluke and despite much searching no planets had been discovered orbiting stars in open clusters. It took some considerable time before the discovery of a pair of planets in such a cluster was announced. With hindsight the doubt is surprising. After all, it is thought that the majority of stars form in clusters or looser associations. Since the majority of stars are known to host planets simple math would suggest the majority of stars in clusters must have planets. Lurking behind this cynicism was the idea that planetary orbits couldn’t remain stable in star clusters because of interactions between the stars. However, typical distances between stars in open clusters are greater than half a light year (Chaps. 1 and 5 ), which makes interactions relatively rare. The two-body relaxation time for a typical open cluster such as the Pleiades is greater than 1 billion years, well beyond the time it would take such a cluster to be ripped apart by tidal forces. So, unless a planetary system was remarkably unlucky, it would almost certainly escape into the general body of the galaxy long before it came perilously close to another star and was disrupted. M67 is roughly 1 billion years old and is a fairly dense open cluster that hosts a few thousand stars. As such, it forms something of a bridge between open clusters such as the Pleiades and the globular clusters. Such a location is somewhat risky for a planet, but not so risky that the possibility that life might evolve reasonable complexity is ruled out.

The Fate of M4’s Pulsar Planet

PSR B1620-26 and its planet were likely formed separately when a Sun-like star, towing the planet, fell into the core of M4. The ancient neutron star almost certainly had another low mass star as a partner. Without an initial low mass partner, it would be unlikely that the interacting star systems could have paired up 294 The Complex Lives of Star Clusters in the manner we observe today. When the neutron star and the Sun-like star approached one another, the neutron star would have exchanged the Sun-like star for its original partner. Somehow, in the confusion the planet remained bound to both, though there was a reasonable chance that the planet would swap stars and end up leaving the binary with the former partner of the neutron star. With some serendipity, instead the planet remained locked to the new partnership. After a few hundred million years the Sun-like star became a red giant and then a white dwarf, losing an appreciable amount of mass to the neutron star. In the process the neutron star was spun up and began to pulse once more. The system has seemingly remained stable ever since. However, its ultimate fate is likely tied up with the structure of its home star system, M4. PSR B1620-26 is now heading back towards the core of M4 and in less than 1 billion years the system will pass through the cluster core. Here, the neutron star will reveal its predilection for infidelity and likely divorce its current partners (both the white dwarf and planet) in favor of a more substantial star. Either separately, or somehow still together, the exchange will eject one or both from M4 into the halo of the galaxy, while the neutron star begins its new life. M4’s planet will continue to stream through the halo of the Milky Way as part of a diffuse moving group of stars (and possibly other divorced planets). Trillions of years hence, when the Milky Way has transformed and gone dark, (as will be discussed in Chap. 9 ), the planet of PSR B1620-26 will continue on its lonely orbit around the core of Milkomeda.

Conclusions

Outside the realm of science fiction, cluster planets were considered an unlikely prospect for existence by astronomers, nevermind habitability. If your mental starting point is a star-lit sky where daytime never ends, where stars that are practically touching one another and frequently colliding, then perhaps inevitably, you are unlikely to believe star clusters could ever host habitable worlds. Any potentially hospitable planet would be scattered from its orbit long before anything resembling complex life could ever arise. However, taking a step back and looking at the numbers there From Science Fiction to the Reality… 295 is little reason to doubt that open clusters will contain stars that host stars with potentially habitable worlds. After all, most solo stars began life in such agglomerations of stars, so why wouldn’t planets be common in these clusters? Globular clusters and the densest rich open clusters are more of a problem, but perhaps not quite as brutal a prospect as once thought. Yes, having complex life arise around a Sun-like star, where planets revolve in fairly wide orbits, might present chal- lenges. However, red dwarf systems and their somewhat showier K-class, orange dwarfs, have planets orbiting close to their stars. These should stick together as the effective target their orbits present to passing stars is small. Interactions are more likely to have one of two effects: to launch the red dwarf and its planets into the galactic halo or to temporarily form a higher order partnership between the red dwarf planetary system and its new neighbor. This is more likely where the interloper is itself part of a binary system. Although such red dwarf systems might end up as free- spirits wandering the halo, their origin would still lie within a globular cluster. Therefore star clusters could certainly host habitable planets, even if in the end the fate of these relationships is likely an inevitable divorce. In the case of open clusters this is through the tidal destruction of the cluster as a whole. In the case of globulars the expulsion from the cluster is more likely to result from interactions between stars in the cluster. In both cases, however, life could and almost certainly would survive in some cases. 9. Milkomeda and the Fate of the Milky Way

Introduction

Few galaxies exist in isolation. Thus, despite the grand tour all take through our expanding universe, most are invariably bound by gravity to one or more others. We like to think that on the scale of the local universe, most galaxies must be far apart, and if we consider it is 2.2 million light years to M31, Andromeda, it does seem a rather long way. Yet the distance between the Milky Way and Andromeda galaxies is roughly 20 times their diameter: just imagine a universe if the distance between every star was only 20 times its diameter. Two million (plus) light years may be big on our scales—or even that of the Solar System, but it is relatively puny compared to the size of the galaxies themselves. With each galaxy bearing a mass measured in hundreds of billions (or even trillions) of solar masses, gravity can turn a seemingly solitary, sedate galaxy into a rampaging train wreck. It is to these cascading galaxy collisions that this chapter turns, examining the impact of this process on star cluster formation and destruction using the example of the ultimate fate of the Milky Way. A very turbulent future awaits our galaxy home: one full of glory, destruction and resurrection. Follow a trail from one collision to the next, before the ultimate expansion of the universe sweeps our future galaxy across the cosmic horizon and away from any other body of light.

© Springer International Publishing Switzerland 2015 297 D. Stevenson, The Complex Lives of Star Clusters, Astronomers’ Universe, DOI 10.1007/978-3-319-14234-0_9 298 The Complex Lives of Star Clusters The Inevitability of Collisions in the Local Group

The Local Group of galaxies consists of around 30 members, mostly low mass. All are bound to one another through gravity, and on larger scales, form an outlying part of the larger Virgo Cluster, centered 50 million light years from us. Over the next few billion years the majority of these galaxies must interact. In a few instances this will be a direct collision or a near miss (Fig . 9.1 ). Currently the Magellanic Clouds are clearly interacting with one another and with the Milky Way. The disc of our galaxy has been warped by the passage of both satellites, sending a giant wave sweeping across it. Moreover, both the Milky Way and Andromeda

NGC 6744 substitutes for the Milky Way

MW Leo I Leo II

UrsaMinor Dwarf Sextans Draco Dwarf Dwarf IC 10 Carina Sag Dwarf Dwarf LMC SMC And VII EGB 0427+63 SculptorDwarf NGC 185 And V FornaxDwarf NGC 142

NGC 6822 NGC 205 M32 Andromeda II And III And I M31 M33 Sagittarius Dwarf Phoenix And VI Irregular Dwarf LGS 3

Aquarius Dwarf

FIG. 9.1 A representation of the galaxies populating the Local Group. The group is dominated by M31 (Andromeda) and the Milky Way. The next largest galaxies are M33 and the LMC. Both of these (with the SMC paired up with the LMC) weigh in at roughly one tenth the mass of the Milky Way. The remaining galaxies, of which M32 is the largest, have masses measured in tens of millions of solar masses at most Milkomeda and the Fate of the Milky Way 299 are wrapped in diffuse tidal streams of stars. Although many of these are derived from shredded globular clusters (seen in Chap. 7 ), others are linked to the destruction of dwarf galaxies that have wandered too close to each giant. In particular, the Sagittarius Dwarf forlornly orbits the core of our galaxy as stars and globular clusters are steadily ripped from it and added to the bulk of the Milky Way. Models of galaxy assembly necessitate such violence. The best-fit model is the cold dark matter lambda model, which has been in favor for nearly 20 years. In essence this model has the visible galaxies as the tips of much more substantial icebergs. Each galaxy, regardless of its apparent mass, small or large, is a small piece of a much larger zone of matter that is primarily dark and of an unknown composition. This material is cold because whatever its constituent particles are, they move relatively slowly and only interact with themselves and visible matter through gravity. Dark matter may also interact through the weak nuclear force, but despite many years of analysis, this remains unconfirmed. The “Lambda” part of the term refers to the more recently discovered dark energy, the mysterious force that is causing the expansion of the universe to accelerate. The cold dark matter lambda model posits that all of the larger galaxies in the universe have been assembled piecemeal from smaller scraps of stars and dust. Most of these collisions occurred when the universe was young and the galaxies lay closer together. What we observe now is merely the crumbs left over from the assembly of our home. Expansion has carried most galaxies further away from one another, making collisions and further assembly less likely. The disassembly of the Sagittarius Dwarf is a reminder that the piecemeal assembly of galaxies continues, albeit at a much lower rate. Within the universe, gravity has sculpted agglomerations of matter that still attract one another. It is in these clusters that collisions and harassment can still occur. This is analogous to an increase in entropy with each cluster being a living thing. The cluster is the organism that continues to grow, while the universe as whole around it increases in entropy (Chap. 8 ). The substantial mass of M31, M33 and the Milky Way places them on intersect- ing orbits around the core of the Local Group and demands that they must interact more strongly. Thus, it is inevitable that over 300 The Complex Lives of Star Clusters the next few billion years all three will encounter one another. As they do they will carry their retinue of dwarfs with them, but their wide orbits will mostly ensure they remain bystanders rather than active participants.

Low Metallicity High Velocity Clouds and Star Formation

Lurking in the halo of our galaxy are large, multi-million solar mass clouds of fairly pristine gas. Such high velocity clouds are free of stars and are in the process of collapsing into the disc of our galaxy. As a fraction of the mass of the Milky Way they are small, but each could form several million stars, either in isolation or as one or more large clusters. Dotted around them are many smaller clouds of gas and dust that spectral measurements suggest are either rising or falling relative to the disc of the Milky Way. The origin of these clouds is rather uncertain. There are three possible origins, each very different from the other. The first is the simplest: they are in-falling clouds of gas left over from the formation of the galaxy. Alternatively, they could be clouds of gas stripped from satellite galaxies of the Milky Way, such as the Large Magellanic Cloud or the Sagittarius Dwarf. Finally, they could be clouds of gas ejected from the plane of the Milky Way through the combined effects of supernovae and stellar winds, most likely from the vicinity of clusters of massive stars. The first hypothesis would predict the gas is almost completely free of heavier elements; the middle hypothesis that it is relatively metal-poor (and matches one or more satellite galaxies in composition); while the final hypothesis would make the clouds relatively metal-rich, matching the overall composition of the galactic thin disc (Chap. 1 ). Spectra of the largest of these high velocity clouds show that they are chemically similar to the neighboring dwarf galaxies, while the majority of the very smallest clouds are relatively metal- rich. Of these, the largest is the Magellanic Stream, a body of low metallicity gas in which both Magellanic Clouds are contained. It extends beyond 100° of the southern sky, arcing around the Milky Way (Fig. 9.2 ). One might readily assume that this was material Milkomeda and the Fate of the Milky Way 301

Magellanic Stream

Low Metalicity HVC

LMC

SMC

FIG. 9.2 Atomic hydrogen map of the Milky Way showing the Magel- lanic Stream and High Velocity Clouds (HVCs) near the Milky Way. The origin of the Magellanic Stream is still contentious despite many years of research. One group contends that this was produced by interactions between both Magellanic Clouds on their first pass of the Milky Way, while another contends that the Magellanic Clouds are long term partners of the Milky Way. Arrow shows the direction of motion of the LMC and SMC. Original image credit: NASA and B. Walker (University of Wis- consin) torn from both Clouds as they orbit the Milky Way. However, the SMC and LMC are also interacting with one another, teasing material from one to the other. The orbits of both galaxies are uncertain, and the mass of the Milky Way prone to change on a seemingly annual timescale. In 2008 a group led by Genevieve Shattow and Abraham Loeb at the Harvard-Smithsonian mapped out the likely orbits of both dwarf galaxies. These modeled orbits were based on contemporary observations of the Stream and the Magellanic Clouds that are embedded within it. Their model had the Magellanic Clouds as permanent partners to the Milky Way, having formed within 350,000 light years of its developing core, some 11–10 billion years ago. Shattow and Loeb proposed that the pair of Clouds orbit the Milky Way with orbital periods measuring 2–4 billion years, depending on the mass of the Milky Way and velocity of the Clouds that was adopted. Then in 2012 Gurtina Besla and colleagues presented evidence that both galaxies were on a once in a lifetime pass of the Milky 302 The Complex Lives of Star Clusters

Way, with an orbital period of at least 6 billion years, changing the picture substantially. Instead of the Stream originating in a tussle between the Clouds and the Milky Way, the likely source of the Stream was infighting between the Clouds themselves. For one there simply wasn’t going to be enough time for the Milky Way to have pulled so much matter from both. Besla’s models suggest that hydrogen was pulled from the underlying discs of both galaxies as they swung close to one another. This was then stretched out into a long streamer that now folds around nearly half of our galaxy. In turn, the Milky Way now controls the movement of the disc. Why is there such a big difference in models, and from groups that share collaborations in the recent past? One significant issue is the mass of the Milky Way. If the Milky Way has a lower mass then the high velocity of the Clouds will only be compatible with both Clouds making a fleeting pass and not being physically bound to our galaxy in a tight orbit. In essence they are like the Voyager probes making a fly-by of the gas giants in our Solar System. However, with the same Cloud velocity and a higher mass for the Milky Way, the higher gravity of our galaxy will bind them into relatively long orbits that stretch from 75,000 to 300,000 light years from the galactic core. The only resolution to this debate will be increasingly accurate and consistent measurements of the mass of our galaxy, which has fluctuated from most massive to second most massive Local Group galaxy every other year for over a decade. If Gurtina Besla’s computer model is correct then neither Cloud will be anywhere near the Milky Way when Andromeda arrives in 3.75 billion years. If, however, Shattow and Loeb are correct, both Clouds will be drawn into an exaggerated conflict, with very profound consequences for the future fate of Milkomeda, the hybrid of a merged Mily Way and Andromeda system.

Harassment and Merging Between M33 and M31

M33 lies 2.3 million light years from the Milky Way, 100,000 light years further than Andromeda. Work by Spencer Wolfe and colleagues shows that the two galaxies are linked by a clumpy stream of hydrogen-rich gas. The stream extends from the disc of Milkomeda and the Fate of the Milky Way 303

M31 towards the orbiting disc of M33, with the clumps having an average density one millionth that at the surface of the Earth. That might not seem much, but it’s a hundred thousand trillion times the density of particles in intergalactic space. The overall structure of the filament is not fully resolved at the moment but most likely it will consist of mostly warm streamers of gas at over 100,000 K, pockmarked by small, cooler condensations that are moving towards or away from each of the spiral galaxies. Some of the clumps that have been seen are rather sizable, with diameters in excess of 6,000 light years. Astronomers have ruled out the possibility that these are clouds that have been expelled from the disc of M31 or M33. Each galaxy has, like the Milky Way, a cluster of High Velocity Clouds (HVCs, above). Unlike the clouds in the stream, HVCs are found much closer to each galaxy disc and have relatively high velocity dispersions (see Chap. 7 ). This implies HVCs are moving fairly rapidly compared to the discs of the galaxy and to one another. Contrast this with the clouds lying within the stream that connects M31 and M33. These clouds have fairly similar velocities to one another and to the bulk velocities of M31 and M33, with the implication that they are moving with the galaxies rather than between them. The origin of the elongated cloud of gas is uncertain, but it is very metal-poor. At the moment there is little evidence that these clouds contain stars, which would suggest it’s not a tidal remnant that’s been pulled from the depths of M31, Andromeda, during some close passage of M33. Nor does it owe its origin to the disc of M33, to which it is not directly linked. Most likely, but still far from certain, is that this is a condensation of atomic hydrogen that is a leftover from the formation of both galaxies. Intergalactic space is known to be filled with such filaments and clouds of fairly hot but low density gas with a mass several times that of all the visible stars and nebulae combined. This represents the leftovers from galaxy formation, which has been modestly polluted with heavier elements blasted into intergalactic space by supernovae and jets from black holes. As the future of the Local Group unfolds this relatively pristine gas will continue to support at least the potential for further rounds of star formation long into the distant future. 304 The Complex Lives of Star Clusters The Fate of M31 and the Milky Way’s Dwarf Satellites

The fate of the little fleet that accompanies the much larger Milky Way and Andromeda galaxies has already been alluded to: in most cases they will evade disaster. The largest of those accompanying the Milky Way are the gas-rich Magellanic Clouds, which collec- tively have one tenth the mass of their host. To put this in perspective this is an equivalent mass to that of the small spiral M33 or Triangulum. Meanwhile, the largest dwarf galaxies accompanying M31 are M32 and NGC 205. Both of these are dwarf elliptical galaxies, each with a few percent the mass of Andromeda, but severely depleted in gas and dust. The remaining dwarf galaxies, such as Draco I or Leo I, (see Fig. 9.1 ) have masses comparable to through to a few hundred times larger the largest globular clusters: a few million to a few billion times the mass of the Sun. These are either faintly dispersed irregular galaxies, or somewhat more compact spheroidal galaxies, similar to M32. With the exception of M32 and NGC 205 all are low in mass. The great distance of these galaxies to the Milky Way and the nature of their orbits implies that none will become deeply embroiled in the coming collision (Fig. 9.3 ). However, there are exceptions. Those dwarfs that are tightly bound to Andromeda are already embroiled in various machinations with it. Over the next few billion years as Andromeda and the Milky Way approach one another, these will likely dissolve into the disc or halo of Andromeda. A notable exception is the surprisingly hardy M32. This dwarf has already punctured the disc of M31, catalyzed a wave of star formation and depleted Andromeda’s disc of much of its gas. It is what lies within the core of M32 that explains its greater influence. A few hundred light years down inside M32 lies a super-massive black hole, with a mass roughly half that of the much more massive black hole that forms the core of the Milky Way. Such a massive black hole for such a small galaxy is unusual and suggests that repeated encounters between M31 and its smaller sibling has stripped M32 of most of its stars and deposited them into M31. However, a black hole of this mass cannot be so readily deflected. Its substantial pull will carry it repeatedly through the Milkomeda and the Fate of the Milky Way 305

FIG. 9.3 A schematic representation of the orbits of different galaxies in the Local Group. Dwarf satellites orbit on relatively steeply dipping paths (blue ellipses ), with the suspected orbit of the Magellanic Clouds shown in purple . Meanwhile, closing in from lower right , are M31, M33 (on a red ellipse ) and the other satellites of M31. Each dwarf galaxy is a mere spec of dust compared to its orbit. The orbital paths are only representations to illustrate that most satellites are likely to miss M31 and Andromeda when the two giants begin their collision in 3.75 billion years. The Magel- lanic Clouds are likely to lie high above Milkomeda as it assembles (indicated by arrow ). With the possible exception of the Magellanic Clouds the orbits shown are illustrative, rather than accurate representations

disc of Andromeda, and each time in doing so will cause it to scatter stars this way and that, further scrambling Andromeda’s disc. Only once a substantial number of orbits have transpired will tidal forces pull it towards the core of Andromeda where it can fuse with M31’s central black hole. The duo will most likely still be embroiled in their final death-dance while Andromeda and the Milky Way are sizing one another up. Excepting the influence of M32, the feeble gravity of these dwarfs will neither deflect the much more massive M31 and Milky Way nor affect the final distribution of their constituent stars. Each dwarf will have its orbit redirected as their parent galaxies, M31, M33 and the Milky Way, converge. However, aside from the small risk that some satellites will collide with one another, most 306 The Complex Lives of Star Clusters will remain largely unchanged in 5 billion years. The fate of these galaxies will be solely determined by their reserves of gas and the gradual and inevitable decline in the rate of star formation. None of the dwarfs have orbits that will ever (or at least in the next 10 billion years) bring them close to the train-wreck, so none will participate directly. However, in the longer term, gas from them may end up in larger structures that can rain into the Milkomeda galaxy (and/or M33). In this regard, it is interesting that researchers have shown, with increasing confidence, that most of the satellites of the Milky Way and Andromeda appear to orbit both on roughly circular orbits and that satellites on opposing sides of these larger galaxies display so called anti-correlations in their velocities. We came across the term anti-correlation in Chap. 4 and there it related to the proportions of oxygen and sodium. Here, the anti-correlation is in the velocities of dwarf galaxies that lie on opposite sides of the larger M31 or Milky Way. The term anti- correlation in velocity is a complicated way of saying that the dwarf galaxies that lie on the opposite sides are moving in the opposite direction—i.e.: they must be orbiting M31 or the Milky Way in the same direction, either clockwise or anti-clockwise but not both (Fig. 9.4 ). This gives the entire entourage of galaxies the appearance of two converging planetary systems. The Milky Way and M31 form the “stars” with the dwarf galaxies corresponding to the orbiting planets. This gives each galaxy an awful lot of angular momentum and will undoubtedly alter the outcome of the collision between the Milky Way and Andromeda. As Fig. 9.4 suggests, these dwarfs may ultimately guide the stars of Milkomeda into something resembling an enormous, lumpy disc, flanked by denuded but still vibrant gas-rich dwarfs. As we have already seen, the precise fate of the Magellanic Clouds depends on both their orbit and the mass of the Milky Way. If they are bound to the Milky Way, as Shattow and Loeb proposed in 2008, then repeated passages will take them into the frothing mass of stars generated by the collision of Andromeda and the Milky Way. As both are likely to retain formidable amounts of gas and dust, the repeated passage of each dwarf could contribute a rain of material from which future generations of stars could form in Milkomeda. Maybe, and there are many unanswered Milkomeda and the Fate of the Milky Way 307

FIG. 9.4 Large-scale motion of galaxies in the Local Group. The dwarf galaxies in the Local Group frequently show “velocity anti-correlations”. This means that dwarfs on the opposite side of the central galaxy (the Milky Way or M31/Andromeda) move in the opposite directions, showing that they orbit the central galaxy. This means that the entire system of galaxies can be thought of as extended discs, surrounding a central mass, much like planets orbiting a pair of stars. The disc of the Milky Way is tilted into the page and appears more dispersed than that of Andromeda (M31). All this extra angular momentum in these gas-rich dwarfs could, in principle, support continued star formation in Milkomeda long after its formation, by seeding its outer portions with star-forming gas

questions here, maybe, they might just power the formation of more youthful clusters, including globular clusters, during the collision of Andromeda and the Milky Way. If not, the firework show from the collision might be rather limp, with both the Milky Way and Andromeda having depleted much of their remaining gas and dust by then.

The Grand Collision

It has been known since Edwin Hubble’s time that Andromeda is moving inexorably closer to the Milky Way. Unlike every other large galaxy, Andromeda displays a blue-shift in its spectrum: spectral bands are shifted towards the blue-end of the spectrum indicating converging motion with us. It was unknown exactly how the eventual collision would play out. Might Andromeda 308 The Complex Lives of Star Clusters simply make a side-ways brush with us, before heading back off into intergalactic space? It has only been in the last few years that precise enough measures of velocity been taken to begin establish- ing the outcome of the crash. In 2007 T. J. Cox and Abraham Loeb (both at the Harvard- Smithsonian) produced a rather dramatic model where the two galaxies collide head on in as little as 2 billion years. Although the initial collision does not obliterate the separate identities of each galaxy, the two return a few hundred million years later to complete their merger. During the first pass, the Sun is most likely scattered into a long tidal tail as it picks up momentum from the incoming Andromeda galaxy. There is a small possibility that the initially retreating Andromeda will then play host to the Sun, albeit briefly, before the two galaxies merge. During the second pass, as the two galaxies swing back towards one another then merge, the Sun will fall back into the morass of stars and likely end up part of a tidal filament, much like those detected within the outer envelope of M87. Unlike many other contemporary models that of Cox and Loeb tried to make sense of the intergalactic medium. At the time they published their model, the filament connecting M31 and M33 was unknown, though the presence of such clouds was suspected from computer modeling of galaxy formation. Cox and Loeb model the intergalactic medium as a diffuse cloud of warm (100,000 K) gas of roughly uniform low density. When Andromeda and the Milky Way approach and begin to interact, momentum is transferred from the violently shocked gas within the galaxies to that surrounding medium. These shocks propagate outwards as large waves that compress and rarify the surrounding gas. As well as generally heating the gas to millions of degrees, some regions are likely to become dense enough to rapidly cool and condense, potentially fueling further star formation. Surrounding the active elliptical galaxy, Hercules A (3C 348) is a cloud of hot gas. Chandra revealed that this has a clear spiral structure, indicating that this gas is likely condensing and falling in towards the nucleus of the galaxy (Fig. 9.5 ). Before you get too excited about the prospect of a massive wave of star formation, there is one rather important compli- cation: the super-massive black holes that reside at the heart o Milkomeda and the Fate of the Milky Way 309

FIG. 9.5 Optical, X-ray and radio images of the active elliptical galaxy Hercules A (3C 348). The image upper left is an optical image). Credit: NASA, ESA, S. Baum and C. O’Dea (RIT), and the Hubble Heritage Team (STScI/AURA). In the X-ray image ( centre ) the visible galaxy is surrounded by a large halo of hot gas, which has a spiral structure centered on the visible galaxy. Lower right is a superimposed radio image from the VLA, showing the pairs of jets, projecting from the central super-massive black hole. Credit: X-ray: NASA/CXC/SAO, Optical: NASA/STScI, Radio: NSF/NRAO/VLA

Andromeda and the Milky Way galaxies. The Milky Way hosts a 4 million solar mass hole in space: large, but one dwarfed by M31’s 30 million solar mass monstrosity. As the two galaxies merge, gas will get channeled into these black holes. Although much may be swallowed whole, a small, yet significant, proportion of gas will get blasted into opposing high-energy jets. These will impact and rapidly heat the surrounding interstellar gas, preventing sufficient cooling to allow condensation and star formation. A cyclical process is evident in many large galaxy clusters whereby gas falling into the black hole triggers the formation of jets, which in turn radically heat the intergalactic gas in the cluster. Now too hot 310 The Complex Lives of Star Clusters to condense, it mills around inside the cluster, broiling at tens of millions of degrees. With its supply now cut off, the black hole is no longer able to launch jets, so the gas can cool down and begin to condense once more. Such cooling gas tends to coalesce into giant cooling flows which funnel into the core of the cluster, normally dominated by one or two giant elliptical galaxies. M87 in Virgo is a nearby example. Within these cooling flows, star formation has been detected in some central galaxies. However, the number of stars forming is relatively small and appears to be dominated by low and intermediate mass stars. Most of the gas (well over 90 %) simply pours inwards to its fate in the central black hole. This is something of a waste of the universe’s remaining resources. The problem with composing any realistic collision scenario is knowing precisely where each of the total 100 trillion stars in the two galaxies will be at the time of the collision. Also, if the velocities of the galaxies are even slightly off, the resulting simulation will not be realistic. One of the major issues was pinning down the precise angle at which M31 and the Milky Way were approaching. Was it directly head on or was a glancing blow most likely? Since Cox and Loeb produced their model in 2008, more analysis has been done which further refines the outcome. Recent models retain much of the detail of Cox and Loeb’s earlier model, but with better data, the timing and detail of the impend- ing collision have been improved upon. Early models neglected the possible impact of the Magellanic Clouds and the substantial Triangulum galaxy, M33, on the course of the collision, which has now been corrected for. In 2012 Roeland P. van der Marel (Space Telescope Institute) and Gurtina Besla (Columbia University) re-examined the collision models and refined them using Monte Carlo simulations. These took into account the role of M33 as well as the uncertainty in the positions of the two galaxies and the Sun. A set of models were produced which clearly defined the likely scenarios that would unfold. A series of three papers were produced which examined different components of the collision. The third paper was perhaps the most illuminating. Although there is a range in the models of Loeb and colleagues, the highest probability simulation has the Milky Way Milkomeda and the Fate of the Milky Way 311 and Andromeda making a close pass in 3.87 billion years, with their cores separated by less than 31,000 light years. The pair then swings past one another, eventually separating by 191,000 light years—or roughly twice the current diameter of the Milky Way— 4.97 billion years from now. The failure of the pair to merge on the first pass is down to their relative velocity: the two galaxies will approach one another at over 580 km/s—or roughly 70 times the speed at which the International Space Station orbits the Earth. At 6.29 billion years from now, as the Sun leaves the main sequence and begins its first ascent of the red giant branch (see Chap. 2 ), the two merge during a second swipe. This is qualitatively the same as the earlier model of Cox and Loeb. However, the timescale is slightly longer and the fate of the aging Sun more precisely mapped. The position of M33 is crucial in this respect. M33 makes it next closest approach to M31 in a little under 1 billion years from present. At this point it is only 80,000 light years from Andromeda. After swinging out to its furthest separation of 219,000 light years at 2.66 billion years, M33 returns at 3.83 billion years, just as the Milky Way and Andromeda are swinging away from one another. This means that M33 will be at least in the line of debris from the collision, if not directly involved. The most likely fate of M33 is that it will settle into a precessing orbit around the pair as the Milky Way and Andromeda make their final approach. When the merger has completed at 6.29 billion years, M33 will lie 100,000 light years distant, placing it within the halo of the merged elliptical galaxy, but not so close that it is completely disrupted. That said the simulations suggests that a good fraction of M33’s stars, and presumably its gas, are stripped off, forming long streamers that populate the halo of Milkomeda (Fig. 9.6 ). In a little less than 10 % of the simulations M33’s orbit around M31 brings it into contact with the Milky Way before M31 hits. Such a collision would likely deliver a sizable amount of gaseous fuel for star formation as this small galaxy, like the Magellanic Clouds, is fairly gas rich. In less than 8 % of the simulation runs, M33 avoids the collision but exchanges enough momentum with M31 and the Milky Way that it is rapidly ejected to very large distances from the centre of the Local Group. In many of these runs M33 is not lost entirely, but instead settles down into a long, 312 The Complex Lives of Star Clusters

M33

FIG. 9.6 A schematic representation of Milkomeda as it is assembling between 3.87 and 6.29 billion years in the future. Milkomeda is the orange oval in the middle of the figure. Surrounding Milkomeda is a pink shaded area , representing stars (and residual gas and dust) that has been scattered into a broad halo and tidal tails. The Sun is highly likely to reside in one of these. Stars within the ends of the tails have sufficient kinetic energy to escape Miklomeda. Black arrows represent stellar motion within this region, while the blue dashed line is the tidal edge of the developing elliptical galaxy. Purple and red solid lines represents the orbits of the Magellanic Clouds and M33, respectively. All are well within the outer stellar morass of stars surrounding Milkomeda’s core. The two central super-massive black holes are also indicated in the heart of the orange oval radial orbit that carries it towards and away from Milkomeda. In some of these, M33 then returns for a head on collision. Where M33 hits the Milky Way first, the two may merge into a chaoti- cally organised elliptical. However, the relatively small mass of M33 might allow much of the original structure of the Milky Way to be retained before M31 arrives, shortly thereafter. Thus in the most likely scenarios, M33 will passively follow M31, on a looping orbit around it, as it collides once then again with the Milky Way. During the first collision, M33 is likely to lie more than 175,000 light years distant to the collision. However, during the second and final crash between M31 and Andromeda, Milkomeda and the Fate of the Milky Way 313

M33 will be moving within a 100,000 light years of the collision. As a result M33 will find itself peppered with stars thrown out of the crumple zone. Being such a close bystander there is a small, yet significant possibility that the Sun will be ejected into a tidal tail that swings it through M33. Although moving far too fast to be captured (at over 100 km/s), it would make for a spectacular journey through not one or two, but three galaxies. Meanwhile, M33 will suffer a number of close encounters with Milkomeda, spilling just over one quarter of its stars into Milkomeda’s halo. Shortly thereafter, what is left of M31 and the Milky Way will settle into an American football (or British rugby ball) shaped conglomeration of stars: an elliptical. Although initially wrapped in long streamers of stars, two-body interactions between the trillion plus montage of these will smooth their orbits out into an amorphous blob of largely random, radial and Keplerian orbits (Chap. 7 ). Sometime later, perhaps 10–15 billion years after the Milky Way and Andromeda collided, M33 will spiral inwards, its orbit steadily decaying as it ploughs through an increasingly dense sea of scattered stars, the Sun included. The inward spiral will, itself, be matched by an outward scattering of stars from Milkomeda, further disrupting its shape. Then, finally, perhaps a billion or 2 years after M33 impacts Milkomeda, the Magellanic Clouds may return for a final, fateful encounter. Again, the return of the Magellanic Clouds could be anytime between 6 and 10 billion years from now—so, realisti- cally, before, during or sometime after M33 spirals inwards. What happens next is anyone’s guess. Milkomeda will have, potentially accreted a lot of gas from its smaller gas-rich companions. This might allow the formation of a disc, or simply fuel smaller burst of star formation. However, by 15 billion years hence, much of the initial fuel supply will be long gone. What remains to power star formation will be gas recycled from dying stars—of which there will be many in Milkomeda.

The Fate of Milkomeda

What would we see if we could look upon the Local Group in 10 billion or even 100 billion years? For one thing very little would probably have changed in terms of the appearance of Milkomeda. 314 The Complex Lives of Star Clusters

At 10 billion years, after the initial fireworks of the grand collision had died down, the scattered tendrils and streams of stars arcing through Milkomeda’s outer districts would have largely dissolved into the great ball of stars. The system, as a whole may not have completely virialized (Chap. 7 ) but most of the body would appear as a largely featureless blob, broadly yellow in color but with a few bright blue sparks. The latter would be subdwarf B and perhaps a few extreme horizontal branch stars (Chaps. 2 , 6 and 7 ). Giving a splash of glamour, perhaps, would be the twisted form of M33, its structure warped and torn through progressive encounters with the bloated Milkomeda. Somewhere nearby might lie the Magellanic Clouds, perhaps merged into one structure, spinning back downwards for another close pass with what was the Milky Way. After 10 billion years, the amount of fuel for star formation will be severely depleted with very little fresh fuel remaining in the universe. Most of the fuel that becomes available will come from dying stars, with a highly unpredictable amount that condenses from the warm intergalactic medium that bathes the cluster. Milkomeda might be served by a cooling flow (Chaps. 4 and 7 ). Here, warm intergalactic gas collects, cools and condenses before pouring into the heart of the elliptical galaxy. Some star formation would persist in this flow, but it would be subordinate to the mass of material blowing out of the galaxy either in a general galactic wind or more energeti- cally powered by jets coming from the central super-massive black hole. Either way, Milkomeda’s glory days will be long gone, and the process of star formation will be slowly subsiding. We leave the other dwarf galaxies alone, still serenely orbiting the heart of Milkomeda. Meanwhile, Miklomeda and the other remains of the Local Group are moving inexorably towards the Virgo Cluster, itself warped, folded and kneaded into new forms through eons of close encounters between its members. It is almost impossible to predict what will happen next. Milkomeda and the other Local Group members will most likely fall into the Virgo cluster, of which, strictly speaking, it is an out-lying member. The heart of the Virgo cluster is dominated by a fleet of elliptical galaxies. Some of these, particularly M84 and M87 at its heart, may be primordial, having formed in roughly their current Milkomeda and the Fate of the Milky Way 315 state. The other elliptical will have two origins. The larger ones may be produced by mergers between spiral galaxies, much like Milkomeda in 6 billion years time. However, others were relatively graceful spirals that suffered a tormented stripping down process called ram stripping. As we’ve seen already the heart of a large galaxy cluster is filled with very hot, ionized gas. Near the outskirts of the cluster, the density of gas is low but the temperature is extremely high, perhaps 30 million Kelvin. Any spiral galaxy moving into this torrid sea experiences a boiling action as highly energetic gas in the cluster medium heats up and begins to boil away the gas in the galaxy’s spiral arms. As the spiral galaxy falls ever inwards through the heart of the cluster, it encounters progressively denser (but only slightly cooler) gas. This denser gas acts like a sand-blaster, stripping the already super-heated gas from the edges of the galaxy disc into the intra-cluster medium. After only two or three passes, most spiral galaxies are completely stripped of gas, leaving a so-called SO galaxy behind. These SO galaxies have a nuclear bulge and disc, but are mostly free of gas and dust, both having been stripped off through repeated close encounters with the hot gas of the cluster. A few spirals emerge, briefly, as ghost spirals, low luminosity spiral galaxies that support a fraction of the star formation rate of the present day Milky Way. Their fate is sealed and soon any capacity to form new stars will be lost altogether. For many of the SO galaxies what happens next is inevitable. Close encounters between each other and the cluster elliptical galaxies strips most of their star away from their outskirts. These lost souls are added to the moribund collection that populates the ellipticals. The nuclei of the galaxies persist as small elliptical galaxies orbiting their larger hosts. If they were particularly small spirals, they may appear as Ultra-Compact Dwarfs (UCDs) , over- sized globular clusters hovering forlornly within the halos of the cluster’s giant elliptical galaxies. At some point perhaps 100 billion years hence the Virgo cluster will settle into a final state. During this interval, all of the cluster’s globular clusters will dissolve into the intergalactic medium, forming a thinly spread sea of low mass stars and stellar remnants. Some may form temporary streams within the halos of their galaxies, but by 30 billion years from now all of the globular clusters will 316 The Complex Lives of Star Clusters be gone, stripped down to individual members through encounters between the component stars. The Virgo Cluster may have merged further with other neighboring clusters, but such hierarchical clustering cannot go on forever. Lurking behind the gravity-driven façade of our universe lies the actual principal driver: dark energy. Clusters of stars and galaxies may be dominated by the invisible had of dark matter, but this has less than one half the power of dark energy. Current estimates have 4 % of the universe in the form of stuff that we might recognize: ordinary matter that is mostly hydrogen and helium. This is what surrounds most of the universe’s galaxies and still contributes star forming material to them. 30 % or so of the universe is dark matter—thoroughly mysterious material that is highly recalcitrant to our attempts to investigate it. The remainder is the universe’s joker: stuff that astronomers can detect the effect of but that is all. Dark energy began accelerating the expansion of the universe around 5 billion years ago and the process continues today. By 100 billion years the descendent of Milkomeda and its Virgoan companions will find themselves cut off from the rest of the universe by its accelerating expansion. Over the next trillion years the visible universe empties out. The space between the galaxies will continue to grow at an accelerating pace. First turning orange, then scarlet, then crimson, galaxies at the edge of the observable universe will steadily dim then wink out not because they are dying but because their light is progressively stretched more and more towards the red end of the spectrum. We will then be alone in our home galaxy or in our cluster. Either way, the next and final process of galaxy evolution will play out, and it is one that we have already encountered.

Galactic Dissolution

Once the galaxies have adopted their final elliptical forms, the same factors that govern the fate of the globular clusters will come into play. While the crossing time for a globular cluster is on the order of a few million years and its two-body relaxation time is on the order of a few hundred million years, the same processes take several orders of magnitude longer for galaxies. A typical elliptical Milkomeda and the Fate of the Milky Way 317 galaxy with 100 billion stars will have a crossing time of roughly 100 million years. Its two body relaxation time will stretch from 10 to 1,000 trillion years: far longer than the current age of the universe. However, given the potential that the universe will last far longer than a trillion, trillion years, even this long two-body time will become important. As stars encounter one another within the elliptical chaotic swirling melee of stars, two-body encounters will drag the heavier stars towards the galaxy core. By the time 1 trillion years have passed the only stars left burning hydrogen will be red dwarfs with less than 20 % the mass of the Sun. Everything else will be cooling embers: neutron stars, white dwarfs or the occasional rogue black hole. More stars will now be dead than are alive. Two body encounters will tend to drag the black holes, neutron stars and the heavier white dwarfs towards the centre of the galaxy. Less massive low mass white dwarfs, made primarily of helium, plus the remaining red dwarfs will tend to migrate outwards in a kind of slow galactic wind. Even at this point, several stars will have sufficient energy to escape into intergalactic space. Here, caught in the rip-tide of the expanding universe, they will accelerate away from Milkomeda or its descendent and be lost forever in the dark open sea of space. Galactic evaporation only really becomes important once the last stars have died out. The term “the last stars” is still something of a misnomer at this stage. What astronomers mean is the last stars formed from the collapse of nebulae. Once the nebulae are reduced to a distant memory, stars will still form for trillions upon trillions of years until the galaxy is no more. Intermittent and rare collisions between brown dwarfs will maintain a population of around 10–50 red dwarfs for 1022 or 10 billion, trillion years. So, although dim, the galaxy will still shine, pockmarked with a faint residue of youth. A few other stars will pop up briefly and perhaps violently when carbon-oxygen white dwarfs collide with other helium- rich white dwarfs or brown dwarfs. A few may produce Type Ia supernovae, but most will blaze for a few million years as R Corona Borealis stars. A few stars, synthesized by the collision of two helium-rich white dwarfs, will burn helium for longer, perhaps 100 million years. This phase persists for as long as the 318 The Complex Lives of Star Clusters galaxy holds together. Such collisions are made more likely as a result of the internal structure of Milkomeda (or its descendant). In a spiral, most stars orbit the galaxy’s core serenely in stable orbits. Such an arrangement produces very few if any collisions as each star rarely encounters another. In an elliptical the orbits of stars are scrambled and most orbits will intersect others, making collisions more probable. Thus, it is to chaos that the future of the galaxy’s light will owe its existence. During this final stellar phase, two body interactions will continue to push the lightest stars away from the galactic nucleus. As these happen productive collisions between stars will become rarer, until none remain. Then, and only then, will the galaxy go dark—almost. By 10 22 years, 99 % of the galaxy’s stars will have evaporated into intergalactic space. These dead, cold objects will pass away into the void as dark and as frozen as the space around them. The remaining 1 % of stars will face a less pleasant fate. Having lost most of their kinetic energy through encounters with neighboring stars, these remaining members of the galactic population will succumb to gravity and fall into the waiting maws of the central super-massive black hole. Perhaps, marking their demise with a small burp of X-rays these final bastions of Milkomeda will vanish from space forever. At 1023 years Milkomeda and all its siblings will be no more (Fig. 9.7 ). At this point, barring exotic physics, the universe will go dark, perhaps for good. All that we know of that can (briefly) break the stygian doom will be the demise of black holes through the steady if still theoretical process of Hawking Radiation. This form of radiation involves a bizarre but utterly real set of objects called virtual particles. As these pop into and out of the universe, they appear as matter-antimatter pairs. This conserves charge, spin and mass. If, however, they were to do this near a black hole, one may get swallowed, while the other survives—and that’s a no-no. The survivor has just broken the universal rule about the conservation of mass and energy. For the universe to get pay-back it must extract an equivalent amount of mass from the black hole: in effect the black hole gives up some of its mass to supply the universe with the deficit caused by the effective creation of a particle of matter. Over time this should whittle down the mass of these beasts until Milkomeda and the Fate of the Milky Way 319

FIG. 9.7 The evaporation of clusters then galaxies from now until 1023 years. In today’s universe globular clusters ( purple spheres ) are evaporating (red arrows ) as stars encounter one another and accelerate, or are ripped out of their parental clusters by galactic tides ( a ). Contemporane- ously, collisions between galaxies increase the mass of the Milky Way and others. In the future, accelerating expansion will prevent galaxies colliding and it will be the turn of galaxies to evaporate ( red arrows , b ). With no new star formation or acquisition, encounters between stars in each galaxy will steadily evaporate them from their homes. Rare galactic encounters within clusters may also strip some stars away through tidal forces. Those few stars that do hang on will be consumed by the growing black hole at their centres (c )

they evaporate. Now, there are a lot of questions here. For one, what exactly are black holes? Do they truly exist as objects with infinitely dense hearts and event horizons that shield our view of their interiors? At best, you can say maybe; but an infinitely dense, infinitely small object sounds a more than a bit like a sea monster or alien abduction. Not impossible, but certainly improbable. So, the eventual fate of black holes, whatever they truly are, remains 320 The Complex Lives of Star Clusters in serious doubt. Either way, Milkomeda is gone; its demise mir- roring that inflicted upon its globular clusters eons before.

Conclusions

There is a certain irony when we consider the fate of our galaxy. In the short term it continues to assemble while gravity plays a con- structive role, pulling matter ever more tightly together. However, over the longer term gravity acts in the opposite manner, driving stars further apart. As stars approach one another gravity acts as a slow-motion sling-shot propelling stars in different directions. Over time, lighter stars are boiled away from clusters, either through the action of tidal forces or through the steady process of evaporation. Most open clusters are shredded by tides acting across them in less than 1 billion years. Globular clusters are reduced by two-body interactions, while tides strip away those stars that have become exposed to the relentless tug of the galaxy. Over longer periods of time these same forces destroy galaxies. Within clusters tidal forces can pull galaxies apart or cause them to merge completely. What remains then falls victim to the same evaporative processes governed by two-body interactions. Long after the last red dwarf that was born in the next trillion years, has died, our galaxy—and every other galaxy—will die. Most of it will boil away into intergalactic space, while an unfortunate but very small percentage of stars will be consumed by our central super- massive black hole. The Sun’s long dead remains will most likely be separated from Milkomeda’s heart by over 100,000 light years after the collision that forms it. This means that it will be more than likely that the Sun’s cold remains will be lost through evaporation, rather than fall into the black hole at our galaxy’s centre. The Sun will simply lie too distantly to the core of Milkomeda to migrate there through two-body encounters. Whether you feel this cold fate preferable to something rather hotter is up to you. It’s a choice of grave versus crematorium. Either way, death is unavoidable. Although this may seem a rather grim way to conclude a book dedicated the universe’s most sparkling of jewels, it is the best portrait we have of our universe. It is all the more intriguing that Milkomeda and the Fate of the Milky Way 321 despite the great depth of time that separates us from our eventual fate, the processes that will decide it are visible for our perusal and dissection now. Astronomers are, in effect, the pathologists that are uncovering the bonds between our mode of construction and ultimately how we will pass. Entropy is the unseen hand. As the universe expands, more and more possibilities emerge. It is the exploration of these possibilities by Milkomeda’s stellar corpses, which will take them ever further from our galaxy into the dark void of future space. Glossary

Accretion disc A disc of material that fl ows onto a star or other compact object under the infl uence of gravity. Because the material arriving at the object carries angular momentum it rarely fl ows directly towards. Instead, friction causes the material to lose energy as it orbits the object, causing it to move ever inwards. Asymptotic giant branch (AGB) star An aging low or intermediate mass giant star that has consumed the hydrogen then helium in its core. Energy now comes from alternating waves of hydrogen and helium fusion, as strong stellar winds remove much of its mass. It is the loss of mass that ultimately destroys these stars. AGB-Manqué star A low mass and somewhat hotter giant star with a meager layer of hydrogen surrounding its degenerate carbon-oxygen core. Like more conventional AGB stars energy is produced by helium fusion in a shell that alternates with hydrogen fusion further out. AGB Manqué stars are believed to originate when red giant stars lose most of their outer hydrogen-rich layers earlier in their lives. Angular momentum A property of matter that is moving in a broadly circular path. Momentum is a product of the mass and its velocity (speed and direction) and, like energy, is always conserved. Anti-correlation When one property or variable changes in the opposite way to another. For example the more cigarettes you smoke the shorter your average lifespan will be. In globular clusters the more oxygen a star has the less sodium it has, and vice versa.

Be stars A type of intermediate or relatively massive star of spectral class B that is surrounded by a disc of material. These stars rotate much more rapidly than the Sun, and several are known to be shaped more like a disc (or the chocolate Minstrel ) Some of these stars are very young and still accreting matter, but the majority are losing mass in a broad disc that surrounds the equator of the star. Some Be stars are known as Shell stars . Beta (β) Cephei stars Fairly massive stars that pulsate with a short period measured in minutes to hours. Pulsations appear to be driven by alternating waves of ionization and recombination of iron (loss and gain of many of their electrons). Blue Straggler star A star produced by the collision of two or more other stars. The straggler is still on the main sequence after all of its siblings have become red giants. When these stars fi nally leave the main sequence they become fi rst yellow stragglers, then red straggler stars (below). Cataclysmic variable A binary system consisting of a low mass, hydrogen-burning star and a white dwarf. The white dwarf acquires material from its low mass companion. In the process the system emits bursts of energy either from the disc or from the surface of the white dwarf. CNO cycles In fairly massive stars energy is produced by hydrogen fusion that involves the addition of hydrogen nuclei in fours to a carbon seed nucleus. After four nuclei have been added, the product oxygen nucleus breaks back down to release the original carbon nucleus and a helium nucleus. The process releases energy at a far higher rate than the pp-chains that power low mass stars like the Sun. Conduction (of energy) Energy transfer between particles through collisions between them. In typical stars the relevant particles are usually electrons. Convection The transport of energy through the bulk motion of a gas or other fl uid. Hot material is less dense and rises; cold material is denser than its surroundings and sinks. Crossing time The time it takes a star to move under the infl u- ence of gravity through a cluster from one side to another. The Glossary 325

crossing time effectively sets the probability that two stars will productively encounter one another during their main sequence lives and merge or interact. The crossing time is very important in young clusters where collisions between massive stars could allow unusually massive stars to form. In older clusters it sets the time for lower mass star to have encounters which could lead to Type Ia supernovae. Cusp A particular distribution of stars within a cluster—typically an old globular cluster where the density of stars rises nearly continuously to the very centre of the cluster instead of rising to a plateau near the cluster core. Disc shocking A term describing the effect of a spiral galaxy disc on a globular cluster passing through it. The gravitational pull of the disc accelerates the stars in different directions as the cluster punches through the disc. This causes the stars to move faster relative to each other, which in turn causes the cluster to expand. Excretion disc A disc of material that is expanding away from a star. Be stars typically show this (above) but they may also form when two stars are spiraling in on one another. The outward movement of material in the disc balances the loss of angular momentum (above) of the two inward migrating stars. Extended clusters Faint and ancient clusters that have been seen in the halos of nearby galaxies but not the Milky Way. These have diameters similar or somewhat larger than typical globular clusters but appear to have similar masses. The half-mass radius (see below) is typically 3–5 times that of a globular cluster. Evaporation (of clusters) The eventual dissolution of a cluster of stars under the combined effect of the pull of surrounding stars and nebulae and the combined motion of the stars within the cluster. Stars acquire enough energy to reach the gravitational edge of the cluster where the pull of the cluster as a whole matches that of the surrounding galaxy. The process is equivalent to water evaporating from a pan of hot water where the particles have acquired sufﬁ cient energy to break the binds that link them to the rest of the liquid in the pan. 326 Glossary

Faint fuzzy Another type of newly discovered cluster with similar nearly identical properties to extended clusters. These have been spotted in the outskirts of neighboring galaxies but not ours. Gamma ray burst A burst of high energy radiation, dominated by gamma rays, released (most likely) when a black hole forms. They are divided into at least three groups: short; long and ultra- long. Short bursts last less than 0.2 s and release harder (higher energy) radiation. These are believed to occur when two neutron stars merge. Long bursts (lasting anything from 0.2 s to several minutes) are softer (less energetic) than short bursts, but appear to happen when the cores of very massive stars implode. Most (maybe all) are associated with the deaths of some but not all Wolf-Rayet stars. Ultra-long bursts are a new discovery, and last for several hours. They may be associated with the implosion of the core of a blue supergiant or luminous blue variable. Globular cluster A ball of 100,000 to a few million stars that is roughly 1–200 light years across. The half-mass radius is typically less than 10 light years and usually less than 5. In the Milky Way all globular clusters are more than 8 billion years old (although the Arches cluster may be analogous). In the LMC, SMC and many other galaxies much younger globular clusters (or globular-like clusters) are known with ages measured in as little as a few million years. Gravothermal collapse The process through which old star clusters, particularly globular clusters are believed to pass. Stars lose so much energy to neighbors that they fall towards the core of the cluster causing the density of stars to increase rapidly. Various effects then come into play, which prevent the cluster’s total implosion. Half-light radius If you look at an image of the cluster it can be divided into an inner and outer region. The half-light radius is the radius of the region from which half the light of the total cluster is emitted. This is a convenient method of determining the bulk property of the cluster including its size and density. Half-mass radius Like the half-light radius (above) the half-mass radius is the radius of an imaginary circle around the core of Glossary 327

the cluster that contains half the cluster’s overall mass. For old clusters, where the stars are low in mass, the half mass and half- light radii are effectively the same. In younger clusters, where there may be a skewed distribution of stars of various masses this may not be true. High mass X-ray binary (HMXB) A binary star system where a black hole or neutron star is paired with another star of high mass—typically more than ten times the mass of the Sun. Material from the stellar wind of the more massive stars is captured into a disc around the compact, dark remnant. These systems can be relatively large, measuring several million miles between the two objects. Keplerian disc A disc of material around (for example) a star that rotates in a relatively stable, elliptical orbit around the star. In globular clusters, many have two populations of stars. The most oxygen-rich (and sodium-poor—see anti-correlation, above) typically have Keplerian orbits around the cluster core, meaning that they are broadly elliptical but not one that is extremely different from a circular path. Low mass X-ray binary (LMXB) A small X-ray binary system consisting of a black hole, neutron star or white dwarf that is orbited closely by a low mass companion. Material ﬂ ows directly from the companion star’s surface onto the compact remnant through an accretion disc. Mass segregation The tendency for stars to separate within a cluster according to their mass. Gravitational interactions between stars cause the heaviest objects to settle towards the centre, while lighter stars are moved towards the outskirts of the cluster. Millisecond pulsar A generally old pulsar that spins at over 100 revolutions per second. These recycled pulsars are spun back up when they accrete material from a nearby companion. Many are partners in low mass X-ray binaries, while others have destroyed their partners and now live alone. The ﬁ rst planets discovered outside the solar system accompany the millisecond pulsar PSR B1257+12. 328 Glossary

Momentum (see angular momentum) Momentum is a property of any moving mass. It is calculated by multiplying the mass of the object (in kilograms) by its velocity (in meters per second). When two moving masses encounter one another—either through a direct collision or by interaction through their mutual gravity—momentum is always conserved. Nebula A cloud of gas and dust. Various types are known including planetary nebulae that are expanding and ﬂ uorescing clouds that are formed from dying red giants; molecular clouds are nebulae that contain predominantly molecular (diatomic) hydrogen. Neutrino A very low mass particle (in the lepton family of particles) emitted during some nuclear reactions. The neutrino balances the change in momentum of the particles involved in the reactions and is an important means through which stars lose energy. Nova A thermonuclear explosion that occurs on the surface of a white dwarf. Hydrogen, which has been accreted from a neighboring star, ignites and burns through the CNO cycles. During these reactions, unstable nuclei build up which undergo radioactive decay. This, further, heats the layer of burning fuel until burning becomes explosive, lifting the entire layer and part of the underlying dwarf off into space. Shocks within the expanding debris and the disc can generate gamma rays. Novae tend to repeat on timescales measured in thousands of years, but a small subset, called recurrent novae such as RS Ophiuci, repeat on decadal timescales. Open cluster A predominantly young cluster of several hundred, to several thousand, stars. There is no longer any hard and fast distinction between open and globular clusters, as many galaxies, including the LMC and SMC, have many large young clusters. A handful of very old open clusters are also known. Pair instability supernova A supernova from a star with 130–260 times the mass of the Sun that is powered by the creation of electron-positron pairs. Energetic gamma rays arise in the core of the star as the temperature exceeds 800 million Kelvin. Following Einstein’s dictate, E = mc 2 , many become electron- Glossary 329

positron pairs, which removes the support from the massive stellar core. This, then, collapses rapidly inwards. In the process carbon and oxygen within the core is fused to iron, releasing suffi cient energy to eviscerate the star. Pulsational pair instability supernova In somewhat less massive stars (95–130 times the mass of the Sun), pair instability is not so extreme that the energy released destroys the star. Instead, the shockwave blasts part of the star off into space at a few thousand kilometers per second. This reduces the temperature and rate of nuclear reactions in the core, allowing the star to “relax”. After pulsing the star may become unstable again if suffi cient mass remains in its core. These later pulses carry away less mass but have the same total energy so move faster. This causes them to catch up with the fi rst pulse. In doing so, much of the kinetic energy of the second pulse (or subsequent supernova) is released as light, powering a long-lasting, highly luminous display. Planetary nebula An expanding low mass cloud of highly ionized gas that surrounds a very young white dwarf star. The gas originates in the outer layers of a former red giant. Strong stellar winds from the giant blow off its envelope which is then heated and ionized by ultraviolet radiation coming from the now exposed stellar core (the proto-white dwarf). Such nebulae are fairly common but last only a few tens of thousands of years until they disperse into the surrounding space. Positron The antimatter equivalent particle of the electron. Positrons have the same mass but the opposite charge and spin of the electron. Primordial binary A binary system that was formed with the star cluster. Radiation (energy transfer by) The transfer of energy from place to place using electromagnetic radiation, rather than with particle motion. Red giant A cool but highly luminous star that is several tens to a few hundred times the diameter of the Sun. They are typically a few hundred to a few thousand times brighter than the Sun. 330 Glossary

Red giants are produced at the end of a star’s main sequence when its core runs out of hydrogen fuel and collapses. Red straggler A red giant star that lies above the main sequence turn off but is also too luminous to have been formed from the stars that are currently leaving the main sequence. These must be the descendents of blue straggler stars, produced when two or more stars collided and merged while still on the main sequence. Red supergiant A very large, cool red star with a very high luminosity, over 10,000 times that of the Sun. These are produced when the core of a much more massive star runs out of hydrogen fuel and then collapses. Ring cluster Odd, ring shaped collections of stars that are a few hundred million years old. These are not known in the Milky Way but are seen in the Magellanic Clouds. r-Process element An element produced by the rapid addition of neutrons to seed elements such as iron. These include the most massive elements in the periodic table. All elements that are more massive than lead-207 are produced through the r-process. The location of the r-process is unclear but almost certainly includes both the environment surrounding young neutron stars in core-collapse supernovae; and also short gamma ray bursts, where two neutron stars collide. rp (and p)-processes A series of nuclear reactions where (predominantly) unstable proton-rich isotopes of elements are produced. The most likely site for their formation is the surfaces of neutron stars undergoing violent hydrogen fusion (Type I X-ray bursts). Novae may also produce some of these elements (the p-process). Sawyer-Hogg classiﬁ cation system A system of classifying globular clusters based on their luminosity, distribution and the density of the stars contained. sdB star A blue sub-dwarf star that appear to be the stripped descendent of red giant stars. Some, but not all are helium burning stars; while many more have too little fuel to do this and will quietly evolve into helium-rich white dwarfs. Glossary 331

Shell star See Be star. Sodium-neon cycle An off-shoot of the CNO cycles where hydrogen is fused into helium. In these reactions, neon picks up four protons, forming fi rst sodium, before the reactions spit the original neon nucleus back out plus the product helium nucleus. s-Process element A series of reactions thought to occur most frequently in AGB stars. Here, a steady but slow supply of neutrons is added to smaller, seed nuclei. This builds them up in mass until they reach Lead-207. At this point the rate of radioactive decay exceeds the rate at which further assembly can occur and the process halts. Larger nuclei are made in a similar set of reactions, called the r-process (see above). Thorne-Żytkow object (TZO) A hypothetical object consisting of a red giant or supergiant with a neutron star-core. The two are brought together when the expanding red (super)giant swallows up the neutron star. Such an object may be the source of rp-elements (above). The lower mass TZOs are thought to die when the outer layers are blown off into space. More massive red supergiant TZOs likely die more violently when their cores implode to form black holes. In the accompanying eruption the black hole may betray its formation with an accompanying long GRB (above). Tidal shocking (See disc-shocking, above.) Trumpler classifi cation system A classifi cation system for open clusters that takes into account the number of stars, their distribution in the cluster and the presence or absence of accompanying nebulosity. Two-body relaxation time In terms of the death of dense (rich) clusters, including globular clusters, this is perhaps the defi ning force. The two-body relaxation time is the time it takes a star to change its velocity (its direction) by 90°. Although sounding rather abstract, it defi nes the number of interactions a star is experiencing on its travels through the cluster. In doing so, the two-body time explains how the velocity of stars becomes randomized through the exchange of momentum between them. In 332 Glossary

the process light stars tend to be accelerated outwards as they gain kinetic energy and momentum, while higher mass stars fall inwards. For globular clusters the two-body time is measured in tens to hundreds of millions of years. For an elliptical galaxy, it is far longer than the current age of the universe. Type Ia supernova An explosion which disrupts a white dwarf in its entirety. These are most likely driven by the white dwarf gaining mass from a companion (or through a collision with another white dwarf) until it approaches and in many cases exceeds the Chandrasekhar mass limit of 1.382 solar masses. It is now apparent that a variety of masses of white dwarf will explode—from 0.9 to nearly double the maximum permitted Chandrasekhar mass, 2.1 solar masses. The variation depends on the type of system in which the white dwarf lives. A few of these supernovae are known as “.Ia” supernovae as they are a tenth as bright as conventional Type Ia events. Type II supernova A type of supernova where the core of a massive supergiant star implodes to form a neutron star (or more rarely a black hole). The spectrum of the explosion reveals abundant hydrogen from the supergiant’s outer layers. When the light output of the supernova is plotted against time, many Type II supernovae also show a plateau in brightness caused by the steady recombination of hydrogen nuclei with neighboring electrons (Type II-P supernovae). Ultra-compact dwarf A small galaxy with a diameter more like a globular cluster (2–400 light years). The stellar density is very high but the half-mass radius is larger. Many UCDs appear to be the stripped-down remnants of larger galaxies that have undergone fatal encounters with larger elliptical and spiral galaxies. Like their much larger elliptical cousins some contain super- massive black holes. The small galaxy M32, in Andromeda, is a close relative, but somewhat larger than most UCDs. Velocity dispersion A measure of the spread in velocity of stars in a cluster. As stars mill around the core of the cluster they have different velocities to one another. This spread is super-imposed upon the velocity of the bulk cluster as it moves through space. Glossary 333

The spread is called the dispersion around the mean velocity of the cluster as a whole. Violent relaxation A thoroughly messy adventure where a young star cluster, rich in massive stars undergoes a process of “realignment”. As massive stars die and expel much of their mass, the distribution of mass changes rapidly, which then violently alters the gravitational fi eld of the cluster. Remaining stars are forced to redistribute themselves within the cluster in order to re- balance the forces acting upon them. In the process any original structure within the cluster is destroyed. The process occurs over the fi rst 5–20 million years as the most massive members of the cluster die. Virialization The process through which the kinetic energy of every star within a cluster (or a galaxy) becomes randomized. Through two-body encounters, stars exchange kinetic energy and momentum with one another causing the appearance of the system to smooth out. White dwarf A small, earth-sized star which does not produce any more energy through nuclear reactions. These stars are supported by electron degeneracy—a repulsive effect between electrons confi ned in very dense environments. They are all the end-products of red giant stars that have completely exhausted their fuel and shed their outermost layers. White dwarfs with less than 44 % the mass of the Sun are made primarily of helium. Those more than a Sun’s mass are mostly made of oxygen, neon and magnesium. Those with intermediary masses (the most abundant currently) are made mostly of carbon and oxygen. X-ray binary A binary system consisting of a small, dense star, such as a white dwarf, neutron star or black hole, paired with another, more conventional star. See low mass X-ray binary and high-mass X-ray binary. X-ray pulsar A binary star system where a neutron star is paired with another (usually high mass) star. Material from the companion streams onto the magnetic poles of the neutron star generating a hot spot which emits X-rays. As the neutron star 334 Glossary

rotates, this hot spot swings into and out of view, generating apparent pulses in the emission. Yellow straggler A yellow star, often a sub-giant, which lies above the main sequence turn-off of a cluster. These are the descendents of blue straggler stars. Further Reading

To improve accessibility to the work that this book is based upon, where possible ArXiV preprints (or equivalent) are provided as references.

Chapter 1—Historical Perspectives

Andres Almeida, A., Forbes, D. A., Spitler, L. R. and Vincenzo Pota, V. Extended Star Clusters in NGC 1023 from HST/ACS Mosaic Imaging. http://www.eso.org/sci/ meetings/2014/FESC14/Andres_Almeida.pdf (2014). This is a more recent refer- ence from the one I used in the body of the text, but is very clear and covers the most recent developments. Brodie, J.P., Romanowsky, A.J., Strader, J. Forbes, D. A. The relationships among compact stellar systems: a fresh view of ultra compact dwarfs. ArXiV http://arxiv.org/ abs/1109.5696 (2011) Andreas Burkert, A., Jean Brodie, J. and Soeren Larsen, S. Faint Fuzzies and the Formation of Lenticular Galaxies. http://arxiv.org/abs/astro-ph/0504064 (2005) King, I. The observational approaches to populations in Globular clusters, in “Globular Clusters”, editors C. Martinez Roger, I. Pérez Fournon and F. Sánchez, Cambridge University Press, ISBN 0 521 77058 0. (1999) Majewski, S. R. Stellar Populations and the formation of the Milky Way, in “Globular Clusters”, editors C. Martinez Roger, I. Pérez Fournon and F. Sánchez, , Cambridge University Press, ISBN 0 521 77058 0. (1999) Penny, S. J., Forbes, D. A., Strader, J., Usher, C., Brodie J.P. and Romanowsky, A. J. Ultra compact dwarfs in the Perseus Cluster: UCD formation via tidal stripping. http://arxiv.org/abs/1402.0687 (2014)

Chapter 2—Adventures in Stellar Evolution

Bate, M. and Bonnell, M. The following link takes you to the University of Exeter’s site covering the modelling of the formation of stars. http://www.astro.ex.ac.uk/ people/mbate/Cluster/ Castellani, V. Globular Clusters as a test for stellar evolution, in “Globular Clusters”, editors C. Martinez Roger, I. Pérez Fournon and F. Sánchez, Cambridge University Press, ISBN 0 521 77058 0. (1999)

King, C. R., Da Costa G. S. and Demarque, P. The luminosity function on the sub- giant branch of 47 Tucanae: a comparison of observation and theory. The Astrophysical Journal, 299, 674-682 (1985) Wu, B. The red giant branch bump and its many uses. This is a pdf ﬁ le, but I have been unable to ﬁ nd it again on the internet, despite being able to track its author. (2012)

Chapter 3—Variable Stars in Clusters

Cassinelli, J. P., Brown, J. C. Maheswaran, M. Miller, N. A. and Telfer, D. C. a magnetically torqued disk model for Be stars. Astrophysical Journal, 578:951–966 (2002) Edmonds P. D. and Gilliland, R. L. K Giants in 47 Tucanae: Detection of a New Class of Variable Star. The Astrophysical Journal, 464:L157-L160 (1996) Feast, M.W Glass, I.S. Whitelock, P.A. A Period-luminosity-colour relationship for Mira Variables. MNRAS 241, 375-392 (1989) Feast, M.W. Pulsating Stars in Globular Clusters and their use, in “Globular Clusters”, editors C. Martinez Roger, I. Pérez Fournon and F. Sánchez, , Cambridge University Press, ISBN 0 521 77058 0. (1999) Hellier, C. Cataclysmic Variable Stars Springer-Praxis, ISBN: 1-85233-211-5 (2001) Whitelock, P. A. The Pulsation Mode and period-luminosity relationship of cool variables in globular clusters. MNRAS, 219, 525-536 (1986) Whitelock, P.A., Pottasch, S.R., Feast M.W. Late stages in stellar evolution, in Kwok, S. and Pottasch, S.R. eds., Springer, I SBN: 978-94-010-8196-2 (1987) Additional information on speciﬁ c variable stars was found on the following Wikipedia pages: ZZ Ceti Stars: http://en.wikipedia.org/wiki/Pulsating_white_dwarf ; Mira Variables: http://en.wikipedia.org/wiki/Mira_variable ; Semi-regular variables: http://en.wikipedia.org/wiki/Semiregular_variable_star ; β-Cephei Variables: http:// en.wikipedia.org/wiki/Beta_Cephei_variable ; δ-Scuti Variables: http://en.wikipedia.org/wiki/Delta_Scuti_variable ; RR Lyrae Variables: http://en.wikipedia.org/ wiki/RR_Lyrae ; RV Tauri Variables: http://en.wikipedia.org/wiki/RV_Tauri_vari- able ; SX Phoenicis variable: http://en.wikipedia.org/wiki/SX_Phoenicis_variable ; Type II Cepheid Variables: http://en.wikipedia.org/wiki/Type_II_Cepheid

Chapter 4—Globular Cluster Formation

Campbell, S. W. et al Sodium content as a predictor of the advanced evolution of globular cluster stars. Nature 498, 198-200 (2013) Dupree, A. K., Strader, J., Smith, G. H. Direct Evidence for an Enhancement of Helium in Giant Stars in Omega Centauri. http://arxiv.org/pdf/1012.4802.pdf (2010) Gratton, R. Early nucleosynthesis and chemical abundances of stars in globular clusters, in “Globular Clusters”, editors C. Martinez Roger, I. Pérez Fournon and F. Sánchez. Cambridge University Press, ISBN 0 521 77058 0. (1999) Gratton, R. G. Carretta, E. Bragaglia, A. Multiple populations in globular clusters: Lessons learned from the Milky Way globular clusters. http://arxiv.org/ abs/1201.6526 (2012) Larsen, S. S., Strader, J. and Brodie, J. P. Constraints on mass loss and self-enrichment scenarios for the globular clusters of the Fornax dSph. http://arxiv.org/ pdf/1207.5792.pdf (2012) Further Reading 337

Maxted, P. F. L. Heber, P U.. Marsh T R. and North R. C. The binary fraction of extreme horizontal branch stars. MNRAS, http://mnras.oxfordjournals.org/content/326/4/1391.full.pdf (2001) Milone, A. P., Marino, A. F. Piotto, G., Bedin, L. R., Anderson, J., Aparicio, A., Cassisi, S. and Rich R. M. A double main sequence in the globular cluster NGC 6397. The Astrophysical Journal, 745:27 (2012) Newsham, G. The Horizontal Branch as a probe of stellar population history. http:// www.astronomy.ohio-state.edu/Dept/Dissertations/Newsham.pdf (2007) Tosi, R. D., Cignoni, M., Matteucci, M., Pancino F. E. and Bellazzini, M. On the origin of the helium-rich population in ω Centauri. MNRAS. 401, 2490-2498 (2010)

Chapter 5—Open Clusters

Portegies Zwart, S. F., Hut, P., Makino, J. and McMillan, S. L. W. On the dissolution of evolving star clusters, http://arxiv.org/pdf/astro-ph/9803084.pdf (1998) Portegies Zwart, S. F. Makino, J. McMillan, S. L. W. and Hut, P. How many young star clusters exist in the Galactic center? http://cds.cern.ch/record/455814/ ﬁ les/0008490.pdf and The Astrophysical Journal, 546:L101-L104 (2001) Portegies Zwart, S. F. Hut, P., McMillan S. L. W. and Makino J. Star cluster ecology V: Dissection of an open star cluster—spectroscopy. http://adsabs.harvard.edu/ full/2004MNRAS.351..473P (2004)

Chapter 6—Stellar Soap Operas

Babu, G.J., Chattopadhyay, T., Chattopadhyay, A. K. and Mondal S. Horizontal branch morphology of globular clusters: a multivariate statistical analysis. The Astrophysical Journal, 700:1768–1778 (2009) Baumgardt, H., Hut, P., Makino, J., McMillan, S., Portegies-Zwart, S. On the central structure of M15 The Astrophysical Journal, 582:L21-L24 (2003) Bachetti, M. et al An ultraluminous X-ray source powered by an accreting neutron star. Nature 514, 202-204 (2014) Sergio Campana et al Indirect Evidence of an Active Radio Pulsar in the Quiescent State of the Transient Millisecond Pulsar SAX J1808.4-3658. The Astrophysical Journal, 614:L49-L52, (2004) Simon Clark’s work on CXOU J164710.2-455216 can be found at: http://www.open. ac.uk/research/main/news/ou-helps-unlock-mystery-supernova-explosions-and- black-holes (2012) Chomiuk L. et al , Binary orbits as the driver of X-ray emission and mass ejection in classical novae. Nature 514, 339–342 (2014) Fiorentino G., b. Lanzoni, B. Dalessandro, E. ferraro, F. R., . Bono, G. Marconi, M. Blue straggler masses from pulsation properties. I: the case of NGC6541. http://arxiv. org/abs/1312.0388 (2013) Fujii, M. S. and Portegies Zwart, S. The growth of massive stars via stellar collisions in ensemble star clusters http://export.arxiv.org/pdf/1210.3732 (1988) Hurley J. R. and Shara, M. M. The Promiscuous Nature of Stars in Clusters The Astrophysical Journal, 570:184-189 (2002) Lifang Li, L. Zhang, F., Han, Z., Jiang D. and Jiang, T. The evolutionary status of W Ursae Majoris-type systems http://xxx.tau.ac.il/abs/0805.4040 (ArXiV) (2009), 338 Further Reading

Lyman, J. D. Levan, A. J. Church, R. P. Davies, M. B. Tanvir, N. R. The progenitors of calcium-rich transients are not formed in situ. http://arxiv.org/abs/1408.1424 (2014), Masgoret R. X-ray sources in globular clusters, in “Globular Clusters”, editors C. Martinez Roger, I. Pérez Fournon and F. Sánchez, Cambridge University Press, ISBN 0 521 77058 0. (1999) Justyn R. Maund, J. R, Smartt, S. J., Kudritzki, R. P., Podsiadlowski, P. and Gilmore, G. F. The massive binary companion star to the progenitor of supernova 1993J. Nature, 427, 129-131 (2003) Maxted, P. F. L., Heber, P U. Marsh, T. R. and North, R. C. The binary fraction of extreme horizontal branch stars, MNRAS, http://mnras.oxfordjournals.org/ content/326/4/1391.full.pdf (2001) Pasquato, M. Raimondo, G. Brocato, E., Chung, C., Moraghan, A., and Lee, Y-W., Core collapse and horizontal-branch morphology in galactic globular clusters. http:// arxiv.org/abs/1305.1622 (2013) A review of Philipp Podsiadlowski’s work on the Eta Carinae system can be found at: www.astro.physics.ox.ac.uk/~podsi/binaries.pdf Portegies Zwart, S. F., Makino, J., McMillan, S. L. W. and Hut, P. Star cluster ecology III: Runaway collisions in young compact star clusters http://arxiv.org/pdf/astro- ph/9812006v2.pdf (1999). Portegies Zwart, S. F., Baumgardt, H., Hut, P., Makino, J. & McMillan S. L. W. Formation of massive black holes through runaway collisions in dense young star clusters. Nature, 428, 724-726 (2004) A PDF covering Robert Quimby’s observations of SN 2006gy can be found at: http:// member.ipmu.jp/robert.quimby/quimby_lsn_talk.pdf Shara, M. M. and Hurley, J. R. Star clusters as type Ia supernova factories, The Astrophysical Journal 571:830–842 (2002) Trenti, M., Ardi E., Mineshige S. and Hut, P. Star Clusters with Primordial Binaries: III. Dynamical Interaction between Binaries and an Intermediate Mass Black Hole. http://arxiv.org/abs/astro-ph/0610342 (2006) Sandquist, E. I. and Shetrone M. D. Time series photometry of M67: W Ursae Majoris systems, blue stragglers, and related systems. The Astronomical Journal, 125:2173- 2187, (2003) with a preprint available at: http://www.as.utexas.edu/~shetrone/ images/ﬁ le6.pdf Thorne, K. and Zytkow, A. Stars with degenerate neutron cores. I - Structure of equi- librium models. The Astrophysical Journal 212 (1): 832–858 (1977) Vesperini E. Star cluster dynamics http://rsta.royalsocietypublishing.org/content/ roypta/368/1913/829.full.pdf (2009) Voss, R. The review of the distribution of low mass x-ray binaries in the Milky Way, M31 and Centaurus A, by Rasmus Voss (Max Planck Institute for Extraterrestrial Studies), can be found at: https://www.imprs-astro.mpg.de/sites/default/ﬁ les/2003- Voss-Rasmus.pdf (2003) Washabaugh, P. C., and Bregman, J. N. The Production Rate of SN Ia Events in Globular Clusters: http://arxiv.org/abs/1205.0588 (2012) S. E. Woosley, S. E., Blinnikov, S. and Heger, A. Pulsational pair instability as an explanation for the most luminous supernovae Nature 450, 390-392 (2007) Yaron, O., Kovetz, A. and Prialnik, D. Hot Helium Flashers – The Road to Extreme Horizontal Branch Stars. The Art of Modelling Stars in the 21st Century Proceedings IAU Symposium No. 252, 2008 L. Deng& K.L. Chan, eds. (2008) Yıldız, M. Origin of WUMa-type contact binaries - age and orbital evolution. (2013) http://arxiv.org/abs/1310.5526 Further Reading 339 Chapter 7—The Complex Life of Globular Clusters

Alamo-Martínez, K. A. et al The rich globular cluster system of Abell 1689 and the radial dependence of the globular cluster formation effi ciency. http://arxiv.org/ abs/1308.1958 (2013) Elson, R. A. W. Stellar Dynamics in Globular Clusters, in “Globular Clusters”, editors Martinez Roger, C. Pérez, I. Fournon and Sánchez, F. Cambridge University Press, ISBN 0 521 77058 0 (1999) Grebel, E. K. The Star Clusters of the Magellanic Clouds. https://astro.uni-bonn. de/~sambaran/DS2014/Modest14…/Grebel.pdf (2014) Hurley, J. R., Aarseth S. J. and Shara M. M. The core binary fractions of star clusters from realistic simulations http://arxiv.org/pdf/0704.0290.pdf (2007) Maxted, P. F. L., Heber, P U., Marsh, T. R. and North, R. C. The binary fraction of extreme horizontal branch stars. http://arxiv.org/pdf/astro-ph/0103342.pdf (2001) Mackey, A. D. and Gilmore, G. F. Surface brightness profi les and structural parameters for 10 rich stellar clusters in the Small Magellanic Cloud. http://arxiv.org/abs/ astro-ph/0209046 (2002) Motch, C. Pakull, M. W. Soria, R., Grisé, F. & Pietrzyński, G. A mass of less than 15 solar masses for the black hole in an ultraluminous X-ray source. Nature 514, 198- 201. (2014) Pasham , D.R., Strohmayer, T. E. & Mushotzky, R. F. A 400-solar-mass black hole in the galaxy M82 Nature 513,74–76 (2014) Portegies Zwart, S. F., Hut, P. & Verbunt, F. Star Cluster Ecology I A Cluster Core with Encounters between Single Stars http://arxiv.org/pdf/astro-ph/9701042.pdf (1997) Portegies Zwart, S. F., Makino, J., McMillan, S. L. W. and Hut, P. The lives and deaths of star clusters near the Galactic center http://arxiv.org/pdf/astro-ph/0102259.pdf (2001) Portegies Zwart, S. F., McMillan, S. L. W., Hut, P. and Makino, J. Star cluster ecology IV: Dissection of an open star cluster—photometry MNRAS, 321 (2002) Portegies Zwart, S. F., Makino, J., McMillan, S. L W. and Hut, P. Star cluster Ecology V: How many young star clusters exist in the Galactic center? http://arxiv.org/abs/ astro-ph/0301041 (2003) Romanowsk, A. R. The ongoing assembly of a central cluster galaxy: phase-space sub- structures in the halo of M87 The Astrophysical Journal. Article. Vol. 748, 29 (2012) Stetson, P. B. The Center of the Core-Cusp Globular Cluster M15: CFHT and HST Observations, Publications of the Astronomical Society of the Pacifi c 106: 250-280 (1994) Strader, J. et al Wide-fi eld precision kinematics of the M87 globular cluster system. http://arxiv.org/abs/1110.2778 (2011)

Chapter 8—One Thousand Rubies in the Sky

Anglada-Escudé, G. et al Two planets around Kapteyn’s star: a cold and a temperate super-Earth orbiting the nearest halo red-dwarf. http://arxiv.org/abs/1406.0818 (2014) Gorby Y. A. et al Electrically conductive bacterial nanowires produced by Shewanella oneidensis strain MR-1and other microorganisms. PNAS 103, 11358–11363 (2005) 340 Further Reading

Zuo, Y., Xing, D., Regan, J. M. and Logan, B. E. Isolation of the Exoelectrogenic Bacterium Ochrobactrum anthropi YZ-1 by Using a U-Tube Microbial Fuel Cell Applied and Environmental Microbiology,74, 3130–3137 (2008) M4 Pulsar Planet: http://en.wikipedia.org/wiki/PSR_B1620-26_b

Chapter 9—Milkomeda and the Fate of the Milky Way

Stellar Dynamics in Galaxies www.maths.qmul.ac.uk/~wjs/MTH726U/chap2.pdf Besla, G., Kallivayalil, N., Hernquist, L., van der Marel, R. P., Cox, T.J. Kereš, D. Simulations of the Magellanic Stream in a First Infall Scenario. http://arxiv.org/ abs/1008.2210 (2010) Bonaca, A., Geha, M., and Kallivayalil, N. A cold Milky Way stellar stream in the direction of Triangulum http://arxiv.org/abs/1209.5391 (2012) Brown, T. M., Ferguson, H. C., Smith, E., Kimble, R. A., Sweigart, A. V., Renzini, A. Rich, R. M. and VandenBerg, D. A. Evidence for a signiﬁ cant intermediate-age population in the m31 halo from main sequence photometry. The Astrophysical Journal, 592 L17-L20 (2003) Chandar, R., Puzia, T. H., Sarajedini, A. and Goudfrooi, P. A genuine intermediate-age globular cluster in M33 http://arxiv.org/pdf/astro-ph/0606419.pdf (2006) Cox T. J. and Loeb, A. The Collision Between The Milky Way And Andromeda http:// arxiv.org/abs/0705.1170 (2008) Robotham A.S.G. et al Galaxy And Mass Assembly (GAMA): Galaxy close-pairs, mergers, and the future fate of stellar mass. http://arxiv.org/abs/1408.1476(2014 ) Shattow, G. and Loeb, A. Implications of Recent Measurements of the Milky Way Rotation for the Orbit of the Large Magellanic Cloud. http://arxiv.org/pdf/0808.0104. pdf (2008) van der Marel, R. P., Besla, G., Cox, T. J. Sohn, S. T. and Anderson, J. The M31 Velocity Vector. III. Future Milky Way-M31-M33 Orbital Evolution, Merging, and Fate of the Sun, http://arxiv.org/abs/1205.6865 (2012) Wolfe, S. A., Pisano, D. J. Lockman, F. J., McGaugh S. S., & Shaya, E. J. Discrete clouds of neutral gas between the galaxies M31 and M33 Nature 497, 224-226 (2013) Index

A Crossing time, 221–225, 230, 244, Accretion disc, 81, 93, 94, 97, 164, 246–248, 269, 316, 324–325 167, 176, 202, 203, 205, 206, 208, 211, 323 AGB-Manqué , 52, 71, 73, 323 D Anti-correlation, 109, 110, 130, 172, 306, Delta Scuti star, 59, 63–64 307, 323 Disc shocking, 170, 231–232, 234, 325, 331 Arches cluster, 15, 148, 149, 225, 226 DQ Hercules, 94

B E Be star , 65, 79, 83–87, 200, 201, 324 EHB. See Extreme horizontal branch (EHB) Binary star, 77, 92, 98, 155–157, 159, Elliptical galaxy, 16, 18, 19, 22, 106, 107, 162, 164, 167, 168, 171, 179, 132, 208, 228, 229, 250–256, 260, 182–184, 201, 221, 226, 233–235, 304, 308–312, 314–316 237, 263 Eta Carinae , 91, 160, 161, 192, 193, 196, Binary system, 30, 31, 52, 85, 92, 98, 199–201 104, 128, 129, 145, 155, 156, 158, Evaporation (of clusters), 234, 235, 160–165, 168, 170, 174, 184, 186, 257–261, 265, 289, 319, 320, 325 195, 197, 200–213, 226, 228, Excretion disc, 167, 325 233–237, 244, 257, 260, 261, 263, Extended cluster. See Faint Fuzzy 264, 295 Extreme horizontal branch (EHB), 51, 52, Black Widow pulsar, 209 73, 109, 117, 124–131, 135, 164, Blazhko effect, 69 167, 168, 172–174, 228, 242, 262, Blue Hook , 8, 9, 49 263, 314 Blue straggler, 64, 86, 145, 163–171, 173, 177, 182, 194, 198, 199, 218, 242, 244, 247, 324 F Faint Fuzzy , 16–20, 25, 122, 325, 326

C Cataclysmic variable, 91–98, 164, 202, G 213, 237, 324 Globular cluster, 2–6, 9, 14–23, 29, Cepheid variable, 20–22, 44, 57, 59, 61, 32–36, 44, 46–51, 55, 57, 61, 72, 80, 83 63–70, 73, 75, 76, 92, 98, 99, CN (CNO) cycle, 37, 110, 186, 324 101–135, 137–142, 144, 147, Color magnitude diagram, 3, 5–11, 13, 149–153, 155, 156, 159, 161–166, 49, 69, 115, 119, 126 168, 169, 172, 185, 190, 204, 208, Core collapse , 38–40, 42, 43, 55, 88, 111, 212–219, 221–265, 267, 284, 285, 187, 190, 191, 195, 198, 217, 226, 287–293, 295, 299, 304, 307, 315, 233–237, 243–245, 261 316, 319, 320, 326

GRAPE-6 , 163, 165, 178, 215, 218 M67, 140, 148, 150–152, 168, 182, 293 Gravothermal collapse, 234, 326 M80, 14, 164 M82, 15, 32, 131, 132, 143, 206, 245–250, 255 H M87, 19, 53, 133, 250–253, 256, 260, Half-light radius, 16–18, 106, 107, 308, 310, 314 223, 326 Magellanic Clouds, 15, 16, 23, 28, 63, Half-mass radius, 17, 101, 143, 223, 228, 231, 242, 244, 264, 298, 300, 301, 246, 260, 326 304–306, 310–314 Herbig Ae/Be star, 79–83 Magellanic Stream, 300, 301 Hertzsprung-Russell diagram, 3, 5–11 Mergeburst, 178, 199 High mass x-ray binary (HMBX), 183, MGG 9 , 132, 143, 245–250 202, 327, 333 MGG 11, 132, 143, 245–248 High velocity cloud (HVC), 300–303 Milky Way, 1, 14, 15, 19–23, 32, 33, 36, HMBX. See High mass x-ray binary 55, 63, 64, 69, 98, 103, 105, 106, (HMBX) 108, 111, 113, 122, 123, 131, 138, Hodge 11 , 70, 168, 238, 241–242 140, 142, 152, 159, 160, 162, 169, Horizontal branch, 10, 11, 51, 61, 66–72, 170, 185, 208, 217, 224, 225, 229, 89, 126, 129, 135, 242, 243 231, 235–238, 240, 242–246, 250, HVC. See High velocity cloud (HVC) 251, 256, 258–260, 264, 265, 289, HVGC-1, 252, 253 292, 294, 297–321 Hyades , 1, 139, 140, 145, 151, 152 Millisecond pulsar, 94, 159, 207, 209, 210, 212, 213, 237, 285, 287, 327 Mira variable , 59, 61, 67, 73–76 K Kapteyn’s star, 288–291 N NGC 188 , 156, 167 L NGC 330 , 86, 198, 238 Large Magellanic Cloud (LMC), 15, 16, NGC 604 , 15, 139, 142–144, 242, 243 28, 31, 70, 73, 91, 122, 130, 132, NGC 3603, 138, 139, 142, 160, 201, 138, 149, 160, 168, 194, 224, 224–226 237–242, 256, 260, 298, 300, 301 Nova, 89, 157, 210, 211, 328 LBV. See Luminous blue variable (LBV) LMBX. See Low mass x-ray binary (LMBX) O LMC. See Large Magellanic Cloud (LMC) OH/IR variables, 75, 76 Low mass x-ray binary (LMBX), 92, 159, Omega (ω) Centauri, 18, 105–109, 114, 202, 203, 207–209, 211, 236, 244, 117, 119, 264, 290 327, 333 Oosterhoff Dichotomy, 69 Luminous blue variable (LBV), 39, 89–91, Open cluster, 1–4, 14–20, 22, 28, 51, 57, 160, 161, 191–194, 199, 201 61, 68, 72, 73, 84, 102, 104, 137–153, 155, 156, 159–162, 164, 166–168, 173, 182, 184, 201, 215, M 217, 218, 222, 223, 228, 232, 237, M15, 46, 69 239, 257, 261, 265, 283, 284, 293, M22, 2, 33 295, 320, 328 M30, 164 M31, 16, 17, 57, 72, 208, 242, 244, 245, 260, 297–299, 302–313 P M32, 298, 304, 305 Palomar 5 , 147, 257, 258 M33 (Triangulum), 15, 142, 143, Palomar 12, 147, 148 242–245, 259, 298, 299, 302–306, Palomar 13, 169, 170, 258, 259 308, 310–314 Pleiades (M45) , 84 M54, 83, 105, 106, 108, 137 PSR B1620-26 , 207, 213, 285, 287, 293, 294 Index 343

R T R136, 28, 142, 145, 149, 198, 224, 226, 30 Doradus , 15, 28, 31, 54, 91, 98, 132, 229, 237, 238, 240, 243 142, 160, 194, 198, 224, 241, 243 R136a, 194, 198, 224, 229 Thorne-Zytkow object (TZO), 174–178, R144, 160 209, 210, 331 Ram stripping, 315 Tidal forces, 107, 147, 148, 150, 232, Recurrent novae, 95, 97 253, 255, 257, 259, 260, 293, 305, Red giant, 6, 10, 11, 16, 43, 45, 46, 48–55, 319, 320 65–68, 71–73, 76, 92, 93, 95, 97, Tidal shocking. See Disc shocking 99, 102, 105, 111, 112, 114, 117, T Tauri star, 63, 79–82 124–129, 137, 138, 146, 152, 157, Two-body relaxation time, 227–231, 260, 158, 163–165, 167–170, 172–176, 269, 283, 293, 316, 331 178, 180, 182, 187, 209, 212, 213, Type Ia Supernova, 96, 97, 102, 179, 191, 217, 224, 228, 236, 241, 242, 262, 215–218, 317, 332 263, 267, 286, 287, 294, 311, 329 Type II Supernova, 35, 55, 102, 111, Red straggler, 169, 182, 330 182–184, 191, 217, 332 Red supergiant, 38–40, 80, 83, 85, 87–89, TZO. See Thorne-Zytkow object (TZO) 161, 174, 175, 183, 184, 198, 245, 330 Rho Cassiopeia, 79, 88–90, 193 Ring cluster, 240–241 U Roche Lobe , 92, 93, 181, 187, 203, 207, 212 Ultra compact dwarf (UCDs), 18, 19, RR Lyrae variable, 21, 22, 57, 59, 61, 62, 101, 105–107, 315, 332 68–71, 242 RV Tauri star, 59, 61, 76–77 V Velocity dispersion, 222, 223, 244–246, S 258, 259, 303, 332–333 Sagittarius Dwarf, 105, 106, 299, 300 Violent relaxation, 144–146, 186, SAX J1808.4-3658 , 212 225–227, 230, 234, 333 sdB star. See Sub-dwarf B (sdB) star Virialization, 228, 229, 314, 333 Semi-regular (SR) variable, 59–61, 66–67, 73, 76, 80, 87 Shell star. See Be star W Small Magellanic Cloud (SMC), 73, 86, Westerlund-1, 15, 184–189, 198 122, 174, 198, 200, 237–240, 242, W Ursa Majoris (W UMa) variable, 256, 298, 301 179–183 SO galaxy , 15, 17, 315 W Virginis star, 59, 61, 72–73, 77 SR variable. See Semi-regular (SR) variable Stellar association, 3, 142, 143, 151, 243 X Sub-dwarf B (sdB) star, 52, 128, 164, 168, X-ray bursts, 203, 211, 247, 249 169, 174, 217, 242, 330 X-ray pulsar, 94, 159, 209, 237, 333 Supernova SN 1987A, 183, 184, 198, 200, 201 Supernova SN 1993J, 182–185, 198 Y Supernova SN 2006gy, 41, 189–200, 224 Yellow straggler, 169, 334 Supernova SN 2006tf, 41, 191, 192, 194, 224 Supernova SN 2008am, 194 Z SX Phoenicis, 64 ZZ Ceti variable, 59, 61, 71, 77–80