Circumstellar Media in Late Stages of Stellar Evolution (Proceedings of the 34th Herstmonceux Conference, Held in Cambridge, July 12-16, 1993)

In the last throes of their lives, how do low- and high-mass stars interact with their immediate surroundings? How do a...

Author: R. E. S. Clegg | I. R. Stevens | W. P. S. Meikle

3 downloads 250 Views 7MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

In the last throes of their lives, how do low- and high-mass stars interact with their immediate surroundings? How do asymmetric stellar winds and the circumstellar medium affect the shape of a nebula? How are supernovae affected by a dense medium? And what do we understand of how stellar winds interact with their environments? These and many other exciting issues are addressed in these proceedings, from the 34th Herstmonceux conference held in Cambridge. Highlights of developments covered include the latest observations (including those with the Hubble Space Telescope) of stellar ejecta in planetary nebulae, novae, ring nebulae and supernovae, and a unified view of the physical processes involved; as well as the latest results on the media around supernovae 1987A and 1993J. This timely volume provides review articles that serve both as an excellent introduction for graduate students, and a handy reference for researchers; and up-to-date research papers for those who want to keep abreast of developments in the field.

Circumstellar Media in the Late Stages of Stellar Evolution

Circumstellar Media in the Late Stages of Stellar Evolution Proceedings of the 34th Herstmonceux conference, held in Cambridge, July 12-16, 1993

Edited by R. E. S. CLEGG Royal Greenwich Observatory I. R. STEVENS University of Birmingham W. P. S. MEIKLE Imperial College, London

CAMBRIDGE UNIVERSITY PRESS

Published by the Press Syndicate of the University of Cambridge The Pitt Building, Trumpington Street, Cambridge CB2 1RP 40 West 20th Street, New York, NY 10011-4211, USA 10 Stamford Road, Oakleigh, Melbourne 3166, Australia © Cambridge University Press 1994 First published 1994 Printed in Great Britain at the University Press, Cambridge A catalogue record for this book is available from the British Library Library of Congress cataloguing in publication data available

ISBN 0 521 46551 6 hardback

Contents

Preface

page ix

Conference Photograph

x

Conference Participants

xiii

Part one: Stellar Evolution and Wind Theory Evolution of massive stars N. Lunger Evolution of AGB stars P. R. Wood Hot star winds J. E. Drew Axisymmetric outflows from single and binary stars M. Livio Flows in clumpy CSM J. E. Dyson & T. W. Hartquist

1 15 27 35 52

Part two: Wolf-Rayet Ring Nebulae Ring Nebulae around LBVs and WR, stars L. J. Smith WR stars in the LMC T. A. Lozinskaya el al. WR Shell Nebulae R. Dufour Three-wind model for WR bubbles G. Garcia-Segura & M. Mac Low Si 19: a new Luminous Blue Variable? A. Notn et al HST images of Eta Carinae D. Ebbets et.nl

64 73 78 85 89 95

Part three: Supernovae Supernovae and their circumstellar environment B. Leibundgut... Radio supernovae and progenitor winds 5. Van Dyk et al Circumstellar interaction in supernovae C. Fmnsson SN progenitor winds J. Blondin Supernovae with dense circumstellar winds Ar. N. Chugai

100 112 120 139 148 vn

Compact supernova remnants R. J. Terlevich 153 The evolution of compact supernova remnants G. Tenorio-Tagle..l66 Massive supernovae in binary systems P. C. Joss et al 179 The progenitor of SN 1993J Ph. Podsiadloivski et al 187 Narrow lines from SN 1993J R. J. Cumming et al 192 UV spectroscopy of SN 1993J G. Sonneborn et al 198 Ryle Telescope observations of SN 1993J D. A. Green & G. G. Pooley 203 SN 1993J - early radio emission K. W. Weiler et al 207 The circumstellar gas around SN 1987A and SN 1993J P. Lundqvist 213 X-ray emission from SN 1987A and SN 1993.1 T. Suzuki et. al, .. .221 The interstellar medium towards SN 1993J in M81 D. L. King et al 227

Part four: Asymptotic Giant Branch stars Mass loss from late type stars /. Cherchneff & A. G. G. M. Helens 232 Kinematics and structure of circumstellar envelopes //. Olofs$on..246 Circumstellar shells of Long-Period Variables A. Ganger et al. ... 262 Observation of circumstellar shells with the IR AM telescopes M. Gue'lin et al 266

Part five: Planetary Nebulae Morphology and kinematics of PNe H. E. Schwarz FLIERs in elliptical Planetary Nebulae M. Perinotto et al Circumstellar dust in PN and PPN S. Kwok et al H-poor ejecta in A30 and A78 ./. P. Harrington et al The neutral envelopes of PNe P. J. Muggins et al Magnetic shaping of Planetary Nebulae /?. Chevalier Aspherical two-wind configurations V. lake

274 291 296 300 304 308 309

Part six: Novae and Symbiotic Stars Novae as tracers of mass loss M. F. Bode Light scattering in symbiotic stars //. M. Schmid & II. Schild

321 331

Poster Papers

335

Author Index

341

Object Index

343

vin

Preface

The historic series of Herstmonceux Conferences was started by the new Astronomer Royal, Sir Richard Woolley, in the late 1950's. The Royal Greenwich Observatory had by then recently finished moving its scientific operations from Greenwich to Herstmonceux Castle in East Sussex. Evidently the first few such conferences were relatively small private affairs, and there are few written records of them, but in later years they grew in size. After the moving of the RGO to Cambridge in 1989, they have been organised jointly with the University's Institute of Astronomy. Our idea for the 34th Conference was to bring together different astronomical communities who study stellar evolution, stellar winds and the physics of circumstellar media, and to bring out the common physics affecting matter around both high and low-mass stars. This volume presents the proceedings. We have included all the invited reviews and the contributed oral talks, and there is a summary listing the titles and authors of all the poster papers. Thanks are due to many people for helping to put together what was the largest-ever Herstmonceux Conference. The Organising Committee were Robin Catchpole, Robin Clegg, Peter Meikle, Jim Pringle, Anne Reynolds and Ian Stevens; Anne did a huge job as Conference Secretary and deserves special mention. The Advisory Committee were John Dyson, Alex Filippenko, Claes Fransson, Harm Habing, Alain Omont and Guillermo TenorioTagle. The RGO and the IoA gave financial support, and the International Science Foundation funded speakers from Russia. We are also grateful to Alec Boksenberg for opening the Conference and to Sir Martin Rees and Jasper Wall for speeches at the Conference Dinner, held in King's College. Finally, thanks to our editor at CUP, Adam Black, for his patient assistance. Robin Clegg, Ian Stevens and Peter Meikle

IX

J 1. O. Hashimoto 2. N. Patat 3. B. Cadwell 4. G. Garcia-Segura 5. R. Corradi 6. R. Oudmaijer 7. K. Borkowski 8. 3. Blondin 9. D. Green 10. J. Solf 11. T. Forveille 12. V. Icke 13. P. Tuthill 14. J. Blommaert 15. A. Gauger 16. P. Podsiadlowski 17. R. Wolstencroft

18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34.

K. Justtanont W. van der Veen P. Joss M. Kolman S. Zhekov F. Fagotto A. Wootten K. Exter C. Fransson R. Tuffs M. Perinotto J. Pringle G. Wynn-Williams Xiao-Wei Liu H. Habing E. Bakker R. Cid Fernandez

35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51.

N. Langer D. King P. Harrington J. Walsh R. Tweedy R. Chevalier R. Cumming S. Scuderi R. Schulte-Ladbeck H. Schwarz F. Pijpers T. Harries M.. Clampin C. Koempe D. Ebbets S. Lorenz-Martins F. de Aravjo

52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68.

P. de Laverney E. Huguet P. Wood L. Stanghellini P. Garcia-Lario W . Maciel P. Pavelin P. Huggins H. Olofsson J. Yates P. Vilchez G. Sonneborn D. Schaerer M . Groenewegen M . Barlow J. Drew P. Williams

69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85.

I. Yamamura D. Hutsemekers P. Lundqvist N. Chugai K Weiler L. Smith A. Skopal A. Harpaz M . Lewis C. Charbonnel S. Van Dyk E. Terlevich R. Terlevich L. Bianchi A. Not a A. Pasquali G. Mellema

86. M. Kato 87. H. Walker 88. J. Pacheco 89. C. Rossi 90. A. Heske 91. R. Dufour 92. A. Manchado 93. A. Michalitsianos 94. B. Pagel 95. M. Wrigge 96. M. Bryce 97. C. Guilain 98. A. Frank 99. N. Berruyer 100. J. Nichols 101. T . Lozinskaya 102. MI. Livio

103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119.

N. Netzer N. Soker R. Clegg Z. Han P. Murdin A. Reynolds I. Stevens P. Meikle U. Munari B. Leibundgut S. Etoka A. Tielens H. Schmid G. Tenorio-Tagle R. Bachiller T. Suzuki B. Hassall


R. Bachiller Yebes, Spain. E. Bakker SRON, Netherlands. J. A. Baldwin MRAO, Cambridge, UK. M. J. Barlow University College London, UK. N. Berruyer Obs. de la Cote d'Azur, France. L. Bianchi STScI, USA. J. Blommaert Inst. d'Asirophysique, Paris, France. J. M. Blondin North Carolina State Univ., USA. M. F. Bode Liverpool John Moores Univ., UK. K. J. Borkowski Univ. of Maryland, USA. M. Bryce Univ. of Manchester, UK. B. C. R. N.

Cadwell Penn State, USA. Charbonnel O6s. de Geneva, Switzerland. A. Chevalier Univ. of Virginia, USA. N. Chugai Russian Academy of Sciences, Moscow, Russia. R. Cid Fernandez RGO, Cambridge, UK M. Clampin STScI, USA. R. E. S. Clegg RGO, Cambridge, UK. R. Corradi Univ. of Padova, Italy. P. Cox MPI f. Radioastronomie, Bonn, Germany. A. Crotts Columbia Univ., USA. R. J. Cumming RGO, Cambridge, UK. R. J. Davis NRAL Jodrell Bank, UK. F. de Aravjo Obs. de la Cote d'Azur, France. P. de Laverny Univ. Montpellier, France. J. S. B. Dick RGO, Cambridge, UK. J. E. Drew Oxford Univ., UK. R. E. Dufour Rice Univ., USA J. E. Dyson Univ. of Manchester, UK. D. Ebbets Ball Aerospace Systems Group, Boulder, USA. S. Etoka 06s. de Paris-Meudon, France. K. Exter Univ. St. Andrews, UK. F. T. A. C.

Fagotto Padova Observatorio, Italy. Forveille 06s. de Grenoble, France. Frank Univ. Minnesota, USA. Fransson Stockholm Observatory, Sweden.

G. Garcia-Segura Univ. of Illinois at Urbana-Champaign, USA. P. Garcia-Lario WE Observatory, Spain. A. Gauger Technische Universitat Berlin, Germany. A. Glassgold New York Univ., USA. D. A. Green MRAO, Cambridge, UK. M. Groenewgen Astronomical Institute, Amsterdam, Netherlands. M. Guelin IRAM, France. C. Guillain CNRS Toulouse, France. H. J. Habing Sterrewacht Leiden, Netherlands. Z. Han loA, Cambridge, UK. A. Harpaz Technicon, Israel. T. Harries University College London, UK. J. P. Harrington Univ. of Maryland, USA. 0 . Hashimoto Seikei University, Japan. B. Hassall RGO, Cambridge, UK. A. Heske ESTEC, Netherlands. R. Hills MRAO, Cambridge, UK. P. J. Huggins New York Univ., USA. E. Huguet 06s. de Paris-Meudon, France. D. Hutsemekers Astrophysical Inst of Liege, Belgium. V. Icke Sterrewacht Leiden,

Netherlands.

P. C. Joss MIT, USA. K. Justtanont Institut d'Astrophysique, Paris, France. M. Kato Keio Univ., Japan. D. L. King RGO, Cambridge, UK. H. Kley IoA, Cambridge, UK. C. Koempe Uni-Sternwarte, Germany. Sun Kwok Univ. of Calgary, Canada. N. Langer MPI f. Astrophysik, Germany. J. Lefevre 06s. de la Cote d'Azur, France. B. Leibundgut UC Berkeley, USA. B. M. Lewis Arecibo Observatory, USA. X-W. Liu University College London, UK. M. Livio STScI, USA, xm

xiv


S. Lorenz-Martin 06s. de la Cote d'Azur, France. T. A. Lozinskaya Sternberg Inst., Russia. P. Lundqvist Stockholm Observatory, Sweden. W. Maciel IAG/USP, Brazil. A. Manchado Univ. of Illinois, USA. W. P. S. Meikle RGO, Cambridge, UK. G. Mellema Sterrewacht Leiden, Netherlands.

A. Michalitsianos NASA/GSFC, USA. U. Munari 06s. di Padova, Italy. P. G. Murdin Royal Observatory Edinburgh, UK. N. Netzer 0RT- Braude College, Israel. J. Nichols-Bohlin NASA/GSFC, USA. A. Nota STScI, USA. H. Olofsson Stockholm Observatory, Sweden. A. Omont Institut d'Astrophysique, Paris, France. R. Oudmaijer Kapteyn Lab. Groningen, Netherlands.

J. Solf MPI f. Radioastronomie, Bonn, Germany. G. Sonneborn NASA/GSFC, USA. L. Stanghellini Obs. di Padova, Italy. E. Stengler RGO, Cambridge, UK. I. R. Stevens IoA, Cambridge, UK. T. Suzuki Univ. of Tokyo, Japan. G. Tenorio-Tagle Univ. de La Laguna, Spain E. Terlevich RGO, Cambridge, UK. R. J. Terlevich RGO, Cambridge, UK. A. G. G. M. Tielens NASA Ames, USA. R. Tuffs MPI f. Radioastronomie, Bonn, Germany. P. Tuthill MRAO, Cambridge, UK. R. Tweedy Steward Observatory, USA. S. D. Van Dyk NRL, USA. W. van der Veen Columbia Univ., USA. S. Vessey MRAO, Cambridge, UK. J. M. Vilchez Inst. Astrofisica de Canarias, Spain.

J. Pacheco IAG/USP, Sao Paulo, Brazil. B. E. J. Pagel NORDITA, Denmark. A. Pasquali Univ. of Firenze, Italy. F. Patat Padova, Italy. P. Pavelin NRAL Jodrell Bank, UK. M. Perinotto Univ. of Firenze, Italy. F. Pijpers Uppsala Obs., Sweden. Ph. Podsiadlowski IoA, Cambridge, UK. J. E. Pringle IoA, Cambridge, UK. G. G. Pooley MRAO, Cambridge, UK.

H. J. Walker RAL, UK. J. R. Walsh ST-ECF Garching, Germany. N. Walton RGO, Spain. K. W. Weiler NRL, USA. P. M. Williams Royal Observatory Edinburgh, UK. R. D. Wolstencroft Royal Observatory Edinburgh, UK. P. Wood MSSSO, Australia. A. Wootten NRAO, Virginia, USA. M. Wrigge Hamburger Sternwarte, Germany. G. Wynn-Williams Univ. of Hawaii, USA.

A. Reynolds RGO, Cambridge, UK. C. Rossi Sapienza. Italy.

I. Yamamura Univ. of Tokyo, Japan. J. Yates NRAL Jodrell Bank, UK.

D. Schaerer Geneva Observatory, Switzerland. H. M. Schmid ETH Zentrum, Switzerland. R. Schulte-Ladbeck Univ. of Pittsburgh, USA. D. Schwarz IoA, Cambridge, UK. H. E. Schwarz ESO, Chile. S. Scuderi STScI, USA. A. Skopal Liverpool John Moores Univ., UK. L. J. Smith University College London, UK. N. Soker Oranim - University Division, Israel.

S. Zhekov Obs. di Arcetri, Italy.

Evolution of massive stars N. Langer MPIf. Astrophysik, Karl-Schwarzschild-Str. 1, D-8511,0 Garching, FRG

Abstract Recent results of the theory of massive star evolution are discussed. We divide the regime of massive stars in a "low" mass and a high mass part, and show that the evolution, the basic theoretical problems in their modeling, and the display of circuinstellar matter are quite different for stars from both parts. For stars in the lower considered mass regime, it is shown that our ignorance about, internal mixing processes is the main source of uncertainty in stellar model calculations. Mixing processes related to thermal convection are discussed, and their effect on the observable stellar parameters and presupernova structure are sketched. The role of mixing induced by differential rotation is also briefly described. We argue that the supernova stage is a good possibility to investigate the circumstellar material of these objects and describe its high diagnostic potential for the whole presupernova evolution. Our understanding of stars in the upper mass range, i.e. the most massive stars, also suffers from uncertainties related to internal mixing. However, we argue that it is the mass loss process which dominates their global evolution. The evidence that those objects do lose the major part of their initial mass before they collapse is discussed, together with the possibility of the display of circumstellar nebulae during their hydrostatic evolution. Finally, the fate of very massive stars is discussed.

1 Introduction As usual in the astrophysical literature, we mean stars sufficiently massive in order not to develop into white dwarfs here (i.e. Mz.4A/s ^ 8 A/©, cf. Iben and Renzini 1983; ZAMS stands for zero age main sequence) when we speak of massive stars. Thus, this review deals with stars in the mass range ~8M0— ~2OOM0, where the upper end of the regime is defined by the upper end of the stellar initial mass function. It is to be mentioned right at the beginning of this paper, that — even at a given constant metallicity Z — stars in different regimes of this mass range have very different evolutionary characteristics. This is true for their internal evolution as well as for the evolution of their surface properties (e.g. their tracks in the HR diagram): Massive stars with masses close to 8 MQ may explode due to carbon deflagration (Type 1^ SN; cf. Iben and Renzini 1983), or collapse due to electron capture (cf. e.g. Nomoto 1984). At somewhat higher mass, stars probably make "standard" Type II SNe due to iron core collapse (see 1

2

N. Langer: Evolution of massive stars

Woosley and Weaver 1986), while still more massive stars may form black holes (cf. Maeder 1992). Further up in the mass range stars may explode again since they end up as low mass objects, but they will appear as Type I SNe as they are devoid of hydrogen (cf. Woosley et al. 1993, Langer 1994; cf. also Sect. 3). And for the uppermost masses, there is (at low metallicity) still the possibility of exploding before the formation of an iron core, clue to the e ± -pair instability (cf. e.g. Woosley 1986, El Eid and Langer 1986). As diverse as the internal evolution of massive stars are their HRD tracks. The less massive of the massive stars evolve always to the right of the main sequence. They stay on that side for their whole life, most of them using up all the available space up to the Hayashi-line. The very massive stars, on the contrary, never get really cold, i.e. their surface temperature always remains hotter than ~ 10 000 K. Their tracks generally cross the main sequence once and lead into the hot part of the HR diagram. The described diversity makes the necessity evident to divide the considered mass range into peaces and to discuss stars in each piece separately. Since this paper is in a book about circumstellar matter, and since we can only perform a qualitative investigation of stellar evolution here, we just give a qualitative discrimination of stars of two different groups: the "just massive stars" (JMS), which are characterized by the fact that their total mass remains basically constant during their whole evolution form the main sequence to the supernova stage, and the "very massive stars" (VMS), which do lose the major part of their initial mass. The JMS, due to the fact that they keep most of their hydrogen-rich envelope, may become red supergiants during their post-main sequence evolution, and as such they may lose several solar masses of their material and develop a rather slow and dense circumstellar wind. This material is, however, unlikely to be directly observable, unless it is illuminated by the supernova outburst of the central star, as in the case of SN 1987A. Otherwise, the circumstellar matter may manifest itself only indirectly, e.g. due to an infrared excess in the spectrum of the central star (cf. Stencel et al. 1989). The VMS do not only accumulate much more circumstellar matter due to their large mass loss rates, but since they develop into very hot and compact stars they eject fast winds which can compress the previously ejected material, and radiate a hard photon field which can ionise it. Consequently, many of them develop nice ring nebulae or spherical circumstellar shells. There is of course no sharp borderline between JMS and VMS, which — according to recent stellar evolution calculations — is somewhere between 25 and 40 Me in our Galaxy (cf. Schaller et al. 1992, Woosley et al. 1993).


3

Note, however, that due to the metallicity dependence of stellar winds (Kudritzki et al. 1987, Leitherer and Langer 1991) this borderline depends strongly on the initial stellar metallicity. In the following Section we will try to illustrate the problems in modelling the JMS, which involves basically internal mixing processes. Sect. 3 deals with the VMS, where the main problem concerns the mass loss. Some final remarks are given in Sect. 4.

2 The "just massive stars" The JMS have been defined as massive stars which keep most of their mass until the end of their hydrostatic evolution. In fact, a. weaker condition may be chosen, though it is a bit harder to understand, namely that a substantial part of the hydrogen-rich envelope remains bound to the star: it ensures that the H-burning shell source is almost unaffected by the mass loss, and even more so is the helium core and thus all the burning stages following He-burning (Maeder 1981, Chiosi and Maeder 1986). With this definition, e.g. a 25 M@ star losing some 5-10 A/,v, during its evolution is still a JMS: it develops a He-core of ~ 8 M 0 (when adopting the Schwarzschild criterion, see below), i.e. it would retain a H-rich envelope of ~7-12 MQ in the above example. Since the topic of this conference concerns late stages of stellar evolution, we only want to mention some problems with JMS main sequence models (cf. Langer 1993, for more details): compared to observed main sequence JMS the models are — on average — too faint, too hot, and lack helium and nitrogen enrichment (cf. Herrero el al. 1992, Langer 1992). Increased mass loss in model calculations leads in principle to larger luminosity-tomass rations, cooler main sequence models, and eventually even to surface enrichment; however, the required mass loss rates are one order of magnitude larger than the observed rates, which are by far not so uncertain. I.e., mass loss is not the right solution. Internal mixing appears to be rather promising. Zahn (1983) and Chaboyer and Zalin (1992) have proposed that differential rotation may lead to substantial mixing in the stellar interior, and Maeder (1987) showed that — for M = 40 MQ — this may even lead to homogeneous evolution. Langer (1992) and Denissenkov (1993) have applied this mechanism to somewhat smaller initial masses and also found a substantial mixing to occur, while Fliegner and Langer (1994), with a selfconsistent treatment of rotating stars, showed that even at 5 MQ rotational mixing may substantially alter the stellar evolution on the main sequence. Certainly, also the post-MS evolution is affected by this kind of mixing (cf. Langer 1992);

4

N. Langer: Evolution of massive, stars

however, systematic studies are still lacking, and the parameters involved in the description of rotational mixing are still weakly restricted. It is to be mentioned here, that models of close binary systems also have the potential to cope with the above problems, but also here a thorough discussion is not yet available (cf. Langer 1992; Joss, this volume; and references therein). When a JMS has finished core hydrogen burning, it develops into a supergiant. Due to the fact that JMS models in the effective temperature regime 4.0 ^ log Tejj ^> 3.6 (the precise numbers depend on the internal composition profile and metallicity) are thermally unstable (cf. e.g. Tuchman and Wheeler 1989ab) — a fact which agrees remarkably well with the observed low population density in that regime of the IIR diagram; cf. Blaha and Humphreys 1989 —, only two long-lived states are possible for post-MS JMS: that of blue supergiants (BSGs) and that; of red supergiants (RSGs). However, despite of this simplicity (especially when compared with VMS; cf. Section 3) it is today by no means clear which of these two states are occupied by JMS of a given mass and metallicity, or in which order they occur during the life of a JMS. We want to emphasize that the situation is much different in the (often better known) case of intermediate and low mass stars (i.e. MZAMS RSG or vice versa (note that due to the dredge-up process in the R.SG stage one may hope to observation ally disentangle the post-RSG BSGs from the others on the basis of their surface abundances; cf. Lennon et al. 1992, 1993; Fitzpatrick and Bohannan 1992), and the case of SN 1987A reminded us that we are not even sure about which of the two possible states (ie. BSG or RSG) is the pre-SN stage. The basic problem which gives rise to the huge uncertainties quoted above exists as long as stellar models are evolved on the computer: which is the correct treatment of convection in the presence of a mean molecular weight (/t) gradient? The extreme answers to this quest ion are: 1) adopting the Schwarzschild criterion for convection, i.e. ignoring the stabilizing effect of

N. Longer: Evolution of massive stars

log -feff

Fig. 1. Tracks of three 20 A/0 sequences of Z = 0.005 in the HR.diagram, computed with different assumptions on the efficiency of convection in regions of varying mean molecular weight. The short-dashed line indicates a track computed with the Schwa rzschi Id criterion, in the dot-dashed track the Ledoux criterion was used, and for the solid track an intermediate case, namely the semiconvection model of Langer el al. (1983), has been applied. Big dots mark core helium ignition and exhaustion, and asterisks mark the pre-SN positions (cf. Langer el al. 1989, for details).

/i-barriers altogether, or 2) adopting the Ledoux criterion, i.e. to prevent any mixing at all in the presence of (sufficiently strong) //-barriers. Following Kato's (1966) stability analysis, an intermediate treatment has been developed (Langer et al. 1983. 1985), and Fig. 1 illustrates the consequences for the evolution of the surface properties at the example of a 20 M.-, star at roughly LMC metallicity. The reason why the evolutionary tracks depend on the treatment of internal mixing are twofold: a) mixing at the H/He-interface during core Hburning (cf. e.g. Chiosi and Summa 1970; Stothers and Chin 1975, 1976) and — even more important — during the contraction phase towards helium ignition (Langer et al. 1985, Bressan et al. 1993) determine the conditions for the hydrogen burning shell during core helium burning, and b) mixing at the He/C+O-interface determines the size of the He-burning convective core and thereby the core He-burning life time, and the C+O-core mass, which in turn determines completely the internal evolution beyond core helium burning (cf. Fig. 2). Since core He-burning stars in the mass range 15 - 30A/(T, are in a. delicate balance between the BSG and RSG states in general, almost any change in the internal composition profile can have a large impact on the predicted BSG/RSG lifetime ratio (cf. Langer 1991). In addition to the problem of the feedback of /t-barriers with convection,

N. Langer: Evolution of massive stars 20M_sun — He — Ne — burning 9.0

•

'

^c^S1

'

log 18c

8.8 8.6 8.4 8.2

Jr

'-•

8.0 4 log

5 rho_c

Fig. 2. Evolution of two 20 MQ tracks in the logTc — \ogpc -plane. The upper track has been computed using the Schwarzschild criterion for convection, in the lower one semiconvection according to Langer et al. (1983) has been applied. The time distance between two neighboring crosses on each track is 5000 yr. further mixing processes are potentially important for JMS. One is the so called convective overshooting, which designates the possibility of convective motions to penetrate into radiatively stable stratified layers adjacent to convection zones. Only the effects of overshooting in stellar convective cores (e.g. Schaller et al. 1992, Bressan et al. 1993) and of convective envelopes (Stothers and Chin 1991; and references therein) have yet been systematically investigated, and at present the importance of this process for JMS is still a matter of debate (see e.g. Langer 1991, Schaller et al. 1992). A second family of mixing processes with almost unknown efficiency are the rotationally induced instabilities, which — at least in rapidly rotating stars — may significantly alter the whole internal chemical structure of the star (see above). Also on this subject, much work remains to be done. It has been mentioned at the beginning of this Section that JMS may well lose appreciable amounts of mass before they explode as supernova. It is important to note here that even for precisely known mass loss laws, i.e. mass loss rates as function of the stellar surface conditions, the amount of matter lost during the life of a star of given initial mass and chemical composition can not be accurately predicted due to the large uncertainty of the effective temperature evolution mentioned above. E.g., we know that the mass loss rate in the RSG stage exceeds by far the BSG mass loss rate. However, since we do not know which star spends which fraction of its lifetime as a RSG, the total amount of lost mass remains very uncertain. In this respect it is the circumstellar matter and the supernova event


7

which provide most useful constraints for the whole pre-SN evolution, and in particular for mass loss rates and total amounts of mass lost; many examples for this can be found in the present book: E.g., the early radio (van Dyk, this volume) and X-ray (Fransson, this volume) emission of supernovae provide stringent limits on the wind density M/v^ of the progenitor star. The same quantity may also affect the early optical light curve of Type II SNe, as shown by Hoflich et al. (1993) for the case of SN 1993J. SN 1987A provides a unique example which is widely discussed in the literature. Let us just mention that the progenitor mass loss history is traced by the famous ring nebula (cf. e.g. Lundqvist, this volume), and the light curve and spectral evolution, which provide envelope and He-core mass, allowed the best estimate of the total mass lost by a. massive star within its hydrostatic evolution (Arnett et al. 1989, Hillebrandt and Hoflich 1989). Also the SN emission at late times may yield information about the circumstellar matter density at moderate distance from the progenitor star and thus probe the stellar mass loss up to some 104 yr before the actual SN explosion (cf. Leibundgut, this volume; Clmgai, this volume). The quoted contributions contain evidence for unexpectedly large wind densities in quite a number of progenitor stars which probably fall into the .IMS mass range since they became Type II SNe. Many interesting results from this area of research are to be expected in the near future.

3 Very massive stars Since a very recent review on VMS evolution is published elsewhere (Langer 1994), we will focus here on topics related to circumstellar matter and late evolutionary stages of VMS. VMS have been defined as stars which do lose the major part of their initial mass before they collapse (cf. Sect. 1). Though such a scenario is largely favored by many observational facts (cf. Langer 1989, Meynet et al. 1994) as well as by theoretical considerations (cf. Langer 1989, 1989a), one would like to have a. direct "proof" for very massive stars evolving into rather low mass objects. If this scenario is indeed correct, then the uncertainties in internal mixing processes discussed for JMS in Sect. 2 play only a minor role for the understanding of VMS (though the VMS may be an ideal tool to investigate them, since the large mass loss allows us to quasi look deep inside the stars; cf. Langer 1991a). Instead, what almost completely determines the JMS evolution is the mass loss process: note that current models predict e.g. a 60 MQ star to evolve into a. star of roughly 5 M 0 (Langer 1989, Schaller et al. 1992, Woosley et al. 1993). Let us concentrate on two possible ways to "proof" this — at first glance

8


surprising — type of scenario, namely that VMS lose some 90% of their initial mass during long lasting evolutionary phases, i.e. core hydrogen and helium burning: one is to show that a correspondingly large amount of freshly ejected matter exists close to the stars which are potential descendants of once very massive stars, and the other is to look for supernovae from such objects.

3.1 VMS ejecta When a VMS — say a 60 M® star with Z = Z@ — ends up as a WC/WO star of ~ 5 M@, it loses some 55 MQ of material during its H- and He-burning phase of evolution. Since the resulting WR. star is very hot and luminous it may be expected that the ejected material appears as a visual nebula. In fact, we know that an appreciable number of WR stars are associated with ring nebulae (Chu et al. 1983, Chu 1991, Lozinskaya, this volume), many of which can be shown to have been ejected by their central WR object on the basis of their kinematical or chemical properties. What kind of nebula can we expect from current VMS models? Figs. 3 and 4 show the time dependence of the mass loss rate and the wind velocity for a 60 MQ sequence which was computed by using the latest available input physics, i.e. OPAL opacities (Iglesias et al. 1992), O star mass loss according to Lamers and Leitherer (1993) and due to pulsational instabilities (cf. Langer 1994, for details), and mass dependent WR mass loss (Langer 1989). The wind velocity has been computed as a function of the escape velocity, using data of Abbott (1982) and Kudritzki and Hummer (1990) for 0 stars and of Hamann et al. (1993) for WR. stars for calibration. We see from Figs. 3 and 4 that, during the first 2 million years, the star — corresponding to a normal 0 star in this period — has a very fast wind and a relatively small mass loss rate. This wind will certainly excavate a large bubble around the central star, which provides the action space for the later nebula formation. Using analytic relations for wind blown bubbles (cf. Shull 1993) and adopting typical parameters for our 0 star (M - 2 10~6 MQ yr~l, VQO = 3500km s'1) and an ISM density of n = lcm~3, the bubble size will be about 80 pc in diameter at an age of 2 106 yr. In the ensuing WNL phase the mass loss rate is by one order of magnitude enhanced, and the wind velocity is somewhat lower. Then, at core hydrogen exhaustion, a large amount of mass (5 —10 MQ) is lost with a rather small wind velocity (some 100 km s" 1 ) in a Luminous Blue Variable phase (cf. Langer 1994). This slow moving gas can be expected to be accumulated and compresses in a shell by the successive fast and energetic WR wind.


Fig. 3. Mass loss rate as a function of time for a. 60 M© sequence at Z=2% (cf. Langer 1994 for details). The first broad maximum corresponds to a first WNL stage. At t ~ 3.35 106 yr, core H-burning is terminated, and an LBV phase occurs. Finally, mass dependent VVR mass loss according to Langer (1989) has been used.

4OOO

3OO0

2OOO -

1ODO -

p

0

Fig. 4. Final wind velocity as a function of time corresponding to the mass loss rate shown in Fig. 3; see text for details. Garcia-Segura and Mac Low (this volume) performed 2D hydrodynamic computations of wind bubbles around massive stars, using a simplified schematic time behavior of the central star wind, which resembles that displayed in Fig. 3. They found a nice agreement of their results with observed WR nebulae like NGC 6888. Note that they assumed a RSG phase instead of the LBV phase in the above example. However, both, RSGs and LBVs, are characterized by large mass loss rates and by wind velocities which are slow compared to the very fast WR winds. I.e., also from post-LBV WRs one may expect the formation of a shell type nebula.

10


It remains to be investigated whether the expected life times of such nebulae agree with the observed number of them (cf. Lozinskaya, this volume), and whether they occur at the evolutionary stage/observed WR subtype where they are expected. However, the basic fact of the presence of many of them, and the success of the theoretical models so far (cf. Garcia-Segura and Mac Low, and references therein) provide a basic support of the idea that VMS do lose the major part of their mass before they explode. In fact, further studies of WR nebulae promise much insight into the details of the progenitor mass loss as a function of time. A second and independent possible evidence in favor of large amounts of mass being lost during the VMS evolution may be through the detection of 7-rays form the radioactive decay of 26Al: 26.4/ is synthesized during core hydrogen burning in massive stars (see e.g. Prantzos et al. 1986), and its long half life of roughly 106 yr provides the possibility that much of the 2eAl decays only after it has been ejected into space so that the produced l.SMeV photons can escape without interaction. Note that galactic 26Al emission has been discovered by HEAO-3 (cf. Mahoney 1982), and — among other sources — WR stars have been proposed to eject radioactive 26Al into the ISM. Fig. 5 shows the expected time dependence of the number of 7-photons from the 26Al decay for the same 60 MQ calculation for which the mass loss rate is shown in Fig. 3 (cf. Langer 1994, for details), assuming a distance to the source of 1 kpc. The flux is turned on in the WNL stage and reaches a maximum in the ensuing LBV phase. During the following WNE and WC stages, the flux declines exponentially with time, since basically all the produced 26Al has been ejected before. Note that at t ~ 4.1 106 yr the star explodes as a supernova, which adds an amount of 5 10~5 MQ of explosively synthesized 26Al according to Woosley et al (1994). Clearly, the detection of a compact source of l.SMeV photons related with a VMS type object would be a strong evidence for significant mass loss. The photon flux, together with the distance, would give hard lower limits to the amount of ejected matter. Note that 7-ray astronomy is in the shape to perform a detailed spatially resolved mapping of the l.SMeV flux at present and individual sources may be recognized in the near future (Diehl et al. 1993); a potential detection of the WC star 7 Velorum has been recently reported (Diehl, priv. communication). Our calculations point out that not only WR stars but also LBVs can be expected to be prominent 1.8 MeV photon sources. In this respect, P Cygni is the most outstanding candidate.


11

26AI — d e c a y — p h o t o n — f l u x

4x1 0 c

2x10

6*10c

time

Fig. 5. 1.8 MeV photon flux from the decay of '-''Al as function of time (in yr), computed for a 60 MQ sequence at Z=2% (c.f. Fig. 3, and Langer 1994, for details). The nuclear reaction rates of Caughlan and Fowler (1988) have been used in that model. Note that the smaller -6Al production rate proposed by Iliadis et al. (1990) would reduce the expected flux by about a factor of 4. For the absolute flux value, a distance of 1 kpc has been assumed.

3.2 Supernovae from VMS ? The question whether or not VMS may develop into supernovae has been investigated recently by Woosley et al. (1993, 1994). Here, we just want to briefly outline the main conclusions (cf. also Langer 1994). Whether or not the core collapse in a given massive star yields a SN explosion depends sensitively on its iron core mass: larger iron cores are less likely to explode (Woosley and Weaver 1986). Though the iron core mass of (non-mass losing) massive stars is not necessarily a monotonic function of the stellar mass throughout, more massive stars develop more massive iron cores in general (cf. Woosley 1986). I.e., from non-mass losing stars one would expect the existence of a critical stellar mass above which core collapse leads to black hole formation and only below a core collapse SN occurs. How does the VMS mass loss affect this picture? The mass dependent WR mass loss leads to a mass convergence, i.e. the final stellar masses of VMS are almost independent of the initial stellar mass for a wide range of ZAMS masses, with typical final masses of the order of 5M© (cf. Schaller et al. 1992, Meynet et al. 1994). Since the pre-SN models are devoid of hydrogen, their total mass is equal to their He-core mass (cf. Table 1). One is thus tempted to conclude that, since the Hecore mass greatly determines the evolution through the final burning stages (cf. Arnett 1978, Sugimoto and Nomoto 1980), a supernova explosion is the

12

N. Langer: Evolution of massive stars Table 1. Core masses as function of the initial mass (in MQ) MzAMS

Mfinal

1

35 40 1 60 1

15.2 11.1 4.3

502

50.0

MC/o

MFe

12.2 11.1 4.3

3.7 3.2 3.0

1.45 1.42 1.40

23

-15

2.45

MHe

Notes: 1) From Woosley et al. (1993). 2) From Woosley (1986). likely result of VMS evolution, since a 5 MQ He-core also corresponds to a 15 MQ ZAMS mass, and SN 1987A showed that 15 — 20 MQ stars do in fact explode as supernovae. However, things are a bit more complicated. Woosley et al. (1993) have shown that the late evolution of low mass VVR. stars descendant of VMS of a certain final mass (4.25 MQ in their example) is quite different from the late evolution of a star of the same final mass which was not very massive in the past. The reason is that the large mass loss during core helium burning of the once very massive star establishes a. very shallow He-gradient, which allows the formation of a much larger C/O-core in this case, compared to the star which was never very massive. Since mass loss beyond core helium exhaustion is negligible, the C/O-core mass is then a good parameter in the sense that its value describes the late evolutionary phases almost independently of the previous evolution (see Table 1). Nevertheless, VMS mass loss reduces the final C/O-core mass to values (~ 3 MQ) which correspond to ZAMS masses of the order of 20 MQ in constant mass calculations, which still makes a SN explosion very likely, at least compared to the case of the evolution of very massive stars at constant mass (cf. Table 1). However, note that not only the iron core mass determines whether or not a SN explosion may occur, but to some extent also the density profile above the iron core (cf. Woosley et al. 1993); i.e., in order to settle the question which stellar models exactly would explode one needs a better understanding of the core collapse supernova mechanism.

4 Final remarks In the previous Sections we have shown that the theory of massive star evolution is still far from being in a final shape. Today, very important


13

parts in such a theory are still lacking or are oversimplified. This concerns points which are generally designated as "input physics" like mixing or mass loss theories, but also genuine stellar structure problems as a theory for the stellar radius evolution or the inclusion of effects of rotation. We have concentrated on current problems of stellar evolution theory since this may be the most efficient way to show its state of the art; it does by no means imply that it has not been also very successful in the recent past. However, it implies that still much work is to be done before massive star evolution models can be used as input for other astrophysical applications without asking for the remaining uncertainties in such models. It has been shown above, and by many contributions to this conference, that the circumstellar media can be considered as one of the best diagnostic tools to test and further improve stellar evolution models.

Acknowledgements This work has been supported by the Deutsche Forschungsgemeinschaft through grant No. La 587/8.

References Abbott D.C. (1982). Astrophys. J., 259, 282. Arnett W.D. (1978). Astrophys. J., 219, 1008. Arnett W.D., Bahcall J.N., Kirshner R.P. & Woosley S.E. (1989). Ann. Rev. Astron. Astrophys., 27, 629. Bertelli G., Bressan A.G. & Chiosi C. (1985). Astron. Astrophys., 150, 33. Blaha C. & Humphreys R.M. (1989). Astron J., 89, 1598. Bressan A., Fagotto F., Bertelli G. k Chiosi C. (1993). Astron. Astrophys. Suppl., 100, 647. Caughlan G.A. & Fowler W.A. (1988). Atomic Data and Nucl. Data Tables, 40, 238. Chaboyer B. & Zahn J.P. (1992). Astron. Astrophys.. 253, 173. Chiosi C. fc Summa C. (1970). Astrophys. Space Sci., 8, 478. Chiosi C. & Maeder A. (1986). Ann. Rev. Astron. Astrophys.. 24. 229. Chu Y.-H. (1991). in: IAU-Symp. 143, 349. Chu Y.-H., Treffers R.R. & Kwitter K.B. (1983). Aslrophys. J. Suppl.. 53, 937. Denissenkov P.A. (1993), Astron. Astrophys., (submitted). Diehl R., et al. (1993). Astron. Astrophys. Suppl., 97, 181. El Eid M.F. & Langer N. (1986). Astron. Astrophys., 1C7, 274. Fitzpatrick E.L. & Bohannan B. (1992). Astrophys. J., 404, 734. Fliegner J. &i Langer N. (1994), in: IAU-Symp. 1C2, (in press). Hamann W.-R., Koesterke L. & Wessolowski U. (1993), Astron. Astrophys., 274, 397. Herrero A., Kudritzki R.P., Vilchez J.M., Kunze D., Butler K. fc Haser S. (1992). Astron. Astrophys., 261, 209. Hillebrandt W. & Hoflich P. (1989). Rep. Prog. Physics, 52, 1421. Hoflich P., Langer N. L Duschinger M. (1993). Astron. Astrophys. Letter, 275, L29. Huang R.Q. & Weigert A. (1983). Astron. Astrophys., 127, 309. Iben I. Jr. &; Renzini A. (1983). Ann. Rev. Astron. Astrophys., 21, 271. Iglesias C.A., Rogers F.J. fc Wilson B.G. (1992). Astrophys. J., 397, 717.

14


Iliadis Ch., et. al. (1990), Nucl. Phys. A, 512, 509. Kato S. (1966). P.A.S.J., 18, 374. Kudritzki R.P., Pauldrach A. k Puls J. (1987). Astron. Astrophys., 173, 293. Kudritzki R.P. k Hummer D.G. (1990). Ann. Rev. Astron. Astrophys., 28, 303. Lamers H.J.G.L.M. &; Leitherer C. (1993). Astrophys. J., 412, 771. Langer N. (1989), Astron. Astrophys. 220, 135. Langer N. (1989a), Astron. Astrophys. 210, 93. Langer N. (1991), Astron. Astrophys. 252, 669. Langer N. (1991a), Astron. Astrophys. 248, 531. Langer N. (1992), Astron. Astrophys. Letter 265, L17. Langer N. (1993), in: lAU-Colloq. 137, 426. Langer N. (1994). In: Evolution of Massive Stars: A Confrontation between Theory and Observations, eds. D. Vanbeveren et al., Space Set. Rev., (in press). Langer N., Sugimoto D. k Fricke K.J. (1983). Astron. Astrophys., 126, 207. Langer N., El Eid M.F. k Fricke K.J. (1985). Astron. Astrophys., 145, 179. Langer N., El Eid M.F. k Baraffe I. (1989). Astron. Astrophys. Letter, 224, L17. Leitherer C. k Langer N. (1990), in: lAU-Symp. 148, 480. Lennon D.J., Dufton P.L. k Fitzsimmons A. (1992). Astron. Astrophys. Suppl., 94, 569. Lennon D.J., Dufton P.L. k Fitzsimmons A. (1993). Astron. Astrophys. Suppl., 97, 559. Maeder A. (1981). Astron. Astrophys., 99, 97. Maeder A. (1987). Astron. Astrophys., 178, 159. Maeder A. (1992). Astron. Astrophys., 264, 105. Mahoney W.A., Ling J.C., Jacobson A.S. k Lingenfeltcr R.E. (1982). Astrophys. J., 262, 742. Matraka B., Wassermann C. k Weigert A. (1982). Astron. Astrophys., 107, 283. Meynet G., Maeder A., Schaller G., Schaerer D. fc Charboimel C. (1994). Astron. Astrophys. Suppl, 103, 97. Nomoto K. (1984), Astrophys. J., 277, 791. Prantzos N., Doom C , Arnould M. & de Loore C. (198(5), Astrophys. J., 304, 695. Schaller G., Schaerer D., Meynet G. k Maeder A. (1992). Astron. Astrophys. Suppl., 96, 269. Shull J.M. (1993). ASP Conf. Set:, 35, Cassinelli et al., eds., p. 327. Stencel R.E., Pesce J.E. k Bauer W.H. (1989). Astron. J., 97, 1120. Stothers R.B. (1991). Astrophijs. J., 383, 820. Stothers R.B. k Chin C.-w. (1975). Astrophys. J., 198, 407. Stothers R.B. k Chin C.-w. (1976). Astrophys. J., 204, 472. Stothers R.B. k Chin C.-w. (1991a). Astrophys. J., 374, 288. Sugimoto D. k Nomoto K. (1980). .Space Set. Rew., 25, 155. Tuchman Y. k Wheeler J.C. (1989a), Astrophys. J., 344, 835. Tuchman Y. k Wheeler J.C. (1989b), Astrophys. J., 346, 417. Woosley S.E. (1986). In: 16th Advanced Course of the Swiss Soc. of Astron. and Astrophys., Saas-Fee Lecture Notes, A. Maeder et al., eds., Geneva Observatory. Woosley S.E. k Weaver T.A. (1986). Ann. Rev. Astron. Astrophys., 24, 205. Woosley S.E., Langer N., k Weaver T.A. (1993). Astrophys. J., 411, 823. Woosley S.E., Langer N., k Weaver T.A. (1994). Astrophys. J., (in preparation). Zahn J.P. (1983). In: 13th Advanced Course of the Swiss Soc. of Astron. and Astrophys., Saas-Fee Lecture Notes, B. Hauck et al., eds., Geneva Observatory.

Evolution of AGB Stars with Mass Loss P. R. Wood Mount Stromlo and Siding Spring Observatories, Private Bag, Weston Creek PO, ACT 2611, Australia

Abstract Observational and theoretical estimates for mass loss rates from AGB stars are discussed. Then models for the evolution of AGB stars including mass loss and the effects of helium shell flashes are presented. Finally, the possibility of mass loss by binary mass transfer is discussed.

1 Introduction It is well established that the bulk of mass loss from low and intermediate mass stars occurs during the asymptotic giant, branch (AGB) stage of evolution, leading to the well-defined sequence of mass-losing stars in the IRAS two-colour diagram (van der Veen and Habing 1988) and the formation of planetary nebulae (Abell and Goldreich 1966; Renzini 1981). However, a reliable theoretical understanding of the causes of mass loss is still not available, although progress is being made. An additional complication is that the time history of mass loss during AGB evolution is quite complex since AGB evolution is modulated by helium shell flashes which control the surface luminosity and thereby the mass loss rate. In this paper, mass loss rates from AGB stars are discussed and the effects of helium shell flashes on the mass loss are described and compared with observations. Time dependent winds produced by AGB stars are reviewed. Finally, the evolution of AGB stars that lose mass in binary mass transfer events is briefly described. 2 A G B mass loss rates IRAS observations of stars in the solar neighbourhood indicate that those stars with substantial circumstellar shells - those with high mass loss rates - are nearly all AGB stars undergoing large-amplitude pulsation (Habing 1990). These stars are the long-period variables (LPVs) with pulsation periods > 1 year. This observational result is in agreement with theoretical calculations which indicate that high mass loss rates from AGB stars can only be produced by the combined effects of pulsation and radiation pressure on grains (Wood 1979; Bowen 1988; see Holzer and MacGregor 1985 for a review of possible mass loss mechanisms). However, the quantitative mass 15

16

P. R. Wood: Evolution of AGB stars

loss rates resulting from theoretical calculations are very uncertain as they depend sensitively on the details of grain formation and the rate of cooling of atmospheric material heated by the passage of the shock front that propagates through the stellar atmosphere once per pulsation cycle. Wood (1979) found that in the limit of no cooling, a. typical Mira would drive a mass loss rate of ~0.01 MQ yr" 1 ! In the opposite limit of instantaneous cooling, but with no dust present, M ~ 10~12 MQ yr" 1 . Allowing dust to form boosted the mass loss rate to M ~ 3 x 10~7 MQ yr" 1 which is perhaps 10 times smaller than the observed mass loss rates typical for the type of Mira star considered. Bowen (1988) introduced a cooling function whose dependence on T and p was adopted as physically plausible, but whose normalizing factor is very uncertain. The results were extended to a range of models of various masses and luminosities to see how mass loss varied with stellar parameters (see also Bowen and Willson 1991). The Bowen models produce quite extended hot post-shock regions whose pressure obviously plays a considerable role in driving the mass flow since Bowen (1988) and Bowen and Willson (1991) find that dust is not necessary to drive significant mass loss (although the dust enhances the mass loss rate). However, observations suggest that there is no hot region above the photosphere in Mira atmospheres. The inverted temperature profile of this region would produce molecular bands in emission (Bessell et al 1989), a. spectral feature that is never observed. Theoretical work on the production of large AGB mass loss rates is continuing. Feuchtinger et al (1993) have developed a code for studying dust formation and pulsation in Mira atmospheres; this code contains a more elaborate treatment of radiative transport than the codes of Wood or Bowen. Other current developments related to the theory of mass loss from AGB stars are further reviewed in these proceedings by Gauger et al and Cherchneff and Tielens. Another approach to estimating mass loss rates in AGB stars is to use observations. Over the last ~ 10 years, a. large number of mass loss rate estimates have been derived for AGB stars from CO microwave emission (e.g. Knapp and Morris 1985), and from infrared emission from dust in the cases of very high mass loss rates where the CO mass loss rate estimates become unreliable (Heske et al 1990). These mass loss rates are shown plotted against pulsation period (P) in Figure 1 (see Vassiliadis and Wood 1993 for the list of sources). Given that theory indicates that pulsation plays a prominent role in the production of the large mass loss rates observed in LPVs, it has been assumed that the mass loss rate is primarily a. function of pulsation period. The mass loss rate shown in Figure 1 shows a remarkably rapid increase

P. R. Wood: Evolution of AGB .stars

17

-3.0

1200 P(days)

1600

2000

2400

Fig. 1. Observationally determined mass loss rates in AGB stars plotted against pulsation period P (dots). The dashed line is the mass loss rate according to Reimers' law (with rj — 1/3), and the solid line is the radiation pressure driven mass loss rate M = L/cvexp, where an expansion velocity vexp = 10 km s" 1 has been assumed.

with P up to ~ 500 days (3 orders of magnitude increase in M for an increase in P of ~ 250 days) but for longer periods, the mass loss rate seems to be relatively constant at large values which have become synonomous with the term 'superwind' (Renzini 1981). Presumably, the transition between the two mass loss regimes corresponds to a change in the physical processes producing the wind, perhaps because the dust formation region moves in to the stellar photosphere rather than lying above it in the optically thin region which is supported and extended by pulsation (Bedijn 1988). A widely used formula, for computing mass loss rates in AGB (and first giant branch) stars is that of Reimers (1975). This formula (M = 4 x 1Q~13T]LR/M, with M in M 0 yr" 1 and L. R. and M in solar units) is also shown in Figure 1, where (Wood 1990) L is obtained from the (M(,o/, P) relation for Mira variables. R. is obtained from r^ejj on the giant branch and the definition I, = 47raR 2 T^ / , and M is obtained from the fit M = 0.013 p0.75 normalizing factor 7/ = 1/3 has been used in Figure 1 as this is the value required to give the amount of total mass loss of ~ 0.2 M,^ observed on the first giant branch of globular clusters (Wood and Calm 1976; Sweigert, Greggio and Renzini 1990).

18


The most significant feature of Figure 1 is that the Reimers' mass loss formula produces a mass loss rate that increases much more slowly with P (or L) on the AGB than is indicated by modern observations. This means that if the AGB is to be terminated at luminosities that agree with observed maximum luminosities in Magellanic Cloud clusters, then the mass loss will occur over a longer time interval, and at lower mean luminosities, than occurs with the formula shown in Figure 1. This may have important implications for the formation of carbon stars, since the ease of dredge-up of 12C is enhanced by a large envelope mass (Wood 1981). Attempts to use Reimers' mass loss law with studies of dredge-up (Groenewegen and de Jong 1993) may result in the requirement for enhanced dredge-up efficiencies to compensate for the excess mass loss at low luminosities. Finally, it should be noted that it has long been realized that a Reimers' law is incapable of producing 'superwind' mass loss rates without an enormous enhancement factor t] (Renzini 1981). Although there is clearly a strong dependence of mass loss rate on P as shown in Figure 1, there are also other factors that influence the mass loss rate. A parameter that has been repeatedly shown to influence the mass loss rate of LPVs is the light curve shape, in particular, the fraction / of the period during which the light is rising. Those LPVs that have rapidly rising light curves always show a higher mass loss rate at a given P than the rate exhibited by LPVs with less rapidly rising light curves (Bowers and Kerr 1977; Onaka, de Jong and Willems 1989; Wood 1990; Jura and Kleinmann 1992). Just which intrinsic property causes the light curve to rise more rapidly in some LPVs than others is at present unknown, but a likely candidate is stellar (envelope) mass. Another factor that clearly influences the mass loss rate is the pulsation amplitude (in K, which is close to the bolometric light amplitude) (Whitelock 1990). Once again, at a given period, mass is likely to be the stellar property that influences the pulsation amplitude. The flat 'superwind' part of Figure 1 seems to correspond to stellar winds in which the momentum has been imparted by radiation pressure on dust grains. The ratio of wind momentum to photon momentum (3 = Mvexpc/L was shown by Knapp (1986) to exhibit a strong preference for a. value around 1.0 for Galactic AGB stars, although a few objects had ratios > 10. In the Galactic Bulge, presumably more metal-rich than the solar vicinity, Whitelock et al (1991) found 1/2 < (3 < 2 while Wood et at. (1992) found a similar result in the LMC where the metal abundance is ~ 1/2 solar. To a first approximation then, the mass loss rate in the superwind phase appears to


19

be given by M = /3L/cvexp, where f3 = 1, although it should be remembered that values of /? > 10 may be possible (Netzer and Elitzur 1993). A final property of AGB winds that needs to be examined is the expansion velocity vexp. The expansion velocity of AGB winds can be readily obtained from wind spectral features such as the width of the microwave CO lines (which originate at radii of a few times 10 17 cm) or the separation of the twin 1612 MHz OH maser lines (which originate at radii of a few times 10 16 cm). When plotted as a function of pulsation period (Sivagnanam et al 1989), the expansion velocities show quite different behaviours for P < 500 days and P > 500 days. For P < 500 d, the expansion velocity increases rapidly with P from a few km s" 1 at 300 days to ~ 15 km s" 1 at 500 days. For P > 500 d, the expansion velocity remains approximately constant at vexp ~ 15 km s" 1 . This behaviour mirrors that of the mass loss rate itself (which increases exponentially to P ~ 500 d and remains roughly constant thereafter), adding further weight to the suggestion that the mass loss mechanism changes as the pulsation period increases through 500 days. Finally, it is worth noting that abundance as well as pulsation period appears to affect the wind expansion velocity. This is shown clearly by the OH/IR stars in the LMC (Wood et al 1992) which have mean expansion velocities of ~ 9 km s" 1 compared to ~ 15 km s" 1 for comparable Galactic OH/IR stars. The most likely source of this effect is the lower abundance in the LMC (~ 1/2 solar). It is interesting to note that if the LMC OH/IR stars have the same ratio of (3 = Mvexpc/L as the Galactic stars, then their mass loss rates will be higher for similar luminosities during the superwind phase. On the other hand, higher luminosities might be required in order to produce a superwind in lower metallicity stars.

3 AGB evolution with empirical mass loss rates Using the mass loss rates shown in Figure 1, Vassiliadis (1992) and Vassiliadis and Wood (1993) have computed the evolution of AGB stars including in full the effects of the luminosity variations associated with helium shell flashes. A typical example of the results obtained is shown in Figure 2. The evolutionary sequence shown begins at the first shell flash; it is found that mass loss up to this time is always insignificant. Indeed, the main feature of the mass loss rates resulting from the empirical formula shown in Figure 1 is that mass loss is insignificant except for the last few shell shell flash cycles, and within those cycles the mass loss is concentrated to the high luminosity quiescent phases which precede each flash. The high mass loss (superwind) phases are clearly discrete, and the potential for multiple shell ejections is

20


"3 QO O

3.6 3.5

3.4 4 >

bo

o

I

I I

I

I I

H—I—h

H—I

-HH

3.5 3 H

h

I—I—I

1000 500

0 15

I—I—I—|—I—I—I—|—I—I—I—|—I—I

10 5 I

1

i

i

=+=+

0.8 0.6

H—i—i—|—i—i—i—|—^

8 6 4 _!__!

1

I

2

I i I

'

>

'

1J L

4 6 Time (10 5 years)

8

Fig. 2. The effective temperature T e /y, luminosity L. pulsation period P (days), expansion velocity v,,rp (km s" 1 ), mass M and mass loss rate M (lO-"'3/V/0 yr" 1 ) plotted against, time on the AGB for a star of mass 1 M 0 , helium abundance Y = 0.25 and metal abundance Z=0.00'4.

clear, although the time interval between superwind phases is large, ~ 105 years'for stars of initial mass ~ 1M 0 and decreasing only slowly with initial mass to ~ 7 x l O 4 yr at 2.5 MQ. The reason for the very rapid onset of the superwind phase is largely the exponential increase in A/ with period shown in Figure 1. However, as can be seen in Figure 2, the pulsation period (taken to be the fundamental mode in these calculations) varies enormously during one shell flash cycle


21

from ~ 150 to ~650 days (taking the last flash cycle as an example). This is because P a R2 <x L/T^JJ and Teff decreases while L increases towards quiescent maximum. Note that optically-visible Mira variables in the solar vicinity mostly have periods in the range ~ 250 to ~ 475 days (Wood and Cahn 1977) so that a given AGB star transits the complete range of Mira periods in one shell flash cycle . It should be noted that the periods shown in Figure 2 are those that the star would have if it were pulsating, but extant pulsation calculations are not capable of determining the stability of AGB stars with enough reliability to say whether the stars will actually be pulsating or non-variable. The superwind mass loss rate used by Vassiliadis and Wood (1993) assumed /? = 1. However, as noted above, (3 is observationally uncertain by a factor of ~ 2 and values of up to 10 are plausible theoretically. If (3 = 2 had been chosen for the AGB calculations, the pre-superwind mass loss rate would be unaffected but the mass loss rate would rise to higher values so that fewer flash cycles would be needed to remove the hydrogen-rich envelope and turn the AGB star into a planetary nebula nucleus. For stars of initial mass ~ 1 M Q , one superwind phase could potentially completely remove the envelope, the actual behaviour depending on the phase of the flash cycle at which the star first reached the superwind limit. In general, however, it would be difficult to avoid at least two superwind phases in most stars, meaning that towards the end of their lives, most ACiB stars should have at least one episode where they regress from being dusty objects in a superwind phase to ordinary optically-visible AGB stars. Many such AGB stars have now been found using the IRAS data base (Willems and de Jong 1986; Zijlstra et al 1992), the signature of these objects being a high 60//, to 25/i flux ratio resulting from the cool dust in the hollow shell formed by the cessation of mass loss. The results of Zijlstra et al (1992) show that the hollow shells are much more common among carbon stars than M stars. The calculations of Chan and Kwok (1988) show that the hollow shells should be observable in the IRAS 2-colour diagram for ~ 104 years. Given an interflash lifetime of ~ 105 years, with ~ 1/2 of that time being a stage when the superwind is not blowing, it is therefore reasonable to expect ~ 20% of optically-visible C stars to show evidence for a hollow shell. Zijlstra. et al (1992) give an estimate of ~ 40% of optically-visible C stars showing evidence for hollow shells, in reasonable agreement with theoretical expectation. It is well known that at helium shell flashes, there is a brief (~600 yr) spike in luminosity which has a peak a factor of 1.5 or more brighter than the quiescent luminosity maximum (see second panel of Fig. 2). This luminosity peak also gives a spike in the mass loss rate (bottom panel of Fig.2 ) but,

22


due to its short duration, the amount of mass lost in the spike is small, being < O.OlM© in the calculations of Vassiliadis and Wood (1993). Olofsson et al (1992) (see also Olofsson, these proceedings) have found a small number of C stars in which a hollow mass loss shell has been spatially resolved with the SEST. Furthermore, the mass loss shells in these stars appear to be thin with a mass of ~ 0.02-0.04 M® and Olofsson et al suggest that they are the result of mass loss occurring during the luminosity spikes associated with helium shell flashes. If this is the case, then the total mass lost is somewhat greater than predicted by the Vassiliadis and Wood calculations. Another place that the mass loss rate enhancement produced by a luminosity spike might be expected is in the Mira variables It Aql and R Hya. These two Miras are currently rapidly decreasing in pulsation period and are in the phase of luminosity decline from a helium shell flash spike (Wood and Zarro 1981). In the case of R Aql, the observed mass loss rate (~ 10~6 M 0 yr" 1 ) is not at superwind levels but it is still some two orders of magnitude higher than expected for a Mira its period of ~ 280 d (Wood 1990). On the other hand, R Hya has a smaller mass loss rate than expected (~ 10~7 M s yr" 1 at ~ 380 days). In neither of these two cases will a significant amount of mass be lost in the luminosity spike interval of < 1000 yr. Figure 2 indicates that this behaviour would be expected if these stars are not undergoing one of their last few shell flashes. The significance of the luminosity spikes as producers of mass loss may well be affected by the abundance of the star. This is shown in Figure 3, where the quiescent luminosity maximum LQ for each shell flash cycle is plotted against core mass Mc. Also plotted in Figure 3 is the peak luminosity Lp in each luminosity spike. Both luminosities increase linearly with Mc but the height of the spike above quiescent maximum is much greater in the low metallicity stars. This may mean that low metallicity stars (for example, population II stars, globular cluster stars, or more massive stars in low metallicity systems) will throw off more mass at shell flashes than more metal rich stars. The larger the mass of an AGB star through the low mass range up to ~ 2.5 M 0 , the greater is the number of shell flash cycles required to remove the envelope in superwind phases (being ~ 3 flashes at 2.5 MQ in the Vassiliadis and Wood 1993 calculations). For even more massive AGB stars (>3.5 M 0 ), mass loss appears to behave quite differently because in these stars the external luminosity is almost unaffected by the shell flashes (except for the first few in each star). While mass loss remains insignificant in these massive AGB stars, the star slowly increases in luminosity. When the star reaches luminosities high enough to produce a superwind, the stellar mass

23

P. R. Wood: Evolution of AGB stars 1

1

0

/

3x10*

"/

y o x

V

in

7~ L P

/ /

/

/

Of Teff > 10000 K'. This has much to do with the way the photospheric radiation field controls the wind ionization state - in effect the line-blocking due to a hot-star wind shifts longward or shortward in wavelength in step with the shift in the wavelength of peak photospheric emission. Encouraged by this, Abbott provided a global fit to the force multiplier in which k — 0.28, a = 0.56 and 6 — 0.09 (cf. the later recommendations of Pauldrach et al. 1986, Pauldrach et al. 1990 which notably tend tdwards higher values of a ) . As this amounts to a massive simplification of a complex problem, the inclination to use an 'off-the-peg' parameterisation of the force multiplier understandably remains strong. It is often the option chosen. To digress briefly, another adaptive pattern of behaviour we can expect of hot star winds is in their thermal balance. Again the root cause is the close link between the photospheric radiation field and the ionization and

30

J. E. Drew: Hot star winds

excitation of the radiation-driven wind. Time-dependent modelling has not changed the expectation that most of a 'typical' hot-star wind is in the form of cool gas maintaining radiative equilibrium. Equilibrium wind temperature profiles, Te(r), derived by Drew (1989) scale in a strikingly simple way as a function of stellar effective temperature (see Figure 1 in Drew 1990). In fact the following numerical fit:

1

(5)

reproduces the model grid of calculated O-star wind temperature profiles to within 10 percent (the grid encompasses dwarfs through to supergiants and a range of velocity laws). Gayley & Owocki (199-1) have examined and discount additional kinematic heating processes, excepting the ion-drag frictional heating discussed by Springmann & Pauldrach (1992) which only interferes substantively with radiative equilibrium in the lowest-density winds (e.g. in r Sco, BO V). In contrast to the effective temperature insensitivity of the force multiplier, there is a marked dependence on chemical composition (Abbott 1982 in the mean). This follows primarily from the obtained M oc (Z/ZQ)0A4 fact that H and He contribute very little line acceleration. So while it can be acceptable to transfer force multipliers derived for one star or group of stars to another maintaining constant abundances, it is dangerous to do this if there is a significant abundance change. Kudritzki et al. (1987) give representative values of k, a and S for LMC, SMC and Galactic abundances.

3 The Wolf-Rayet wind momentum problem — a solution A longstanding problem in the study of Wolf-Rayet mass loss is the apparent shortfall in the available supply of momentum from the i-a.dia.tion field: it has been found that empirical estimates of the wind momentum rate Mv^ are often an order of magnitude higher than the radiant momentum rate L\>o\lc (Barlow, Smith & Willis 1981; and more recently - Schmutz, Hamann & Wessolowski 1989). If radiation pressure is to drive such winds, every stellar photon has to undergo repeated scatterings so that its momentum can be exploited repeatedly. Is this possible? A recent paper by Lucy and Abbott (1993) suggests quite plausibly that it is. The reason lies in another example of adaptive behaviour, like the effective temperature insensitivity of the force multiplier noted above. In contrast to the winds of normal OB stars, Wolf-Rayet winds are stratified such that ionization decreases outward. This is a response to what may be thought

J. E, Drew: Hot star winds

31

of as the radial cooling of the ambient radiation field (there is no classical photosphere in a WR star). Lucy & Abbott demonstrate how the ionization gradient causes the line-blocking to shift longward in wavelength with increasing radius (see their Figure 5). This creates the conditions needed for multiple-scattering to work as required, since the effect of many scatterings is a progressive redshift. In short, the wind opacity moves to where the photons are to be found. The specific calculation performed by Lucy & Abbott (1993) is for a WN5 star for which L/iMv^c) ~ 10 is achieved.

4 Successes and failures of modified CAK theory Since the inclusion of the finite disk correction, the time-independent theory of radiation driven winds has done a good job of matching the observed run of mass loss rate against bolometric luminosity for normal OB stars. Predicted and observed mass loss rates vary from around 10~ 8 M© yr" 1 for the latest 0 dwarfs up to as much as 10" 5 M© yr" 1 for the earliest 0 supergiants. However, nothing is perfect and some dispute continues (e.g. the recent reassessment by Lamers & Leitherer 1993). The current situation is not quite as satisfactory when it comes to the comparison between measured and predicted terminal velocities. On the basis of a relatively modest sample of OB stars, Groenewegen, Lamers & Pauldrach (1989) showed that predicted wind terminal velocities are on average too high by around 40 percent. This picture was substantially confirmed by Prinja, Barlow & Howarth (1990) whose sample included more than 200 stars. What is wrong? An obvious candidate for the blame is the force multiplier. Perhaps the recent drift toward a ~ 0.7 is inappropriate - much better agreement would be achieved for a between 0.5 and 0.6 (see equation 3). Another factor that may be at work has to do with the influence of wind-shocking on the kinematic structure (see below). Another way of testing the validity of modified CAK theory is to compare the predicted wind ionization with that implied by ultraviolet observations of wind-formed resonance line profiles. Cassinelli & Olson (1979) were the first to point out that 0 VI AA1O32,38 absorption, found to be widespread in O-star spectra obtained by the Copernicus satellite, was too strong to be the product of O 4 + valence shell ionization - they proposed instead Auger ionization of the dominant O 3 + ion by a modest flux of X-rays. Soon after, soft X-ray emission from OB stars was detected by the Einstein satellite (Lx ~ 10~7i/bob s e e Cassinelli 1985). However, Cassinelli k Olson's conjecture was apparently contradicted by the modelling of Pauldrach (1987) wherein sufficient O 5 + was obtained just from valence shell ionization. Mak-

32


ing similar assumptions about the photospheric radiation field and wind density profile, Drew's (1989) modelling supported Cassinelli & Olson's interpretation. Drew suggested that Pauldrach's models might overestimate the 0 5 + fraction because of the use of elevated wind temperatures. Since then, Groenewegen & Lamers (1990) and MacFarlane et al. (1993) have identified Pauldrach's omission of dielectronic recombination from his models as a second factor leading to too high a predicted O 5 + abundance. The work of MacFarlane et al. shows for the case of £ Pup (04 If) that mixing X-ray emission of the magnitude observed inward to at least r ~ 2i?» solves the 0 vi problem. The observed strength of the Si iv A1397 P-Cygni feature also poses a problem - Si 3+ is easily destroyed by photoionization in 0 star winds and so the Si IV line should not be as prominent as it is observed to be. There are several examples in the literature of fits to UV spectra of hot stars that notably fail to match this line (e.g. Kudritzki et al. 1991). Drew (1989) suggested that, just as the wind-shocking invoked to explain the soft X-ray emission could solve the 0 vi problem, it could also make good the Si iv deficit. This is because wind-shocking provides both the source of X-rays needed for 0 5 + production and compressed zones within which the Si 4+ recombination rate can be enhanced by one or two orders of magnitude. However it should be noted that CAK theory does explain the trend that gives rise to the 'Si IV luminosity effect" (Walborn & Panek 1984: discussed by Drew 1989 and Pauldrach et al. 1990). It became apparent through the 1980s that line-driving is naturally unstable and that instabilities may amplify, eventually steepening into full-blown shocks (see Owocki, Castor & Rybicki 1988 and references therein). Thus the basic radiation-driven wind concept contains within it the physics required to explain both the observed X-rays and and also a range of otherwise troublesome spectroscopic phenomena, including those just described.

5 Wind-shocking and time-dependent theory In performing the first 1-D hydrodynamical simulation of the growth of instabilities in a radiation-driven wind, Owocki, Castor & Rybicki (1988) found that the resultant wind structure was one swept by a sequence of reverse shocks. In other words, the shocks were found to expand outward more slowly than the host fluid - the outflow stumbles over itself. This carries the implication that the thin post-shock shells of cool compressed gas, which will contain most of the integrated wind column density, travel at speeds below those predicted by the time-independent theory (see Fig-


33

ure 1 of Puls, Owocki & Fullerton 1993). Simulation of unstable outflow is not yet advanced enough to reliably quantify this effect, but it may be a partial explanation for the remaining discrepancy between observed and predicted terminal velocities (discussed in section 3). So far no two- or three-dimensional simulations have been published, nor have the 1-D examples been continued on to the large radii sampled by radio emission. The details of the early scenario for wind-shocking proposed by Lucy (1982) have not been borne out by either the recent hydrodynamical simulations or by the shock temperatures deduced from fits to the observed X-ray emission (the greater sensitivity of ROSAT to very soft X-ray emission has not changed this: e.g. Cassinelli et al. 1991). Lucy envisaged a tightly-spaced train of soft shocks within hot star winds, radiating at a few times 105 K. Simulation and observation favour fewer, harder and more widely-spaced shocks radiating at a few times 10(:> K. However, Lucy (1984) has argued compellingly that the typically high optical depth presented by the inner wind to scattered line radiation must defer shock growth to larger radii. Hence, for more luminous stars with substantive winds, acceleration through the outflow critical point (whose properties fix M) will be untroubled by the kinematic and radiative effects of shocking of the outer wind. However, at around Snow & Morton's (1976) bolometric magnitude limit of ~ —6, the line-driving instability causes a qualitative change in the wind physics. For Mboi ~ - 6 , modified CAR" theory would suggest M ~ 10~ 8 M 0 yr" 1 . For mass loss rates of this order, the flow timescale begins to drop below the typical post-shock cooling time. Once heated to T^ilO 6 K by shocking, such winds stay hot and radiate accordingly. An outline discussion of the physical effects involved may be found in Castor (1987). We may picture lower-luminosity OB stars near the Snow & Morton limit as enclosed within hot tenuous envelopes of soft X-ray emitting gas created by windshocking. This inability to recover from shocking can explain the anomalously low terminal velocities associated with late-0 main sequence and lower luminosity stars (Abbott & Friend 1989). Their winds, once shocked, are too highly-ionized to yield the ultraviolet opacity needed for continued acceleration. Relatively extreme cases have now been identified in which windshocking occurs near enough to the critical radius to significantly irradiate it and so lower the mass loss rate: for the B stars, ft and e CMa, the derived mass loss rates are ~5 times below the predictions of modified CAK theory (Drew et al. 1994). This is apparently how radiation driving turns off at the lower bolometric luminosity limit - not gradually by simple extrapolation of the Lbol - M relation, but more suddenly as the consequences of the

34


line-driving instability overwhelm. Qualitatively this amounts to another success of a very fruitful wind theory. Acknowledgements JED is presently in receipt of an Advanced Fellowship funded by the Science & Engineering Research Council of the United Kingdom. References Abbott, D. C , 1982, Astrophys. J., 259, 282 Abbott, M. J. k Friend, D. B., 1989, Astrophys. J., 345, 505 Barlow, M. J., Smith, L. J. k Willis, A. .]., 1981, Mon. Not. R. ash: Soc, 196, 101 Cassinelli, J. P., 1985, in The origin of nonradiative healing/momentum in hot stars, eds. A. Underhill k A. Miclialitsianos, NASA Conf.Publ. 2:158, 2 Cassinelli, J. P., Cohen, D. H., MacFarlane, J. J., Sanders. VV. T. k Welsh, B. Y., 1994, Astrophys. J., 421, 705. Cassinelli, J. P. k Olson G. L., 1979, Astrophys. J., 229, 304. Castor, J. I., Abbott, D. C. k Klein, R. I., 1975, Astrophys. J., 195, 157 (CAK) Castor, J. I., 1987, in Instabilities in Luminous Early-Type Stars, eds. H.J.G.L.M. Lamers k C.W.H. de Loore, D. Reidel, The Netherlands, pl59 Drew, J. E., 1989, Astrophys. J. Suppl., 71, 267 Drew, J. E., 1990, in Properties of hot luminous stars, ed. CD. Gannany, ASP Conf. Series, 7, 230 Drew, J. E., Denby, M. k Hoare, M. G., 1994, Mon. Not. R. ash: Soc, 266, 917. Friend, D. B. k Abbott, D. C , 1986, Astrophys. J., 311, 701 Garmany, C. D. fcConti, P. S., 1984, Astrophys. J., 284, 705 Gayley, K. G. k Owocki, S. P., 1993, Astrophys. J.. (in press). Groenewegen, M. A. T. k Lamers, H. J. G. L. M., 1990, Astron. Astrophys., 243, 429 Groenewegen, M. A. T., Lamers, H. J. G. L. M. k Panldrach, A. W. A., 1989, Astron. Astrophys., 221, 78 Kudritzki, R. P., et al., 1991, in Massive stars in starbursts, eds. C. Leitherer, N. R. Walborn, T. M. Heckman k C. A. Norman, CUP, Cambridge, p59 Kudritzki, R. P., Pauldrach, A. W. A. k Pnls, J., 1987, Astron. Astrophys., 173, 293 Lamers, H. J. G. L. M. k Leitherer, C , 1993, Astrophys. J., 412, 771. Lucy, L. B., 1982, Astrophys. J.,255, 286 Lucy, L. B., 1984, Astrophys. J., 284, 351 Lucy, L. B. k Abbott, D. C , 1993, Astrophys. J., 405, 738 Lucy, L. B. k Solomon, P. M., 1970, Astrophys. J.. 159, 879 MacFarlane, J. J., Waldron, W. L., Corcoran, M. F., Wolff". M. .1., Wang, P. k Cassinelli, J. P., Astrophys. J., 419, 813. Owocki, S. P., Castor, J. I. k Rybicki, G. B., 1988. Astrophys. J.. 335, 914 Pauldrach, A. W. A., 1987, Astron. Astrophys., 183, 295 Pauldrach, A. W. A., Kudritzki, R. P., Puts, .]. k Butler, K., 1990, Astron. Astrophys., 228, 125 Pauldrach, A. W. A., Puls, J. k Kudritzki, R. P., 1986, Astron. Astrophys., 164, 86 Poe, C. H., Friend, D. B. k Cassinelli, J. P., Astrophys. J., 337, 888 Prinja, R. K., Barlow, M. J. k Howarth, I. D., 1990, Astrophys. J., 361, 607 Puls, J., Owocki, S. P. k Fullerton, A. W., 1993, Astron. Astrophys., 279, 547. Schmutz, W., Hamann, W.-R. k Wessolowski, U., 1989, Astron. Astrophys., 210, 236 Snow, T. P. k Morton, D. C , 1976, Astrophys. J. Suppl., 32, 429 Springmann, U. k Pauldrach, A., 1992, Astron. Astrophys., 262, 515 Walborn, N. R. k Panek, R. J., 1984, Astrophys. J., 280, L27

Axisymmetric Outflows from Single and Binary Stars Mario Livio Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA and Dept. of Physics, Technion, Haifa 32000, Israel

Abstract Mechanisms that can produce axisymmetric planetary nebulae are reviewed. It is suggested that the "interacting winds" model, in the presence of a "density contrast" between the equatorial and polar directions, can produce most of the observed morphologies. Mechanisms that can produce a density contrast are examined and it is suggested that binary stellar companions and even brown dwarfs and massive planets may play an important role in the generation of such a contrast, either via common envelope evolution or by spinningup the envelope of the AGB star. It is shown that the statistics of planetary nebulae of different morphological types are consistent with models which rely on the presence of binary companions.

1 Introduction Axisymmetric outflows are associated with many nebulae (e.g. He 2-36, BI Cru, My Cn 18, IC 4406, K 3-72, Corradi k Schwarz 1993a, b, c; OH 17.72.0, La Bertre 1986; R Aquarii, Burgarella k Paresce 1992) and with Be stars. I will concentrate in the present review mainly on planetary nebulae (PNe). An examination of the catalogue of narrow band images of Schwarz, Corradi and Melnick (1992) and other images in the literature reveals a few interesting morphologies. In some cases, almost perfect rings are observed (e.g. ScWe 3, ScWe 2, Schwarz, Corradi k Melnick 1992; Hen 1357, Bobrowsky 1993). In others, a pronounced bipolar structure is extending from a ring (e.g. He 2-104, Schwarz et al, 1992; NGC 2346, Bond k Livio 1990). Some PNe exhibit an exact point symmetry (e.g. IC 4634, J 320, Schwarz et al. 1992). Finally, in some PNe highly col Unrated outflows ("jets") are observed (e.g. NGC 7009, Schwarz et al. 1992; Kl-2, Bond k Livio 1990). In the present work, I examine different potential mechanisms for the formation of axisymmetric outflows and the different morphologies that are observed. 35

36

M. Livio: Axisymmeiric

outflows from single and binary stars

2 The Shaping of Planetary Nebulae PNe are very probably bubbles, formed when a tenuous but fast (V ~ 1000 km s" 1 ) wind, emitted by the exposed hot nucleus, catches up with the slowly moving (V ~ 20 km s" 1 ) wind, ejected by the AGB star, and shocks it. This "interacting winds" model has been suggested for the shaping of PNe around single stars (Kwok 1982; Kahn 1982). Balick (1987) proposed a classification scheme, in which all PNe are divided into "spherical," "elliptical" and "butterfly," with the different morphology classes being the consequence of an increasing "density contrast" (in the slow wind) between the equatorial and polar directions. Balick suggested that "spherical" PNe form when there is no "contrast" in the AGB wind, while "butterfly" PNe ensue when material near the equator is considerably denser than material near the poles (thus making it much easier for the fast wind to penetrate in the polar direction). The exploratory work of Soker and Livio (1989) and the more recent, highresolution gas dynamics simulations of Icke, Balick and Frank (1992), Frank et al. (1993; and see Icke, these proceedings) have demonstrated that bipolar flows are indeed obtained by the "interacting winds" model, in the presence of a "density contrast." Furthermore, the work of Frank et al. (1993) has shown that a very wide range of PNe morphologies, ionization structures and kinematics can be reproduced by this model, when the inclination angle at which the nebula is viewed is taken into consideration. The main questions that need to be answered are therefore: (i) what is the mechanism that produces the density contrast in the slow wind? (assuming that this represents the correct model) or (ii) are there any other shaping mechanisms that can produce the observed axially symmetric morphologies?

3 Mechanisms that Can Produce a Density Contrast or Axially Symmetric Outflows I will now review different mechanisms that can, in principle at least, generate a density contrast or produce by themselves an axially symmetric outflow. Some of these mechanisms involve single stars and some binary systems.

3.1 Common Envelope Evolution The term common, envelope (CE) is usually used to describe a configuration in which a binary system (typically consisting of the core of a giant and a

M. Livio: Axisymmetric outflows from single and binary stars

37

secondary star) revolves inside an envelope which is not corotating with the binary and not even necessarily in hydrostatic equilibrium. The development of a CE usually accompanies the event of mass being transferred from one star to the other on a dynamical timescale (see e.g. Iben & Livio 1993 for a recent review), this being the consequence of the mass losing star being unable to contract as rapidly as its Roche lobe. Such mass transfer events can occur when mass is being transferred from a star which possesses a deep convective envelope (such as an AGB star; in which case it tends to expand upon mass loss) onto a less massive companion (this leads to a shrinkage of the orbit). A dynamical mass transfer event may ensue also as a consequence of the instability discussed by Lai, Rasio & Shapiro (1993a, b; this is probably most applicable for double degenerate systems). Since the timescale of mass transfer in these cases is much shorter than the timescale on which the secondary star is able to adjust thermally, the outer layers of the secondary start expanding, and the system rapidly evolves to a configuration in which the core of the mass losing star and the original secondary are embedded inside a common envelope. Due to frictional drag, the internal binary starts spiralling-in (e.g. Taam and Bodenheimer 1989, 1991; Livio and Soker 1988). The final outcome of the CE phase can be a dramatic reduction in the binary separation (or even a merger of the two components in some cases) accompanied by the ejection of the envelope due to the deposition of orbital energy (other sources of energy may contribute, see Iben & Livio (1993)). This course of CE evolution is responsible for the formation of PNe with close binary central stars (Bond & Livio 1990; Livio 1993; Yungelson et al. 1993), cataclysmic variables (Paczynski 1976), double degenerate systems (Iben & Tutukov 1984; Webbink 1984) and in fact all the binaries containing at least one compact component, with an orbital period of less than a few days. One of the most important results of multi-dimensional hydrodynamic studies of the CE phase, has been the demonstration that mass ejection (due to the spiralling-in binary) occurs preferentially in the orbital plane (Livio & Soker 1988; Taam & Bodenheimer 1989; Term an, Taam & Hernquist 1994, see Fig. 1). Thus, the generation of a density contrast between the equatorial (orbital plane) and polar directions is an almost inevitable consequence of CE evolution. A higher contrast can be expected for more massive secondaries (for a given primary configuration, see Livio & Soker 1988). It is interesting to note that two of the well known PNe which exhibit bipolar outflows (NGC 2346 and Kl-2, Bond & Livio 1990) are known to contain close binary nuclei.

38

M. Livio: Axisymmetric outflows from single and binary stars 1

'

I

I Tto.i. o. I

1

1

1

TfSmoTT

1

1

1

inm. ia,.»

Tlma - 1S0.4

Fig. 1. The particle distribution in the CE in the (x- z) plane. About 80% of the ejected matter was found to be confined to within 30 degrees from the equatorial plane. From the calculation of Terman, Taam & Hernquist (1994).

3.2 Magnetic Fields Magnetic fields are often invoked to explain the formation of configurations which possess axial symmetry. In the case of the shaping of planetary nebulae, magnetic fields could operate (in principle) in at least three different ways: (i) by centrifugally accelerating material in the equatorial plane (in the slow wind), to form a density contrast (to the best of my knowledge, this has actually not been previously suggested for PNe), (ii) by the evolution of a large scale azimuthal field that is embedded in an equatorial torus (in the


39

slow wind), and (iii) by exerting magnetic tension in the equatorial plane in the fast wind (and thus inducing a bipolar flow). In order to obtain a significant centrifugal acceleration of the slow wind, the minimal surface field strength that is needed is (e.g. Michel 1969; Blecher & MacGregor 1976)

M V

/ 2

/

VSw 15 km s

w

-

-l

UU

f

10

where "*" denotes surface values, Vsw iS the speed of the slow wind, VTOt is the equatorial rotation velocity and all other symbols have their usual meaning. Pascoli and co-workers (e.g. Pascoli 1990; Pascoli et al. 1992) suggested that the time evolution of a large-scale azimuthal magnetic field naturally leads to a bipolar morphology. These authors, however, have been somewhat unclear (at least to the present author) about the mechanism generating their initial configuration (in fact, Pascoli et al. 1992 assumed the ejection of an equatorial torus). In any case, for their model to work, Pascoli et al. require at the base of the convective zone and azimuthal field of at least Bv ~ 1000 G. From a dynamo model of fully convective stars (Tout k Pringle 1992) we can estimate the surface field to be

where rj ~ 3R*/lc (lc is the mixing length) and 7 measures the efficiency of the dynamo regeneration term. The azimuthal field component in the convective region can be estimated to be

*

77

L

\2/9

f R. y1*'* \AOORZ)

{

0.1

(3)

where Slcrit is the critical angular velocity, ilcru = (GM./ A comparison of eqs. (2) and (3) with the above requirements (eq. (1) and the required Bv) reveals that the minimum field strength required

40

M. Livio: Axisymmetric


by the model of Pascoli et al. may be difficult to achieve in AGB stars. For centrifugal acceleration to have any effect it is required that the AGB star will rotate at (il*/£lCrit) > 0.02. This requirement is not easy to satisfy for single AGB stars, because of their large moments of inertia. Even when the increase in central condensation is taken into account, typically (n,/£lcrh)AGB ~ (0.01 - Q.l)(n,/ncrit)MS (see also Eriguchi et al. 1992), MS and for low mass stars ($l/SlCrit) ~ 0.01. However, the above requirement can be satisfied if the AGB star is spun-up by a companion (even of very low mass, see discussion on rotation in 3.4 below). In a recent work, Chevalier and Luo (these proceedings) suggest that the magnetic field may become important in the shocked wind bubble. They then obtain (under some simplifying assumptions) that a bipolar flow can be generated due to the presence of magnetic tension in the equatorial plane (and its absence in the polar direction). Chevalier and Luo show that a significant effect is obtained for a surface magnetic field strength of the central star larger than

1/2

/ RN r 1 ( y

m ( ) 2000kms-V U O " cm/ V 0.1 / ' where a is the ratio of the field energy density to the kinetic energy density in the wind, Rjy is the radius of the central star, Vp\y is the speed of the fast wind and Vrot is the equatorial surface rotational velocity of the central star. Again we notice (as in the case of centrifugal acceleration), that the magnetic tension can become more effective if the rotation velocity of the star is a non-negligible fraction of the wind velocity. 3.3 Rotation and Be Stars In a recent important work, Bjorkman and Cassinelli (1992, 1993) proposed that the winds from rapidly rotating (early type) stars are very effectively focused towards the equatorial plane. This is essentially a consequence of conservation of angular momentum in a case in which the gravitational force exceeds the radiation pressure force over a sufficiently large distance. If the velocity with which the gas arrives to the equator is supersonic, a wind compressed equatorial outflow ("disk") forms (see Fig. 2). Bjorkman and Cassinelli (1993) used simplified analytic solutions to show that the formation of the equatorially compressed outflow occurs for Vrot/Vcrit ~ 50% (for



41

Equatorially Compressed Outflow

Fig. 2. Diagram of the compressed equatorial outflow produced by the wind from a star rotating at half the critical velocity. From the calculations of Bjorkman k Cassinelli (1993).

B2 stars; close to 90% for 0 stars). The analytic results of Bjorkman and Cassinelli were essentially confirmed (small differences exist) by the twodimensional hydrodynamic simulations of Owocki, Cranmer and Blondin (1994; and see also Blondin, these proceedings). An examination of the results of Bjorkman and Cassinelli (1993; see also Owocki et al. 1994) shows that the important parameter in determining whether or not an equatorially compressed outflow forms, is the ratio of the (equatorial) rotation velocity to the wind velocity. The threshold values of Vrot/V<x> (where V^ is the wind velocity at large distances), above which the compressed "disk" occurs are shown in Fig. 3. Since V^ rapidly increases for earlier spectral types than B2 (e.g. Prinja, Barlow & Howarth 1990; Bjorkman 1989), these stars must rotate faster in order to form an equatorial "disk." This may explain the decrease in the frequency of Be stars towards earlier spectral types than B2. Bjorkman and Cassinelli (1993) showed (using some simplifying assumptions) that when the dependence of the force multiplier on the ionization balance in the wind is included, a minimum

42



Voo/V. esc

Fig. 3. The threshold values of KW^co (see text) above which an equatorially compressed outflow is obtained. Calculated on the basis of the results of Bjorkman & Cassinelli (1993). in the threshold (for "disk" formation) rotational velocity is obtained near spectral type B2. They suggested that this may explain why the maximum frequency of Be stars occurs near B2 (however, this problem is far from being settled, since there exists still a large discrepancy between the theoretical and observed terminal wind speeds at late spectral types). At any rate, the work of Bjorkman and Cassinelli (1993) and Owocki et al. (1994) has shown that in the case of radiatively driven winds in early type stars, once the ratio Ko«/Kx> exceeds some critical value (which depends on the ratio of V^ to the escape speed from the stellar surface, see Fig. 3), an equatorially compressed outflow forms.

3.4 Rotation and AGB Stars—Common Planets and Brown Dwarfs

Envelopes,

It is important to examine the question of whether an equatorially compressed outflow of the type obtained for early type stars (3.3 above) can also form in AGB stars, since this will clearly generate an equatorial to polar density contrast. The first thing to note is that unlike in the case of early type stars, for AGB winds, typically V^/Vesc^l (e.g. Reimers 1977; Weyman 1962; Likkel et al. 1992; Knapp & Morris 1985; Zuckerman & Dyck 1986). An examination of Fig. 3 then suggests that the threshold rotational velocity for the equatorial outflow to form is V ^ / V ^ ~ 0.25 (which corresponds approximately to ft*/ftcrtt ~ 0.35; or it could be as low as fi*/ficrtt ~ 0.2 if V^,IVesc ~ 0.5). Secondly, while the exact mechanism of wind acceleration in AGB stars is


43

not fully understood (radiation pressure on dust is probably important; see e.g. poster paper by Netzer and review by Wood, these proceedings), the following is very probably true. If the AGB star were to rotate sufficiently fast (for centrifugal support to be important), then the distance between the sonic point in the wind and the location of loss of centrifugal support can be expected to be sufficiently large, for gravity to focus the flow towards the equator. In the absence of direct observational material, the main question that needs to be answered of course is: can stars be expected to rotate as fast as ft*/Merit ~ 0.35 at least during some stage of their AGB phase"! As discussed in Section 3.3 above, such high rotation rates are highly unlikely for truly single AGB stars. However, it is interesting to examine the possibility of the star being spun-up by a companion, either tidally or via a CE phase. 3.4-1 Spin-up in a CE phase During the spiralling-in process, the internal binary deposits orbital angular momentum into the CE and it spins it up (see e.g. Iben and Livio 1993 for a review). For example, in a three-dimensional calculation of the CE phase of a 4.67M0 red giant with a 0.94MQ companion, Terman, Taam and Hernquist (1994) found that material in the vicinity of the double core has been spun-up to ~ 40% of the angular velocity of the two cores. Furthermore, we can use the fact that the density in the convective region of AGB configurations (which constitutes most of the envelope mass) behaves approximately as p ~ r~2 (e.g. Soker 1992). Tf we then assume that the convective region can be brought to nearly solid body rotation (due to strong turbulent viscosity coupling), then a secondary of mass A/2 depositing even only 20% of its angular momentum into the envelope (corresponding roughly to the minimum efficiency of energy deposition obtained in hydrodynamic calculations, e.g. Taam & Bodenheimer 1989; Livio k. Soker 1988; Terman et al. 1994), can generate an angular velocity of fi»/QCrtt ~ Q.%M2/Menv). Therefore, AGB stars with stellar companions that are sufficiently close to go through a CE phase can certainly be brought in many cases to the rotation rates that are necessary for the production of the equatorially compressed outflow (and thereby for the generation of a density contrast in the slow wind). 3.4-2 Spin-up by brown dicarfs and planets I shall now examine the question of the spin-up of a single AGB star by brown dwarfs or planetary companions. The potential role of planets and

44

M. Livio: Axisymmeiric


brown dwarf is very relevant, since recent observations of many young stellar objects (e.g. in the Taurus-Auriga dark clouds, in the Orion nebula, in the L1641 molecular cloud) reveal that perhaps most (if not all) solar type stars are initially surrounded by disks (see e.g. Beckwith et al. 1990; O'Dell, Wen k Hu 1993; Strom, Strom k Merrill 1993 and references therein). While some recent searches (e.g. Murdoch, Hearnshaw k Clark 1993) find brown dwarfs to be rare in orbits closer than 10 AU, brown dwarfs (and high-mass planets) could still be found in wider orbits. In a series of recent works, Tassoul and Tassoul (1990,1992 and references therein) suggested that a hydrodynamic spin-down (up) mechanism can be much more efficient in synchronizing close binary systems than equilibrium tides (which are based on diffusive transport of angular momentum). The hydrodynamic mechanism operates by large-scale meridional transport of angular momentum, which is regulated by an Ekman-type suction layer (in a somewhat similar manner to the spin-down of a stirred cup of tea). On the basis of the results of Tassoul and Tassoul (1992) and Zahn (1977) it is easy to show (see also Soker 1994) that the orbital decay time of a companion around an AGB star is given approximately by Tdec ~

1.2 x 105 yr

MxV 1 / 8 M&)

\ Mi )

V300/

where 6 is the ratio of the synchronization time to the spin-down time, q = M2/M1 is the mass ratio, a is the initial separation and all other symbols have their usual meaning. I now make the following assumptions: 1. The luminosity on the AGB is given approximately by (Eggleton, private communication) x

1O 5 - 3 (A/ C /M 0 ) 6

_ f>*/

+ 10°-5(A/c/A/®)5'

where Mc is the core mass. 2. The radius of the AGB star is related to its luminosity and mass by (Eggen & Iben 1991) p

/ T

\ 0.68 / 1/

\ -0.1(5

A ~ (h.) (EL)


45

3. Mass loss rate is approximately given by a Reimers (1977) type formula

While this formula clearly does not represent the exact mass loss rate at all phases (see e.g. the review by Wood, these proceedings), it is adequate for our present limited purposes. 4. The relation between the total mass and the core mass is given by the results of Iben and Tutukov (1985, their Fig. 30). Under these assumptions, I calculated the evolution of a system consisting of an AGB star (with an initial mass on the AGB of A/° = IMQ and a core mass of Mc = 0.6MQ) and a low mass companion. The calculation showed that for an initial separation cio satisfying oo/R® ~ 5.5 (corresponding to do ~ 2020RQ), the orbit decayed completely during the AGB phase. For an initial separation cio ~ 2OOO.R0, the deposition of the orbital angular momentum into the envelope, could spin the envelope up to ftro 1 and j3 = 4/3 if Mp < 1. We first make a general point regarding the effects of mass loading on flows in the context of steady 1-D flows, but one which turns out to have widespread implications. The continuity and momentum equations (cf. equations (5) of H86) give a schematic equation for the variation of the flow Mach number Mp with the distance coordinate r, dMF _ \ [TERMS] - [MASS - LOADING TERM] ]

~dT~\

(Ml - 1)

/•

()

In this equation, [TERMS] would include the effects of geometrical divergence, etc. Irrespective of whether a flow is subsonic or supersonic, mass

54

J. E. Dyson & T. W. Hartquist: Flows in clumpy CSM

loading always forces Mp towards unity. Thus H86 speculated that once mass loaded flows reach a Mach number of around unity, they would remain there. We argue on observational and theoretical grounds that this speculation has widespread validity.

3 The Dynamics of RCW58 The low (~ 5MQ) mass, clumpy structure and observed He and N enrichment imply that RCW58 is largely composed of stellar ejecta from the red supergiant phase preceding the present WR stellar phase (Smith et al. 1988— S88). Much less than 1% of the WR. wind kinetic energy appears to have been converted into kinetic energy of nebular material (Smith et al. 1984S84). 'Classical' wind-blown bubble theory predicts a. roughly 20% conversion efficiency for the pressure driven bubble which should have been produced by a 2000 km s" 1 WR wind (e.g. Dyson 1989). IUE absorption data on features formed in the bubble (S84) are clearly at variance with the 'classical' structures. The sense of the observed ionization potentialvelocity correlation is such that the only possible site for the formation of the absorption features is in the shocked wind zone. We assume that the flow behind the stellar wind shock is plane parallel and isobaric. In the observers' frame, the wind velocity and shock velocity are respectively —Vw and — V5. The observed ionic species with the highest and lowest ionization potentials are C 3 + and Fe + which exist respectively at temperatures T\ (~ 105K) and T2 (^ 104 K) and at corresponding densities p\ and p2.

Since the flow is isobaric, hP\T\

hp2T2

, pw(nv

=

.2 - Ks)

f0\ (^)

// //. where kf> is Boltzmann's constant, fi is the mean-mass per particle and pw is the immediate pre-shock density of the stellar wind. If we assume that the subsonic flow behind the shock mass-loads by a factor cf> between the post-shock wind gas and gas at Tj, and that the loading between gas at T\ and T2 is negligible, the continuity condition is -P1V1 - -P2V2 = -<j>Pw{V\v - Vs).

(3)

From equations (2) and (3), in the observers' frame, the velocity difference between the gas at T\ and T2 is kb

d> ^ ~

^

T

^

(4) (4)

J. E. Dyson & T. W. Hariquist: Floivs in clumpy CSM

55

Since Vw — Vs — 1800 km s" 1 (from the known wind speed), and the observed AV is —45 km s" 1 (S84), then cf> ~ 50, i.e. the post-shock flow is dominated by mass addition from clumps. H86 gave a phenomenological explanation of why takes the required value. They argued that the post-shock flow was driven to a Mach number of about unity (Section 2) and remained there until radiative cooling became important, and showed that under these circumstances, the ratio of entrained mass to wind mass is (25/8)(7V/2T,. ac |) 2 , where Tw is the initial post-shock wind temperature (w 4 107 K) and T r a d(« 105 K) is the temperature at which the radiative cooling timescale becomes shorter than the timescale for significant mass addition. Hence the mass ratio is about 44, in good agreement with that required. The absorption features are formed in a very thin layer (T ~ 105 K) in which negligible mass loading occurs. The strong radiative cooling removes most of the wind kinetic energy, thus accounting for the low conversion efficiency. Arthur, Dyson & Hartquist (1993) have constructed a spherically symmetric time dependent numerical model to describe R.CW58. A stellar wind mass loss rate of 1O~5M0 yr" 1 and wind velocity of 2000 km s" 1 were adopted. The mass loading rate was \0~34 Mp gm cm" 3 s" 1 . For a uniform ambient density distribution of 1 cm" 3 (the interclump medium) and an age of 13000 yr, they showed that the observed nebular radius and expansion velocity (S88) and, critically, the velocity spread in gas at temperatures between 105 K and 104 K, were well matched. However, the model temperature-velocity relationship is neither linear nor monatomic in that temperature range in contrast to the relationship inferred in S84 and H86 on the assumption that ionization equilibrium held. It is likely that this discrepancy can be resolved only if the non-equilibrium ionization structure is followed simultaneously with the hydrodynamics. This presents conceptual problems, in particular, the effect that the ablated material has on the ionization structure of the flow. In this flow, the interclump medium determines the position of the wind shock, implying that the clumps are neither too numerous nor massive. However, mass injection from the clumps totally dominates the shocked wind flow. As in the 'classical' case, the wind shock and that driven into the interclump medium move systematically outwards from the star. This model is probably too simplistic in its assumption of distributed mass loading sources. The morphology of the nebula shows that it contains a limited number of relatively large clumps. The mass loading zone may be located behind a stationary large scale bow shock around such a clump.

56

J. E. Dyson & T. W. Hartquisi: Flows in clumpy CSM

However, the overall energetics of RCW58 must be dominated by the effects of these localized zones. 4 Planetary Nebula Dynamics If clumps are numerous, massive and dense enough, the interclump medium plays little role in establishing the radius at which most of the wind is shocked. The wind is decelerated primarily in bow shocks around the clumps. Provided that the clumps are not substantially ablated, the positions at which the wind is shocked either do not move systematically outwards, or move more or less with the clump velocity, in clear distinction to the situation obtaining in RCW58. There is a substantial body of observational evidence showing that the AGB ejecta. wliich on pliotoionization give the major contribution to the mass of PNe, is clumpy clown to all resolvable scales. We assume therefore that clumps totally dominate the flow. This wind percolating through the clumps becomes mass-loaded and will be headed by a shock driven into the interclump medium. Initially, the mass-loaded wind also shocks and the flow consists of a supersonic mixture of ablated and wind gas which passes through the inner shock. The shocked mixture is separated from shocked interclump gas by a contact discontinuity. Implicit in this description is the assumption that the energy generated as the wind is decelerated in individual bow shocks can be lost radiatively, most likely in the interface regions. The temperature of the gas in theflowregions will be maintained at a temperature of about 104 K by pliotoionization induced by the radiation field of the central star. Provided the outer shock velocity is low enough, the supersonic interior flow is steady. At radii r > rjy = (SM./STT*/) 1 / 3 , the flow velocity v oc r~3 and the density p oc r4. M* is the stellar wind mass-loss rate and q the mass loading rate per unit volume. If the flow is isothermal, the Mach number is Mp — (K/c,)(rjvf/r)3 and approaches unity when r = r\ « (V./c,) 1 / 3 ^/ ~ 6rjv/ characteristically. Provided that the mass loading zone extends out further than radius r\, the flow should tend to a constant Mach number of around unity (H86). The inner shock becomes very weak or may vanish. Arthur, Dyson & Hartquist (1994) have simulated flows where the mass loading zone extends out further than r\. They assume that the gas behaves isothermally. They find that indeed the flow Mach number drops to a more or less constant value, but that the constant value is about 0.6-0.7. This is because of spherical divergence. An important application of these ideas is to the study of the extended haloes of PNe. PNe morphology supports the idea that the PNe clumps are

J. E. Dyson & T. W. Hartquist: Flows in dumpy CSM

57

localized around the central star as a result of the finite lifetime of the AGB ejecta phase. They may be surrounded by a lower density extended red giant envelope. The transsonic mass loaded wind eventually exits from the mass loading zone into a lower density region (which may itself be clumpy). The pressure in a transsonic flow is non-negligible, and its expansion is partially pressure driven. The flow can accelerate provided that its temperature is maintained by photoionization. The flow gently accelerates and the Mach number has a roughly logarithmic increase with radial distance. Hence Mach numbers of a few at most would occur in the halo regions.

Multiple shell PNe (Chu, Jacoby & Arendt 1987) are likely examples of bubbles with clumpy cores and gently accelerating haloes. Although few spectroscopic data for the haloes are available, we would expect them to show flow speeds corresponding to Mach numbers of about unity near their inner edges and up to a few further out. Indeed some. PNe seem to show this behaviour (Chu 1988).

Meaburn et al. (1991) made echelle studies of the halo of the PNe NGC 6543 and showed that the electron temperatures in an arc-like feature in the halo was about 15000 K, whereas the velocity dispersion in the gas was only 5-6 km s" 1 . They attributed this feature and its elevated temperature to a bow shock formed around a clump in the halo by the impact of a gently accelerating transsonic flow as described above. The gas flowing against the clump cannot have a speed much greater than 20-25 km s - 1 relative to the clump, otherwise the velocity dispersion would be too high. A shock decelerating gas at the core electron temperature of about 9000 K would heat the gas to the observed temperature of about 15000 K. The flow Mach number of 2-3 is in harmony with, and thus supports, the description above. Comparison of the required volume mass-loading rate here (Meaburn et al. 1991) with that required for RCW58 (H86) shows that the former is a factor of about 40 greater than the latter. This large difference in the mass loading rate is the reason why, in RCW58, the inner shock decelerating the wind is in the inner parts of the mass loading zone, whereas the wind in a PNe is decelerated in bow shocks around clumps distributed over a large part of the mass loading zone. Hot haloes have been observed in other PNe (Middlemass et al. 1990), but individual bow shocks have not been resolved, possibly due to too low an angular resolution. Their presence may still be responsible for the elevated halo temperatures (Dyson 1992).

58

J. E. Dyson & T. W. Hartquist: Flows in clumpy CSM

5 Intermediate Length-Scale Structures-Tails In general, clumps will be embedded in global flows with systematically directed velocity fields. Thus mass ablated from clumps will be preferentially accelerated in the radial direction and produce elongated structures or 'tails'. Dyson, Hartquist & Biro (1993) give simple analytic and semi-analytic models of tail shapes and appearances expected from the various categories of stream-obstacle systems resulting from the interaction of streams and sources. The advantage of these simple models is that they can be used to infer quite general yet robust conclusions about tail structures. In the following discussion, the stream can ordinarily be identified with a wind (mass loaded or not) and the source with the evaporative flow from an obstacle. (a) Supersonic stream interactions with sources. If the source is supersonic, the flow contains shocks in both the stream gas and the source gas. The post-shock regions are separated by a contact discontinuity, E, the shape of which can be calculated approximately from the balance of the normal components of momentum flux in the pre-shock flows. If the stream is plane parallel, the shape of E is invariant with respect to the relative momentum fluxes in stream and source (Dyson 1974). If both originate from point sources, S depends on these relative fluxes, but if, say, the stream dominates, either case in principle can give a, surface E which has a low width to length ratio. This does not imply that a long thin tail would be observed. If, for example, on S the area! emission power is proportional to the areal shock dissipation rate of mechanical energy near S, the tail will appear short and stubby because the areal dissipation rate strongly peaks towards the stagnation line. If the emission comes from gas which has cooled to a photoionization maintained equilibrium state, the areal emission power varies roughly as the gas pressure squared (% normal component of momentum flux on S) 2 . Again the fall-off is rapid. The slowest fall-off of areal emission power occurs if emission comes predominantly in a narrow temperature range considerably below the temperature to which gas on the stagnation line is heated. But even so, geometric divergence of wind flows always results in a rapid fall-off of areal emission away from the stagnation line and thus tails formed by two interacting supersonic winds appear to be short and stubby. If the source is subsonic, only the stream shocks and the shape of S is determined by the balance between the normal component of the wind momentum flux and the thermal pressure of the subsonic source gas. If the ratio of the pressure at some general position in the source gas to the stagnation pressure in the stream is (1 - e), where f < 1 for very subsonic

J. E. Dyson & T. W. Hariquist: Flows in clumpy CSM

59

flow, the tail width is about a factor e" 1 / 2 times the head width (Dyson et al. 1993), i.e. this interaction produces physically fat stubby tails. (b) Subsonic stream interactions with sources. If the source is subsonic with respect to S, the tail shape is determined by the balance of the thermal pressures of the stream and source gases. The tail width-to-head diameter ratio is about Z? 1 / 2 ^" 1 / 4 (Dyson et al. 1993) where the ratio of the downstream confining pressure to the stream stagnation pressure is (1 - 6), where 6 < 1 and /?(~ 1) is the Mach number at which the head empties. Since /3 w 61/2 can be expected, tail widths comparable to head diameters are possible even for very long tails. If the source gas is not highly supersonic and/or does not radiate well, the description is adequate and the effective source size is the characteristic shock radius. If the source gas is highly supersonic and the shock is nonspherical, the post-shock gas can remain supersonic and the shock shape is fixed by the balance of the normal component of the pre-shock wind momentum flux and the sum of the ambient gas pressure and the centrifugal correction due to flow along a curved surface. Canto (1980) calculated cavity shapes; in general they close in on themselves in the direction of decreasing ambient pressure and are crudely spherical abut some point coaxial with the source and displaced from it in the same direction. Long, thin cavities (i.e. long thin tails) cannot be produced by the small pressure gradients existing in subsonic streams. In summary, long thin tails are produced only when the wind is subsonic (i.e. the mass loss is a 'breeze' or the wind has been shocked far upstream) and the mass loss from the obstacle is subsonic.

6 Tails in the PNe NGC 7293 Spectacular high resolution images of the most prominent cometary globules in the highly evolved PNe NGC 7293 have been obtained by Meaburn et al. (1992). The clump gas may originate in SiO maser knots in the atmosphere of the parent red giant or AGB star (Dyson et al. 1989). These authors, on the basis of consideration of the survival times of clumps, predicted that they would be shown to possess molecular cores, as was later confirmed observationally. The tails behind the globules have lengths ~ 1016 cm, tail width-to-head diameter ratios of about unity, and tail length-to-width ratios of 5-10. Meaburn et al. (1992) suggested that the 'wind-swept' morphology was due to the impact of a supersonic wind from the central star on the globule. The discussion above shows that the impacting stream must be subsonic and the source must be no more than mildly supersonic. In

60

J. E. Dyson & T. W. Hariquist: Flows in clumpy CSM

any case, the central star is highly evolved with a low (as 600 LQ) bolometric luminosity and shows no signs of supersonic mass loss (Cerruti-Sola & Perrinoto 1985). A suitable subsonic stream could be produced if hot shocked stellar wind gas remains 'bottled up' in the interior cavity for a long enough period after fast wind activity from the nuclear star has ceased. Small pressure gradients would be expected in this gas as it gradually percolates outwards into the very clumpy PNe shell. Since the stream gas and globule gas will be approximately in pressure equilibrium, then (Dyson et al. 1989), nsTs ~ 107 cm" 3 K, where ns and Ts are the stream density and pressure. If the stream gas consists of hot shocked wind gas uniformly distributed throughout the central cavity, the mass of stream gas is Ms « 4 10~7 (7?L-/1017 cm) 3 (T s /10 8 K)" 1 MQ, where Rc(za 6 10 17 cm) is the radius of the central cavity and Ts « 108 K for a fast wind speed % 3000 km s" 1 . Hence Ms « lO" 4 A/ 0 , and since the fast wind mass loss rates of PNe central stars are ~ 10~' - 1 0 ~ 8 MQ yr" 1 , roughly a 103 - 104 yr supply of stellar wind gas needs to be bottled up. This is between 3-30% of the fast wind output for the dynamical age ( « 4 104 yr) of this PNe. We assume that the tails originate after the time when the fast wind has died away. The maximum tail length is C^ w viy:ite, where v^ is the tail gas velocity and te is some emptying time for the hot cavity gas. If the tail flow is isothermal at sound speed ct, application of Bernouilli's equation shows that its velocity is v^ « Mpct, where the hot shocked wind flows at Mach number M/r( pc, a 5 - 10 MQ ejected shell would fade away at a. radius of 4-7 pc if the shell's geometrical thickness A r / r ~ 0.1 - 0.2. For an expansion velocity of 50 - 100 km s" 1 the duration for detectability is only a few x 104 yr, which is about 10% of the WR. phase lifetime. For the same limiting emission measure and shell thickness, the minimal radius of a detectable swept interstellar shell corresponds to 100 - 200 pc for an ambient density 7?n = 0.1 cm" 3 or to 5 - 10 pc for no = 0.5 cm" 3 . Therefore, first, in a low-density area we should see ejecta-type nebulae around only about 10% of WR stars even if every one has ejected a shell in the course of its evolution. (In the dense ambient gas, the ejected and swept shells should merge.) And, second, we can try to discriminate ejected and swept up interstellar shells around WR. stars in low-density regions. To try to make this distinction we selected 31 WR stars in the LMC which appear to be located in low density ambient gas. The criteria we used for the selection were as follows: (i) (ii) (iii) (iv)

Low Hi-column density N(HI)< 1021crrr2, according to Rolilfs et al. (1984) Low color excess E(B-V) < 0.08 according to Breysacher (1986) Location outside bright CO clouds as delineated by Thaddens (1987) Location outside bright HI] complexes as delineated by Davies et al. (1986)

The results of the search for ring nebulae associated with these "selected" WR stars are as follows: Large faint ring nebulae of a typical size from 50 to 300-400 pc are found to be related to 64% of "selected" WR. stars. These big rings most probably display bubbles blown by the winds at Main Sequence and by the common winds of parent OB associations.

T. A. Lozinskaya et al.: WR stars in the IMC

75

Table 1. Statistics of big shells related to "selected" WR stars. "Selected" WRs N° "selected" WRs N° big rings around "selected" WRs Percentage of "selected" WRs with big rings

in LH ass.

nearby LH ass.

no OB ass.

10

4

17

31~~

10

2

8

20

100%

50%

47%

65%

total

Table 2. Statistics for all Br WR stars. All WR stars N° WRs N° big rings around WRs Percentage of WRs with big rings

in LH ass.

nearby LH ass.

no OB ass.

total

35 30

14 6

30 6

79 42

86%

43%

20%

53%

Winds of associations seem to dominate for the majority of big rings. Indeed, in Table 1 we show the number and percentage of "selected" WR stars with big shells separately for stars which belong to Lucke and Hodge (1970, hereafter LH) associations; for stars which are located nearby (less than 5' from the edge as delineated by LH) and for stars which neither belong to nor are in vicinity of any OB association. Table 2 shows the same statistics for all WR. stars, excluding those located in the overcrowded 30 Dor area. One can see that the majority of WR stars which belong to OB associations are related to big shells; the percentage of big shells around stars outside associations is less than 50%. Small ring nebulae are found to exist around eleven of the "selected" WR stars, all of them bar one inside big swept shells. The inner "small rings" might probably display circumstellar material. Tables 3 and 4 show the distribution of the small WR-rings between spectral types for "selected" WR. stars and for all WR stars in the LMC (excluding members of LH 100 in the core of 30 Dor).

76

T. A. Lozinskaya et ai: WR stars in the IMC Table 3. Statistics of the small WR-rings related to "selected" WR stars of different spectral types. spectral type

N(WNE)

N(WNL)

N(WN8)

N(WC)

19

5

2

4

7

1

1

2

37%

20%

50%

50%

N° "selected"WRs N° small rings around "selected" WRs Percentage of "selected" WRs with small rings

Table 4. Statistics of small WR-rings related to all WR stars of different spectral types. spectral type N° WRs N° small rings around WRs Percentage of WRs with small rings

N(WNE)

N(WNL)

N(\VN8)

N(WC)

58

18

3(4)

21

16

2

2

4

28%

11%

67 (50)%

19%

Though the sample is poor for WN8 stars and for "selected1' WC stars, three facts seem to be meaningful: (i) the scarcity of small rings around WNL stars, confirming data by Chu for the Galaxy, (ii) the low percentage of WC stars with small ring nebulae (again like in the Galaxy), (iii) the very high percentage of WN8 stars with small ring nebulae. The only WN8 nebula in the sample of "selected" stars is of Ejecta type. This third result confirms the previous data for the Galaxy and most probably indicates that first, these stars do differ from other WNL types and second, they do eject shells. The sizes of small ring nebulae related to "selected" stars of different spectral types are as follows. Ring nebulae around WNE stars have sizes in the range 3 to 30 pc; two WC ring nebulae are among the largest (sizes 30-45 pc); and two stars WN7 and WN8 are among the smallest (less then 10 pc). The statistics are poor (size strongly depends on the ambient density and we used only "selected" stars), though it is consistent with an evolutionary

T. A. Lozinskaya et a/.: WR stars in the LMC

77

sequence from the WNL classes to the WC types, as suggested by theory. The wide range of sizes of WNE-rings probably reflects different ways of evolution in the LMC: According to Breysacher (1986) both young and old stars belong to WNE types. Therefore, the statistical evidence for the LMC does not contradict the suggestion that all WR stars may eject a stellar shell of a typical mass about 5 - 1 0 M 0 at a velocity of about 50 to 100 km s"1; the strong stellar wind sweeps-up the ejected material. However, since all these estimates are very uncertain we'll refrain so far from making more definite statements. References Azzopardi, M. fc Breysacher, J. 1979, Astr. Astrophys., 75, 120. Breysacher, J., 1981, Astr. Astrophys. SuppL, 43, 20:5 . Breysacher J., 1986, Astr. Astrophys., 160, 185. Chu, Y.-H. 1982, Astrophys. J., 254, 578. Chu, Y.-H., Treffers, R. R. fc Kwitter. K. B. 1983. Astrophys. J. SuppL, 53, 937. Chu, Y.-H. 1981, Astrophys. J., 249, 195. Chu, Y.-H. 1991, in IAU Symposium 143, Wolj-Rayet Stars and Interrelations with Other Massive Stars in Galaxies, eds. K. A. van der Hiicht fc B. Hidayat, Dordrecht: Kluwer), p. 349. Davies, R.D., Elliott, K.H. fc Meaburn, J., 1976, Mem. R. Astron. Soc, 81, 89. Dopita, M.A., Bell J.F., Chu Y.-H. fc Lozinskaya T.A., 1994 Astrophys. J. SuppL, (in press). Lortet, M.-C. 1991, in IAU Symposium 143, Wolj-Rayet Stars and Interrelation with Other Massive Stars in Galaxies, eds. K.A. van der Hiicht. and B. Hidayat, Dordrecht: Kluwer, p.513. Lozinskaya, T.A., 1982, Astrophys. and Space Sci., 87, 313. Lozinskaya, T.A., 1983, Pis 'ma Astron. Zh., 9, 469. [Sov. Astron. J. Letters], 9, 247. Lozinskaya, T.A., 1992, "Supernovae and stellar wind in the Interstellar Medium", (Second Edition): American Institute of Physics, New York. Lucke, P.B. & Hodge, P.W., 1970, Astrophys. J., 75, 171. Miller G.J. k. Chu Y.-H., 1993, Astrophys. J. SuppL, 85, 137. Morgan, D.H., Vassiliadis, E. & Dopita, M.A. 1991, Mon. Not. R. astr. Soc, 251, 51P. Rohlfs K., Kreitschmann J., Siegman B.C. & Feitzinger J.V., 1984, Astr. Astrophys., 137, 343. Thaddeus P., 1987, "A CO survey of the LMC" Lecture Notes in Physics, 306 The other Galaxy, p. 241-246

Morphology & Physical Conditions in WR Shell Nebulae Reginald J. Dufour1'2 1

Department of Space Physics & Astronomy, Rice University, Houston, Texas, USA 77251-1892 2 Investigador Visitante, Instituto de Astronomia, UNAM, A.P. 70-264 Cd. Universitaria 04510 Mexico, D.F.

1 Introduction Wolf-Rayet Shell Nebulae (WRSN) provide a "quick look" at an intermediate stage of evolution of massive stars between the main sequence 0 stage and their ultimate demise as SNII. During this evolutionarily brief epoch, the 0 star develops a strong wind which affects the surrounding ISM, and can even have significant mass loss which enriches the ISM with H-burning products —specifically He and N (Maeder 1990). Therefore, studies of these objects are both interesting and important regarding the physics of windshock effects on the ISM and in the role they have in galactic chemical evolution. In this short contribution I will present some of the results of two recent students of mine who completed Ph.D. theses studying the morphology and spectra of the WRSN NGC 6888 (Mitra 1990) and NGC 2359 (Jernigan 1988). A more comprehensive review of the literature on WRSN is given by the fine paper by L. Smith in this volume. The theses studies incorporated CCD imagery mapping of the ionization structure of the nebulae in the emission lines of H/3, [OIII]A5007, Ha, [NII]A6583, & [SII]A6717+30; followed by spectroscopy of parts of the two nebulae that were of special interest from the imagery. Herein I will note some of the spectroscopic results regarding the hot wind-driven gas; the imagery mapping is available in their theses and moreso in a forthcoming Atlas of CCD Imagery of Galactic HII Regions (Hester et al. 1994).

2 NGC 6888 NGC 6888 is the prototype of the "wind blown" WRSN (Chu 1981). Our imagery and spectroscopy indicates that it also contains 5M e of ionized material, enriched in He and N, and therefore a composite "wind+ejecta" classification is more appropriate. This is supported by the two photographs on the next page: Figure 1 shows the nebula in Ha and Figure 2 shows the 78

R. Dufour: WR shell nebulae

79

Fig. 1. CCD image of NGC 6888 in Ha. Field diameter is 16'.

same field in [OIH]/Ha+[NII] (from Mitra 1990). For an adopted distance of 1.45kpc, the size of the nebula in Ha is 7.6x5.Ope, with the WN6 star (HD 192163) offcenter to the NW of the ellipsoid (in Ha). However, with respect to the boundaries of [OIII] (Fig. 2), the nebular perimeter is almost circular with the WN6 star centered. Therefore, while the nebula is bipolar in the "mass-loaded" stellar ejecta (best seen in Ha and [Nil]), it is spherical in the wind-driven bubble (as indicated from [OIII]). Spectra of 30 positions in NGC 6888 were taken in 1989 July & Sept. using the IRS on the KPNO 0.91m telescope. The spectra covered 36006800A with a resolution of ~12A and were taken simultaneously through two rectangular 35.6"x7.8" apertures separated by 61.2". Temperatures were derived from [OIII] & [Nil] lines for 12 knots in NGC 6888, as well as densities from [SII]. For the knots, the T e 's from [Nil] range from 650011,500 K and 11,900-29,100 K from [OIII] - with the [OIII] T e always higher for the knots where both ions could be observed. In addition, spectra were taken of several of the [OIII] rims or bubbles to the NE and N of the WN6 star which showed very high excitation and surprisingly high T e 's, with T e =55,000±20,000 K for the [OIII] rim (called NO3) on the NE edge of the nebula. Figure 3 shows the spectrum covering 3600-6800A. The ratio

80


Fig. 2. [OIII]A5007/Ha+[NII] ratio map of NGC 6888 showing the windblown bubble boundaries (dark = strong [OIII]). of [OIII]A5007/H/3 is ~20 and the line near 4350A is ~80% [OIII]A4363 (blended with H7) The spectra of NO3 and several other strong [OIII] features in NGC 6888 indicates that the entire nebula is filled with hot wind-shocked low density gas of T e ~50,000 K. While the existence of such a medium in WRSN comes as no surprise given theoretical expectations, the study of Mitra (1990) was the first successful measurement of pure wind-shock spectra in a WRSN. The ramifications of this medium is significant in interpreting the spectra of the stellar ejecta knots of NGC 6888 since the wind-shocked medium affects the observed spectra and derived physical properties of the higher ionization species. Using a "two-zone" model for the spectra of the knots seen in Ha and [Nil], Mitra demonstrated that the wind-shock medium contributes 1436% of the emission in [OIII]A5007 seen in the knots, as well as most of the [OIII]A4363 emission. When the wind-shock contributions are accounted for, the [OIII] T e of the knots become ~9000 K, comparable to the [Nil] T e observed and consistent with the knots being photoionized by the WN6 star. Even when the wind-shock emissions contaminating the knots' spectra are removed, abundances of He-N-0 for the knots in NGC 6888, derived


81

NGC6888-H03 5.00E-15

1

4.00E-15

-

1

•

-

3000s

-

1

1

1

1

1

3.00E-15

-

2.00E-15

-

I.00E-15

lib 1 .

1 lAdhiyjMiMilllnPl \ur

0

1 3500

1 4000

1 4500

5000

1

1

1

5500

6000

6500

Fig. 3. Spectrum of the shock on the NE rim of NGC 6888.

from both photoionization models and direct emission line diagnostics, are peculiar. For the eight best observed knots in NGC 6888 (those for which T e 's were derivable from both [OIII] & [Nil] and Ne from [SII]), average elemental abundances in terms of 12-f [X/H]±l. 130 -D

O

120 -

>

110 --

• •

D n

100 90 --

80 -5

-4

-3

-2

-1

Position from star in arcsec (East-West) Fig. 1. Radial velocity map of the SI 19 nebula obtained from the [Nil] 6583 A line, where in ordinate we report the nebular velocities, referred to the LSR and expressed in km/sec, and in abscissa the nebular positions, measured along the East (left) - West (right) direction. The zero point in the abscissa is the location of the star.

2.1 The Systemic Velocity Radial velocity measurements of the stellar lines of Si 19 indicate a systemic velocity VLSR between +100 and +140 km/sec. This is in excellent agreement with the value obtained for the surrounding ring nebula. These velocities are totally at odds with systemic velocities of LMC member stars, which normally range from +240 to +300 km/sec. The rectified spectrum of S119 in the region of the interstellar Nal 5890,96 doublet is shown in Figure 2. The Galactic absorption features are clearly visible (with VLSR = +4 km/s), as well as an intermediate velocity component (VLSR - +110 km/s), probably originating in the Galactic Halo. However, no components are detected (ew < 8 mA) at velocities corresponding to those expected for the LMC (between +240 and +300 km/s). For comparison purposes, Figure 2 shows the same spectral region in the spectra of two nearby LMC Ofpe/WN9 supergiants (BE 381 and BE 294), taken with an identical instrument configuration, where the LMC interstellar components are prominent (especially in BE 381). The peculiar systemic velocity and the absence of LMC interstellar features suggest that, unless SI 19 is not a member of the LMC (a very doubtful hypothesis in view of its mag-

A. Nota et al.: SU9: a new Luminous Blue Variable?

91

X

N •—* t—t cd

fi m 5890

5892

5894

5896

5898

5900

5902

Wavelength (Angstroms) Fig. 2. The rectified spectrum of S119 in the region of the interstellar Nal 5890,96 doublet is here shown. Notice the absence of absorption components at the LMC expected velocity (between +240 and + 300km/s). Compare the same spectral region in the spectra of two nearby LMC Ofpe/WN9 supergiants (BE 381 and BE 294), where the LMC interstellar component is prominent. nitude and spectrum, characteristic of an Of supergiant), it has probably been ejected from its birthplace and is now located outside the main body of the LMC. This ejection could have occurred as a result of a supernova explosion in a close binary or, more likely, after interactions with other stars in the early phases of a cluster formation. Assuming that SI 19 was ejected from an LMC cluster with an average velocity of 150 km/s towards the Sun shortly after its formation, it could have travelled a distance of 600-800 pc during the 4-5 Myrs required for a 40 M 0 star to reach the Of/WN stage. This distance represents a non-negligible fraction of the depth of the LMC, and is probably large enough to put the star in front of most LMC neutral gas clouds.

3 Coronographic Imaging Si 19 was observed with STScI NTT Coronograph mounted on the New Technology Telescope, ESO, La Silla. Images were obtained on February 12, 1993, in excellent seeing conditions (FWHM ~ 0.75") in the light of [Nil] (Ae/y = 6584.6 A, AA = 23.8 A) and Ha. With this technique we resolve,

92

A. Nota et al.: S119: a new Luminous Blue Variable?

Fig. 3. [Nil] image of the S119 nebula taken with the STScI NTT Coronograph. The nebula is 7" x 9" in size. The central star is not occulted in this image, still the contrast in the circumstellar region is enhanced by pupil apodization. North is up and East to the left.

for the first time, the circumstellar nebula which is shown in Figure 3. The nebula has an extension of 7" X 9" which translates into 1.9 x 2.1 parsecs at the LMC distance of 51.2 kpc (Panagia et al. 1992). The morphology of the nebula is clearly axisymmetric, with a very bright lobe extending towards the NE up to ~ 4" from the star. Towards the SW, more complex structures can be discerned. We find from the spectral data that the nebula is expanding with a velocity of ~ 25 km/sec; the linear size then implies a dynamical timescale of ~ 5 x 104 yrs for the nebula. The integrated, dereddened [Nil] 6584 A flux is derived from the images to be 1.7xlO~ 12 ergs cm" 2 s" 1 , in the assumption of E(g_y) — 0.16 (Bohannan and Walborn, 1989). We can calculate the ionized mass of the nebula from the integrated Ha emission luminosity. From the spectra, we obtain a ratio ricvnef,u;ar/[NII] ~ 1. We adopt an average electron density for S119 of 800 cm" 3 , obtained from the [SII] 6716/6731 A line ratio. Assuming a temperature T e ~ 7500 K, the nebular mass of S119 is ~ 1.7 M 0 .

A. Nota et al.: S119: a new Luminous Blue Variable? Table 1. Comparison Star

category

of nebular

AG Car R 127 S119

Gal. LBV LMC LBV LMCOf

properties

size (PC)

1.1 X 1.0 1 . 9 x 2 .2 1. 9 x 2 .1

93

(km/s) -70 -281 -30

t (yr) - 104 ~ 4 x 104 ~ 5 x 104

P (cm" 3 )

M (M 0 )

500 1000 800

-4.2 -3.1 - 1.7

1: Walborn, 1982.

4 Is S119 a new Luminous Blue Variable? The similarity with two well known Luminous Blue Variables, the galactic AG Carinae (Nota et al. 1992, Robberto et al. 1993) and R127 (Clampin et al. 1993) in the LMC, is remarkable. They both exhibit axisymmetric gaseous shells, with comparable properties in terms of size, expansion velocity (v), density (p) and mass (M) as we show in Table 1. Their nebulae are most probably the relic of a massive outburst, which occurred some 104 yrs ago. From a morphological point of view, their similarity would point to a common mechanism for the outburst (Figure 4). This finding would support the assumption that a close relationship exists between Ofpe/WN9 stars and Luminous Blue Variables. After all, one of the original members of the Ofpe/WN9 class, R.127 (Stahl et al. 1983, Walborn, 1984) became a Luminous Blue Variable (LBV) following an increase in brightness in 1982, as its spectrum evolved from an O-type to a B-type star and then to an A-type star (Wolf 1992). Similar spectral evolution had been noted for other LBVs which had been observed during their brightening phases [e.g., AG Carinae (Caputo & Viotti 1970), S Doradus (Leitherer et al. 1985; Wolf 1989), and R71 (Wolf cl. al. 1981)]. Throughout these brightness and spectral variations, the bolometric luminosity of R127 remained constant, a very well know characteristic of LBVs. The behaviour of R127 provides indications that there is a close relationship between the Ofpe/WN9 stars and the LBVs. In addition, Of features have been observed in the LBV AG Carinae during a light minimum phase, with the possible implication that the Ofpe/WN9 stars are LBVs in a quiescent state (Bohannan and Walborn, 1989). The similarity of the nebulae displayed by two classes of objects is striking, and suggests the possibility that a common mechanism has generated the outburst. More observational coverage of the currently known Ofpe/WN9 stars is required to investigate a) if light and spectral variations are present b) the structure of the circumstellar environ-

94

A. Nota et ai: S119: a new Luminous Blue Variable?

Fig. 4. Circumstellar nebulae around the two LBVs AG Carinae (galactic - left) and R127 in the LMC (right). Notice the similar morphology of the two axisymmetric shells. The scale is different, in the two images: the AG Car nebula is 32" x 36 " in size, while the R127 nebula is 8" x 9". North is up and East to the left.

ment, in order to establish on a statistically significant basis the relationship between 0fpe/WN9 stars and LBVs. References Bohannan, B., fc Walborn, N. 1989, P.A.S.P. 101, 520. Caputo, F., & Viotti, R. 1970, A. &A.,7, 266. Clampin, M., Nota, A., Golimowski, D., Leitherer, C , fc Durrance, S. 1993, Ap. J. Letters, 410, L35. Leitherer, C , Appenzellar, I., Klare, G., Lamers, H.J.G.L.M., Stahl, O., fe Waters, L.B.F.M. 1985, A. & A., 153, 168. Nota, A., Leitherer, C , Clampin, M., Greenfield, P. fc Golimowski, D. 1992, Ap.J., 398, 61. Panagia, N., Gilmozzi, R., Macchetto, F., Adorf, H.-M., & Kirshner, R.P. 1992, Ap.J. Letters, 380, L23. Robberto, M., Ferrari, A., Nota, A., & Paresce, F. 1993, A. & A., 269, 330. Stahl, O., Wolf, B., Klare, G., Cassatella, A., Krautter, J., Persi, P., & Ferrari-Toniolo, M. 1983, A. & A., 127, 49. Walborn, N. 1982, Ap.J., 256, 452. Walborn, N. 1984, in IAU Symp. 108, Structure and Evolution of the Magellanic Clouds, ed. S. van den Bergh and K.S. de Boer, p.239. Wolf, B. 1989, A. & A. Suppl., 217, 87. Wolf, B., Appenzeller, I., & Stahl, O. 1981, A. & A., 103, 94. Wolf, B., 1992, in Nonisotropic and variable outflows from stars , ed. L. Drissen, C. Leitherer and A. Nota, (A.S.P. Conf. series Vol. 22), p. 327.

New HST images of Eta Carinae and its surrounding nebulosity Dennis Ebbets 1 , Harry Garner 1 , Rick White 2 , Kris Davidson 3 , Eliot Malumuth 4 , and Nolan Walborn 2 1

Ball Aerospace and Communications Group Space Telescope Science Institute 3 University of Minnesota 4 Computer Sciences Corporation 2

1 Introduction Eta Carinae is a hot, massive, very luminous star which has erupted with episodes of greatly enhanced mass loss several times during the past few centuries. It is surrounded by an expanding shell of material, commonly known as the Homunculus, which was ejected during an outburst between 1830 - 1860. Fainter condensations are visible at greater distances. Their composition and radial proper motions suggest clumps of ejecta, some of which are contemporaries of the Homunculus, while others may have been expelled during earlier outbursts. Eta Carinae represents a dramatic although possibly brief phase of the interaction between a massive star in an advanced stage of evolution and its environment. The spatial and temporal characteristics of the nebulosity make it an attractive target for observations with the Hubble Space Telescope. The smallest structures currently known have sizes between 1/10 to 1/4 arc second, beyond the reach of ground based instruments, but well matched to the resolution of HST. Proper motions are typically 0.05 to 0.1 arc seconds per year. Displacements of several pixels per year will be seen with HST. Over the planned 15 year HST mission the kinematics of the expanding debris should be clearly revealed.

2 Observations and Data Processing Our observations were made on February 25, 1993 using the first generation Planetary Camera. Filter F336W isolated a violet spectral region which is free of bright emission lines in Eta Carinae. We avoided saturation problems caused by the large contrasts between the central regions and the nebulosity seen in nebular lines. We acquired one short and three longer exposures. The telescope was moved slightly between the subexposures to translate the images by approximately eight pixels on the camera. This sequence 95

96

D. Ebbets et ai: HST images of Eta Carinae

; ' • • - '

*

K*~,.

•

*

1

•

Fig. 1. A Planetary Camera image of the Ilomiinculus

of exposures was repeated with a nearby star to provide a. point spread function. As a reference for future archival users, the root names of our observations are W13M0101 - W13M0108 for Eta Carinae, and W13M0201 - W13M0208 for the PSF star. The data products provided by the STScI were combined to produce composite images of Eta Carinae and the PSF star. The individual images were shifted into registration and coadded, with cosmic ray events identified and withheld from the sums. The few saturated pixels near the core were "repaired" by scaling the data values in the short exposure and inserting them into the composite. An "adaptive histogram equalization" algorithm was applied to improve the visibility of low contrast features, and a "damped Lucy" algorithm was used to deconvolve the point spread function while controlling noise in regions of low signal (White, 1993). The processed images were digitally rotated to the correct orientation (north up, east left), and displayed with a linear or square root grey scale, inverted to produce a "negative" image.

D. Ebbets et al: HST images of Eta Carinac

97

3 Morphology of the Homunculus Figure 1 shows the Homunculus displayed using the AHE algorithm. Note that because there has been no deconvolution performed on this image the regions immediately surrounding the bright central star are dominated by the psf. The scale bar is 10 arc seconds long. As Hester et al. (1991) showed with the first HST observations of Eta. Carinae, the Homunculus has a bipolar appearance, rather than an "ovoid" or "ellipsoidal" shape suggested by ground-based pictures. Two nearly circular lobes extend to the southeast and northwest, each of which is eight arc seconds in diameter. They are tangent to each other at the central star. The centers of the lobes and the central star lie along a line. There is nothing obviously at the "centers" of either lobe. The southeast component is more conspicuous, and is presumably closer to us and less obscured. Both lobes are resolved into a large number of small knots, typically 1/4 arc second wide and up to one arc second long. Many of these knots can be associated with features identified photometrically by Burgarella and Paresce (1991). Their shapes and orientations give the appearance of clumps along the walls of a hollow spherical cavity (cf. Meaburn et al. 1993a). If the material which is visible today was ejected around 1830 the proper motions of the outermost knots must be approximately 0.05 arc seconds per year. HST observations over a decade or so will trace the trajectories of many of the resolved knots, and will help discriminate between possible physical models.

4 The Central Region Figure 2 shows the central 5 arc seconds of our image processed with the Lucy algorithm. It is dominated by a single bright source, but shows dozens of discrete bright spots. Most of these objects have sizes around 1/4 to 1/2 arc second, and show elongated or irregular shapes. They appear to be knots of diffuse matter rather than stars. A group of knots on the northwest side form a linear structure with a position angle near 316 degrees. They lie along the major axis of the Homunculus, and appear to project about four arc seconds into the NW lobe. A cluster of knots to the northeast form an arc that defines the northern rim of the SE lobe. There is a conspicuous dark lane on the west side of the central star which appears to be obscuring brighter regions behind it. It arcs from the north to the southwest, and has a small extension which gives the appearance of the Greek "lambda".

98

D. Ebbets et al.: HST images of FA a Carinae

Fig. 2. The region surrounding the bright central star

Fig. 3. The core of Eta Carinae in violet light

D. Ebbets et al.: HST images of El a Carinae

99

5 The Core Glimpses of the core were obtained using ground based speckle imaging by Weigelt and Ebersberger (1986) and by Hofmann and Weigelt (1988). Figure 3 shows our best image of the inner one arc second enlarged from the Lucy restored violet data. In addition to the dominant bright star, Eta Car A, there are numerous fainter features with diameters less than 1/4 arc second. The locations of speckle components B, C and D are indicated. Components C and D correspond to the brightest features in our image (after A). Their positions relative to A agree to within 0.05 arc seconds between 1985 and 1993. We see no evidence that these two components have proper motions comparable to those measured in the Homunculus and outer condensations. Speckle component B is not visible in our data. Either it is much fainter in our violet continuum than in the broad band red (which contained bright nebular emission lines), or it has moved or dissipated in the intervening eight years. One of the other components visible here is particularly interesting. The knot at a position angle 28 degrees (northeast) and 0.3 arc seconds from component A is aligned with the more distant, features which comprise the NN and NS condensations, and the intriguing jet like structure which they define (Hester et al. 1991, Meaburn et al. 1993b). The NN and NS condensations have proper motions of 0.1 arc seconds per year. If this new knot is kinematically related, and not just a coincidence in position angle, its motion away from the central star will be measurable with HST over periods of months, and questions about uniform vs decelerated motion, and its interaction with the ambient medium, should be fruitfully tackled with the HST. Acknowledgements This work is based on observations with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science Institute, which is operated by AURA, Inc. under NASA contract NAS5-26555. References Burgarella, D. & Paresce, F. (1991). Astron. Astrophijs., 241. 5. Hester, J. et al. (1991). Astron. Journal, 102, 654. Hofmann, K. & Weigelt, G. (1988). Astron. Astrophys.. 203. L21. Meaburn. J., Walsh, J. & Wolstencroft, R. (1993a). Astron. Aslrophys., 2C8, 283. Meaburn. J. et al. (1993b). Astron. Astrophys., (preprint) Weigelt, G. k Ebersberger, J. (1986). Astron. Astrophys., 103, L5. White, R. (1993), Newsletter of the STScI Image Restoration Project, 1, 11.

Observations of Circumstellar Media Around Supernovae Bruno Leibundgutf Astronomy Department University of California Berkeley, CA 94720 USA

Abstract Some supernovae are visible for several years past explosion. The main energy source for this sustained emission conies from the supernova shock interacting with the remnant of the stellar wind of the progenitor star. We review the available evidence for this picture and exclude other power sources on the basis of the radiated energies. We also discuss a group of supernovae which display narrow emission lines with high fluxes in their spectra and very slowly declining optical light curves. These observations can most readily be explained as being clue to interaction with a very dense medium close to the supernova.

1 Introduction A variety of supernova interactions with circumstellar material (CSM) has been observed to date. The best, and most direct, example is the ring of emission around SN 1987A (Jakobsen et a.l. 1991). This material has been ionized by the UV and soft X-ray flash of the shock breakout at the surface of the supernova (Fransson et al. 1989, Lundqvist & Fransson 1989). The density enhancement in the ring is caused by the interaction of the fast blue supergiant wind colliding with the slow red supergiant wind of a. previous epoch (Blondin & Lundqvist 1993). In the case of SN 1993.1, the early detection of radio and X-ray emission, in combination with narrow emission lines in the UV and optical, are indicative of interaction with the CSM. Blue optical continua, X-ray detection at early phases, as well as the UV emission have been proposed as characteristics of a shock in the CSM around SN 1979C (Fransson 1984). Infrared light echos observed around SN 1979C and SN 1980K are indicators of cold CSM (Dwek 1983). The most frequently observed signatures of CSM around supernovae, however, are radio emission (Chevalier 1984, 1990) and optical detection several years past explosion. The structure of the stellar interior is exposed when the explosion becomes transparent and the circumstellar environment is excited by the supernova t Present Address: European Southern Observatory. Karl-Sdiwar/.schilcl-Stiasse 2, D-85748 Garching bei Miinchen, Germany

100

B. Leibundgut: Supernovae and their circumstellar environment

101

Table 1. Supernovae interacting with circumstellar media SN 1957D 1961V 1970G 1978K 1979C 1980K 1981K 1983N 1984L 1986J 1987A 1988Z 1989R 1990B 1993J

Type

radio

Ref.

?

V V V V V V V V (V)

1 2 3-7

II II ?

II II Ib Ib II II II II II II

v/

V

8 9 10 6 6,11 6 12-15 16,17 18 — 19 20,21

optical Ref. (>3 years)

V v/

V V V V o —

V V V 0

—

22-24 25,2,26 27 8,28 29 30-32 — 22 — 31,32 33,34 35,36 37 38 still young

X-rays

Ref.

_ — —

_ — —

V

8 —

— —

(V) — — v/

39 — — — 40

41,42 — — —

43,44

Refei'ences: 1 Cowan k Branch 1985; 2 Cowan et al. 1988; 3 Gottesman et al. 1972; 4 Allen et al. 1976; 5 Brown k Marsclier 1978; 6 Weiler et al. 1986; 7 Cowan et al. 1991; 8 Ryder et al. 1992; 9 Weiler et al. 1991; 10 Weiler et al. 1992; 11 Panagia et al. 1986; 12 Rupen et al. 1987; 13 Weiler et al. 1990; 14 Sukumar k Allen 1989; 15 Bartel et. al. 1991; 16 Turtle et al. 1987; 17 Staveley-Smith et al. 1992; 18 Sramek et al. 1990; 19 Van Dyk et al. 1993a; 20 Pooley k Green 1993; 21 Van Dyk et al. 1993b; 22 Long et al. 1989; 23 Long et al. 1992; 24 Turatto et al. 1989; 25 Fesen 1985; 26 Goodrich et al. 1993; 27 Fesen 1993; 28 Dopita k Ryder 1990; 29 Fesen k Matonick 1993; 30 Fesen k Becker 1990; 31 Leibundgut et al. 1991; 32 Uomoto 1991; 33 Bouchet et al. 1991; 34 Suntzeff et al. 1991; 35 Stathakis k Sadler 1991; 36 Turatto et al. 1993; 37 this contribution; 38 A. C. Porter, private communication; 39 Canizares et al. 1982; 40 Bregman k Pildis 1992; 41 Dotani et al. 1987; 42 Sunyaev et al. 1987 43 Zimmermann et al. 1993; 44 Tanaka et al. 1993

shock. Investigation of the stellar properties, however, has to await fading of the radiative display from the explosion. In a few cases the CSM is dense enough to dominate the emission from the explosion. The list of supernovae (SNe) still optically observable several years after outburst includes now at least ten objects (Table \). With one exception, SN 1989R which has not been observed, all these objects have been detected at radio wavelengths. Conversely, all SNe II with a radio detection have been recovered optically. All SNe Ib/c with radio observations faded in the optical a few (10 years is SN 1961V (Fesen 1985, Cowan et al. 1988), but its classification as a supernova has been questioned (Goodrich et al. 1989, Bower et al. 1993). Three objects were discovered by a dedicated search program of Robert Fesen, who recovered SN 1980K (Fesen & Becker 1990), SN 1979C (Fesen & Matonick 1993), and SN 1970G (Fesen 1993). At the same time a search for supernova remnants in nearby galaxies disclosed SN 1957D more than two decades after explosion (Long et al. 1989). SN 1978K and SN 1981K have very little optical information and will not be discussed here.

2.1 SN 1957D in M83 This supernova in the southern sky has no spectral classification. Nonthermal radio emission was found by Cowan k. Branch (1985). The same

B. Leibundgut: Supernovae and their circumstellar environment

103

radio map also shows non-thermal sources near the sites of SN 1950B and SN 1983N, but neither has been recovered in the optical (Long et al. 1989). At optical wavelengths SN 1957D was first detected in the light of the [0 III] (AA 4959, 5007 A) doublet lines in 1987. The narrow-band imaging was verified by spectra obtained a year later (Long et al. 1989, Turatto et al. 1989). Strong emission in [0 III] but only weak lines of Ha and [O I] (AA 6300, 6364 A) were found. A remarkable drop in optical luminosity occurred between 1987 and 1991, when the supernova faded by a factor of ~5 (Long et al. 1992). This fading has been attributed to the supernova shock reaching the edge of the wind zone of the progenitor star. Another interpretation is that in 1987 an excited state was observed when the shock ran into a condensation. In this case the emission would vary on time scales of years. Careful monitoring of SN 1957D is expected to provide further clues as to the nature of this object.

2.2 SN 1910G in M101 The location of SN 1970G only 0.5 arcseconds away from a stellar-like source, which is probably part of the giant HTI complex NGC 5455 in M101 (NGC 5457), makes observations of this object exceedingly difficult (Fesen 1993). SN 1970G has been clearly recovered in [0 I] images and a spectrum has been secured. Broad emissions of [0 I] and Ha are detected, with narrow lines superposed. The latter are most likely due to the nearby HII region. The spectral identification of SN 1970G has profited greatly from comparison with SN 1980K, which shows similar line widths (Fesen 1993). While [O I] and Ha lines are observed, no [0 III] emission is detected in SN 1970G, a remarkable distinction from all other supernovae at similar ages. Fesen (1993) points out that SN 1970G exhibited a normal light curve and spectral evolution around maximum (Barbon et al. 1973, Kirsliner et al. 1973) and does not constitute a peculiar case, which might be argued for SN 1979C and SN 1980K (see below).

2.3 SN 1979C in M100 This supernova in the Virgo cluster of galaxies is the most distant in the sample of decade-old SNe II. It was exceptionally luminous at maximum (de Vaucouleurs et al. 1981, Branch et al. 1981) and its light curve did not display the characteristic plateau observed in many SNe II. The spectral evolution around maximum was also peculiar, exhibiting narrow Ha emission and weak absorption in the P Cygni profile of Ha (Branch et al. 1981).

104

B. Leibundgut: Supernovae and their circumstellar

environment

Interaction with CSM was inferred from the UV and blue spectral shape (Fransson 1984) and the strong radio emission, which turned on less than 13 months after explosion (Weiler et al. 1986). This supernova is relatively luminous even at these very late phases (Fesen 1993). The optical emission of SN 1979C is again mostly in [0 I] and Ha lines but also with a sizable [0 III] emission (Fesen & Matonick 1993). In addition, [0 II] or [Ca II] near A 7300 A is observed. The spectrum appears unchanged over the course of one year (Fesen & Matonick 1993). Narrow lines are probably due to a nearby HII region, which is situated less than 2 arcseconds from SN 1979C.

2.4 SN 1980K in NGC 6946 The most extensive data set is currently available for SN 1980K. It was recovered optically in 1987 (Fesen & Becker 1990) and has been monitored since (Leibundgut et al. 1991, 1993, Uomoto 1991, Fesen 1993). Around maximum this supernova is one of the best studied SN II. Its light curves declined steadily for the first 13 months (Barbon et al. 1982, Buta 1982). The spectral evolution was similar to SN 1979C (Barbon et al. 1982). It developed a nebular spectrum after about 8 months with strong Ha, [0 I], and [0 III] emissions (Uomoto & Kirshner 1986). Although the decline rate of Ha matched the half-life of 56 Co decay, Chugai (1990) showed that the emission had to be powered by an additional energy source. Note that this differs from the bolometric light curve (Schmidt et al. 1994), where the decline indeed reflects the characteristic lifetime of the radioactive source. More than a decade after maximum SN 1980K is still emitting in the same lines. The decline, however, has stopped completely and all lines are emitting at a constant flux level (Fig. 1). Typical line widths are ~5000 km s" 1 (Fesen & Becker 1990, Leibundgut. et al. 1991). Broad-band photometry established SN 1980K at V=22.8±0.2 mag. and R=21.9±0.1 mag. (Leibundgut et al. 1993). The total luminosity in the BVRI wavelength range amounts to 1.2-10"1'1 erg cm" 2 s" 1 . Given the distance to NGC 6946 the total emitted flux is then 8-1037 • (yyrs—) 2 erg s" 1 . Such high fluxes exclude long-lived radio isotopes as the power source since 10~2 M 0 of ''''Ti would be needed, in contradiction with nucleosynthesis yields in SNe II (Woosley et al. 1989). Since the line fluxes did not change subsequent to the recovery of SN 1980K, it is safe to integrate over this time span. The emitted energy is then ~ 10 46 • ( 7 5 ^ ) 2 erg and, if the flux had not changed since 1982 (the last observation of Uomoto & Kirshner), this number is doubled. This rules


10"

* io

12 -12

1

1 t

i

i

-13

-14

1

L-

10" 1 3 ^ CO

'

E

environment

1

i

1 ' i SN 1980K ^ —=

*

—=

2> 10 -15

^

10

•

i-

10-16 4000

105

A

O i

,

, 1

1

1

,

i

, -

6000 8000 JD (24440000+)

Fig. 1. Ha (filled symbols) and [O I] (open symbols) line fluxes for SN 1980K. The data are taking from the following references: Uomoto & Kirshner 1986 (circles), Fesen & Becker 1990 (triangles), Uomoto 1991 (hexagons), Leibundgut et al. 1991 (squares), and this contribution (diamonds).

out a light echo from the shock breakout as the power source, since at least 10% of the UV flash would have had to be converted into optical emission and emitted over a period of >10 years. Because the flash ionizes the volume just once, the number of Ha photons (3 • 10 5 ') would have required >3 MQ of hydrogen, assuming Case B recombination. The flux in SN 1980K is remarkably close to what, is expected from a powerful pulsar (like the Crab). Chevalier & Fransson (1992) have investigated the influences of powerful pulsars inside old SNe and young supernova remnants. They predict line velocities of a few hundred km s" 1 , which is not observed in SN 1980K. The observed line widths of ~5000 km s" 1 exclude this power source for SN 1980K. Shock interaction with CSM taps the kinetic energy pool (~10 51 erg) by converting a fraction of it into radiation. The forward shock ionizes the CSM and X-rays from the reverse shock produce emission from the ejecta (Chevalier & Fransson 1994). The observed energies and the line widths strongly favor circumstellar interaction. While the presence of the [O III] lines is probably due to ejecta heating, the [0 I] and Ha emissions more likely arise from the CSM (Chevalier & Fransson 1994).

106


2.5

environment

Summary

While only a few supernovae at very late epochs have been detected and observations of them are still scarce, a trend seems to be emerging. All objects show emission lines of the same elements, namely hydrogen and oxygen with a possible contribution from calcium to the 7300 A feature. This fits nicely with the CSM model presented by Chevalier & Fransson (1994). The ionizations differ among the individual SNe: SN 1957D shows mainly [0 III] emission, while this is not detected in SN 1970G. The constant flux in SN 1980K, however, is not predicted by the models.

3 Strong Interactors In recent years a few SNe II have been observed displaying some distinct properties. A first example was presented by Dopita et al. (1984) with observations of SN 1984E at maximum. This object had narrow lines of hydrogen and helium on top of strong P Cygni profiles. The narrow emissions were attributed to an extended, optically thick, outflow. For the derived density a very strong wind phase had to bo invoked (Dopita et al. 1984). Schlegel (1990) proposed a subclass for SNe II consisting of events with narrow emission lines. His list included a few objects which are clearly influenced by nearby HII regions and the class as a whole has no distinct characteristics apart from the narrow lines. We will discuss here three objects with narrow emission lines and particularly slow decline rates, which are not affected by contaminating light.

3.1 SN 19863 in NGC 891 The longest-observed supernova of this group, SN 1986J, was first discovered in the radio in 1986, but is found on radio maps and optical images as far back as 1984 (Rupen et al. 1987). Estimates of the explosion date have large uncertainties, partly due to the peculiar nature of the supernova, but late 1982 seems generally accepted (Rupen et al. 1987, Weiler et al. 1990). The optical spectrum of SN 1986.1 is remarkable for its narrow, unresolved (Av10 >80 >10

>le >10 >10 >10

4 22 11

100-10 100-10

2.3

75

20-1

>100 >1000 >100

>10 3 -10 2 >10 5 -10 4 >10 4 -10 3

10 5 2

2

erg s

in 1045 erg Period of integration for the total emitted energy Fluxes from the [0 III] lines

supernovae are much more luminous than regular old SNe. This indicates a much stronger interaction with a very dense medium as was inferred for SN 1986J and SN 1988Z (Filippenko 1991, Turatto ct al. 1993, Chugai & Danziger 1994), and is supported by the X-ray detection of SN 1986J (Bregman k Pildis 1992, Clvugai 1993). The detection of [N II] indicates nitrogen-enriched CSM, possibly from the dredge up in the convection near the surface of a rather massive star. The total energies emitted in Ha are remarkably uniform at >10 4 6 erg (over a decade time span) for the old SNe II. The other group has energies of about one to two orders of magnitude higher over much shorter periods. Other distinctions include the different radio luminosities, radio light curve shapes (Van Dyk, this conference) and decline rates of the optical light curves. A further distinction lies in the widths of the emission lines. The narrow lines arise from H, He, N, and Fe, while O appears in extended lines. For the old supernovae only broad 0 and Ikv lines are observed. Narrow-line supernovae display all signatures of strong interaction with their circumstellar environment. The optical emission is dominated by the CSM at all times, while for regular supernovae the CSM interaction becomes visible only a few years past explosion. Observations of CSM around supernovae have become available only recently and the sample is still small. Nevertheless, the shock interaction with the remnant of the stellar wind seems clearly confirmed in all cases. Although we are not yet able to derive detailed parameters of the CSM from

110


environment

these observations, they reinforce our picture of SNe II resulting from core collapse in massive stars that suffered appreciable mass loss during their lives. Eventually, we should be able to map such mass loss histories as the supernova shock moves through the CSM. The variations in SN 1957D might be a hint of clumping in the wind or pulsations of the progenitor. Likewise, flux variations might disclose the presence of companion stars modulating the mass loss as proposed from radio data for SN 1979C (Weiler et al. 1992).

Acknowledgements I would like to thank W. Vacca for helpful comments on the manuscript. Financial support by the Swiss National Science Foundation and through NSF Grant AST-9115174 is gratefully acknowledged.

References Allen, R. J., Goss, W. M., Ekers, R. D., and de Bruyn, A. G. (1976). A&A 48, 253. Barbon, R., Ciatti, F., and Rosino, L. (1973). A&A 29, 57. Barbon, R., Ciatti, F., and Rosino, L. (1982). .4CM 116, 35. Bartel, N., Rupen, M. P., Shapiro, I. I., Preston, R. A., and Ruis, A. (1991). Nature 350, 212. Blondin, J. M. and Lundqvist, P. (1993). , 4 p J 4 0 5 , 337. Bouchet, P., Danziger, I. J., and Lucy, L. B. (1991). .47, 102, 1135. Bower, G. C , Filippenko, A. V., Ho, L. C., Stringfellow, G. S., Goodrich, R. W., and Porter, A. C. (1993). BAAS 25, 819. Branch, D., Falk, S. W., McCall, M. L., Rybski, P., Uomoto. A. K., and Wills, B. J. (1981). ApJ 244, 780. Bregman, J. N. and Pildis, R. A. (1992). ApJ 398, L107. Brown, R. L. and Marscher, A. P. (1978). ApJ 220, 467. Buta, R. J. (1982). PASP04, 578. Canizares, C. R., Kriss, G. A., and Feigelson, E. D. (1982) ApJ 253, L17. Chevalier, R. A. (1984). ApJ 259, 302. Chevalier, R. A. (1990). Supernovae, ed. A. G. Petschek (Springer-Verlag, New York), p. 91. Chevalier, R. A. and Fransson, C. (1992). ApJ 395, 540. Chevalier, R. A. and Fransson, C. (1994). .4;)./, 420, 208. Chugai, N. N. (1990). SovALlG, 457. Chugai, N. N. (1992). SovA 36, 63. Chugai, N. N. (1993). ApJ 414, L101. Chugai, N. N. and Danziger, I. J (1994). MNRAS, in press. Cowan, J. J. and Branch, D. (1985). ApJ 293, 400. Cowan, J. J., Goss, W. M., and Sramek, R. A. (1991). ApJ 370, L49. Cowan, J. J., Henry, R. B. C , and Branch D. (1988). ApJ 329, 116. Dopita, M. A. and Ryder, S. D. (1990). IAU Circ. 4950. Dopita, M. A., Evans, R., Cohen, M., and Schwartz, R. D. (1984). ApJ 287, L69. Dotani, T. et al. (1987). Nature 330, 230. Dwek, E. (1983). ApJ 274, 175. Fesen, R. A. (1985). ApJ 297, L29. Fesen, R. A. (1993). ApJ 413, L109. Fesen, R. A. and Becker, R. H. (1990). ApJ 351, 437. Fesen, R. A. and Matonick, D. M. (1993). ,4p./407, 110.

B. Leibundgui: Supernovae and their circumstellar environment

111

Filippenko, A. V. (1991). SN 1987A and Other Supernovae, eds. I. J. Danziger and K. Kjar (ESO, Garching), p. 343. Fransson, C. (1984). A&A 133, 264. Fransson, C , Cassatella, A., Gilmozzi, It., Kirshner, R. P., Panagia, N., Sonneborn, G., k Wamsteker, W. (1989). ApJ 336, 429. Goodrich, R. W., Stringfellow, G. S., Penrod, G., D., and Filippenko, A. V. (1989). ApJ 342, 908. Gottesman, S. T., Broderick, J. J., Brown, R. L., and Balick, B. (1972). ApJ 174, 383. Jakobsen, P., et al. (1991). ApJ 369, L63. Kirshner, R. P., Oke, J. B., Penston, M. V., and Searle, L. (1973). ApJ 185, 303. Kirshner, R. P., Leibundgut, B., and Peters, J. (1989). IAU Circ. 4873. Leibundgut, B. (1991). SN 1987A and Other Supernovae, eds. I. J. Danziger and K. Kjar (ESO, Garching), p. 363. Leibundgut, B. (1994). IAU Colloquium 145: Supernovae and Supernova Remnants, ed. R. McCray, (Cambridge University Press, Cambridge), in press. Leibundgut, B., Kirshner, R. P., Pinto, P. A., Rupen, M. P., Smith, R. C , Gunn, J. E., and Schneider, D. P. (1991). ApJ 372, 531. Leibundgut, B., Kirshner, R. P., and Porter, A. C. (1993). B.4.4S25, 834. Long, K. S., Blair, W. P., and Krzeminski, W. (1989). .4pJ 340, L25. Long, K. S., Winkler, P. F., and Blair, W. P. (1992). ApJ 395, 632. Lundqvist, P. and Fransson, C. (1989). ApJ 380, 575. Panagia, N., Sramek, R. A., and Weiler, K. W. (1986). .4p./300, L55. Pooley, G. G. and Green, D. A. (1993). MNRAS, 264, L17. Rupen, M. P., van Gorkom, J. H., Knapp, G. R., Gunn, J. E., and Schneider, D. P. (1987). 4 / 9 4 , 61. Ryder, S. D., et al. (1992). IAU Circ. 5615. Schlegel, E. M. (1990). MNRAS 244, 269. Schmidt, B. P., Eastman, R. G., and Kirshner, R. P. (1994). ApJ, submitted. Sramek, R. A., Weiler, K. W., and Panagia, N. (1990). IAU Circ. 5112. Stathakis, R. A. and Sadler, E. M. (1991). MNRAS 250, 786. Staveley-Smith, L., et al. (1992). Nature 355, 147. Sukumar, S. and Allen, R. J. (1989). ApJ 341, 883. SuntzefF, N. B., Phillips, M. M., Depoy, D. L., Elias, .1. 11., and Walker, A. R. (1991). AJ 102, 1118. Sunyaev, R. A., et al. (1987). Nature 330, 227. Tan'aka, Y., et al. (1993). IAU Circ. 5753. Turatto, M., Cappellaro, E., and Danziger, I. J. (1989). ESO Messenger 56, 36. Turatto, M., Cappellaro, E., Danziger, I. J., Benetti, S., Gouiffes, C , and Delia Valle, M. (1993). MNRAS 262, 128. Turtle, A. J., et al. (1987). Nature 327, 38. Uomoto, A. (1991). AJ 101, 1275. Uomoto, A., and Kirshner, R. P. (1986). ApJ 308, 685. Van Dyk, S. D., Sramek, R. A., Weiler, K. W., and Panagia, N. (1993a). ApJ 409, 162. Van Dyk, S. D., Weiler, K. W., Sramek, R. A., Rupen, M. P., and Panagia, N. (1993b). IAU Circ. 5776. de Vaucouleurs, G., de Vaucouleurs, A., Buta, R., Abies, II. D., and Hewitt, A. V. (1981). PASP 93, 36. Weiler, K. W., Panagia, N., and Sramek, R. A. (1990). ApJ 364, 611. Weiler, K. W., Sramek, R. A., Panagia, N., van der Hulst, J. M, and Salvati, M. (1986). ApJ 301, 790. Weiler, K. W., Van Dyk, S. D., Panagia, N., Sramek, R. A., and Discenna, J. L. (1991). ,4p7 380, 161. Weiler, K. W., Van Dyk, S. D., Panagia, N., and Sramek, R. A. (1992). ApJ 398, 248. Woosley, S. E., Pinto, P. A., and Hartmann, D. (1989). ApJ 346, 395. Zimmermann, H. U., et al. (1993). IAU Circ. 5766.

Radio Supernovae as Probes of Progenitor Winds Schuyler D. Van Dyk1-2, Kurt W. Weiler1, Nino Panagia 3 , and Richard A. Sramek4 1

Remote Sensing Division, Code 7215, Naval Research Laboratory, Washington, DC, 20375-5351, USA 2 Naval Research Laboratory/National Research Council Cooperative Research Associate Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA 4 National Radio Astronomy Observatory, P. O. Box O, Socorro, NM 87801, USA

Abstract Radio supernovae (RSNe) are an excellent, means of probing the circumstellar matter around, and therefore the winds from, supernova (SN) progenitor stars or stellar systems. The observed radio synchrotron emission is best described by the Chevalier model, which involves the generation of relativistic electrons and enhanced magnetic fields through the SN shock interacting with a relatively high-density circumstellar envelope, which is presumed to have been established through mass loss in the late stages of stellar evolution. From the Chevalier model, modified to include a mixed, internal, nonthermal emission/thermal absorption component, we can use the observed radio emission from these SNe to derive physical properties, including the ratio of the mass-loss rate to the stellar wind speed, which determines the circumstellar matter density. Assuming a value for the wind speed then allows us to determine a mass-loss rate for the star responsible for the circumstellar matter and to estimate its mass. For Type II RSNe, this mass loss is assumed to originate from the presupernova star itself, while for Type Ib/c RSNe, the stellar wind is assumed to be from the binary companion to the SN progenitor. Extreme examples of progenitor winds are found for unusual Type II RSNe, for which radio properties indicate that the matter around these SNe resulted from very high mass-loss rates in the late stages of the evolution of very massive stars. Additionally, we have observed deviations from the standard model, probably providing evidence for inhomogeneities in the circumstellar matter density and possibly indicating the presence of stellar pulsations or an interacting binary companion.

1 Introduction SN 1970G was the first instance of a SN with detectable radio emission (Gottesman et al. 1972), and SN 1979C was the first SN to be observed in detail over a lengthy time span (Weiler et al. 1986, 1991). Since the detection of radio emission from SN 1979C in April 1980, we have carried out an 112

S. D. Van Dyk et al.: Radio supernovae & progenitor winds

113

ongoing monitoring program using the Very Large Array (VLA). We have observed about 50 SNe for radio emission (see Weiler et al. 1989, Van Dyk et al. 1994), and of these, eight examples of RSNe have been detected and studied in detail: SN 1980K, SN 1981K, SN 1983N, SN 1984L, SN 1986J, SN 1988Z, SN 1990B, and SN 1993J. In addition, Ryder et al. (1993) have recently discovered radio emission from the unusual SN 1978K, which appears to have radio properties similar to SN 1986J and SN 1988Z. Another unusual RSN is the very luminous Mk 297A recently discovered by Yin & Heeschen (1991). We will not discuss here the SNe 1957D in M 101 (Cowan & Branch 1985) and 1987A in the LMC (Turtle et al. 1987; Staveley-Smith et al. 1992), which have also been detected in the radio, since these are, respectively, poorly studied and of a very unusual SN type. However, we do draw the reader's attention to an interpretation by Ball & Kirk (1993) of SN 1987A's current radio emission. We find that most RSNe share the following common properties: 1) nonthermal emission with high brightness temperature; 2) light curve "turn-on" at shorter wavelengths first and longer wavelengths later; 3) a rapid increase in flux density with time at each wavelength, with a power-law decline, with index /3, after maximum; and, 4) a decreasing spectral index between two wavelengths, as the longer wavelength emission goes from being optically thick to thin, with the spectral index, a, asymptotically approaching an optically thin, nonthermal, constant negative value (see Weiler et al. 1986). Over the years we have also noted several systematic trends in radio emission from SNe (see Weiler et al. 1986, 1989): Type la SNe are not radio emitters to the sensitivity limit of the VLA; Type Ib/c SNe (e. g., SN 1983N, 1984L, 1990B) have a steep spectral index, a rapid "turn-on" at 6 cm before optical maximum, a rapid decline after maximum, and homogeneity in all radio properties, particularly spectral luminosity: and, Type II SNe (e. g., SN 1979C, 1980K, 1981K) show more diversity, a flatter spectral index, a. slower "turn-on" at 6 cm after optical maximum (1 month to more than 1 year), and a slow decline after maximum . Extreme examples of Type II RSNe are SN 1986J (Weiler et al. 1990), SN 1988Z (Van Dyk et al. 1993b), and, possibly, SN 1978K (Ryder et al. 1993). In Figure 1 we illustrate the difference in behavior between the three groups of RSNe by showing the radio light curves for the Type Ic RSN 1990B (Van Dyk et al. 1993a), the "normal" Type II RSN 1979C (Weiler et al. 1991), and the peculiar Type II RSN 1988Z (Van Dyk et al. 1993b). The radio emission from SNe can best be described by the Chevalier (1981, 1982, 1984) "mini-shell" model, with the radiation interpreted as synchrotron emission from relativistic electrons/positrons and enhanced mag-

1 .5

:' I — -

ii I, 1.8

2

2.2

i

| in

2.4

MI

|

m

-

-—

rI

[//

2.6

1.7

5. D. Van Dyk tt al.: Radio supernovae & progenitor winds

r

114

I J haft

b)

ml

IMI

2.8

3

Mil I I I I L

3.2

3.4

3.6

log (Days Since Explosion)


.4

2.8

2.9

3

3.1

3.2

3.3


Fig. 1. The radio light curves for: a) SN 1990B, b) SN 1979C, c) SN 1988Z. Data at 20 cm are represented by open triangles; data at 6 cm, by open squares; data at 3.6 cm, by filled triangles; and, data at 2 cm, by filled squares.

netic fields external to the SN photosphere produced through interaction of the SN shock wave with high-density circumstellar matter arising from mass loss in a red supergiant stellar wind before explosion. It is through this interpretation that RSNe are an excellent means of probing the circumstellar matter, and therefore the winds from, massive presupernova stars or star systems.

2 Modelling the Radio Emission Using the Chevalier model for the radio emission from SNe, we can derive model radio light curves to fit the observed data. The formulation for this model is based on several assumptions. First, the external circumstellar medium is assumed to have a uniform density PcWcum = (M/iv)/4nr2 from

5. D. Van Dyk et al.: Radio supernovae & progenitor winds

115

a red supergiant wind of constant speed w and mass-loss rate M. Second, the SN shock radius evolves with time during the interaction with the circumstellar medium as R oc tm (Chevalier 1981, 1982). Third, the SN ejecta have a density profile of dejecta oc r~n, where m and n are related by m = (n — 3)/(n — 2). And, fourth, the nonthermal radio emission from circumstellar interaction can be formulated as 5 oc i/ a ^-(T+ 5 - 67n )/ 2 = vatP, where 7 = —2a + 1 is the index of the power law of the relativistic electron energy spectrum and a is the optically thin radio spectral index (Chevalier 1984). The progression in "turn-on" time of the radio light curves with wavelength, as well as the very steep initial rise of flux density, are due to the absorption of the synchrotron radiation by the intervening circumstellar medium. We assume that the medium is completely ionized, such that the absorption is purely thermal, free-free, with frequency dependence v~2A. For free-free absorption external to the emitting source, r a R~3, so that r oc i~ 3m = ts. In addition to the external absorbing matter, it has been found that for some RSNe for which the circumstellar medium is very dense, e. g., SN 1986J (Weiler et al. 1990), the addition of an internal, expanding, mixed nonthermal emitting/thermal absorbing medium is necessary. Assuming constant ejected mass, constant temperature, and homologous expansion, the electron density for this internal emitting/absorbing region is ne oc R~3 oc t~3m, and for absorption by internal matter, r' oc EM oc n2eR oc

r 5 m = ts>.

Combining these components, the generalized formulation for the model flux density, S, the external absorption, r, and the internal absorption, r', for a RSN is:

where -2.1

and -2.1

/.i

* \ S'

(The mixed absorber attenuation term (1 - C~T')T'~X is appropriate for a plane-parallel approximation.) The scaling parameters Ii\, A'2, and A'3 formally correspond to the flux density, external absorption, and internal absorption, respectively, at 5 GHz, one day after explosion. We solve for

116

5. D. Van Dyk et al.: Radio supernovae & progenitor winds Table 1. Model Parameters

for Radio

SN

a

A2

Type Ib/c RSNe: 1983N 4.4 x 10 3 -1.03 1984L 2.8 x 1O2 -1.01 1990B 2.0 x 102 -1.12

-1.59 -1.48 -1.27

5.3 x 102 6.9 x 102 1.5 x 10"

Type II RSNe: 1970G 1.5 x 105 1979C 1.5 x 103 1980K 7.4 x 101 1981K 1.9 x 101 1986J 6.7 x 105 1988Z 1.5 x 105

4 -1.58 8.7 x 10 -0.78 3.7 x 107 -0.66 3.4 x 105 5 -0.50 < 1 . 5 x 10 -1.18 3.0 x 105 4.0 x 1012 -1.50 5.8 x 104 1.0 x 10 12

A'i

a

-0.81 -0.74 -0.52 -0.74 -0.67 -0.80

Supernovae A:3

to 1983 Jun 29 1984 Aug 12 1989 Dec 15

1979 Apr 1980 Oct 1981 Jul 1982 Sep 1988 Jan

04 01 31 13 23

the "best" values for the fitting parameters a, ft, I\\, A'2, A'3, and to ( t h e date of explosion) from each RSN's dataset using a minimum reduced x2 procedure. To reduce the number of fitting parameters we assume from t h e Chevalier model t h a t 6 = a - (3 - 3 and 6' = 5tf/3 (see VVeiler et al. 1990). The model provides a good fit t o the d a t a for all RSNe, indicating t h a t t h e RSNe generally behave in a regular manner (although deviations from this model are observed for some RSNe; see §5). T h e fitting parameters for t h e well-studied RSNe are listed in Table 1. We give t h e calculated values for 6 and 6', as well as the indices m and n for the SN, in Table 2. T h e fitting parameters and calculated properties for SN 1993J are not well determined at this time (see Weiler et al., this volume).

3 Mass-Loss Rates from the Progenitor Star Systems We can use the fitting parameters to derive M / w , which is essentially a measure of the circumstellar density pcircnm, using the following expression appropriate for free-free absorption (Equation 16 from Weiler et al. 1986): M

- P ~ 3.02 x 10-6 r?^ Hz m"1-5 [——^

w/10 km sec"- 1

1.5

A

\10 4 kmsec" 1 / (4) 1.5 / f \1.5ra /

rp

\ 0.68

where for / = 1 day, we assume r 5 GHZ(1 da.v) = A'2. If we further assume

S. D. Van Dyk et ah: Radio supernovae & progenitor winds

117

Table 2. Calculated Properties for Radio Supernovae m

n

( = - with n « 7 — 12. A useful similarity solution can then be found (Chevalier 1982a, b; Nadyozhin 1985). Here we sketch a simple derivation. More details can be found in these papers, as well as in the review by Chevalier (1990). Assume that the shocked gas can be treated as a thin shell with mass Ms, velocity Vs and radius Rs. Balancing the ram pressure from the circumstellar gas and the impacting ejecta, the momentum equation for the shocked shell of circumstellar and ejecta is Ms^-

= 4nR2s(pejV2ev -

PcsV

2

)

= 4rrR][pcj(V - Vs)2 -

2 PcsV }.

(2)

Here, Ms is the sum of the mass of the shocked ejecta. and circumstellar gas. The swept up mass behind the circumstellar shock is MC3 = MRa/uw, and that behind the reverse shock Mrev = 4ir t^V"[t/Ra)n~3/(?i — 3), assuming that Rs >> Rp, the radius of the progenitor. With V = Rs/t we obtain M D„ , 4vpot3o Von r - 3 l d2Rs 1 : uw s (n (71 - 3) 6) Rs Ks " J dt* df . [n. li3 VVn l f"-3

,2 Po o

o

J

/

Rsft t

This equation has the power law solution l/(n-2) 7^TT n - T~ 1/ " II

R.(t) = l(n - 4)(n - 3) M\

V2 )

,11) \ 2 dR s' dt

\,f /,}/? (dR M s 2 4TT uw R \ dt

\ 2"|

(3)

C. Fransson: Circumstellar interaction in supernovae

123

The form of this similarity solution can be written down directly by dimensional analysis from the only two independent quantities available, p0 t\ Von and M/uw. This solution applies after a few expansion times, when the initial radius has been 'forgotten'. To justify the neglect of the initial conditions, n > 5. More accurate similarity solutions, taking also the structure within the shell into account, are given by Chevalier (1982a) and Nadyozhin (1985). In general, these solutions differ by less than ~ 30% from the thin shell approximation. The maximum ejecta velocity close to the reverse shock depends on time The velocity of the circumstellar shock, dRs/dt, as V = Rs/t oc rlHn-2\ in terms of V is Vs — V(n — 3)/(rc — 2) and the reverse shock velocity, Vrev = V — Vs = V/(n — 2). Assuming equipartition between ions and electrons, the temperature of the shocked circumstellar gas is

and the reverse shock T — "(n-3)»The time scale for equipartition between electrons and ions is ~ 2.5 X 107 (T e /10 9 K)1-5 (n e /10 7 c m " 3 ) - 1 s. One finds that the reverse shock is marginally in equipartition, unless the temperature is Z 5 X 108 K. The ion temperature behind the circumstellar shock is ^ 6 x 109 K for V4 j£ 1.5, and the density a factor ^ 4 lower than behind the reverse shock. Ion-electron collisions are therefore ineffective, and Te 109 K relativistic effects become important and increases the cooling considerably (Lundqvist & Fransson 1988). One can estimate the free-free luminosities of the two shocks from

U = t* J MTMtdTHJssWtj}^.

. (10)

The density behind the circumstellar shock is pcs = 4 po = M/{•KUWR2S). The swept up mass behind the circumstellar shock is Mcs = MRs/uw and that behind the reverse shock is Mrev = (n — 4)M c s /2. With jjj — 2.4 x r e 0 5 , we get

£,-« 3.0 X 1039 gfJ Cn [^lYf-L-Y

ergs"1,

(11)

where gjj is the free-free Gaunt factor, including relativistic effects. For the reverse shock Cn = (n — 3)(??. — 4)2/4(??. — 2), and for the circumstellar shock Cn — 1. This assumes electron-ion cqnipa.rtit.ion. which is highly questionable for the circumstellar shock. Because of occultation by the ejecta only half of the above luminosity escapes outward. At Te ^ 2 x 107 K line emission increases the cooling rate, j w 3.4 X 23 10~ T~70'67 erg s~ 1 cm 3 . If the reverse shock temperature falls below 2 X 107 K a thermal instability may occur and the gas cool to ^ 104 K, where photoelectric heating from the shocks balances the cooling. Most important, the cool gas may absorb most of the emission from the reverse shock. If cooling, the total energy emitted from the reverse shock is Lrev

=

47Ti( s -2 P eiVrr,, ^3'rev-

=

(n - 3)(n - 4) 4

(

n

_ 2TT^ ) 3

J

3 7MV

126


Because V oc /- 1 /( n ~ 2 ) ) Lrev a ^~3/(n~2) in the cooling case. For high M/uw the luminosity from the reverse shock may contribute appreciably, or even dominate, the bolometric luminosity. 4 Early circumstellar interaction The earliest form of circumstellar interaction occurs at shock breakout. As the shock approaches the surface, radiation leaks out on a time scale of less than an hour. The color temperature of the radiation is ~ (1 — 5) X 105 K and the energy ~ (1 - 10) x 1048 ergs (Falk 1978; Klein & Chevalier 1978; Ensman & Burrows 1992). This burst of EUV and soft X-rays ionizes and heats the circumstellar medium on a. time scale of a few hours. In addition, the momentum of the radiation may accelerate the circumstellar gas to a high velocity. Most of the emission at energies £ 100 eV is emitted during the first few hours, and after 24 hours little ionizing energy remains. After about one expansion time (~ Rp/V) the reverse and circumstellar shocks are fully developed, and the radiation from these will dominate the properties of the circumstellar gas. The fraction of this emission going inward is absorbed by the ejecta and there re-emitted as optical and UV radiation (Fransson 1982; 1984). 4.1 SN 1981A The primary example of the influence of the soft X-ray burst is the echo from the ring of SN 1987A. The properties of the circumstellar medium and the emission from the ring are discussed in detail in the contribution by Lundqvist in this volume. The UV light curves observed with IUE up to day 2000 are presented in Fransson & Sonneborn (1994). General reviews of SN 1987A can be found in Chevalier (1992), Lundqvist (1992) and McCray (1993). Here only the basic scenario is summarized. Ensman & Burrows (1992) have calculated the burst in detail, and find a peak effective temperature of ~ 6 x 105 K and a color temperature of ~ 1 X 106 K. In total ~ 7 X 1046 ergs were emitted above 13.6 eV during the first day. Photons above ~ 100 eV, capable of producing e.g. N V, are emitted during less than ~ 500 sees. At shock breakout the radiation ionized the pre-existing ring on a time scale of seconds (Lundqvist & Fransson 1991). The temperature in the gas was then ~ 105 K. With a ring geometry the gas is optically thick in the continuum, and distinct ionization zones develops. Subsequently, the gas cooled and recombined. After a year the temperature was ~ 5 X 104 K, in agreement with temperatures determined

C. Fransson: Circumstellar interaction in siipernovae

127

from the [0 III] ratio (Wampler & Richichi 1989). The distance of the ring from the supernova is R — 6.3 x 10 1 ' cm, and its inclination i as 43° (Panagia et al. 1991). Therefore, after 2 sin i R/c w 410 days the full ring will be seen. This marks the maximum in the light curves. The recombination/cooling scenario, together with the light echo effects, explain the shape of the observed light curves. By modelling the state of ionization Fransson & Lundqvist (1989) found that the radiation temperature of the burst should be in the range (3 - 6) x 105 K. The ejecta are predicted to collide with the ring in the year ~ 2000 ± 3, with mainifestations in radio, IR, optical, UV and X-rays (e.g., Luo, McCray & Slavin 1994).

4.2 SN 1993J The recent SN 1993J is a good example where emission from the shock breakout has observable consequences. The first detection of the supernova was on March 28.30 at magnitude 13.6 (Merlin & Neely 1993). Hydrodynamical models (Shigeyama et al. 1994) show that shock breakout should have occurred at approximately March 28.0. The first spectra taken with IUE (Wamsteker et al. 1993; Sonneborn et al. 1993; Fransson & Sonneborn 1994; Sonneborn this volume) on March 30.21, gave a photospheric temperature of 22,500 K. One day later on March 31.2 the temperature had decreased to 14,500 K. Such rapid cooling is expected just after shock breakout, and agrees well with light curve calculations by Shigeyama et al. (1994). Shigeyama el al. 's model shows that immediately after breakout T e // « 3 x 105 K, decreasing in 24 hours to ~ 4 x 104 K. The total energy during the first day was ~ 6 X 1018 ergs, most coming out as UV and soft X-rays. The optical spectrum displayed a very rapid evolution (cf. Meikle et al. 1993; Filippenko, Matheson, & Ho 1993). From having a nearly featureless blue continuum, very broad lines became apparent after ~ 15 days. On April 11, expansion velocities of at least 19,000 km s" 1 can be seen for Ha (Meikle et al. 1993). In early May the spectrum changed character from an H-dominated Type II spectrum to a spectrum dominated by lines of He, resembling a Type Ib. From this it was apparent that only a small fraction of the hydrogen envelope remained on the progenitor at the time of the explosion. Based on the bolometric light curve several groups have concluded that the mass of the hydrogen envelope was ;$ IMQ (Nomoto et al. 1993; Podsiadlowski et al. 1993; Utrobin 1993; Woosley et al. 1994). Most of these models invoke a. binary scenario, although Hoflich, Langer, &

128

C. Fransson: Circumsiellar interaction in snpernovae

Duschinger (1993) propose that the mass loss is by a stellar wind from a single star. With a ZAMS mass of ~ 15M 0 , about 1OM0 must have been lost from the progenitor, either in the form of a wind, to a companion star, or by outflow in an excretion disk. Besides the hot continuum, the most obvious feature in the UV spectrum during the first week was the N V AA 1238.8 - 1242.8 doublet in emission. High resolution observations showed a line width ~ 103 km s" 1 , implying a circumstellar origin. The lines are very strong in the first spectra, but weaken rapidly with time. On March 30.2 the luminosity was ~ 2x 1040 erg s" 1 . Between March 30.2 and April 4.5 the N V flux decreased to ~ 5 x 10 38 erg s" 1 . By April 22 the N V emission was undetectable. The presence of the N V line shows that dining the first few days there was a high flux of photons above 77 eV. Models by Fransson, Lundqvist & Chevalier (1994) (henceforth FLC94) show that the burst from the shock breakout ionizes the circumstellar gas nearly completely. Because of the high density, gas inside ~ 1015 cm has time to recombine before the shock hits the gas, explaining the presence of the line. Once the reverse and circumstellar shocks are formed, radiation from these re-ionize the gas completely, effectively cutting off the emission. The N V line, as well as Ha (Cumming et al. , this volume), shows evidence for circumstellar velocities of ~ 1000 km s" 1 . Fransson & Sonneborn (1994) and FLC94 propose that this can be explained as a result of pre-acceleration of the wind by the strong UV burst at shock break-out. The UV and optical were the first wavelength ranges to show evidence for a circumstellar medium. A few days later radio emission was also detected from SN 1993J (Weiler et al. 1993; Pooley & Green 1993), as well as X-rays (Zimmermann et al. 1993a; Tanaka 1993). The radio emission showed the same characteristic pattern as had previously been observed for other Type II and Type Ib supernovae. First, emission was seen at short wavelengths, and later at longer. The radio flux at a given wavelength increased roughly linearly with time up to a maximum, and remained nearly constant up to at least 100 days (Pooley & Green 1993). From the radio spectrum of Van Dyk et al. (1993b) and Phillips & Kulkarni (1993) on April 22.5 FLC94 estimate that TJJ = 0.3 at 2 cm on day 25.5. Because of the heating of the circumstellar medium by the outburst of the radiation, as well as by the shocks formed by the interaction of the ejecta and the circumstellar medium, FLC94 find from photoionization models of the circumstellar medium that the temperature was ~ 3 x 105 K. The expansion velocity on day 25.5 was ~ 20,000 km s" 1 . From equation (1) one finds M/uw « 3 x 1O" 6 M 0 yr" x ( km s" 1 )" 1 , or M = 3 x 1O~ 5 M 0 yr" 1 if

C. Fransson: Circumstellar interaction in siipcrnovae

129

uw = 10 km s" 1 , within approximately a factor two. This is of the same order as derived for other 'normal' Type II supernovae (§2). Although qualitatively correct, the form of the observed light curves do not agree with the expectations (Lundqvist et al. in Wheeler & Filippenko 1994; FLC94; Lundqvist, this volume). In particular, the rising part is nearly linear, while models predict a more abrupt turn-on. The rising part of the light curves depends mainly on the density distribution of the circumstellar medium. FLC94 find that while a standard p oc r" 2 wind does not fit the observations by Pooley & Green, a better fit is obtained for p oc r~ 1 5 . Exploiting this idea, Weiler et al. (this volume) find that this also fits the observations at other wavelengths. The departure from p oc r~2 may be a result of variations in the mass loss rate during the last ~ 1000 years before the explosion, or due to the influence of the binary companion. SN 1993J has been observed by ROSAT (Zimmermann et al. 1993a,b) in the 0.1 - 2.4 keV band and by ASCA (Tanaka et al. 1993) between 1 10 keV. The luminosity at the first observations, ~ 7 days after explosion, was 1.6 X 1039 erg s" 1 between 0.1 - 2.4 keV, and 5 x 1039 erg s" 1 between 1 - 1 0 keV. There is evidence for a decrease of ~ 45% from April 3.4 to May 4 in the ROSAT flux (Zimmermann et. al. 1993c). On November 2 - 3 1993, on day 220, ROSAT observed a remarkably soft flux with kTx « 0.5 keV (Zimmermann et al. 1993d). At the same time the column density had increased by a factor ~ 6. The November observation strongly indicates that the flux was then coming from the reverse shock. Equation (6) can be written as k,Trev — 117 F 4 2 / ( n - 2 ) 2 keV. With V4 > 1.5 the density gradient must have n > 25. This implies that the reverse shock is radiative, and therefore has a dense cool shell, absorbing most of the X-rays from the reverse shock. The X-rays during the first months must therefore come from the circumstellar shock, explaining the much higher temperature, kTx ^ 10 keV, at these epochs. Without collisionless heating Te « 2 x 109 K. Using equation (11) with gjj « 2 the outgoing luminosity at 10 days is ~ 3 x 10 39 (A/_. 5 /v u ,i) 2 erg s" 1 , with a fraction 2 X 109 K, indicating that collisionless heating of the electrons is probably necessary. A possible problem is that the high temperature gives a Comptonized flux in the ROSAT and ASCA ranges higher than observed, and with too steep a spectrum. The total mass swept up during this first half-year by the supernova is only M t V/uw « 0.02 MQ. Emission from the circumstellar interaction may therefore be observable for a long time, as is the case for a number of other Type II supernovae (§5).

4-3 Other

supernovae

SN 1979C was the first supernova for which circumstellar interaction was found to be important. IUE spectra showed strong emission lines from He II, C III-IV, N III-V, and O III (Panagia et al, 1980; Fransson et al. 1984). In addition, a strong UV excess was found. Fransson (1982; 1984) interpreted the UV excess as Comptonized emission of the photospheric radiation by the circumstellar shock. The lines were explained as a result of the inwardgoing Compton emission in the EUV, photoionizing the outer ejecta, and possibly the cooling, shocked ejecta. The decay of the UV lines followed the decrease in the photospheric flux feeding the Comptonization, in agreement with the observations. The modelling of the radio observations implied a mass loss rate of ~ 1.5 x 10~4MQ yr" 1 (Lundqvist & Fransson 1988). In spite of this high mass loss rate, no X-rays were observed (Palumbo et al. 1981), probably because most of the X-ray emission was absorbed by the cool shell between the reverse shock and contact surface. The Type II-L SN 1980K was the first supernova to be detected in X-rays (Canizares, Kriss, & Feigelson 1982). About 40 days after outburst it was detected by Einstein, with a luminosity of ~ 2 x 1039 erg s" 1 between 0.2 - 4 keV. A month later it was again observed but only marginally detected with less than half the previous luminosity. This supernova was also well observed in radio (Weiler et al. 1992), and from the radio turn-on a mass loss rate of ~ 3 x 1O~5M0 yr" 1 has been estimated (Lundqvist & Fransson 1988). The X-ray luminosity derived for this mass loss rate agrees well with that observed.


131

IR emission has been observed from several Type II supernovae. Both SN 1979C and SN 1980K were detected (Bode & Evans 1980; Dwek 1983). Dust is assumed to be the main accelerating component in late type winds, and dust emission is therefore not surprising. The radiation from the supernova will evaporate the dust inside a radius ReVap ~ 10 17 — 10 18 cm, depending on the luminosity of the supernova and the evaporation temperature, 1000 — 2000 K (Dwek 1983). The heated dust grains outside Revap produce a light echo of a duration ~ 2 Revap/c. The mass loss rates inferred from the models agree well with those obtained from the radio. Graham & Meikle (1986) modelled the IR observations of the Type la SN 1982E (Graham et al. 1983) with a dust echo and find that a good fit to the data can be obtained for a dusty wind with M ss 1O~5M0 yr" 1 , and suggest that the binary companion was a low mass AGB star with heavy mass loss. A narrow Ha component in SN 1984E was observed by Dopita et al. (1984), who estimated a mass loss rate as high as ~ 3 x 1O~5M0 yr" 1 . This estimate is, however, uncertain, since it assumes pure recombination for the Ha line. This is clearly not the case, as is indicated by the P-Cygni profile. Also this object showed evidence for pre-acceleration.

4-4 Abundances From the UV lines it has, in two cases, been possible to determine the relative abundances of the CNO elements. In SN 1979C the ratios of the broad N III] A 1750 and C III] A 1909 lines gave a relative abundance ratio N/C « 8 by number (Fransson et al. 1984). For the O/N ratio only an upper limit of ~ 0.5 could be determined. The relative fluxes of the narrow lines from the ring of SN 1987A resulted in N/C « 7.8 and N/O « 1.6 (Fransson et. al. 1989). The nitrogen enrichments implied by these observations clearly indicate that the observed gas has undergone CNO-processing. This in turn, implies that substantial mixing must have occurred in the progenitor. The large nitrogen enhancement found for SN 1987A has turned out to provide one of the main constraints on the evolution of the progenitor. For reviews of various models see e.g. Podsiadlowski (1992).

5 Very late emission from supernovae The persistent radio emission shows that many supernovae, even after several years, are interacting with a circumstellar medium. The radio flux of SN 1979C has, for example, only decayed by a factor ~ 2 over the more than

132


ten years it has been observed (Weiler et al. 1991). Recent observations of several supernovae, reviewed by Leibundgut in this volume, show that circumstellar interaction is also important for the optical emission at late stages. Here I only mention some of the implications, discussed in Chevalier k Fransson (1994), henceforth CF94. As the supernova ejecta expand they sweep up the circumstellar medium. At the same time the reverse shock propagates further into the ejecta. The maximum ejecta velocity therefore decreases continuously. Hydrodynamical calculations show that the density profile flattens considerably for ejecta velocities less than ~ 5000 km s" 1 . As a first approximation we may use a simple power law model, but with an n that may be different from that inferred from the early epochs. The maximum ejecta velocity, i.e. the widths of the line profiles, is then /

v = Vn

\WT)

\ V(i-2)

^/{n~2)

km8 1

"'

where V6 = 12.1 x 103, V8 = 9.61 x 103, Vi0 = 7.44 x 10 3 , and V12 = 6.51 X 103 km s" 1 . Here po-\e is a reference density at one year and at 5000 km s" 1 in units of 10~ 16 g cm" 3 . Hydrodynamical models show that n » 8 and p o _ 1 6 » 1. With M_ 5 = 5.0, we find V % 5000 km s" 1 at 10 years, in agreement with the observed velocities. The shock emission is dominated by the reverse shock with velocity ~ 5000/(n—2) « 700—1500 km s" 1 . The temperature behind the reverse shock is therefore 1 0 6 - 107 K. Below ~ 4 x 106 K the shock spectrum is dominated by far-UV and X-ray lines below ~ 100 cV. At higher temperatures Xrays between 0.5 - 1 . 0 keV dominate. The low temperature makes cooling important for the reverse shock. This can lead to a thermal instability, cooling the gas to ^ 104 K, where photoelectric heating may balance the cooling. The result is a cool, dense shell between the reverse shock and contact discontinuity, opaque to all X-rays below 0.2 — 1 keV. Therefore, most emission from the reverse shock will be absorbed, half in the ejecta and the other half in the cool shell (fig. 2). Photoionization models show that the energy absorbed by the ejecta is transformed into Lycv, Ha and emission from highly ionized species, like [0 III]AA 4959 - 5007, C III] A 1909, and C IV AA 1550. The emission from the cool shell is dominated by Lycv, Ho and Mg II A 2800. The temperature in the ionized ejecta is ~ (15 - 20) X 103 K, while that of the cool shell is only ~ 8000 K. The ionized zone in the ejecta is confined to the region just inside the reverse shock with AR/R a 0.1.


133

The most interesting objects to compare the model to are SN 1979C and SN 1980K. Both are strong radio supernovae, and are fairly typical Type II's. SN 1980K was the first to be recovered optically, and among the 'normal' cases is the best observed (Fesen & Becker 1990; Leibundgut et al. 1991). The optical spectrum was dominated by Ha, [0 I] AA 6300 — 64, and [0 III] A A 4959 — 5007. These lines are also the strongest in the models. In addition, both in the UV and IR the models show several lines with similar or larger fluxes. Observations in these ranges should therefore be rewarding. The [0 III] AA 4959 — 5007/Ha ratio shows an interesting evolution (fig. 3). Up to ~ 2 years the [O III] lines are collisionally de-excited. As the ejecta density decreases the lines become increasingly strong. At an age of ten years or more they dominate the cooling of the ejecta. This is consistent with observations of SN 1980K, SN 1979C (Fesen & Matonick 1993) and the 30-year-old SN 1957D. In the last of these, the [O Til] lines are by far the strongest (Turatto et al. 1989; Long et al. 1989). The decline of the Ha light curve of SN 1980K was consistent with radioactive decay from 56 Co during the first year or so. At the time of the last observation by Uomoto & Kirshner (1986) at 670 days, Chugai (1988) found a clear excess, and proposed that the origin of the excess was circumstellar interaction, based on a suggestion by Chevalier & Fransson (1985). Subsequent observations at 7.8 years and later show a nearly constant Ha flux. This indicates that the energy source is nearly constant. As we saw in §3.2, this is expected if the reverse shock is radiative, Lrev oc i~ 3 /( n ~ 2 ). Whether the reverse shock is radiative or not is sensitive to both the density gradient, n, and the mass loss rate. Most emission originates either in the cooling shell or from the outer parts of the ejecta close to the reverse shock, and the lines are expected to have flat, box-like profiles, with extent V « (4 - 5) x 103 km s" 1 , as is seen in SN 1980K and SN 1979C. The observed profiles, however, tend to be asymmetric with a stronger blue edge. A possible reason for this may be dust absorption in the ejecta. While SN 1957D, SN 1979C and SN 1980K are fairly well behaved, and fit well into the scenario discussed above, there is a number of objects which have a more peculiar behavior. These include SN 1978K, SN 1986J, and SN 1988Z, and are discussed by Leibundgut (this volume). This class probably only represents a small fraction of the total number of Type II supernovae. Most of them have been discovered through a variety of peculiarities; SN 1978K and SN 1986J due to their extraordinary radio emission, and SN 1988Z because of its peculiar optical spectrum (in spite of cz ~ 6000 km s" 1 !).

134


shocked ejecta T * 1CT K cool shell

Lya, Ha, Mg U ionized wind

shocked wind T * 3x1 Q8K

Fig. 2. Schematic figure showing the different regions of the supernova - circumstellar medium interaction, and the most prominent lines from the various regions (from CF94).

J

5

0

1

'

1

' ' "1

4959-5007

IS

10

1

-

-

J

y .

i

, ,,

5 t (years)

-

11

10

i 20

,-

Fig. 3. Evolution of the [ 0 III] A A 4959 —5007/ITo ratio as a function of time (from CF94).

C. Fransson: Circumsiellar interaction in supernovae

135

SN 1986J was discovered as a radio source, with a radio luminosity of ~ 3000 times Cas A. Modelling the light curve Weiler, Panagia & Sramek (1990) estimate the explosion date to have been September 1982. VLBI observations (Bartel et al. 1991) show the source as an irregular shell, whose velocity is ~ 8000 km s" 1 , with 'protrusions' reaching out to ~ 18,000 km s" 1 . The optical spectrum shows no evidence of lines with these kinds of velocities. Instead, two systems of lines are present, with Ha having a FWHM velocity of ~ 530 km s" 1 , and [0 I] - [0 III] lines having higher velocities, 1000 - 2 0 0 0 km s" 1 . A likely explanation for this apparent inconsistency is that the low velocity H lines originate in shocked circumstellar gas, while the high velocities inferred from the radio may be due to the expanding blast wave. These aspects can be combined if the circumstellar medium is very clumpy, with high density clouds immersed in alow density wind (Chugai 1993). Alternatively, the circumstellar gas may be asymmetrically distributed, with low density in the polar direction and high density in the equatorial plane, possibly in the form of a circumstellar disk. In either case, the II lines originate from shocked circumstellar gas. To obtain the low velocities, the circumstellar gas density has to be higher than the ejecta by a factor pCs/Pej ~ [V /Vcs)2 * 100. The high 0 I density estimated by Leibundgut, el al. (1991), ~ 109 cm" 3 , makes it likely that the 0 lines originate in the oxygen-rich regions of the ejecta. SN 1986J has also been detected as an X-ray source with ROSAT, with a luminosity of (1.6 - 7) x 1040 erg s" 1 between 0.1 - 2.5 keV, and kTx = 1.0 - 3.9 keV (Bregman & Pildis 1992). The high column density found is consistent with absorption by cooling gas. SN 1988Z showed similar characteristics: strong radio emission (Sramek, Weiler & Panagia 1990), and narrow optical lines at late stages (Stathakis & Sadler 1991; Turatto et al. 1993). This supernova was discovered shortly after explosion. During the first weeks SN 1988Z showed a broad Ha emission with a velocity width of ~ 20,000 km s" 1 , as well as a narrow component with FWHM ;$ 200 km s" 1 . The broad component faded rapidly, and instead an intermediate component with FWHM % 1500 km s" 1 appeared. Stathakis & Sadler (1991) find that the bolometric luminosity was higher than for normal Type IPs, and suggest that this is due to the circumstellar interaction. Chugai (1992) found that the light curve could be fitted for a mass loss rate M/uw\ « 3 x 10~3Mi) yr" 1 . This is much higher than is likely, given the turn-on time of the radio flux. In the same way as for SN 1986J, the high mass loss rate, the broad Ha and the narrow components can, however, be explained qualitatively by an aspherical circumstellar medium.

136


Recently, SN 1978K has been added to the same category (Ryder et al. 1993), having a strong radio flux, ~ 300 times Cas A, an X-ray luminosity of ~ 1040 erg s"1 with kTx ~ 0.5 keV, and narrow emission lines, FWHM ~ 450 km s~l. In January 1980 an Einstein observation gave an upper limit of 0.9 x 1039 erg s" 1 , indicating strong X-ray absorption. The mass loss is estimated to be ~ 4 x 1O~4M0 yr" 1 . An alternative scenario for the late emission is excitation by a central pulsar (Chevalier & Fransson 1992). A fundamental difference between these two models is that while the pulsar scenario predicts line profiles with a maximum width of < 1500 km s"1, increasing with time, the circumstellar interaction scenario predicts line velocities of ~ 5000 km s"1, decreasing with time. Chevalier (1987) proposed that SN 1986.J was powered by pulsar excitation. However, the line widths, the presence of strong radio emission, and X-rays all indicate circumstellar interaction in this case also. In addition, VLBI observations show that the emission originates in a shell. Acknowledgements The modeling of the observations of SN 1993J has been done in collaboration with Roger Chevalier and Peter Lundqvist. I am also grateful to Robert Cumming and Peter Lundqvist for comments. This work is supported by the Goran Gustafsson Foundation for Research in Natural Sciences and Medicine. References Bartel, N., Rupen, M. P., Shapiro, I. I., Preston, R. A., k. Rius, A. 1991, Nature, 350, 212. Bode, M.F. & Evans, A. 1980, AW 193, 21 P. Bregman, J. N., & Pildis, R. A. 1992, ApJ, 398, L107. Canizares, C. R., Kriss, G. A., & Feigelson, E. D. 1982 ApJ, 253, LIT. Chevalier, R. A. 1982a, ApJ, 258, 790. Chevalier, R. A. 1982b, ApJ, 259, 302. Chevalier, R. A. 1987, Nature, 329, 611. Chevalier, R. A. 1990, in Supernovae, ed. A. G. Petschek, Berlin, Springer, p. 91. Chevalier, R. A. 1992, Nature, 355, 691. Chevalier, R. A., & Fransson, C 1985, Supernovae as Distance Indicators, ed. N. Bartel,Berlin, Springer, p. 123. Chevalier, R. A., & Fransson, C. 1987, Natwe, 328, 44. Chevalier, R. A., & Fransson, C. 1992 ApJ, 395, 540. Chevalier, R. A., & Fransson, C. 1994, ApJ, 420, 268. (CF94) Chevalier, R. A., Blondin, J. M., & Emmering, R. T. 1992, ApJ, 392, 118. Chugai, N. N. 1988, Ap&SS, 146, 375. Chugai, N. N. 1992, Sov. Astr., 36, 63. Chugai, N. N. 1993, ApJ, 414, L101. Dopita, M., Evans, R., Cohen, M., & Schwartz, R. 1984, ApJ, 287, L69.


137

Dwek, E. 1983, ApJ, 274, 175. Ensman, L. k Burrows, A. 1992, ApJ, 393, 742. Falk, S. W. 1978, ApJ, 226, L133. Fesen, R. A., k Becker, R. H. 1990, ApJ, 351, 437. Fesen, R. A., k Matonick, D. M. 1993, ApJ, 407, 110. Filippenko, A. V., Matheson, T., k Ho., L. C. 1993, ApJ, 415, L103. Fransson, C. 1982 ABA, 111, 150. Fransson, C. 1984 A&A, 133, 264. Fransson. C , k Lundqvist, P. 1989, ApJ, 341, L59. Fransson. C , k Sonneborn, G. 1994, in Frontiers of Space and Ground-based Astronomy, eds. W. Wamsteker, M.S. Longair k Y. Kondo, Kluwer, in press. Fransson, C , Benvenuti, P., Gordon, C , Hempe, K., Palumbo, G.G.C., Panagia, N., Reimers, D., k Wamsteker, W. 1984, A&A, 132, 1. Fransson, C , Cassatella, A., Gilmozzi, R., Kirshner, R.P., Panagia, N., Sonneborn, G., k Wamsteker, W. 1989, ApJ, 336, 429. Fransson. C , Lundqvist, P., k Chevalier, R. A. 1994, in preparation (FLC94). Graham, J.R. k Meikle, W.P.S. 1986, MN, 221, 789. Graham, J.R. et al. 1983, Nature, 304, 709. Hoflich, P., Langer, N., k Duschinger, M. 1993, A&A, 275, 1,29. Jura, M. k Kleinmann, S. G. 1990, ApJ Suppi, 73, 769. Klein, R. I., k Chevalier, R. A. 1978, ApJ, 223, L109. Knapp, G.R. k Woodhams, M. 1993, in Massive Stars: Their Lives in the Interstellar Medium, eds. J. P. Cassinelli k E. B. Churchwell, ASP Conf. Series, Vol. 35. Leibundgut, B., Kirshner, R. P., Pinto, P. A., Rupen, M. P., Smith, R. C , Gunn, J. E., k Schneider, D. P. 1991 ApJ, 372, 531. Leising, M. 1993, priv. comm. Long, K. S., Blair, W. P., k Krzeminski, W. 1989, ApJ, 340, 1,25. Lundqvist, P. 1992, PASP, 104, 787. Lundqvist, P., k Fransson, C. 1988, A&A, 192, 221. Lundqvist, P., k Fransson, C. 1991, ApJ, 380, 575. Luo, D., McCray, R. k Slavin, J. 1994, ApJ, in press. McCray, R. 1993, Ann. Rev. Ast. Astr., 31, 175. Meikle, P., et al. 1993, Gemini, 40, 1. Merlin, J.-C. k Neely, A. 1993, IAU Circ, No. 5740. Nadyozhin, D. K. 1985, Ap&SS, 112, 225. Nomoto, K., Suzuki, T., Shigeyama, T., Kumagai, S., Yamaoka. H.. k Saio, H. 1993, Nature, 364, 507. Palumbo, G. G. C , Maccacaro, T., Panagia, N., Veltolani. G., 1: Zamorani, G. 1981 ApJ, 247, 484. Panagia, N.,e« al. . 1980 MN, 192, 861. Panagia, N., Gilmozzi, R., Macchetto, F., Adorf, H.-M.. fc Kirshner, R. P. 1991, ApJ, 380, L23. Phillips, J. A., k Kulkarni, S. R. 1993, IAU Circ, No. 5763. Podsiadlowski, P., Hsu, J. J. L., Joss, P. C , k Ross, R. R. 1993, Nature, 364, 509. Podsiadlowski, P., 1992, PASP, 104, 717. Pooley, G. G., k Green, D. A., 1993, MN, 264, L17. Ryder, S., Staveley-Smith, L., Dopita, M., Petere, R., Colbert, E., Malin, D., k Schlegel, E. 1993, ApJ, 416, 167. Shigeyama, T., Suzuki, T., Kumagai, S., Nomoto K., Saio, H., k Yamaoka, H., 1994, ApJ, 420, 341. Sonneborn, G. et al. 1993, IAU Circ, No. 5754. Sramek, R.A., k Weiler, K.W. 1990, in Supernovae, ed. A. G. Petschek, Berlin, Springer, p. 76. Sramek, R.A., Weiler, K.W., k Panagia, N. 1990 IAU Circ. No. 5112. Stathakis, R. A., k Sadler, E. M. 1991, MN, 250, 786.

138


Suzuki, T., Kumagai, S., Shigeyama, T., Nomoto, K., Yamaoka, H., k Saio, H. 1993, ApJ, 419, L73. Tanaka, Y. k the ASCA team 1993, IAU Circ, No. 5753. Turatto, M., Cappellaro, E., k Danziger, I. J. 1989, ESO Messenger, 56, 36. Turatto, M., Cappellaro, E., Danziger, I. J., Benetti, S., Gouifles, C , k Delia Valle, M. 1993, MN, 262, 128. Turtle, A.J. et al. 1987 Nature, 327, 38. Uomoto, A., k Kirshner, R. P. 1986 ApJ, 308, 685. Utrobin, V. 1993, A&A, 270, 249. Van Dyk, S. D., Sramek, R. A., Weiler, K. W., k Panagia, N. 1993a, ApJ, 409, 162. Van Dyk, et al. 1993b, IAU Circ, No. 5776. Wampler, E. J., k Richichi, A., 1989, A&A, 217, 31. Wamsteker, W. et al. 1993, IAU Circ, No. 5738. Weiler, K. W., Panagia, N., k Sramek, R. A. 1990, ApJ, 364, 611. Weiler, K. W., Van Dyk, S. D., Panagia, N., fc Discenna, J.L. 1991, ApJ, 380, 161. Weiler, K. W., Van Dyk, S. D., Panagia, N., k Sramek, R. A. 1992, ApJ, 398, 248. Weiler, K. W., Sramek, R. A., Van Dyk, S. D., k Pana R ia, N., 1993, IAU Circ, No. 5752. Wheeler, J.C. k Filippenko, A.V. 1994, in JA U Colloquium No I.{5. eels. R. McCray k Z. Wang, CUP, in press. Woosley, S. E., Eastman, R. G., Weaver, T. A., k Pinto. P. A.. 1994, ApJ, in press. Yamaoka, H. k Nomoto, K. 1991, in Supernova 1981'A and Oilier Supernovae, eds. I. J. Danziger k K. Kjar, Garching, ESO, p. 193. Zimmermann, H. U. et al. 1993a, IAU Circ, No. 5748. Zimmermann, H. U. et al. 1993b, IAU Circ, No. 5750. Zimmermann, H.-U. et al. 1993c, Astr. Ges. Abstract Series No. 9, 103. Zimmermann, H. U. et al. 1993d, IAU Circ, No. 5899.

2D Hydrodynamic Models of Supernova Progenitor Winds John M. Blondin Department of Physics North Carolina State University Raleigh NC 27695-8202

Abstract The conventional wisdom that a Type II supernova explosion occurs inside a spherical stellar wind bubble blown by the wind of the red snpergiant progenitor misses two important points: the progenitor wind may be time-dependent, and it may be asymmetric. These two features of SN progenitor winds have been well illustrated by the ring observed around SN 1987A. The existence of this circumstellar shell directly implies a time-dependence in the wind on time scales less than about 10,000 years. Also, the shell is undeniably asymmetric, implying some form of asymmetry in the progenitor wind(s). Some of the theories for an asymmetric circumstellar medium include gravitational focussing in a wide binary, rotationally deformed wind, colliding winds in a binary system, and asymmetric mass ejection in a common envelope or accretion phase of a close binary system. The wind dynamics of these various theories will be reviewed with an eye toward understanding the true history of Sk -69°202 .

1 Introduction The standard picture of a Type II SN progenitor star is a red supergiant (RSG) that has evolved from a massive star with an initial main-sequence mass above ~ I O M Q . These RSGs are observed to have very massive, slow winds with terminal speeds in the range of 10 — 50 km s" 1 , and mass loss rates in the range of 10~7 - lO~ 5 M0yr" 1 . These slow winds will gradually blow a stellar wind bubble of RSG wind into the relic main-sequence stellar wind bubble, building up a shell of shocked RSG wind at the edge of the expanding bubble. Given the typical lifetimes for this RSG stage, this RSG wind bubble is expected to reach a radius on the order of 10 pc. If one assumes that the RSG wind remained at constant velocity and mass loss rate throughout its lifetime, the blastwave from the SN explosion will therefore propagate through a circumstellar medium with p oc r~2 for roughly 5,000 years. This simple picture is not always appropriate: The progenitor stellar wind may be time-dependent, asymmetric, or both. The resulting circumstellar 139

140

J. Blondin: SN progenitor tvinds

medium (CSM) would then be non-spherical, and/or would not follow the p oc r~2 radial profile. The clearest evidence against a steady, spherical wind is the CSM around SN 1987A. Optical observations of SN 1987A have revealed a narrow ring surrounding the SN (Jakobsen et al. 1991), and slightly more extended lobes above and below this ring (Wampler et al. 1990). The inner ring has a radius of ~ 6.3 x 10 17 cm, a. radial velocity (either expansion or contraction) of 10.3 km s" 1 (Crotts & Heathcote 1991), and an inferred density of approximately 2 x 10~ 4 cm~ 3 (Lundqvist & Fransson 1991). The radius and low velocity of the ring implies a change in wind (and hence stellar) properties only some 104 years before the SN explosion. The nonspherical shape of the ring implies some sort of asymmetry in the stellar wind from the progenitor star in at least one phase of it's late evolution. In fact, a circumstellar shell at roughly the radius of the ring was anticipated by Chevalier and Fransson (1987). This conclusion was based on two observations. First, the pre-SN star was a blue supergiant (BSG) rather than a RSG as expected. Second, optical observations of the light echo revealed a circumstellar shell at roughly 3-4 pc (Crotts & Kunkel 1991). If this shell were interpreted as the edge of a stellar wind bubble created by a RSG, then one would be lead to the conclusion that the progenitor evolved from a RSG into a BSG before it's fateful explosion as a SN. This evolutionary scenario is consistent with at least some theoretical models for the evolution of the progenitor of SN 1987A (e.g., Arnett 1991). In this scenario, the fast wind of the BSG must have overtaken the slow wind of the RSG, sweeping it up into a shell. The surprise is not that there is a. circumstellar shell at a distance of ~ 10 18 cm, but that it is not spherical.

2 Interacting Winds Model Models for the formation of aspherical circumstellar structures (in planetary nebulae) in fact predate SN 1987A (Kwok 1982; Kahn 1982). In this "interacting winds" model, a fast, low density wind from a proto-white dwarf blows into the relic slow wind emitted from the previous asymptotic giant branch (AGB) stage of evolution. Because the later wind is faster than the previous AGB wind, it overruns and sweeps up the gas of the AGB wind forming a dense, expanding shell. If the AGB wind was not spherically symmetric, that asymmetry would be reflected in the asymmetry of the swept up shell. This interacting winds model was applied to the circumstellar shell of SN 1987A by Luo & McCray (1991) and Wang k Ma.zza.li (1992), and later by Blondin & Lundqvist (1993). In this case the shell was formed as the

J. Blondin: SN progenitor winds

141

fast wind of the BSG progenitor overtook the slow wind of the RSG phase. In order to calculate the evolution of the shell, Wang & Mazzali and Luo & McCray used two assumptions: that the shell of swept up RSG wind was infinitely thin, and that the shocked BSG wind was isobaric. The first assumption is valid as long as the shocked RSG wind has time to radiatively cool, collapsing to high density and very narrow width. The second assumption is valid in the limit that the sound speed in the shocked BSG wind is much larger than the expansion velocity of the bubble, cs ^ vexp. Luo & McCray (1991) pointed out that under this assumption any asymmetry in the BSG wind would be wiped out by subsonic motions in the shocked BSG wind, and would not affect the expansion of the pressure-driven shell. Hence the asymmetry of the shell must relate to an asymmetry in the RSG wind. While these models produced a nice explanation for the ring around SN 1987A, there remained a few problems. One of these dilemmas was the inconsistent conclusions: Wang & Mazzali (1992) claimed a. very small asymmetry was needed, while Luo & McCray (1991) employed a wind asymmetry of 5. Second, both of these models predated the reported velocity mapping of the shell by Crotts & Heathcote (1991), which found a very small expansion of the ring of only 10.3 km s" 1 . And third, the assumption of an isobaric interior turns out not to be a valid approximation for the parameters used in their models. This third point was illustrated by Blondin & Lundqvist using 2D hydrodynamical simulations. For the standard parameters used by previous authors, they estimated a ratio of cs/vexp ss 4. Although this ratio exceeds unity, it does not do so by much. Hydrodynamic simulations using these standard parameters showed that the interior pressure varied by more than two orders of magnitude, with a high pressure region formed on the equator at the point where the previous authors found the formation of a cusp. On the other hand, by the time the velocity ratio cs/vfrp reaches a value of 10, the interior pressure is relatively uniform. Note that a direct result of dropping the isobaric approximation is that an asymmetry in the BSG wind could give rise to an asymmetric shell. The principle conclusions of the hydrodynamical models of Blondin & Lundqvist (1993) were that a strong cusp was not formed as seen in the thin shell models, and that a very strong density asymmetry in the RSG wind was required to produce the observed ring/lobe structure around SN 1987A. A reasonably good model was constructed in this work by assuming some rather extreme parameters: a very slow, dense RSG wind with v = 5 km s" 1 and M = 2 x lO~ 5 M 0 yr~ 1 , a density in the equatorial wind of the RSG some 20 times higher than in the polar wind, and a very steep density

142


Fig. 1. A hydrodynamic model for the circumstellar shell surrounding SN 1987A (from Blondin & Lundqvist 1993). The density contours are spaced logarithmically. The highest density gas is shaded in black to highlight the location of the equatorial ring. profile in the RSG wind such that roughly 50% of the RSG wind mass was collimated within 10° from the equatorial plane. The circumstellar bubble resulting from the interaction of a spherical BSG wind with this aspherical RSG wind is shown in Figure 1.

3 Wind Asymmetry The hydrodynamical modeling of the ring around SN 1987A suggests that the interacting winds model can produce the observed structures if there is a significant asymmetry in the progenitor stellar wind. Why might the stellar wind have been asymmetric? At least three possibilities come to mind; rotation of a single star, interaction in a close binary system, or dynamically significant stellar magnetic fields. Due to lack of space I will


143

not review the possible role of magnetic fields, but simply refer the reader to the contributions in these proceedings by Livio and by Chevalier & Luo.

3.1 Stellar Rotation Let us consider the simplest case first: rapid rotation of a single star. Eriguchi et al. (1992) pointed out that although the radius of the final BSG progenitor was larger than the radius of the original main sequence star, the moment of inertia of the BSG was smaller than that of the main sequence progenitor due to the increased mass concentration toward the center of the star. Taking into account mass and angular momentum loss at each evolutionary stage, their work suggests that a main sequence star rotating at 15% of its critical rotation rate would end up as a BSG rotating at 35% of its critical rotation rate. This may be sufficient to produce a significant asymmetry in the BSG wind. Bjorkman & Cassinelli (1993) have recently presented a simple but powerful explanation for Be stars, of how a radiatively-driven wind from a rapidly rotating star will be strongly focussed into the equatorial plane. This effect is particularly strong in B-type stars where the terminal wind velocity is only of order the escape velocity from the surface of the star, as is likely for the immediate progenitor to SN 1987A. Although this model has yet to be directly applied to the progenitor of SN 1987A, 2D hydrodynamic simulations of a B2.5 star show that a rotation rate of 40% the critical rotation produces a density contrast from equator to pole of 2.5 (Owocki, Cranmer, & Blondin 1994). This density contrast rises sharply as the rotation rate nears the critical rotation rate. This theory of rotating winds may also apply to late-type stars as suggested by Livio (these proceedings). While the direct application of this theory to RSG winds must await a better understanding of such winds, it is at least interesting to speculate that this theory may suggest the existence of strongly asymmetric winds from rotating RSGs. Given that the winds from late-type stars are relatively slow compared to the escape velocity from the stellar surface, one might even speculate that this dynamical effect will be more pronounced in evolved stars than in early-type stars. Any confirmation of such speculation must await a theory for the origin of winds from late-type stars.

144

J. Blondin: SN progenitor xirinds 3.2 Interacting

Binary

There are a variety of possible ways in which the interaction of a binary companion could have affected the CSM of the supernova progenitor star. To simplify matters, we can separate these processes into two categories, interacting binaries and wide binaries, where here we use the term wide binary to refer to systems in which the binary separation is sufficiently wide, and the secondary mass sufficiently small that no mass is transferred between the stars and they remain two separate stars throughout the evolution. In the former category of interacting stars there are several processes that might lead to an asymmetric CSM shortly before the SN explosion. In the absence of an observed massive companion, these scenarios all come down to one common final stage in the binary evolution: a common envelope (CE). During the CE evolution angular momentum is transferred from the binary orbit to the rotation of the stellar envelope. The final outcome of the CE evolution depends critically on how much energy is transferred from the binary orbit into the CE. If sufficient energy is transferred to the CE that the envelope will be ejected from the system, the binary will end up with the companion star in a compact orbit around a BSG. The BSG will presumably have gained angular momentum during this process, but may not be in synchronous rotation with the binary orbit. If this is the case, tidal torques will continue to increase the rotation rate of the BSG star. In the absence of sufficient energy transfer during the CE evolution the binary companion will merge with the core of the SN progenitor while the CE is still partially intact. The final state of the system would then be a rapidly rotating RSG. Subsequent evolution must then produce a. BSG star, as constrained by observations. Conservation of angular momentum then implies that this BSG progenitor would be rotating near the critical rotation rate. Thus any CE evolution is likely to end with a rapidly-rotating BSG. If the initial binary separation of the progenitor system is larger than ~ 7 X 1013 cm the system does not evolve into a common envelope phase (Chevalier & Soker 1989) and we are left with our latter category of wide binary systems. In this case we would expect the companion star to still be in the vicinity of the supernova. Current observational constraints limit the mass of this companion to under 2.5 MQ (Crotts & Kunkel 1993). In this case there are at least three possible sources of wind asymmetry: tidal spin up of the progenitor star, tidally enhanced mass loss, and gravitational focussing of the stellar wind. Chevalier & Soker (1989) point out that for a. relatively narrow range of initial separations, the system will not enter a. CE phase, but will be close


145

enough to spin up the progenitor star during its RSG phase. This would give the RSG a value of Q/Q.cr ~ 0.5. This may or may not produce an asymmetric RSG wind, but it would certainly lead to a rapidly rotating BSG, which we have seen above implies an asymmetric BSG wind. It has also been suggested that a companion in a wide orbit could produce enough tidal distortion to generate an enhanced wind in the equatorial plane (Tout & Eggelton 1988). This effect would likely occur only for the same small range in binary separation that tidal spin up would be important. As in the case for rotationally-enhanced mass loss in RSGs, a quantitative estimate of tidally-enhanced mass loss must await a better understanding of the origin of winds from RSGs. If a low-mass companion is in too wide an orbit to tidally affect the progenitor star, it can still affect the RSG wind through its gravitational field. Because the RSG wind is moving relatively slowly, even a. small companion can gravitationally focus the wind, concentrating the mass of the wind into the equatorial plane.

4 Ring Formation Given these different possible sources of wind asymmetry, let us consider how they might give rise to the observed ring. Most of the sources of asymmetry discussed in the previous section rely either directly or indirectly on stellar rotation. In fact, this is the only suggested source of asymmetry (other than magnetic fields) for a single star progenitor. For the binary models, rotation is almost inevitable. If rotation of the RSG progenitor is able to focus the wind into the equatorial plane, perhaps by the Bjorkman-Cassinelli mechanism as in early-type stars, then the hydrodynamical simulations of Blondin & Lundqvist (1993) suggest that a compact ring can be formed in the interacting winds model. However, conservation of angular momentum implies that if rotation is important in the RSG wind, it is very important in the BSG wind. The interacting winds model must then include both an asymmetric RSG wind and an asymmetric BSG wind. If the shell expansion is sufficiently slow that the isobaric approximation is valid, then the asymmetric BSG wind becomes an unimportant feature of the model. If the isobaric approximation does not hold, then it becomes possible, if not likely, that an asymmetric BSG wind (due to rapid rotation) will lead to an asymmetric shell. Whether the asymmetry in such a shell matches the morphology of the ring around SN 1987A must await the results of further hydrodynamic simulations. We discussed a few scenarios for asymmetric CSM in §3 that did not in-

146

J. Blondin: SN progenitor luinds

volve a wind from a rapidly rotating star. These included mass ejection in a CE phase, tidally enhanced winds, and gravitationally focussed winds. Although the mass ejection in the CE phase has been shown to be strongly confined to the equatorial plane (Livio & Soker 1988), the typical outflow velocity is of order the orbital velocity of the merging companion. This is much higher than the 10.3 km s"1 expansion of the ring, and thus the identification of the ring material with the outflow from a CE stage must be considered tentative. One might envision that the mass outflow at a later stage in the CE evolution might not be so fast, but yet still be asymmetric. In the limit that this mass outflow is really just a stellar wind from a rapidly rotating RSG, we are back to the rotation models discussed previously. More work must be done on the dynamics of mass loss from CE systems to verify the validity of this class of models. In any case, the end result of a CE evolution must be a rapidly rotating star, which again leads us to the conclusion that the immediate BSG progenitor to SN 1987A had an asymmetric stellar wind. A tidally-enhanced RSG wind would presumably produce a mass concentration in the equatorial plane as assumed in the hydrodynamical models of Blondin h Lundqvist (1993). But again, we do not have an accurate description of such a model. In particular, would such a wind have a faster or slower velocity in the equatorial plane (a very slow velocity being required for the ring around SN 1987A)? Would the binary be close enough to also spin up the RSG, leading eventually to a rapidly-rotating BSG progenitor? These questions could be answered with hydrodynamic simulations if we had a good model for the origin of winds in RSG stars. Gravitational focussing appears to be the only source of asymmetry that is not accompanied by the ultimate conclusion that the BSG progenitor was rapidly rotating. However, it is not yet clear that the strong density asymmetry required in the models of Blondin k. Lundqvist (1993) could be produced by a low-mass companion. 5 Conclusions Although the existence of a circumstellar shell around SN 1987A is in fact expected based on the assumed evolutionary history of the progenitor, it is much harder to explain the severe asymmetry of the observed ring. The most likely source of asymmetry in the progenitor wind is rapid rotation in either the RSG or BSG phase of evolution. While a rapidly rotating BSG may occur in a single star system, a rapidly rotating RSG necessitates a binary system. In all binary scenarios (except perhaps a CE evolution in


147

which all the orbital angular momentum went into the ejected envelope), the final BSG star must have been rotating rapidly, and hence must have had an asymmetric wind in accordance with the Bjorkman-Cassinelli model. There are, however, other scenarios that could work, but have not been sufficiently explored. These include a tidally enhanced wind from the RSG, a gravitationally focussed RSG wind due to a solar mass companion, or perhaps a gentle form of envelope ejection in a CE evolution. Finally, a few words on future observations. Unfortunately, we will not be able to observe a possible companion star as small as a solar mass (Crotts, private communication), so we may never be able to rule a binary system out. We can, however, expect to get more information on the circumstellar structure when the young supernova remnant, begins to interact with the ring and lobes, and material exterior to the ring. In particular, any sort of asymmetry in the RSG wind will become readily apparent. Also, high resolution observations of the SNR may be able to determine the extent of any asymmetry in the progenitor BSG wind. I would like to thank many of the Herstmonceux conference participants, including Roger Chevalier, Vincent Icke, Mario Livio, and Noam Soker, and especially my collaborator in this field, Peter Lundqvist, for enlightening discussions on this subject. References Arnett, W. D. (1991), Astrophys. J., 383, 295. Bjorkman, J. E., & Cassinelli, J. P. (1993), Astrophys. J., 409, 429. Blondin, J. M., k Lundqvist, P. (1993). Astrophys. J., 405. 337. Chevalier, R. A., k Fransson, C. (1987), Nature, 328, 44. Chevalier, R. A., &; Soker, N. (1989), Astrophys. J., 341, 867. Crotts, A. P. S., k Heathcote, S. R. (1991). Nature, 350, 683. Crotts, A. P. S., k Kunkel, W. E. (1991), Astrophys. ./., 3G6. L73. Crotts, A. P. S., k Kunkel, W. E. (1993), IAU Circular, 5691. Eriguchi, Y., Yamaoka, H., Nomoto, K., k Hashimoto, M. (1992), Astrophys. J., 392, 243. Jakobsen, P., et al. (1991), Astrophys. J., 3C9, L63. Kahn, F. D. (1982), in IAU Symp. 103, Planetary Nebulae, ed. D. R. Flower (Dordrecht: Reidel), p. 305. Kwok, S. (1982), Astrophys. J., 258, 280. Livio, M., k Soker, N. (1988), Astrophys. J., 329, 764. Lundqvist, P., k Fransson, C. (1991), Astrophys. J., 380, 575. Luo, D., k McCray, R. (1991). Astrophys. J., 379, 659. Owocki, S. P., Cranmer, S. R., k Blondin, J. M. (1994). Astrophys. J., in press. Tout, C. A., k Eggelton, P. P. 1988, M.N.R.A.S., 231, 823. Wampler, E. J., Wang, L., Baade, D., Banse, K., D'Odorico, S., GouifFes, C , k Tarenghi, M. (1990). Astrophys. J., 362, L13. Wang, L., & Mazzali, P. A. (1992). Nature, 355, 58.

Supernovae with dense circumstellar winds N. N. Chugai Institute of Astronomy, Russian Academy of Sciences Pyatnitskaya 48, 109017 Moscow, Russia

Abstract The circumstellar (CS) wind around a type II supernova (SN II) can be revealed through the optical emission induced by the collision of SN ejecta with the wind. The optical manifestations of the ejecta-wind interaction provide an excellent tool for the study of the mass-loss history of pre-SN II at the final red supergiant stage. There is strong evidence that pre-SN II with an extraordinarily high mass-loss rate, M > 10~ 4 MQ yr" 1 , originate from the low-mass end of the massive star range (Mm, ~ 8 - IOMQ), while pre-SN II-P originating from Mmi > 12A/C, are characterized by a very low mass-loss rale, M < \()~bMQ yr~'.

SN 1979C (a type II-L), known for its powerful radio emission, was the first SN II where the late-time Ha luminosity was attributed to the ejecta-wind interaction (Chevalier & Fransson 1985). Yet the success of the radioactive model for the late-time luminosity of SN 1987A raised the problem of choosing between radioactive and shock-wave mechanisms in SN II. One possible solution was prompted by the observed excess in the Ha luminosity of SN 1980K (also type II-L and a strong radio emitter) at t — 670 days, relative to the predictions of the radioactive model (Chugai 1988). The interpretation of the excess in terms of the ejecta-wind interaction was supported by the strong radio luminosity and wide flat-top profile of Ha. In a sample of six SN II whose Ha fluxes were available (69L, 70G, 79C, 80K, 87A, 87F), three were found to possess significant excesses in their Ha luminosities (Chugai 1990). As mentioned already, two of these (79C and 80K) were radio emitters, but the third, SN 1987F, had not been detected at radio wavelengths. Nevertheless, the strong Ha luminosity of SN 1987F on day 150, L{Ra) « 1.5 1041 ergs s" 1 , indicates a dense wind with a parameter w = M/uw « 10 17 g cm" 1 , compared to 8 1015 g c m ' 1 for SN 1979C. Remarkably, the two SN II with the strongest excesses in their \la luminosities (79C and 87F) show a broad emission profile in this line with no absorption component. This property is consistent with the emission of the line from a narrow over-excited outer layer of the SN envelope. A similar pure-emission Ha profile is characteristic of the supernovae selected by Schlegel (1990) in the new subclass SN Tin. This property seems to be an indication that the luminosity of all SN Iln is strongly affected by the ejecta-wind interaction. This is particularly convincing in the case of the 148

N. N. Chugai: Supernovae with dense circnmstcllar winds

149

SN Iln SN 1988Z. Strong, variable, narrow, nebular lines in its spectrum are indicative of a dense CS wind (Filippenko 1991), while the broad Ha emission, lacking an absorption component, is consistent with the ejecta-wind interaction mechanism (Chugai 1990). This suggestion is confirmed both by the tremendously high Ha luminosity compared to that of SN 1987A and by the pronounced excess over the radioactive model (Turatto et al. 1993). In addition to attaining similar maximum Ha luminosities, SN 1987F and SN 1988Z exhibited similar light curves, characterized by an unusually slow decline in the B, V, R magnitudes. This resulted in extraordinarily high bolometric luminosities at the stage t > 100 days (Filippenko 1989; Stathakis and Sadler 1991; Turatto et al. 1993). This peculiar behaviour in their luminosities places these supernovae out. of the standard photometric subclasses II-P and II-L. As indicated above, the unusually strong excess in the bolometric luminosities of SN 1987F and SN 1988Z can be attributed to the energy released during the collision of the SN ejecta. with a dense wind. The required wind density parameter is w ss 101' g cm" 1 (Chugai 1992). This estimate may be reduced to w ss 210 1 6 g cm" 1 if the efficiency of the transformation of the kinetic energy into radiation is close to 100%. Nevertheless, even in this case, the CS winds around SN 1987F and SN 1988Z would be the densest ever detected around SN II. To emphasise the unusually strong effects of the ejecta-wind interaction on the Ha and broad-band emission, we tentatively designate such supernovae SN IIsw, with "sw" standing for "superwind" or "shock wave". Their key properties are (i) a strong Ha luminosity > 1041 erg s" 1 ; (ii) a, slow decay in broad-band optical luminosity relative to normal SN II; (iii) the lack of an Ha absorption component. Whether or not all SN Iln are identical to SN IIsw is unclear, since properties (i) and (ii) have been established only for SN 1987F and SN 1988Z. The details of the formation of the optical spectrum of SN IIsw are poorly understood. In the case of a sufficiently dense CS wind, w > 10 16 g cm" 1 , one expects that the reverse shock wave propagating into the ejecta. is radiative for approximately one year after the explosion, and that a cool dense shell forms (Chevalier & Fransson 1985). The cool dense shell, irradiated by X-rays, could be responsible for the bulk of the optical radiation in SN IIsw. However this shell is liable to fragmentation clue to Ray leigh-Taylor instability (Chevalier & Fransson 1985). If certain conditions are met (relatively large geometrical thickness compared to Sobolev length; large covering factor) the line profile, even from a clumpy shell, can be formally described in the Sobolev approximation. For SN 1987F on day 150 the Ha profile can

150

N. N. Chugai: Supernovae with dense circumstellar winds

be reproduced by an optically thick outer shell with zero velocity gradient attached to the partially opaque inner ejecta with rp ss 1 (Chugai 1991). However, this simple model fails in the case of SN 1988Z . In addition to the broad component of Ha (FWHM « 104 km s" 1 ) and the unresolved line from the photoionized wind, a third, intermediate component with FWHM « (1 - 2) x 10 3 km s" 1 is seen (Filippenko 1991; Turatto et al. 1993). A promising model for the origin of this spectrum invokes a clumpy wind, with the intermediate component originating from a radiative shock wave being driven into the dense clumps (Chugai & Danziger 1994). In a similar model, the interaction of the SN ejecta with a clumpy wind may account for the optical and X-ray emission from the strong radio/X-ray supernovae SN 1986J and SN 1978K (cf. Chugai 1993). The proposed dumpiness of the pre-SN II wind is supported by the increasing evidence for clumpy structure in RSG winds, revealed by molecular radio line observations (cf. Olofsson, this volume). SN II-P constitute roughly two thirds of all SN II but, so far, have never shown any of the signatures of a CS wind. An upper limit for the wind density around SN II-P can be obtained from the maximum velocity, vmax, determined from the blue wing of the Ha absorption. This velocity is obviously lower than the velocity of the ejecta at the reverse shock wave, vmeLX < Rc/t, where Rc is the radius of the contact discontinuity, and t is the expansion time. To estimate the maximum wind density, a constant density in the inner part of the SN II, p oc v~k in the outer part, and a wind density profile p = w(4nR2)~1 is adopted. The self-similar solution for Rc (cf. Chevalier 1982) is used, and M = 12M© for the ejecta mass, 1051 ergs for the kinetic energy and k = 8 are assumed. From observation, vmax = 11900 km s" 1 in SN 1969L on day 59 (Benetti 1991) and vmaK = 13000 km s" 1 in SN 1990E on day 32 (Schmidt et al. 1994). The resulting upper limit is to < 1015 g cm" 1 for both SN II. This limit decreases if k or M increases. The derived upper limit shows that the wind density around SN II-P is relatively low compared with that of SN 1980K (type SN II-L) where i » « 3 1015 g cm" 1 (Lundquist & Fransson 1988). The origin of the dramatic difference in the mass-loss rates among preSN II, ranging from M < 10~5 M© yr" 1 for SN II-P to ~ 5 10" 4 M Q yr" 1 for SN IIsw, is unclear. Basically, two factors may affect the mass-loss rates of pre-SN II viz. main-sequence mass, and the separation of the components if the SN II progenitor is in a binary system. With some suitable initial values for the main-sequence mass and binary separation, the common envelope regime at the RSG stage might stimulate a robust mass-loss at just the appropriate epoch viz. ~ (several)xlO 4 yrs prior to the explosion. This

N. N. Chugai: Supernovae with dense circumstellar winds

151

possibility, while plausible, has not yet been given detailed consideration (A.V.Tutukov, private communication). Alternatively, the range of main-sequence masses alone could be responsible for the observed variation in mass-loss rates among pre-SN II. Although standard mass-loss rates for massive stars (de Jager & Nieuwenhuijzen 1988) are insufficient to account for the rates inferred in pre-SN IIsw, nevertheless some facts favour this alternative. Firstly, the low ejecta mass of SN 1988Z indicates a progenitor main-sequence mass in the range 8 — 10 MQ (Chugai & Danziger 1994). Secondly, SN II-P always exhibit a strong saturated [01] 6300, 6364 A doublet, suggesting that supernovae of this subclass originate from stars which produce large amounts of oxygen. Such stars have masses M m s > 12 MQ (cf. Woosley & Weaver 1986). Thirdly, the masses of SN II-L seem to be generally lower than those of SN II-P (cf. Chugai 1990). All these facts may be reconciled with the data on the pre-SN II winds if the main-sequence mass increases along the sequence SN IIsw, SN II-L, SN II-P, while the mass-loss rate at the final R.SG stage decreases with main-sequence mass in the range 10-12 MQ. This scheme suggests that preSN II-P (Mms > 12 MQ) lose their mass according to the de Jager law, while the mass-loss rates of SN II-L (Mms « 1 0 - 1 2 MQ) and SN IIsw (Mms ss 8 — 10 MQ) are substantially higher. It is tempting to identify the tremendously high mass-loss rates of preSN IIsw with a superwind phenomenon in stars M m s « 8 — 10 MQ. This superwind might have an origin similar to that seen in WD producers (Mms < 8 MQ). However, given the relatively low rate of SN IIsw, only 10-20% of all the stars in the range 8-10 MQ should result in such supernovae. Other supernovae arising from this mass range must be unobserved. This could happen if e.g. (i) the hydrogen envelope is completely lost before the supernova explosion, so that the pre-SN becomes a blue helium star, and (ii) the amount of ejected 56 Ni is very low (< O.OIMQ). At explosion, such a supernova would be faint with a narrow (5-10 days) maximum of absolute magnitude only « —14. The hydrogen lost by the progenitor would lie not far from the supernova, at a distance R = vwtss ~ 10 17 cm, where ^BS * 10 3 ~ 4 yr is the duration of the final blue stage and vw w 10 km s" 1 is the RSG wind velocity. Such faint SN II may be recoverable in optical, radio and X-ray observations after several years, due to the collision of the supernova ejecta with the dense wind (e.g. SN 1978K and SN 1986J?).

152

N. N. Chugai: Supernovae with dense circumslellar winds

References Benetti, S. (1991). SN 1987A and other supernovae, eds. I.J. Danziger & K.Kjar (ESO, Garching), p. 339. Chevalier, R.A. (1982). Astrophys. J., 258, 790. Chevalier, R.A. & Fransson, C. (1985). Supernovae as Distance Indicators, ed. N. Bartel (Springer, Berlin), p. 123. Chugai, N.N. (1988). Astrophys. Space Sci., 146, 375. Chugai, N.N. (1990). Sov. Astr. Lett., 16, 457. Chugai, N.N. (1991). Mon. Not. R. astr. Soc, 250, 513. Chugai, N.N. (1992). Sov. Astr., 36, 63. Chugai, N.N. &; Danziger, I.J. (1994). Mon. Not. R. astr. Soc.,, in press. Chugai, N.N. (1993). Astrophys. J. Lett., 414, L101. Filippenko, V.A. (1989). Astron. J., 97, 726. Filippenko, V.A. (1991). Supernovae, ed. S.E. Woosley (Springer, New York), p. 467. de Jager, C. & Nieuwenhuijsen, R. (1988). Atmospheric Diagnostics of Stellar Evolution, ed. K.Nomoto (Springer, Berlin), p. 122. Lundqvist, P. &; Fransson, C. (1988). Astron. Astrophys., 192, 221. Schlegel, E.M. (1990). Mon. Not. R. astr. Soc, 244, 269. Schmidt, B.P., Kirshner, R.P., Schild, R. et al. (1994). Astrophys. J., (in press). Sramek, R.A., Weiler, K.W. & Panagia, N. (1990). IA U Circular No. 5112. Stathakis, R.A. & Sadler, E.M. (1991). Mon. Not. R. astr. Soc, 250, 786. Turatto, M., Cappellaro, E., Danziger, I.J., Benetti, S., Goiiiffes, C. & Delia Valle, M. (1993). Mon. Not. R. astr. Soc, 262, 128. Woosley, S.E. & Weaver, T.A. (1986). Ann. Rev. Astron. Astrophys., 24, 205.

Compact Supernova Remnants Roberto J. Terlevich Royal Greenwich Observatory, Madingley Road, Cambridge CB3 OEZ, U.K.

Abstract Two new kinds of peculiar type II supernovae (SNe) have been observed recently: namely the very luminous type II radio supernovae (RSNe) and the so-called Seyfert 1 imposter. I will show that a simple model of interaction of supernova (SN) ejecta with a high-density homogeneous circumstellar medium (CSM), combining analytic and numerical hydrodynamic simulations together with static photoionization computations, can describe their observed emitted spectrum, optical light curve, X-ray luminosity and emission line widths. I suggest that these two new kinds of SNe are not peculiar type Us, but are, in fact, the optical or radio manifestation of the same phenomenon, i.e. the interaction of the SN ejecta with a high density CSM. During the interaction with a high density CSM a young remnant can radiate most of its kinetic energy and outshine the SN event itself; therefore to emphasize the unique aspects associated with this type of event, I suggest calling this group of small, luminous and rapidly evolving remnants, compact supernova remnants (cSNRs).

1 Introduction The defining characteristic of type II SNe is the presence of very broad Ha emission with a strong P-Cygni profile. A small number of peculiar type II SNe have been found in recent years which are either very bright in the optical continuum with very strong and broad Ha emission without a P-Cygni profile, or are strong radio sources then called radio supernovae. Probably significant is the fact that these SNe tend to be associated with regions of active star formation. The optical light curve of peculiar type II SN, such as SN 1987F (also know as Seyfert 1 imposters; see Filippenko 1989) and SN 1988Z, is characterized by a high luminosity maximum followed by an extremely slow decay. The decay is so slow that the broad Ha emission can be observed for several years after discovery. SN 1988Z is also very bright at radio frequencies. In the past 10 years, since the discovery of radio emission in SN 1979C in M100 in the Virgo cluster, several luminous RSNe have been discovered and studied in detail (Weiler et al. 1989, 1990). Some appear to be very luminous at optical wavelengths and with optical spectra dominated by relatively broad Balmer lines showing no P-Cygni profiles. The most luminous examples so far are SN 1986.1 and SN 1988Z. The radio spectrum of RSNe 153

154

R. J. Terlevich: Compact supernova remnants

is inverted at low frequencies at early times, with a turnover above a critical frequency that decreases with time. This behaviour is interpreted in terms of time-dependent free-free absorption by circumstellar material (Weiler et al. 1989). Both the high Ha luminosity with no P-Cygni profiles, and the strong radio emission can be interpreted as resulting from an interaction between the expanding SN ejecta and a homogeneous, dense circumstellar shell created by the interaction of the slow wind from the progenitor star with the high ambient pressure of the surrounding HIT region (Chevalier 1982; Terlevich et al. 1992; Chugai & Danziger 1994). Supernova remnants evolving in a dense and homogeneous CSM (n > 105 cm" 3 ) reach their maximum luminosity (L > 107 L@) at small radii (R < 0.1 pc ), soon after the SN explosion (t < 20 yr) and while still expanding at velocities of more than 1000 kms~1(Shull 1980; Wheeler et al. 1980; Draine and Woods 1991; Terlevich et al. 1992). In these compact SNR.S, radiative cooling becomes important well before the thermalization of the ejecta is complete, making the remnant miss the Sedov track. As a result, the shocked matter undergoes a rapid condensation behind both the leading and the reverse shocks. Two concentric, high-density, fast-moving thin shells are then formed. The cool, dense shells, the freely expanding ejecta, and a section of the still dynamically unperturbed interstellar gas, are all irradiated and ionized by the photon field produced by the radiative shocks. This contribution describes an attempt to explain the most luminous optical and radio SN events observed in the past few years. 2 cSNRs, or SNRs evolving in a high density medium The interaction of the SN ejecta with a dense CSM causes a shocked region of hot gas enclosed by two shock waves: on the outside the leading shock, and on the inside the inward facing "reverse" shock. The leading shock (V ~ 104 kms" 1 ) encounters dense circumstellar material and raises its temperature to ~ 109 K. The reverse shock, which is initially substantially slower (V ~ 103 kms" 1 ), begins to thermalize the supernova ejecta to temperatures of about 107K. Early analytical and numerical computations of the evolution of SNRs in a dense medium (Chevalier 1974; Shull 1980; Wheeler et al. 1980) showed a speeded-up evolution compared with the "standard" solution in a medium of no — 1 cm"3 . All evolutionary phases (free expansion, thermalization of the ejecta, the quasi-adiabatic Sedov phase, the radiative and the pressure modified snow-plough phases)


155

which have been thoroughly studied for the standard case, are substantially speeded up. For SNe evolving in circumstellar densities of the order of no ~ 107 cm" 3 , the cooling-time (tc) and cooling-length (r c ) scales for the post-shock temperatures are very small. f

f. ^ n 9

rv

LQ —

°

w»i^^^—^^—

1

14

vr

(A\

y i«

i j. i

U3

r c ~-* c t, s ~1.8xlO —f- cm, 4

(2)

TJ7A23

where v$ = u/10 8 km s" 1 and A23 = A/10" 2 3 erg cm 3 s""1. Thus, radiative losses become important at very early times when the shock velocities and temperatures are vs > 103 km s" 1 and Ts > 107 K, well before the ejecta is even thermalized. This means that a large flux of ionizing photons will emerge from the shocked gas at X-ray energies. For a supernova remnant which injects 1051 ergs into a medium of constant density n-j = n o / 1 0 7 c m - 3 the onset of the radiative phase behind the leading shock (assuming f-f cooling only) causing the formation of a dense outer shell, (Shull 1980, Wheeler et al. 1980, Draine and Woods 1991), begins at a time tsg given by tsg = 230 E\l%n~3lidays

(3)

where £51 is the energy deposited by the SN in units of 10 51 ergs. At this stage, the shock is at a radius of Rshock = 0.01 £ 5 ( iij '

(-—J

pc

(4)

with velocity,

Vshock = 4600 EH8 A1' (j-) ' ^ km s"1

(5)

temperature, tsg

and luminosity, Lshock = 2 x 10 43 E75l8 n 3 / 4 ( / - ) ~ n / < ergs" 1 .

(7)

156


These approximate formulae assume that the ejecta has already been fully thermalized. However, for the case of interest in this work (i.e. for no > 105 cm" 3 ) strong radiative losses occur before thermalization is completed. Because the cooling processes radiate the thermal energy at the same rate as thermalization proceeds, the Sedov phase is totally inhibited and thus there is no self-consistent analytic treatment for the evolution of such remnants. It is therefore necessary to follow the detailed time evolution of the gas flow. In particular, special care should be given to the post-shock structure which is sensitive to the details of the ambient density distribution and to the temperature dependence of the radiative cooling function. The cooling time scale, tc, for an optically thin plasma, is proportional to the inverse of the gas density and the evolution proceeds faster at higher ambient densities. On the other hand, the temperature dependence of the cooling function is different for different temperature ranges. Adiabatic shocks are stable but cooling instabilities can develop over a wide range of radiative shock conditions (Avedisova 1974; Falle 1975, 1981; McCray, Stein and Kafatos 1975; Chevalier and Imamura 1982; Imamura 1985; Bertshinger 1986). The details of the transition from a nearly adiabatic to a strongly radiative shock depend on the ability of the gas to readjust to the cooling rate. Pressure gradients tend to be smoothed out in a. sound-crossing time, td, and the ratio tc/td provides an estimate of the conditions prevailing in the cooling gas. For tc/td > 1, at moderate cooling rates, the gas elements are continuously compressed as their temperature falls and the cooling process operates quasi-isobarically at the pressure attained by the gas immediately behind the shock. For i c /i^ < 1, however, the cooling rate dominates over any pressure readjustment and the process becomes quasi-isochoric at the post-shocked density of the cooling gas elements. A large pressure imbalance then develops in the flow, and new additional shocks are generated which end up compressing the cooled gas. This process, termed "catastrophic cooling" (Falle 1975, 1981), appears during thin shell formation and the instabilities continue to operate during the rest of the radiative shock evolution (Chevalier and Imamura 1982; Bertshinger 1986; Cioffi et al 1988; Tenorio-Tagle et al 1990). The catastrophic cooling acquires a central role in the case of supernovae evolving in high density media due to the strength of the radiation produced upon cooling, and the rapid variations inherent in the shock propagation. These features imply that a large flux of ionizing photons will emerge from the shocked gas. The wide range of gas temperatures in the cooling region results in a power-law-like spectrum at UV and X-ray frequencies.


157

3 Numerical hydrodynamic models I present here a short summary of the calculations that follow the evolution of a SNR caused by the release of EQ = 1051 ergs into a constant density medium of n0 = 107 cm" 3 (Terlevich et al. 1992; Tenorio-Tagle et al. 1994). The results of numerical simulations in one and two dimensions indicate that the onset of radiative cooling starts before thermalization is completed for values of no > 105 cm" 3 . Given the large densities and thus the strong radiative cooling, the combination of values of thermal (Eth ~ 0.7 X Eo) and kinetic (E^n ~ 0.3 x EQ) energies which characterizes the Sedov solution is never achieved. Instead, the thermal energy content is radiated away within the first 5 years of evolution, before the thermalization of the ejecta is completed. The calculations also show how both energies continuously drop after strong cooling occurs, and values of Eth much smaller than 0.1 x £o are soon achieved. Such low values of Eth occur in the standard case some 106yr after the explosion. The present solution thus indicates a speeded up evolution and, at t = 10 years, most of the injected energy has been radiated away. The remaining kinetic energy, a. major fraction of which is stored in the remnant outer shell, also decays rapidly. Clearly, given the similar rate of change of both Eth a n d Ekin towards the end of the calculated evolution, the kinetic energy is radiated as soon as it is thermalized behind the shock waves. Strong radiative cooling leads to the development of a thin-shell at the edge of the remnant as well as the collapse of the shocked ejecta behind the reverse shock. The 2-D calculations, which also allow for the development of cooling instabilities, clearly show, as in 1-D, the steady approach of the two shells of cool matter, promoted by the weakening and withdrawal of the reverse shock while the unshocked ejecta fills the remnant interior. The various numerical solutions, despite different resolutions and initial conditions, all show a small (Rsnr ~ a few times 1016cm) rapidly radiating remnant, with two massive and geometrically thin concentric shells moving at large speeds and bounded by radiative shock waves. A small fraction of the swept-up mass lies between the two shocks and remains at very high temperatures (about 108K). On the other hand, in the central regions of the remnant, the ejected matter continues to expand homologously, and thus presents a density and velocity distribution which reflect the initial conditions, until it is thermalized at the reverse shock. This central region, together with a section of the still unperturbed background gas and the two thin, cool shells, are clearly subjected to the ionizing radiation provided by the radiative shocks.

158

R. J. Terlevich: Compact supernova remnants Table 1. Shock evolution for no — 10 7 cm~3 Time Ug 1 2

3 4

6 10 25 50

Time years

Log(fl)

0.63 1.26 1.89 2.52 3.78 6.30 15.7 25.6

16.5 16.6 16.65 16.7 16.75 16.8 16.9 17.0

cm

v,h kms *

Log(T) K

4600 3700 2760 2250 1680

8.3 7.8 7.6 7.4 7.2

1170 610 370

6.9 6.3 5.9

Log(L) ergs" 1 43.3 42.8 42.6 42.4 42.1 41.7 41.1 40.6

Log nshell

cm'3 12.0 11.6 11.3 11.1 10.9 10.6 10.0 9.6

The emitted spectrum is similar for both the main shock moving into the interstellar medium and the reverse shock thermalizing the ejecta. The details, viz. the specific temperature structure of the cooling region and the resulting photon spectrum, cannot be resolved with the numerical time dependent calculations and have been computed with analytical solutions (see Terlevich et al. 1992). The wide range of temperatures in the cooling region results in a power-law-like spectrum at UV and X-ray frequencies. In a strict sense the analytical formulae, described in the previous section, can only be applied when tsg > tth, and this occurs only for no < 5.0 X 10 5 cm~ 3 . Comparing our numerical and analytical results we found that at densities n 0 > 5.0 x 10 5 cm~ 3 , the analytical formulae still give a good description of the run of luminosity, shock velocity, size and shock temperature with time. One important difference is that for these densities, thin shell formation and peak luminosity occur at about 2.5tsg. Table 1 shows the results for a. shock with a total energy of 3 x 1051 ergs" 1 , interacting with a medium with initial density nQ = 107 cm" 3 for several units of tsg = 230 days. Table 1 does not include the luminosity of either the hot cavity or the reverse shock. In the homogeneous case the dominant flux is that of the leading shock (Terlevich et al. 1992).

4 Photoionization models Typical values for the density, size, column density, velocity of the cooled regions of gas, and ionizing flux of the radiative shock, were obtained from the analytical and numerical hydrodynamical models described in the previous sections. These parameters were used as input to the photoionization code CLOUDY (Ferland 1990). Table 2 lists the results of the photoionization


159

Table 2. cSNR line luminosities Time t.g

Log L(Ha ) ergs" 1

1 41.55 2 3 4 6 10 25 50

41.47 41.38 41.29 41.08 40.73 39.81 39.12

Balmer

Log L(H0) ergs" 1 41.04 40.86 40.67 40.48 40.07 39.42 38.71 38.42

Decrement 1.9 2.8 4.2 6.3 13.5 44 89 63

3.3 4.1 5.1 6.5 10.2 20.4 12.6 5.1

models for the same set of parameters listed in Table 1 and solar abundance gas. Values of Ho luminosity in Table 2 for tsg < 2.5 should be treated with caution because before thin shell formation they represent an upper limit to the total Ha luminosity emitted by the cSNR. Values are reliable after thin shell formation, i.e. for t > 2.5 x tsg. General trends can be recognized, i) Both Ha and H/? luminosities decrease steadily with time ii) The Balmer decrement increases reaching a maximum of around 20; only in the early phases is the value of the Balmer decrement close to the Case B recombination value of about 2.85, and iii) Lya is very strong at later times becoming the main coolant in the photoionized region. These results apply only to the broad component with line widths similar to the shock velocities listed in Table 1.

5 Comparison with observations The nearest example of a RSN is probably SN~1955 in the core of the nearby starburst galaxy M82. This RSN, with an age between 20 and 50 yr, has a radio luminosity equivalent to 200 times that of the most luminous galactic SNR, Cass A. This remnant and SN 1986J are the only ones for which VLBI radio images are available. The interpretation of the radio data indicates high CSM densities (~ 107 cm" 3 ) and large progenitor masses (M> 20 M 0 ) for these two RSNe. The most dramatic example of bright RSN was discovered in the distant irregular galaxy Mk297. Mk297A reached a maximum luminosity in 20cm equivalent to 30,000 Cass A.

160

R. J. Terlevick: Compact supernova remnants Table 3. cSNR

candidates

SN name

Galaxy

a 1950. 0

Mrk 297A SN 1988Z SN 1986J SN 1980K SN 1987F SN 1979C SN 19881 41.9+58 SN 1978K

Mrk 297 Zw095-049 NGC 891 NGC 6946 NGC 4615 NGC 4321 (M100)

16 03 10 49 02 18 20 34 12 39 12 20 10 18 09 51 03 17

— NGC 3040 (M82) NGC 1313

01.72 10.62 22.6 26.7 07.7 26.71 17 41.96 38.62

6 1950.0 20 40 43.6 16 15 56.4 42 06 18.9 59 55 56.5 26 20 49 16 04 29.5 35 54 .7 69 54 57.4 -66 33 03.4

Notes: * Coordinates are for galaxy nucleus. SN was 5" E and 1" N.

Other notable examples are the "Seyfert 1 imposters" SN 1987F and SN 1988Z. The enormous Ha and bolometric luminosities of these two objects have been interpreted as resulting from the interaction of the ejecta with a high density (n ~ 107 cm" 3 ) CSM. The case of SN 1988Z is particularly important because strong radio emission has been reported and the CSM density was measured, using the [OUT] narrow lines ratio, to be about 107 cm" 3 . SN 1988Z is therefore particularly interesting in that it shows a link between luminous RSNe and cSNRs. Almost all the listed cSNR candidates are reported to be associated with HII regions or regions of active star formation. 5.1 SN 1988Z At a distance of 133 Mpc (cz - 6670 kms" 1 ; for HQ - 50 kms" 1 Mpc"1) this is one of the most distant examples of a cSNR. SN 1988Z is one of the most extreme cases of bright type II with very slow rate of decay and lack of P-Cygni profiles in the emission lines. The rate of decay is so slow that at an age of 568 days (this is a lower limit to the age) SN 1988Z was more than 3 magnitudes brighter than a typical plateau type II (type IIP) (Stathakis and Sadler 1991). Good spectroscopy has been reported up to day 1149 (Stathakis and Sadler 1991; Turatto et al. 1993). Narrow emission lines from the CSM were detected in spectra, taken on day 73. Electron densities in the range 4 X 106 to 2 X 10' cm" 3 are determined from the ratio of [OIII]5007A to [OIII]4363A for an assumed electron temperature in the region of 20,000-7,000 K. This is the best determination to date of the

161


42

i

i

i

1

••

1

1

!

1

1

!

1

•

O' •

•

•

•

•

•

•

«

.

"•Q

• •

o...

•

o

*

c

E u

'

41

_l en

:

O

•

O

/\c\

i

1

200

400

i

600

800 Time

1000

1200

1400

Fig. 1. Observed light curve of SN 1988Z in Ho (Dots) compared with the prediction of the simple homogeneous model with no = 10' cm" 3 (Open circles). Luminosity in ergs" 1 time in days.

pre-shock circumstellar density and coincides with the assumed density of the canonical model in Tables 1 and 2. Figure 1 shows the observed Ha luminosity of SN 1988Z as a. function of time. The data were taken from Stathakis and Sadler (1991) and Turatto et al. (1993). The origin of the time axis is the one adopted by the observers, viz. an explosion epoch of 1988 December 1. The predicted Ha luminosity curve from Table 2 is also shown; again the origin of the time axis corresponds to the SN explosion epoch. The predicted light curve gives a reasonable but not excellent prediction of the observed light curve. Both the value and epoch of the maximum luminosity in Ha seem well predicted, but the decay of the model is too slow compared with the observations. This suggests either a CSM density slightly higher than the value adopted here or a radial gradient in the CSM density. Integration of either the Ha light curve or the blue light curve indicates, after applying a moderate bolometric correction, that the total energy radiated by SN 1988Z in the first 4 years was in excess of 2 X 1O50 ergs. This

162


suggests that a substantial fraction of the total kinetic energy was radiated in this phase, supporting the classification as a compact remnant rather than a supernova (see also Chugai 1992) Radio emission at the level of about 1 mJy at 6cm was detected in 1990 July (Sramek et al. 1990). This detection indicates that SN 1988Z is among the brightest RSN observed; in fact it is even brighter than SN 1986J. This suggests a link between RSNe and cSNRs, or between strong radio emission and strongly radiative shocks. This is confirmed by the analysis of the optical and radio data of SN 1986J and comparison with the model predictions. 5.2 SN 1986J Powerful radio emission was detected in the disk of NGC891 in 1986 August by Rupen et al. (1987). With a radio luminosity equivalent to 2,000 times Cass A, this was, at that time, the most luminous RSN discovered. The optical spectra was dominated by broad Ha emission with a FWHM= 1070 km s"1 a very high Balmer decrement Ho /H/3 ~ 60, and broad [OIII] lines. The location of this SN near the minor axis of an edge-on spiral galaxy and the large observed Balmer decrement seem to suggest a large amount of reddening. The estimated time of the explosion is 1982 September, giving an age of about 4 years at the time of the spectral observations. The fact that it was possible to see the SN 4 yr after explosion is apparently in contradiction with the expected large extinction. The observed average Ha luminosity in 1986 September/October is 5.1 x 1038 ergs" 1 and the observed average Balmer decrement is about 60. The age of the remnant was, at the time of the observations, more than 5 years or almost 8 tsg. For this age we obtain from Table 1 a shock velocity of 1400 kms" 1 , an Ha luminosity of 8 x 1040 ergs" 1 and a Balmer decrement of 14. Assuming 14 to be the intrinsic Balmer decrement, an extinction coefficient of c = 1.9 would be needed to explain the observed Balmer decrement. Correcting the observed Ha flux for this value of c gives an intrinsic luminosity for SN 1986J of 4.2 x 104Oergs~1 . This result is within a factor of two of the predicted value of the Ho flux (Table 1.) Bregman and Pildis (1992) detected X-ray emission from SN 1986J in 1991 August. The thermal emission model fits show two \ 2 minima: a low column density x 2 minimum giving a best fit with a luminosity 2.5 X 1O40 ergs" 1 between 0.1 and 10 keV, for T = 3.9 keV and a high column density X2 minimum indicating a softer spectrum (T = 0.35 keV) and an X-ray luminosity of 1042 ergs" 1 . When fitted by a power law over a bandpass of 0.1 - 2.5 keV, the spectrum of SN 1986J is described by a very soft

R. J. Terlevick: Compact supernova remnanis

163

spectral index of a = —2.1. This again suggests the presence of a very soft component in the X-ray spectrum. At the time of the X-ray detection the age of the cSNR was 10 years or I6tsg. The predicted X-ray luminosity at that time is about 2.5 x 10 41 ergs" 1 , ten times larger than the best fit solution but smaller than the value for the soft spectrum case. At this age the shock spectrum is very soft with a temperature of about 0.4 keV, similar to the best fit for the second X2 minimum in the thermal solution.

5.3 SN~1955 in M82 (41.9+58) This RSN is the most luminous radio source in M82. At a distance of 3.3 Mpc this is also the nearest example of a RSN. Its spectral evolution has been followed since 1973 and it shows increasing optically thick flux at low frequencies and decreasing flux at high frequencies (Kronberg et al. 1985). This RSN has a radio luminosity equivalent to more than 200 times that of the most luminous galactic SNR, Cass A, and together with SN 1986J, are the only two RSNe for which VLBI radio images are available. The measured diameter and the inferred column density (from the observed critical frequency of thermal absorption) requires that the shell of pre-supernova ejecta had a thickness of about 3 x 1016 cm and a density of no ~ 107 cm"3 at the time of the Kronberg et al. (1985) observations. These values are remarkably similar to the expectations of our canonical model in Tables 1 and 2. All the evidence points to a massive (M> 20 M@) progenitor. A compact and variable X-ray source with a luminosity of about 3 X 1039 ergs" 1 in the band 0.1-2.5 keVhas been identified at the radio position of this RSN (Kronberg et al. 1985, Collura et al. 1994). The X-ray flux seems consistent with the prediction in Table 1 and the detection of rapid X-ray variability on time scales of less than a day raises very important questions relating to the origin of such variability (Terlevich and Fabian in preparation). 5.4 SN 1981F SN 1987F was discovered in 1987 March. Early spectra revealed broad Ha superimposed on a nearly featureless continuum. The Ha profile did not have the characteristic P-Cygni profile of type II SNe (Filippenko 1989). The observed light curve was extremely flat (although steeper than that of SN 1988Z), fading by about 2.5 magnitudes in 400 days in the R. band; a typical type IIP will fade about 5 magnitudes in V in the same interval of

164


time. A few months after discovery the spectrum was dominated by broad hydrogen, Fell and Call emissions. The estimated density of the ejected envelope at this late time was > 107 cm" 3 . The maximum in the observed Ha luminosity was 2.0 X 1041 ergs" 1 at an age of about 185 days. The FWHM of the Ha emission was about 7000 km s - 1 at that time. Comparing the results of the observations with the models of Table 1 and 2 we see that again the simple homogeneous models for n0 = 107 cm"3 give a reasonable description of the line widths and Ha luminosities. 6 Conclusions and future prospects We have seen that the optical and radio observations of cSNRs indicate a high circumstellar density of around 10" cm"3 and remnant sizes of a few times 1016cm with line widths (FWHM) between 1,000 and 10,000 kms" 1 . The size, Ha luminosity, X-ray luminosity and time scale of evolution predicted by the simple spherical homogeneous model of a radiative SN shock in a medium with density 107 cm"3 shows good agreement with the observations of cSNRs. This gives support to the idea that these systems are indeed strongly radiative shocks in the circumstellar environment. The study of strongly radiative astrophysical shocks is at an early stage. These shocks provide a unique laboratory capable of yielding valuable insights into physical processes in extreme conditions. They give unique information about the CSM and the ejecta of massive stars and can provide clues about some important aspects of galaxy evolution and of active galactic nuclei. Much work is needed, on the one hand to identify cSNR candidates early enough in their evolution, on the other to obtain good quality data over the widest possible spectral range. Radio and X-ray frequencies are, besides the optical, the most important spectral windows for observations of cSNRs. High quality and high resolution spectral information is particularly needed in the X-ray region where most of the energy is radiated. UV monitoring with high time-resolution will be very valuable if cSNRs prove to be variable on time scales of weeks. Due to their high luminosities, the study of cSNRs can be performed even at very large distances. The sample of cSNRs listed in Table 3 is perhaps the best data set for the study of shocks in a high density environment. Future developments in the theory should explore the effects of cooling out of equilibrium in strongly radiative shocks, the generation of cooling instabilities and luminosity variability and the effects of magnetic fields in the evolution of these cSNRs.


165

Acknowledgements I would like to take this opportunity to thank Elena Terlevich for her support and help with this paper. Most of the work reported here is the result of a long-term collaboration with Guillermo Tenorio-Tagle, Pepe Franco, Michal Rozyczka and Jorge Melnick. I would like to thank Nino Panagiafor invaluable discussions, and Gary Ferland for the use of his code CLOUDY. References Avedisova, V. S., 1974, SvA, 18, 283 Bertshinger, E., 1986, ApJ , 304, 154 Bregman, J. N. and Pildis, R. A., 1992, ApJ , 398, L107 Chevalier, R. A., 1974, ApJ , 188, 501 Chevalier, R. A., 1982, ApJ ,259, 302 Chevalier, R. A., and Imamura, J. N., 1982, ApJ , 2C1, 543 Cioffi, D. F., McKee, C. F., and Bertshinger, E., 1988, ApJ , 334, 252 Collura, A., Reale, F., Schulman, E. and Bregman, J.N., 1994, ApJ , 420, 163 Chugai, N.N., 1992, SvA, 36, 63 Chugai, N.N. and Danziger, I.J., 1993, MNRAS, (in press) Draine, B.T. and Woods, D.T., 1991, ApJ , 383, 621 Falle, S. A. E. G., 1975, MNRAS, 172, 55 Falle, S. A. E. G., 1981, MNRAS, 195, 1011 Ferland, G., 1990, OSU Astronomy Dep. Internal Report (90-02) Filippenko, A., 1989, AJ, 97, 726 ' Imamura, J. N., 1985, ApJ , 296, 128 Kronberg, P.P., Biermann, P. and Schwab, F.R., 1985, ApJ , 291, 693 McCray, R, Stein, R. F., and Kafatos, M., 1975, ApJ , 196, 565 Rupen, M.P., van Gorkom, J.H., Knapp, G.R. and Gunn, J.E., 1987, AJ, 94, 61 Shull, J. M., 1980, ApJ , 237, 769 Sramek, R.A., Weiler, K.W. and Panagia, N., 1990, IAU Circ No. 5112 Stathakis, R.A. and Sadler, E.M., 1991, MNRAS, 250, 786 Tenorio-Tagle, G., Terlevich, R., Franco, J., and Rozyczka, M., 1994, (in preparation) Tenorio-Tagle, G., Bodenheimer, P., Franco, J., and Rozyczka, M., 1990, MNRAS, 244, 563 Terlevich, R., Tenorio-Tagle, G. Franco, J. and Melnick, J., 1992, MNRAS, 255, 713 Turatto, M., Cappellaro, E., Danziger, I.J., Benetti, S., Gouiffes, C. and Tarenghi, M., 1993, MNRAS, 262, 128 Weiler, K.W., Panagia, N., Sramek, R.A., van der Hulst, J.M., Roberts, M. and Nguyen, L., 1989, ApJ , 336, 421 Weiler, K.W., Panagia, N. and Sramek, R.A., 1990, ApJ , 3C4, 611 Wheeler, J. C , Mazurek, T. J., and Sivaramakrishnaii, A., 1980, ApJ , 237, 781

The evolution of compact supernova remnants Guillermo Tenorio-Tagle Instituto de Astrofisica de Canarias, 38200 La Laguna, Tenerife, Spain.

1 Introduction This is a short summary of several calculations of the evolution of supernova remnants in a constant high density medium no > 106~8 cm"3 and an abundance in the range O.O1Z0< Z < IOZQ. The main difference found when comparing them with the standard calculation of a supernova evolving into a constant density medium HQ — 1 cm"3 is that radiative cooling becomes important very early in the life of the remnants. The radiative phase starts well before the ejecta is fully thermalized and while the expansion velocities are still in the range of several thousands of kms" 1 . Consequently, the remnants miss their Sedov evolutionary phase and, unlike the standard case, in these calculations full thermalization of the ejecta is only completed long after the moment of thin shell formation (see Terlevich et al. 1992, 1994a; hereafter referred to as papers I and II). The cooling event leads to large luminosities (> 109 LQ) in spans of time of only a few years, causing a major rapid depletion of the supernova's stored thermal energy, in only a few weeks or months. Strong radiative cooling leads to an ionizing spectrum (see paper I) and thus to an HII region with multiple components, as it photoionizes the recombining, rapidly-moving swept-up gas and the outer unperturbed matter. The ionizing radiation is also absorbed by the still unshocked and dense expanding ejecta. Such remnants, hereafter termed "compact SNRs", are capable of producing the strong ionizing flux that makes them appear as Seyfert I impostors (Filippenko 1989) when occurring in dense regions far away from the nucleus of galaxies. Also, their rapid hydrodynamical evolution leads to the dimensions and luminosities as well as to the expansion velocities detected in radio supernovae (Weiler et al. 1989). Their enhanced radiation is primarily promoted by the high background densities, which may have been established during a slow and massive wind prior to the supernova explosion and contained as circumstellar matter by the high pressure of the surrounding gas. Here we describe the basic time evolution of compact SNRs looking also at particular moments that, when thoroughly analysed, lead to an alternative sound explanation to the phenomena observed in active galactic nuclei. 166

G. Tenorio-Tagle: The evolution of compact supernova remnants

167

2 Numerical Calculations and Results 2.1 General Description Several two dimensional calculations of the evolution of supernova remnants in a constant high density medium (n > 106 cm"3) have been performed with the help of the hydro code described by Rozyczka (1985). Radiative losses, allowing in every run for a different Z metal abundance, were calculated from the basic interstellar cooling law of Raymond et al. (1976). The interstellar cooling law for the range of metal abundances considered here {\Q~2ZQ< Z < 10Z@) presents an increasingly larger cooling rate in the range 6 X 107 K > T > 3 x 105 K, as well as a shift of the minimum value of the function towards higher temperatures, for larger values of Z. The minimum value of the cooling function marks a definite change in the slope of the cooling curve implying that gas cooling from very high temperatures (say T > 4 x 107 K) enters, sooner or later, a cooling regime that causes a stronger cooling the further the gas cools down. This, together with the shift towards higher temperatures as a function of Z, marks the onset of strong radiative cooling in compact SNRs. 2.2 Boundary and Initial Conditions In all runs, the supernova ejected matter is at time t = 0 inserted in a. small portion of the computational grid. The procedure assumes an homologous growth and thus, the ejecta has a density proportional to R~fecta and velocity distribution proportional to Rejecta, as prescribed by the models of Arnett (1988). The insertion procedure ensures that 1051 erg are deposited in the form of kinetic energy with the largest speeds of about lC'kms" 1 . Given the homologous growth and consequent temperature drop of the ejecta, at t = 0 the amount of thermal energy in the grid is of the order of a few times 1048 erg. 2.3 The Time Evolution of Compact SNRs Figure 1 shows the run of luminosity versus time where a density no = 107cm~3, and an abundance Z = 10^0were assumed. The total luminosity of the remnant increases steadily at first, as t 0 8 , until strong radiative cooling becomes important leading to a total maximum luminosity of the order of a few times 1O9L0. The remnant then rapidly fades (L proportional to t~ n / 7 ) while exhausting its remaining energy. The main features, or main and secondary maxima, in the luminosity-time diagram (see Figure 1) correspond to major structural changes experienced by the compact

168


remnant as a function of time. The absolute maximum corresponds to rapid cooling behind the leading blast wave. This maximum occurs as soon as the temperature of the shocked gas decays to values T < 5 x 107 K. Before then, radiative losses from gas cooling from even higher temperatures, have only led to a steady growth of luminosity because of the larger size acquired by the remnant as a function of time. However, as soon as the temperature drops below 5 x 107 K the interstellar cooling function changes slope, leading to a larger cooling rate the further matter cools down, and this is sufficient in the high density regime to cause the run-away cooling described in detail in papers I and II. This rapid cooling causes a major loss of pressure behind the outer shock leading to its deceleration, while promoting the condensation of the swept up gas (see paper II). The second maxima occurs almost immediately afterwards. It is produced during the process of thin shell formation, when the cool swept up matter is collected together by the passage of several secondary shocks into a thin, cool and dense outer shell. Following this, conservation of momentum in the shell restores the leading shock velocity to values comparable to the ones it had before strong radiative cooling took place. Strong cooling behind the reverse shock produces the third apparent maximum in the total luminosity curve. This promotes the collapse of the swept up ejecta into a secondary thin and cool inner shell. The consequent loss of pressure behind the reverse shock leads to its withdrawal, and while favouring the approach and merging of the two shells, it delays the complete thermalization of the ejecta (see paper I). The last feature in the total luminosity curve produced mainly by an enhanced emission from gas cooling from T < 2 x 105K, is caused by the collision and merging of the two remnant shells.

2.4 Comparison with

Observations

A detailed comparison of the model results with the observations of AGNs can only be made with sources monitored for a. long period of time. The best example of these is perhaps NGC 1566 which has been followed for 15 years by Alloin et al. (1986). In that time four major periods of activity were detected. Each of these lasted about 1500 days while releasing 1051 erg. We associate these four events with four supernova explosions and their rapidly cooling compact remnants. The last one of these energy outbursts was observed with a higher temporal resolution to show a series of three (maybe four) rapid bursts (see Alloin et al. 1986, their Figure 4). Each of these bursts presents a steep rise time of about 20 days and a much longer


169

decay of about 400 days. Similar features are present in the results displayed in Figure 1. Alloin et al. (1986) noted the spectral resemblance that the NGC 1566 energy bursts from January and November 1982 have with Type II supernovae, and in particular with SN 1970g. All rapid energy bursts within a major period of activity could however be better matched if interpreted as produced by a rapidly evolving compact SNR. In this way, the FWHM of the emitting lines will reflect the remnant expansion speed at the time that strong radiative cooling sets in, which is of the order of a few thousands of k m s " 1 , in agreement with the 2000km s" 1 observed in NGC 1566. Also as shown in the previous section, the maximum intensity of consecutive bursts decays steadily but rather slowly as a function of time. Factors of 4 - 10 between first and last luminosity maxima are indicated in the various runs. This is in good agreement with NGC 1566 and thus t.lie decay should not be interpreted in terms of the light curve of a. supernova, which as noted by Alloin et al. (1986), fades orders of magnitude earlier than the luminous events expected during the time evolution of compact SNRs.

3 T h e n a t u r e of t h e lag A further detailed analysis of the events promoted by strong radiative cooling had led to an important result concerning the applicability of the starburst model to the realm of AGNs. The lag, the observed delay between abrupt changes in the continuum ionizing radiation followed after some time by changes in the intensity of the emission lines from the broad line region (BLR) of AGNs have been accurately matched by the cooling events during maximum luminosity of compact SNRs (see paper II). The lag, usually interpreted as a result of the geometry of the accretion disk sources, implying a measure of the distance at which the BLR sits away from the ionizing source, and that has also been used to deduce the mass of the putative central black hole (see e.g. Netzer 1993), has a different explanation in the starburst model of AGNs. Here, the lag results from the time-dependent changes in the ionization parameter within the layer of gas swept by the supernova blast wave, as matter adjusts itself to the drastic drop in pressure suddenly promoted by strong radiative cooling, and consequently in this model the lag has nothing to do with the size of the emitting source! Detailed calculations of radiative shocks evolving in a high density medium performed with a greater time resolution have shown in detail the sequence of events that take place as remnants approach and reach maximum luminosity. The matter involved in the process suddenly has to readjust to the

170


0.1

TIME ( y r )

10

Fig. 1. The luminosity output, of CSNRs. The integrated luminosity produced by CSNRs throughout their evolution, versus time.

large pressure imbalance promoted within the shocked gas by the onset of strong radiative cooling. The final outcome is the formation of a. thin shell at the edge of the remnant, several orders of magnitude denser than the original background medium. Gas condensation however, does not happen immediately, as it requires of the passage of secondary shocks through the cool region for this to acquire the appropriate density (and thus pressure). The secondary shocks emanate from the hottest section of the remnant interior, overtaken earlier during the evolution when the shock speed was larger, and that had taken longer to cool. The shocks follow the blast wave in an attempt to communicate the interior pressure, however, given the increasingly larger densities behind them and thus the correspondingly shorter cooling distances, inevitably become also rapidly radiative. Meanwhile cooling proceeds and moves as a wave, ahead of the secondary shocks, into gas more recently overtaken by the progressively slower blast wave to eventually catch up with it. The blast wave then slows clown for two reasons: because of ge-

G. Tenorio-Tagle: The evolution of compact supernova remnants ometrical dilution as the remnant grows and because it has now suddenly lost its piston pressure due to strong cooling behind it. The gas steadily overtaken by the cooling front is the source of ionizing continuum radiation, to be observed as a variation in the continuum of the AGN. This radiation is immediately absorbed by the reshocked matter. By the cool layer of gas continuously changing density after the passage of secondary shocks. The combination of a steadily denser layer constantly irradiated, as the cooling wave progresses through the layer of shocked gas, leads to a continuous decrease in the effective ionization parameter and results into a rapidly changing ionization structure of the fast-moving photoionized gas. The numerical calculations show that the width of the photo-ionized shocked region, traversed by the cooling front and continuously swept by secondary shocks to condense it into a cool thin shell, is only about 1013cm (with a light travel time of 103 sec) and yet, lags of up to several days, weeks, and even months, are generated for different lines. In general, the calculations predict shorter delays for high ionization lines than for low ionization ones. A detailed comparison with the results from the NGC5548 extensive monitoring campaign, agree both with the time delays for different lines, and with the intensity values reach by the various lines (Paper II). These results have also been independently corroborated with a different computational scheme (see Plewa et al. 1994). The calculations thus show that the compact supernova remnant model is capable of giving an accurate and detailed description of the temporal behaviour of the BLR, as well as accounting for all of its intrinsic properties with a, minimum of free parameters. 4 Rapid X-Ray Variability in Compact SNRs The role played by inhomogeneities in the supernova ejecta on the generation of rapid X-ray variability has also been recently explored (see Terlevich et al. 1994b). The interaction of supernova, fragments with the internal structure of a compact supernova, remnant has been followed with analytical approximations and 2-D hydrodynamical simulations. Here we focus on those interactions expected to lead to the largest energy bursts and the shortest durations. The model calculations show the evolution of fragments into thin, dense and cold "tortillas" as they encounter the reverse shock. These tortillas eventually cross the hot cavity and collide at large speeds with the remnant outer shell, causing luminous and short lived bursts that radiate most of their energy at X-ray frequencies. The range of predictions of the models has been compared with the observations, focusing particularly

171

172


on time scales, luminosities and light curve properties of rapid X-ray variable AGN. The models account for the complete observed range, including the highest luminosities and the shortest X-ray flare time scales, as well as for the power spectrum observed in AGN. In the past few years evidence for the existence of dense and fast moving fragments of SN ejecta has grown substantially. Metal rich high velocity knots moving at thousands of kms" 1 , have been found in Cas A (Braun et al. 1987 and references therein), Tycho (Seward et al. 1983), Puppis A (Winkler et al. 1988), Kepler (Bandiera k van den Bergh 1991) and in a number of extragalactic remnants (Lasker & Galinowski 1991, Hanuschik et al. 1993, Fesen & Matonick 1993). Theoretical work, in particular for the case of SN1987A, strongly suggests that Rayleigh-Taylor instabilities are the most likely origin of the condensations in the ejecta (Arnett et al. 1989). The effects of a fragmented ejecta in the evolution of "normal" remnants, i.e. those evolving in a low density interstellar medium, have been investigated for a wide variety of conditions (see review by Franco et al. 1991). In particular, Tenorio-Tagle et al. (1991) and Franco et al. (1993) have studied the interaction of SN fragments with the internal structure of single and multiple SNRs. In small remnants, high density fragments move almost unimpeded through the hot cavity to be rapidly thermalized as they impact the outer remnant shell. The multitude of fragment-shell interactions result into well localized strong shocks that partially disrupt the shell while causing a larger X-ray remnant radius. Fragments form soon after the explosion, via Rayleigh-Taylor instabilities, deep in the inner parts of the ejecta, at the interfaces between layers of different chemical compositions (Arnett et al. 1989). The resultant fragment properties, thus depend on details of the structure of the ejecta (see Franco et al. 1993). For simplicity we have assumed that the fragments (f) _

n/(r)

have a density contrast e = . L\ > 1 with respect to the interfragment medium (IFM) which is constant with r, and a velocity characteristic of their location. Fragments of different sizes, and at different locations have been inserted within the ejecta for a number of calculations. In all cases the flow has been assumed to undergo an homologous diverging expansion and thus the value of e remains constant for each fragment and its neighbouring gas, until the fragments meet the reverse shock. The resolution of the numerical grid limits the minimum initial dimensions of a properly resolved fragment, although the homologous expansion soon leads to large fragment sizes. Note that the evolutionary features and the physics of the interaction are independent of the fragment size, and thus the calculations accurately


173

illustrate the time evolution. Figure 2 shows isodensity contours and the velocity field at selected evolutionary times for one of the calculated cases. As described in Section 2, the remnant caused by the violent release of the ejecta presents two well-defined shocks: the leading blast wave and the reverse, inward-facing, shock. At first, before t = 4yr, strong cooling is in full operation, but it is slow and the flow remains isobaric. During this time the density of the cooling gas changes only slightly. Soon afterwords, at t = 4 yr, a well defined thin, very dense cool shell is present right behind the blast wave. This also happens at t = 6.3yr behind the reverse shock where matter condenses also into a secondary inner shell. Note that at that time the ejecta is not yet fully thermalized and it is still impacting the disrupted shell with large velocities. At that time however, the energy left over amounts to about 1/10 of the initial value, and the luminosity of the remnant has fallen by an order of magnitude from the maximum emission (see Figure 1). The bulk of the thermal energy stored in the remnant was in fact rapidly radiated away before t = 3-5yr. From then onwards, a continuous and immediate release of energy occurs as soon as the remaining kinetic energy is thermalized. Figure 2 also shows the initial conditions, the homologous expansion, as well as the initial fragment interaction with the reverse shock. These are followed by frames showing the motion of the fragment through the hot cavity, as it is overtaken by a weaker reverse shock to evolve into a condensed "tortilla" with its smallest dimension along the flow direction, while catastrophic cooling also causes the formation of a thin dense shell at the edge of the remnant. Note that the lateral walls of the fragment are also condensed by a strong shock that develops as soon as the fragment sides enter into contact with the hot cavity. The final frames display the tortilla-outer shell interaction.

4-1 Analytical Description

of the X-Ray

Bursts.

All ejected fragments eventually meet the reverse shock which tends to decelerate them while thermalizing their kinetic energy. The shock strength is inversely proportional to the square root of the incoming gas density and thus, high density contrast fragments (e > 10) are compressed by weaker sections of the reverse shock. Given the high densities, and the correspondingly smaller cooling times these weaker shocks are regarded as isothermal, leading to large compression factors. Meanwhile, the shocked fragments acquire a tortilla shape with its smallest dimension in the radial direction. These cold tortillas move with high velocities through the hot cavity generated by the reverse and outer shocks. The tortillas eventually reach the edge

174


'•».. = I.IS x 10* cir

fragment

' =• 1.44 X 10' 1 " - . . = 8.C5 x 10* cm I"'

I = l.2< X 10* 1 "n«» = 3.94 x 10' cm •"

I = J.12 x 10' >

1 =r 1.02 X 10* •

« - . , = 6.00 x 10* cm I"1

t ' ^ , o 3.80 x 10* cm s"

' « 6 67 x 10' • ••-.. - 4.65 x 10« cm i - 1

Fig. 2. The evolution of cSNRs. Two dimensional representation of the density distribution and velocityfieldwith scales indicated at various times. The contours are logarithmically spaced with 6\ogn = 0.33. The crowding of contours indicates shocks and/or density discontinuities while sudden velocity changes also help to identify the location of the various shocks developed throughout the evolution.


175

of the remnant and either catch up with the leading shock or the thin shell of compressed swept up matter depending on whether or not catastrophic cooling has already taken place behind the leading shock. The interaction of the fast moving tortillas with the outer structure generate bursts of X-ray emission promoting the rapid and strong X- ray variability of the source. The thin shell bombardment phase can last several years during which the remnant is highly variable in X-rays on time scales down to one hundred seconds. The parameters of the tortillas (denoted by superscript 't1) can be estimated from Papers I and II. In the cSNR model of Paper I, at t = tsg = 230 days, the leading shock velocity is V3hock = 5000 km s" 1 and its radius is Rshock — 3.0 X 1016cm, while the outer thin shell has a density nshell ~ 1012 cm~ 3 and thickness ARsheii ~ 1011 cm. At the same time the density of the IFM just in front of the reverse shock is very similar to that of the circumstellar medium, i.e. ~ 10 7 cm~ 3 . A fragment with e = 100 has a density of nj = 10 9 cm~ 3 just before it meets the reverse shock. The reverse shock effective velocity in the IFM is about 5000kms - 1 , and thus the shock velocity into the fragment is 909km s ^ a n d the cooling time is about 2.2 x 10'1 sec. A fragment with initial size / = 10 14 cm, has a total shock crossing time of 1.1 x 106 sec and therefore is completely shocked before reaching the thin outer shell. As the cooling time is much shorter than the crossing time, the shock is isothermal, and the compression factor equals the square of the Mach number M. For the parameters considered here, M2 — (909/10) 2 = 8100 (assuming that the sound speed is 10 km s~afor the fragment gas in front of the shock). The final density of the fragment is about nj = 8 x 1012 cm" 3 and as the compression is only in the radial direction the final result is a "tortilla" 10 14 cm in diameter and with thickness A/' = 1.2 X 10 10 cm. The tortilla leaves the reverse shock with a space velocity of F* = 9900 km s" 1 , corresponding to a relative speed (with respect to the shocked IFM) of vt~IFM = 4900 k m s " 1 . After about t ~ 1.0 x 10 1 6 /4.9 x 108 = 2 x 107sec, the tortilla reaches the edge of the remnant. There, due to the deceleration of the outer shell, the relative velocity between the tortilla and the shell is about 7000 km s - 1 a t the time of the collision. The densities of the thin shell and the tortilla are very similar and around n* ~ nsheii — 3 x 1012 cm" 3 . The collision leads to two similar shocks, one into the tortilla the other into the shell. Given the similar densities the velocity of both shocks is about 3500 km s" 1 . For the velocities and densities involved in this interaction the cooling time is, rcoo\ ~ vg / n.\3 ~ 1.8 minutes while the shock crossing time is Tlxi = .,?'' ~ V2*-1,1?^ ~ 0.5 minutes. shock

This implies that the kinetic energy of the tortilla is thermalized and emit-

176


ted in about a cooling time, i.e. ~ 2 minutes. The mass of the tortilla is Ml = (/') 2 A/*n* m; ~ 2 x 10 27 gm ~ 1O~ 6 M 0 , and its kinetic energy, E

k = \Mt

(y1-3^")

~ 5 x 1044 erg. The maximum luminosity reached

by a flare of duration Atfiare

is Lx — h-AT^— — 2 x l O ^ e r g s " 1 with

&tflare = \ Tcool — 1 min, and typical energy about 15keV. In general, larger tortillas produce more energetic flares and last longer. The above estimates of duration and peak luminosity corresponds to the "face-on" case when the observer sees the whole event simultaneously, case that corresponds to an interaction at the central part of the remnant shell as viewed by the observer. Interactions in other positions produce a light-travel time delay between different points in the tortilla. In the case of the /' = 1014 cm tortilla the maximum delay corresponding to an edge-on view is 3330 sec.

5 Concluding remarks In the starburst model of AGNs, sometimes viewed as exotic and/or unconventional, the applied physics are in fact most, conventional, as it uses the little, or the lot, that we know about real events: the physics of stars and stellar evolution and their interaction with the surrounding gas, and with these sound predictions are made. Under the assumption of a normal IMF, the supernova rate detected in NGC 1566 (about 1 every 4.5 years) should result from a coeval major burst of stellar formation, leading to a nuclear cluster with a mass of 5 x 108 MQ. Every supernova explosion from the evolving cluster that occurs in a high density medium (highly likely produced by slow winds from the progenitors) will lead to a compact SNR. The photoionizing radiation arising from the cooling shocks of each remnant will immediately be absorbed by the rapidly expanding recombined gas, producing as shown in papers I and II the broad emission lines with the intensities, relative ratios and lags with respect to the ionizing contunuum, in excellent agreement with the BLR of active galactic nuclei. The supernova rate will lead to detectable bursts of energy each releasing, a total of about 1051 erg in only a few years (< lOyr). By then the size of the remnants will be of the order of a few times 1016 cm. Each remnant will lead to several distinct energy bursts which have here been identified with: 1) the onset of strong radiative cooling behind the blast wave; 2) the completion of thin shell formation; 3) The onset of strong cooling behind the reverse shock; and 4) the collision and merging of both shells. The energy burst caused by these events agrees well with detailed observations of the


177

long-term variability in AGNs. The remnants will effectively end up their evolution as rather fragmented shells in a time longer than the SN rate of the nuclear star cluster, thus providing at all times a minimum intensity value of the emitting region. During the evolution of each compact SNR (< 10 years), more than 90% of the initial energy of the explosion (1051erg) is radiated away, while the remnants acquire dimensions of only a few times 1016cm. The high circumstellar densities thus provoke the rapid onset of strong radiative cooling and with it the various hydrodynamical events that lead to the broad line region, to the continuum and line variability and also to the lag typical of these sources. The original problem of processing 1051 erg into the medium surrounding an exploding star acquires a. number of new facets when the early fragmentation of the ejected matter is taken into consideration. This has been shown here to be directly related to the rapid X-ray variability of AGN if the remnants evolve in a high-density circumstellar medium, as postulated in the Starburst model of AGN. Under such conditions, strong radiative cooling rapidly begins to drain the thermal energy of the remnant even before the ejecta is fully thermalized. Therefore, most of the ejected fragments, particularly denser ones like the oxygen-rich, fast-moving (5000 km s"1) knots of Cas A, experience a complex evolution to end up colliding at large relative speeds with the remnant outer shell. This is one of several possibilities capable of generating rapid X-ray variability in r.SNRs. Perhaps the one that leads to the most energetic and shortest duration energy bursts, reason that motivated this first exploration.

Acknowledgements. GT-T acknowledges support from the EEC grant for international collaboration No CI1*-CT91-O935. References Alloin, D., Pelat, D., Phillips, M.M., Fosbury, R.A.E. k Freeman, K., 1986, Ap.I, 308, 23. Arnett, W. D., 1988, ApJ, 331, 377. Arnett, W. D., Fryxell, B. & Miiller, E., 1989, ApJL, 341. L63. Braun, R., Gull, S. F. & Perley, R., 1987, Nature, 327, 395. Bandiera, R. fc van den Bergh, S., 1991, ApJ, 374, 186. Fesen, R. A. fc Matonick, D. A., 1993, ApJ, 407, 110. Filippenko, A., 1989, AJ, 97, 726. Franco, J., Tenorio-Tagle, G. Bodenheimer, P. & Rozyczka, M., 1991, PASP, 103, 803. Franco, J., Ferrara, A., Rozyczka, M., Tenorio-Tagle, G. fc Cox, D., 1993, ApJ, 407, 100. Hamilton, A. J. S., 1985, ApJ, 291, 523.

178


Hanuschik, R., Sypromilio, J., Stathakis, R., Kimeswenger, S., Gochermann, J., Seidensticker, K.J. k Meurer, G. 1993, MNRAS, 261, 909. Lasker, B. M. k Golinowski, D. A., 1991, ApJ, 371,563. Netzer, H. 1993 in The Nearest Active Galaxies eds J. Beckman, L. Colina k H. Netzer. Madrid. Coleccion Nuevas Tendencias Vol 22, p. 219. Plewa, T., 1994 submitted to Violent Star Formation, from SODoradus to QSOs, ed G. Tenorio-Tagle, Cambridge University Press. Raymond, J., Cox, D. P. k Smith, B. W., 1976, ApJ, 204, 290. Rozyczka, M., Franco, J., Miller, W., Tenorio-Tagle, G. &: Terlevich, R., 1994, (in preparation). Rozyczka, M., 1985, A&A, 163, 59. Seward, F., Gorenstein, P. k Tucker, W., 1983, ApJ, 266, 287. Stathakis, R. A. k Sadler E. M., 1991, MNRAS, 250, 786. Tenorio-Tagle, G., Rozyczka, M., Franco, J. k Bodenheimer, P., 1991, MNRAS, 251, 318. Terlevich, R., Tenorio-Tagle, G., Franco, J. k Melnick, J., 1992, MNRAS, 255, 713 (Paper I) Terlevich, R., Tenorio-Tagle, G., Rozyczka, M., Franco, J. k Melnick, J., 1994a, MNRAS, (in press) (Paper II). Terlevich, R., Tenorio-Tagle, G., Cid-Fernandes, R. Franco, J. k Rozyczka, M., 1994b, MNRAS, submitted (Paper III). Winkler, P. F., Tuttle, J. H., Kirshner, R. P. k Irwin, M. J., 1988, Supernova Remnants and the Interstellar Medium (IAU Coll. 101), ed. R. S. Roger and T. L. Landecker (Cambridge University Press: Cambridge), p. 65. Weiler, K. W., Panagia, N., Sramek, R. A., Van der Hulst, J. M. Roberts, M. S. k Nguyen, L., 1989, ApJ, 336, 421.

Massive Supernovae in Binary Systems P. C. Joss 1 , J. J. L. Hsu 2 , Ph. Podsiadlowski3, and R. R. Ross4 1

2 3 4

Department of Physics, Center for Space Research, and Center for Theoretical Physics, Massachusetts Institute of Technology, Cambridge, MA 02139, U.S.A. Department of Astronomy, University of California, Berkeley, CA 94720, U.S.A. Institute of Astronomy, Madingley Road, Cambridge CB3 OHA, U.K. Department of Physics, College of the Holy Cross, Worcester, MA 01610, U.S.A.

Abstract The presence of a close binary companion can affect the evolution of a massive star through one or more episodes of mass transfer, or by merger in a common-envelope phase. Monte Carlo calculations indicate that ~ 20 - 35% of all massive supernovae are affected by processes of this type. The duplicity of the progenitor may be revealed by the illumination, in the supernova event, of axially symmetric material that had previously been ejected during the mass-transfer phase or by the expulsion of a common envelope. Moreover, the properties of the progenitor star, the peak supernova luminosity, and other observable features of the supernova event can be affected by prior binary membership. Binary interactions may be the cause of much of the variability among Type II supernova light curves, and may result, in Type Ib or Ic events in cases where the entire hydrogen-rich envelope has been stripped from the progenitor. Many of the peculiarities of SN 1987A and SN 1993J may well have resulted from the prior duplicity of the progenitor.

1 Introduction A large fraction of all stars are members of binary systems. It is therefore reasonable to consider the possibility that the properties of many massive supernovae (i.e., supernovae whose progenitors had initial main-sequence masses, Mms, greater than ~ 8 MQ) are influenced by prior interactions of the progenitor with a binary companion star. This possibility was brought into focus in recent years by the nearby Type II supernovae SN 1987A and SN 1993J, many of whose properties differed markedly from theoretical expectations. As a result, several studies have been undertaken to estimate the frequency of massive supernovae in binaries and the unique properties of the progenitors and the resultant supernova events that result from the evolution of a massive progenitor in a binary system (Podsiadlowski, Joss & Hsu 1992; Tutukov, Yungelson & Iben 1992; Hsu et al. 1993). Of particular interest, in the context of these Proceedings, is the possibility of mass ejection from the presupernova binary in an axially symmetric pattern, with 179

180

P. C. Joss et al.: Massive supernovae in binary systems

the axis of symmetry coinciding with the orbital axis, and the subsequent illumination of the ejected matter by the supernova event. We here describe the main results of recent theoretical work on massive supernovae in binaries and briefly discuss the application of this work to SN 1987A and SN 1993J.

2 Evolution of Massive Supernova Progenitors in Binaries The principal effects of evolution in a close binary system on the progenitor of a massive supernova can be broadly divided into three categories (Podsiadlowski, Joss & Hsu 1992; hereafter PJH): (1) loss of part or all of the stellar envelope to the companion star, (2) accretion of matter from the companion star, or (3) merger of the two stars in a common-envelope phase. (In addition, a star in a close but detached binary may lose a large fraction of its envelope in an enhanced stellar wind [Vanbeveren 1987; Tout & Eggleton 1988] whose time-averaged morphology will display axial symmetry.) In cases (1) and (2), it is highly likely that the mass-transfer process will entail the loss of a. significant amount of matter from the system, while in case (3) the common envelope itself may well be ejected prior to the supernova event. On the basis of Monte Carlo calculations, PJH concluded that ~ 20-35% of all massive stars experience binary interactions of one of the above types before undergoing a supernova explosion. This is consistent with the findings of Tutukov et al. (1992), who concluded, by somewhat different means, that ~ 25-45% of all supernovae (including those involving low-mass progenitors) originate in initially close binaries. In the following paragraphs, we describe the salient features of each of the three modes of presupernova binary evolution described above.

2.1 Mass-Loss

Models

If the supernova, progenitor was originally the more massive of the binary components, it can lose mass to its companion via Roche-lobe overflow. This scenario has been considered in detail by Joss et al. (1988) and PJH. If the star first fills its Roche lobe while it is still on the main sequence, a contact system and eventual merger of the binary components is likely to result; the merged star should then have the properties of a rejuvenated main-sequence star. (Any mass that is lost from the system as a byproduct of this process is likely to dissipate before the merged star is able to reach the supernova stage.) Of greater interest, in the present context, is the possibility that the primary first fills its Roche lobe during the course of its post-main-sequence


181

evolution. If the masses of the binary components are not too different, the resultant mass transfer can take place on a sufficiently long time scale that a common envelope does not form. In cases where the entire hydrogen-rich envelope is lost, the progenitor will become a helium star and will likely end its life as a Type Ib or Ic supernova. If, however, the mass-transfer process terminates when the progenitor still retains at least a few tenths of a solar mass of its envelope, its final presupernova radius and effective temperature will be nearly the same as those it would have had in the absence of mass loss. Monte Carlo calculations (PJH) indicate that a few percent of all massive stars (perhaps up to ~ 5% if systems with binary-enhanced winds are included) become supernovae of this latter type.

2.2 Accretion

Models

The original secondary in a close binary system with a Roche-lobe filling primary should accrete a substantial fraction of the mass lost by the primary. If the mass transfer commences before the original secondary has completed core hydrogen burning, the subsequent evolution of the secondary should mimic that of a more massive main-sequence star (Hellings 1983; PJH). Analogously to the case of mass-loss models, of greater interest here is the situation where mass transfer commences only after the original secondary has left the main sequence (Podsiadlowski & Joss 1989; de Loore & Vanbeveren 1992; PJH). Due to the accreted mass, the original secondary will generally become the more massive of the two stars, and its concomitantly accelerated evolution may cause it to reach the supernova stage prior to the original primary. If the orignal secondary is the first star to become a supernova, it will have a normal post-main-sequence companion at the time of the explosion; if, instead, the original primary readies the supernova, stage first, it should leave a neutron-star or black-hole remnant that will remain gravitationally bound to the original secondary until it, too, becomes a supernova. In either case, however, the explosion of the original secondary will eject more than half of the residual mass of the system, generally causing it to become unbound. Nevertheless, the companion object may become detectable after the supernova photosphere has receded sufficiently. Another diagnostic of supernova events of this type is the color of the immediate supernova progenitor; if the original secondary accretes a sufficient amount of mass, it will end its life as a blue supergiant rather than a red supergiant, which is the generally expected precursor for a Type II supernova that has evolved in isolation (Falk & Arnett 1977; Woosley k. Weaver 1985).

182

P. C. Joss et al,: Massive supernovae in binary systems

2.3 Merger Models If the initial masses of the two stars are sufficiently different, the time scale for mass transfer, once it commences, will be much less than the Kelvin time of the secondary. As a result, the mass transfer will be unstable, and the system will develop a common envelope (Paczyriski 1976; Kippenhahn & Meyer-Hofmeister 1977; Podsiadlowski, Joss & Rappaport 1990; PJH); the primary should lose its entire hydrogen-rich envelope to the common envelope. Thereafter, dynamical friction between the secondary and the common envelope will cause it to spiral in toward the system center-of-mass. It is uncertain whether or when the common envelope will subsequently be ejected (see Hsu et al. 1993 for a discussion). If the envelope is ejected before either the secondary is dissolved or the binary components merge to form a single star, and if the core mass of the primary is greater than ~ 1.4 MQ at the time of the ejection, a Type Ib or Ic supernova may result; however, if the supernova explosion strips off a significant amount of the hydrogen-rich envelope of the secondary, the supernova event may be misclassified as Type II. In cases where mass transfer commences when the primary is still on the first red giant branch (case B transfer) and the common envelope is not subsequently ejected, the spiral-in time scale should be much shorter than the remaining evolutionary time for the primary; the binary components should therefore merge before a supernova, event occurs. If, instead, mass transfer does not commence until the primary has reached the asymptotic giant branch (case C transfer) and the common envelope is not ejected, it is uncertain whether merger will occur before the the primary becomes a supernova. When the binary components merge before the occurence of the supernova event, the net effect is very similar to that of the accretion scenario described in §2.2, and the merged star may well end its life as a blue supergiant. If the merger is not yet complete by the time of the supernova event, the immediate progenitor (i.e., the common envelope itself) may have the appearance of either a. red or a blue supergiant, depending on the values of various parameters for the initial binary system and the details of the common-envelope evolution. 3 Hydrodynamics of Massive Supernovae in Binaries We have recently completed a series of hydrodynamic calculations to explore the consequences of mass-loss and accretion/merger scenarios for the observational properties of the resultant supernova events (Hsu et al. 1993). We restricted our attention to cases where the progenitor retained at least a portion of its hydrogen-rich envelope, so the the event would be of Type


183

II; the hydrodynamics of events in which the entire hydrogen-rich envelope had been stripped from the progenitor, leading to Type Ib or Ic events, has previously been explored by Ensman & Woosley (1988) and Shigeyama et al. (1990). We here briefly summarize the results of some of our calculations. To investigate the effects of mass loss from a massive supernova progenitor, either through mass transfer to a close-binary stellar companion (see §2.1) or via a strong intrinsic stellar wind, we followed the explosion of a star with an initial main-sequence mass of 12 MQ. In the absence of mass loss, such a star would have a hydrogen-rich envelope of mass M env ~ 8.7 MQ; we considered cases with residual envelope masses of 0/1, 1.9, and 4.9 MQ, as well as a case with no mass loss. The visual light curves of our four mass-loss models are shown in the lefthand panel of Figure 1. The most dramatic effects of a reduced envelope mass are (1) a much more rapid rate of decline of the light curve, (2) a higher peak luminosity (by as much as two magnitudes in the V band), and (3) peak photospheric velocities that are higher by as much as a. factor of ~ 2 (~ 2 X 104 km s" 1 for the model with the smallest residual envelope, compared to ~ 1 X 104 km s" 1 for the model with no mass loss). In order to explore the effects of an increase in the envelope mass of the progenitor via accretion from or merger with a binary companion during the course of its post-main-sequence evolution, we calculated the explosion of three stars, each of which had a final presupernova mass of 20 MQ. The first star had an initial main-sequence mass of 20 MQ and underwent no mass loss or gain during the course of its evolution; the other two stars had M ms = 17 and 15 M 0 and gained 3 and 5 MQ, respectively, during their post-main-sequence evolution. Both of the models that gained mass were blue supergiants, rather than red ones, at the time of the supernova event (see §§2.2 and 2.3, Podsiadlowski & Joss 1989, and PJII). The visual light curves for these three models are shown in the right-hand panel of Figure 1. The principal effects of the addition of mass are (1) a peak luminosity that is fainter by as much as 3.5 magnitudes in the V band (if we exclude the initial flash near / = 0, which is not modeled very accurately in our calculations), (2) a more rapid decline of the light curve, and (3) a reduction of the peak photospheric velocity by as much as a factor of ~ 2.5, from ~ 1.6 X 104 km s" 1 for the constant-mass model to only ~ 6 X 103 km s" 1 for the model that has gained 5 MQ during the course of its evolution.

184

P. C. Joss et al.: Massive supernovae in binary systems -19

i

i

1

-18 -17

MV

1 ' 1 '

•

•

1t>

-

"v".

in '

\

\

1 \

-16

\

-15

\\ "l i i i i \

~

V, vi

-14 -13

' i ' i ' i ' i ' i '

a _

i 20

40

_l1 I

i

i

i

i.

60

80

~"

~\

•*

'

1

Ik

\ •••

. - • • "

\ --1

-

'••••I

",\ , . i . 1 , _[0VTv. 20 40 60 80 100 120 140 160 1 !,

t (days) Fig. 1. (a) Light curves (absolute visual magnitude, My, as a function of elapsed time, t, since core collapse) for four supernova models whose progenitors had initial main-sequence masses, A/ ms , of 12 M© but had lost their hydrogen-rich envelopes to varying degrees during the course of their post-main-sequence evolution. All four hydrodynamic calculations assumed an explosion energy, E, of 1 foe and no energy input from the decay of radioactive material. Solid curve, M env — 0.4 MQ. Long-dashed curve, M env = 1-9 MQ. Short-dashed curve, A/env = 4.9 MQ. Dotted curve, M e n v = 8.7MQ (corresponding to no mass loss), (b) Same as (a), but for three models whose progenitors underwent accretion from, or merger with, a binary stellar companion during the course of their post-main-sequence evolution; the final presupernova mass was 20 M© in all cases. All calculations again assumed E = 1 foe and no energy input from radioactive decay, except where otherwise noted. Solid curve, M m s = 15 MQ. Long-dashed curve, M m s = 17 MQ. Dotted curve, Mms — 17 MQ, with additional energy from the radioactive decay of 0.071 MQ of Ni 56 and its decay product, Co 56 , deposited in the innermost layers of the ejecta; this light curve comes closest to matching the general properties of the light curve of SN 1987A, although it does not fit the observed light curve in detail (see Hsu el al. 1993 for a discussion). Short-dashed curve, M m s = 20 MQ, with no mass gained or lost by the progenitor during the course of its presupernova evolution, shown for comparison.

4 Application to Recent Supernovae 4.1 SN 1987A A number of authors (Fabian & Rees 1988; Joss el. al. 1988; Barkat & Wheeler 1989; Hillebrandt & Meyer 1989; Podsiadlowski & Joss 1989; Podsiadlowski, Joss & Rappaport 1990; de Loore & Vanbeveren 1992; PJH; Rathnasree 1993) have explored the possibility that Sk -69°202 , the progenitor of SN 1987A, had been a member of a binary system prior to the


185

supernova event. Among the various binary scenarios that have been proposed, the most promising appears to be one in which Sk —69°202 underwent merger with a binary stellar companion in a common-envelope phase (Hillebrandt & Meyer 1989; Podsiadlowski, Joss & Rappaport 1990; PJH). (The plausibility of accretion scenarios has been somewhat diminished by the lack of evidence for either a prior supernova event or a normal or neutronstar companion following the recession of the photosphere of SN 1987A.) The major lines of evidence in support of a merger scenario include (1) the blue color of Sk — 69°202, which was in contrast to most prior theoretical expectations (see §§2.2 and 2.3, Podsiadlowski & Joss 1989, and PJH), (2) chemical peculiarities in the progenitor and in the supernova ejecta, which may result from the dredge-up of nuclear-processed material during the merger process (see Hillebrandt & Meyer 1989, PJH, and references therein), (3) the low peak luminosity of SN 1987A and the exceptionally strong effect of energy input from radioactive decay upon its light curve, which is in accord with the results of hydrodynamic calculations by Hsu et al. (1994) for accretion/merger models of Type II supernovae (see §3 and Fig. lb), and (4) the approximate axial symmetry of the circumstellar material (Wampler et al. 1990). In regard to this last point, it is intriguing to speculate that the ring structure around the supernova represents relatively dense, low-velocity material from the common envelope that was ejected in the orbital plane of the original binary, while the "Napoleon's Hat" nebulosity is material from a high-velocity stellar wind that was emitted by the hot progenitor star after the completion of the common-envelope phase and gained its axial symmetry by interaction with the pre-existing ring structure.

4.2 SN

1993J

The likelihood that the progenitor of SN 1993J lost, most of its hydrogen-rich envelope by transfer to a close binary companion, in the manner discussed in §2.1, has already been noted by a number of authors (Nomoto et. al. 1993; Podsiadlowski et al. 1993a; Ray, Singh k Sutaria 1993; Woosley et. al. 1994) and is discussed in some detail elsewhere in this volume (Podsiadlowski et al. 1993b). Here, we only observe that there may be an interesting evolutionary link between SN 1993 J and SN 1987A. If the companion of SN 1993 J accreted several solar masses of material during the mass-transfer process, at a, time when it had already evolved off the main sequence, it should end its life (~ 10 5 -10 6 years hence) as a blue supergiant, and this second supernova event should resemble SN 1987A. It is remarkable that both SN 1987A and SN 1993J, the two nearest known

186

P. C. Joss el ai: Massive supernovae in binary systems

supernovae of the past century, have both displayed substantial evidence for origin in massive close-binary systems. Of course, evidence of prior duplicity becomes easier to obtain with increasing proximity of the supernova event. It is distinctly possible that, for supernovae of all types, prior duplicity will turn out to be the rule rather than the exception. Acknowledgements This work was supported in part by the U.S. National Aeronautics and Space Administration under grant NAGW-1545. References Barkat, Z. k Wheeler, J. C. (1989). Astrophys. J., 342, 940. de Loore, C. k Vanbeveren, D. (1992). Astr. Astrophys., 2G0, 27.$. Ensman, L. M. k Woosley, S. E. (1988). Astrophys. J., 333, 754. Fabian, A. C. k Rees, M. J. (1988). Nature, 335, 50. Falk, S. W. k Arnett, W. D. (1977). Astrophys. .1. Suppl., 33. 515. Hellings, P. (1983). Astrophys. Space Sci., 9G, 37. HillebrancU, W. k Meyer, F. (1989). Astr. Astrophys.. 219, 1,3. Hsu, J. J. L., Joss, P. C , Ross, R. R. k Podsiadlowski, Ph. (1994). Astrophys. J., (submitted). Joss, P. C , Podsiadlowski, Ph., Hsu, J. J. L. k Rappaport, S. (1988). Nature, 331, 237. Kippenhahn, R. &; Meyer-Hofmeister, E. (1977). Astr. Astrophys., 54, 539. Nomoto, K., Suzuki, T., Shigeyama, T., Kumagai, S., Yamaoka, H. k Saio, H. (1993). Nature, 364, 507. Paczyriski, B. (1976). In 7.4 U Symposium 13, Structure and Evolution of Close Binary Systems, ed. P. P. Eggleton, S. Mitton, k J. Whelan, pp. 75 (Dordrecht: Reidel). Podsiadlowski, Ph., Hsu, J. J. L., Joss, P. C. k Ross, R. R. (1993a). Nature, 364, 509. Podsiadlowski, Ph., Hsu, J. J. L., Joss, P. C. k Ross, R. R. (1993b). This volume. Podsiadlowski, Ph. k Joss, P. C. (1989). Nature, 338, 401. Podsiadlowski, Ph., Joss, P. C. k Hsu, J. J. L. (1992). Astrophys. J., 391, 246 (PJH). Podsiadlowski, Ph., Joss, P. C. k Rappaport, S. (1990). Astr. Astrophys., 227, L9. Rathnasree, N. (1993). Astrophys. J., 411, 848. Ray, A., Singh, K. P., k Sutaria, F. K. (1993). J. Astrophys. Astr., 14, 53. Shigeyama, T., Nomoto, K., Tsujimoto, T. k Hashimoto, M. (1990). Astrophys. J. Lett., 361, L23. Tout, C. A. k Eggleton, P. P. (1988). Astrophys. J., 334, 357. Tutukov, A. V., Yungelson, L. R. k Iben, I. (1992). Astrophys. J., 38G, 197. Vanbeveren, D. (1987). Astr. Astrophys., 182, 207. Wampler, E, J., Wang, L., Baade, D., Banse, K., D'Odorico, S., Gouifles, C. k Tarenghi, M. (1990). Astrophys. J. Lett., 362, L13. Woosley, S. E., Eastman, R.. G., Weaver, T. A. k Pinto, P. A. (1994). Astrophys. J., (in press). Woosley, S. E. k Weaver, T. A. (1985). In Nucleosynthesis and Its Implications On Nuclear and Particle Physics, Proc. 5th Moriaud Astrophys. Cotif., ed. J. Audouze k T. van Thuan, pp. 145 (Dordrecht: Ilciclel).

The Progenitor of SN 1993J and its Mass-Loss History Ph. Podsiadlowski1, J. J. L. Hsu 2 , P. C. Joss 3 and R. R. Ross4 1 2 3 4

Institute of Astronomy, Cambridge CB3 OHA, UK University of California at Berkeley, CA 94720, USA Massachusetts Institute of Technology, Cambridge, MA 02139, USA College of the Holy Cross, Worcester, MA 01610, USA

Supernova 1993J in the spiral galaxy M81 is the brightest supernova since SN 1987A and, like the latter, appears to be another peculiar type II supernova. Its early light curve is characterized by a very sharp initial peak (lasting for less than ten days) followed by a less rapid secondary brightening, which was qualitatively similar to the secondary brightening observed in SN 1987A. Humphreys et al. (1993) have identified a candidate progenitor consistent with the position of the supernova. Combining their UBVR. photometry with the I magnitude obtained by Blakeslee & Tonry (1993), they concluded that the colors of the apparent progenitor require the presence of at least two bright stars. One star is an early-type supergiant (most likely a lateB to early-A supergiant), the other a late-type supergiant (most likely a G to early-K supergiant). The bolometric magnitudes of both stars are in the range of - 6 to - 8 , with best-fit values of - 7 to - 7 . 5 (for an assumed distance of 3.3 Mpc). We have performed our own fits to the photometric data and obtained similar results. These best-fit magnitudes imply mainsequence masses of ~ 15 MQ, but the masses could be as low as 8 MQ or as large as 20 MQ. The image of the candidate progenitor appears extended on some plates (Blakeslee & Tonry 1993). This suggests that, at the distance of M81, the two stars do not form a close binary (although either star could have an undetected binary companion). Neither of the two inferred stars is a theoretically expected progenitor for a. typical type II supernova. However, a G or early-K supergiant can account for the early light curve, provided that it had lost almost all of its hydrogen-rich envelope before the supernova explosion (i.e., provided that it was a "stripped" supergiant). To constrain the properties of the progenitor, we performed a series of hydrodynamical calculations (see Podsiadlowski et al. 1993; Hsu et al. 1994). We found that the visual light curve can be well fitted with the explosion of a progenitor star with a main-sequence mass of 15 MQ, which suffered severe mass loss and had a residual hydrogen-rich envelope mass of ~ 0.2 MQ, and 187

188

Ph. Podsiadlowski tt al: The progenitor of SN 1993J

with an explosion energy of ~ 1051 erg. The secondary peak of the light curve and the subsequent exponential decay can be understood as the result of ~ O.IJV/Q of 56Ni that was produced in the supernova, and confined to the inner region of the ejecta (the exact amount of 5

Circumstellar Media in Late Stages of Stellar Evolution (Proceedings of the 34th Herstmonceux Conference, Held in Cambridge, July 12-16, 1993)

The nature of compact objects in active galactic nuclei: proceedings of the 33rd Herstmonceux conference, held in Cambridge, July 6-22 i.e. 16-22 , 1992

Category Theory: Proceedings of the International Conference held in Como, Italy, July 22-28, 1990

Category Theory: Proceedings of the International Conference Held in Como, Italy, July 22-28, 1990

Category Theory: Proceedings of the International Conference held in Como, Italy, July 22-28, 1990 (Lecture Notes in Mathematics)

Logic Colloquium '87: Proceedings of the Colloquium Held in Granada, Spain July 20-25, 1987

Stellar Structure and Evolution

Stars and Stellar Evolution

Proceedings of the Logic Colloquium held in Aachen, July 18-23, 1983

Logic Colloquium '87: Proceedings of the Colloquium Held in Granada, Spain July 20-25, 1987

Stellar structure and evolution

Evolution of Stars and Stellar Populations

Evolution of Stars and Stellar Populations

Evolution of Stars and Stellar Populations

Evolution of Stars and Stellar Populations

Evolution of stars and stellar populations

Logic Colloquium'76: Proceedings of a conference held in Oxford in July 1976 (Studies in Logic and the Foundations of Mathematics, 87)

Progress in Turbulence III: Proceedings of the iTi Conference in Turbulence 2008 (Springer Proceedings in Physics)

Trends in Ring Theory: Proceedings of a Conference at Miskolc, Hungary July 15-20, 1996 (Conference Proceedings (Canadian Mathematical Society))

Stellar Rotation (Cambridge Astrophysics)

Foundations of Software Technology and Theoretical Computer Science: 13th Conference, Bombay, India, December 15-17, 1993. Proceedings: Proceedings of ... 1993

The Stages of Meditation

The Cambridge History of Philosophy in Late Antiquity, Volume I

Advanced Metallization for Ulsi Applications in 1993: Proceedings of the Conference Held October 5-7, 1993, San Diego, California, U.S.A., and

Stellar Rotation (Cambridge Astrophysics)

Stellar Rotation (Cambridge Astrophysics)

Late, Late in the Evening

Differential equations in Banach spaces: proceedings of the Bologna conference

The Evolution of Morphology (Studies in the Evolution of Language)

Mathematical Results In Quantum Physics: Proceedings of the QMath11 Conference

Mathematical Results In Quantum Physics: Proceedings of the QMath11 Conference

Circumstellar Media in Late Stages of Stellar Evolution (Proceedings of the 34th Herstmonceux Conference, Held in Cambridge, July 12-16, 1993)

The nature of compact objects in active galactic nuclei: proceedings of the 33rd Herstmonceux conference, held in Cambridge, July 6-22 i.e. 16-22 , 1992

Category Theory: Proceedings of the International Conference held in Como, Italy, July 22-28, 1990

Category Theory: Proceedings of the International Conference Held in Como, Italy, July 22-28, 1990

Category Theory: Proceedings of the International Conference held in Como, Italy, July 22-28, 1990 (Lecture Notes in Mathematics)

Logic Colloquium '87: Proceedings of the Colloquium Held in Granada, Spain July 20-25, 1987

Stellar Structure and Evolution

Stars and Stellar Evolution

Proceedings of the Logic Colloquium held in Aachen, July 18-23, 1983

Logic Colloquium '87: Proceedings of the Colloquium Held in Granada, Spain July 20-25, 1987

Stellar structure and evolution

Evolution of Stars and Stellar Populations

Evolution of Stars and Stellar Populations

Evolution of Stars and Stellar Populations

Evolution of Stars and Stellar Populations

Evolution of stars and stellar populations

Logic Colloquium'76: Proceedings of a conference held in Oxford in July 1976 (Studies in Logic and the Foundations of Mathematics, 87)

Progress in Turbulence III: Proceedings of the iTi Conference in Turbulence 2008 (Springer Proceedings in Physics)

Trends in Ring Theory: Proceedings of a Conference at Miskolc, Hungary July 15-20, 1996 (Conference Proceedings (Canadian Mathematical Society))

Stellar Rotation (Cambridge Astrophysics)

Foundations of Software Technology and Theoretical Computer Science: 13th Conference, Bombay, India, December 15-17, 1993. Proceedings: Proceedings of ... 1993

The Stages of Meditation

The Cambridge History of Philosophy in Late Antiquity, Volume I

Advanced Metallization for Ulsi Applications in 1993: Proceedings of the Conference Held October 5-7, 1993, San Diego, California, U.S.A., and

Stellar Rotation (Cambridge Astrophysics)

Stellar Rotation (Cambridge Astrophysics)

Late, Late in the Evening

Differential equations in Banach spaces: proceedings of the Bologna conference

The Evolution of Morphology (Studies in the Evolution of Language)

Mathematical Results In Quantum Physics: Proceedings of the QMath11 Conference

Mathematical Results In Quantum Physics: Proceedings of the QMath11 Conference

Recommend Documents