Massive Stars: Life and Death

Home , 17 Monocerotis, 18 Monocerotis, 20 Monocerotis, 27 Monocerotis, 28 Monocerotis, 3 Monocerotis

Dissertation

Presented in Partial Fulﬁllment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Jos´eLuis Prieto Katunari´c

Graduate Program in Astronomy

The Ohio State University 2009

Dissertation Committee: Professor Krzysztof Z. Stanek, Advisor Professor Christopher S. Kochanek Professor John F. Beacom Copyright by

Jos´eL. Prieto

2009 ABSTRACT

Although small in number, massive stars are critical to the formation and evolution of galaxies. They shape the interstellar medium of galaxies through their strong winds and ultra-violet radiation, are a major source of the heavy elements enriching the interstellar medium, and are the progenitors of core-collapse supernovae and gamma-ray bursts, which are among the most energetic explosions in the Universe and mark the death of a massive star. Still, our understanding of the connection between massive stars and supernovae from observations is fairly limited.

In this dissertation, I present new observational evidence that shows the importance of metallicity, mass-loss, and binarity in the lives and deaths of massive stars.

We investigate how the different types of supernovae are relatively affected by the metallicity of their host galaxy. We take advantage of the large number of spectra of star-forming galaxies obtained by the Sloan Digital Sky Survey and their overlap with supernova host galaxies. We find strong evidence that type Ib/c supernovae are occurring in higher-metallicity host galaxies than type II supernovae. We discuss various implications of our findings for understanding supernova progenitors and their host galaxies, including interesting supernovae found in low-metallicity hosts.

ii We present the discovery of the progenitors of SN 2008S and the luminous transient in NGC 300 in archival data obtained with the Spitzer Space Telescope.

They are deeply dust-enshrouded massive stars, with extremely red mid-infrared colors compared to other massive stars, and relatively low bolometric luminosities

4 (≈ 5 × 10 L⊙). We discuss the implications of these ﬁndings for the evolution and census of “low-mass” massive stars (i.e., ∼8 − 12 M⊙), and we connect it with theoretical discussions of electron-capture supernovae near this mass range, explosive birth of massive white dwarfs, and massive star outbursts.

We present a Spitzer low-resolution mid-infrared spectrum of the luminous transient in NGC 300. The spectrum shows that the transient is very luminous in the mid-infrared and most of the pre-existing progenitor dust survived the explosion.

Furthermore, the spectrum shows strong, broad emission features that are observed in Galactic carbon-rich proto-planetary nebulae. These observations support our conclusions of an explosive event on a massive carbon-rich AGB or post-AGB star as the origin of the transient in NGC 300 and SN 2008S.

We present extensive ugrizY HJKs photometry and optical spectroscopy of

SN 2005gj. These data show that SN 2005gj is the second possible case, after

SN 2002ic, of a thermonuclear explosion in a dense circumstellar environment. The interaction of the supernova ejecta with the dense circumstellar medium is stronger than in SN 2002ic.

iii Finally, we present the discovery of a peculiar eclipsing binary in a variability survey of the dwarf galaxy Holmberg IX undertaken with the Large Binocular

Telescope. The binary has a period of 271 days, and is composed of two yellow supergiants that are overﬂowing their Roche lobes. Such systems must be rare, and indeed we only note one similar system in the Small Magellanic Cloud. We propose that these systems may be the progenitors of supernovae that appeared to have yellow-supergiant progenitors.

iv A mis padres, Alicia y Rodolfo.

v ACKNOWLEDGMENTS

I am very grateful of so many people that in diﬀerent ways have helped me these years during grad school and before. In particular, I don’t think I would have made it without the continuous support and encouragement of my adviser, Kris

Stanek, my family, and Linda. First let me say Gracias to Kris, I will come back to my family and Linda at the end. I feel very fortunate that I had the chance to meet

Kris and work with him over the last three years. Kris has shaped my way to see, think about, and do astronomy and science in general. He has always been the most supportive person, almost in unthinkable ways by (for example) letting me continue working in supernova-related topics and encouraging me to pursue exciting ideas and projects that came up along the way. We had many discussions about astronomy and science, new ideas for research projects, and his yard (among other topics). I enjoyed very much all of them. He also shared with me his vast experience and knowledge of gamma-ray bursts, the distance scale, variable stars, and photometry

(to name just a few). His stories and experiences of other people working in the

ﬁeld and how “not to do” science were always very enriching. I also enjoyed playing ping-pong and sledding at his house. I will miss him much.

vi I am also very grateful of John Beacom, Chris Kochanek, and Todd Thompson.

They were always happy to help, support, and encourage me. I learned a lot from many extensive discussions we had, including Kris, about massive stars and supernovae. Chris, with his amazing fundamental knowledge of astronomy and physics, always had time to share his thoughts and help with papers and new ideas.

I appreciate very much the advice and help of John and Chris with my English and scientiﬁc writing. I also thank Chris and John for being part of my defense committee. I think it will be hard to ﬁnd a person with more energy and excitement about astronomy than Todd, and I hope I can at least have a fraction of that.

Most of the theoretical interpretation in this dissertation, particularly about the dusty transients, were Todd’s ideas. Mi amigo Todd, thank you for sharing all that knowledge about several areas in astrophysics and for so many discussions about science and Espa˜nol. I hope that Todd at least learned this from me, and doesn’t forget it: Pablo Neruda was Chilean.

I thank Oleg Gnedin for advising me in my first project at OSU. It was hard to work in a purely theoretical project on globular clusters, but with a lot of help, advice and patience of Oleg, we could finally finish it and publish a paper on the results.

My big thanks to Darren DePoy for letting me work in the SDSS-II Supernova

Search, for sharing his observing experience and stories, and for organizing the supernova follow-up campaigns with the MDM 2.4m telescope. Here I have to give special thanks to all my fellow graduate students that went observing to MDM and

vii participated in the 3 campaigns, to all the MDM staﬀ for maintaining the telescopes and instruments and also for their help, and to Rick Pogge for being always there to help when something came up with the instrument, telescope or control system. It wouldn’t have been possible to carry out this extensive follow-up project without them. I thank the SDSS-II Supernova Search and the Carnegie Supernova Project groups for contributing most of the photometric data for SN 2005gj. I learned a lot from working in the SDSS-II Supernova Search, and I am especially grateful to Masao Sako and Peter Garnavich. I thank Kris Sellgren for sharing part of her experience and knowledge of infrared astronomy with me. Kris was very enthusiastic to teach me about mid-infrared spectra of proto-planetary nebula and other infrared sources.

Going all the way back to high school, I thank my physics teacher Jimena

Scarich, she was very important in my early science education and interest in physics and astronomy. I would like to thank also my undergraduate adviser at

Universidad Católica, Alejandro Clocchiatti, who introduced me to the exciting field of supernovae, always supported me, and helped me get a position doing supernova research at CTIO after undergrad. At Católica, I also enjoyed very much working with Felipe Barrientos, and I thank him for teaching me how to use Perl and PDL, which had been very useful tools since. At CTIO, working with Nick Suntzeff and

Armin Rest was a really great experience. I am indebted to Nick and Armin (and also others of the supernova group in Tololo, especially Kevin Krisciunas) for giving

viii me the great opportunity to work in very exciting scientiﬁc projects during that year and for introducing me to participate in two big scientiﬁc collaborations, Essence and Supermacho.

The environment at the Astronomy Department was always nice and friendly, and this made a big difference in the day-to-day life. I thank the Department office staff, especially Kristy Scheckelhoff, for helping so much with all the paper work. Also David Will and Michael Savage for so much help and assistance with computers. I am also grateful of all the faculty for organizing morning coffee, I enjoyed very much many discussions and learned quite a bit about very different areas in astronomy. I also appreciate all the work put by faculty to make the graduate courses useful and interesting (including order of magnitude) and I thank them for that. I thank many conversations over lunches and dinners with fellow graduate students and friends, especially Julio Chanamé, Roberto Assef, and Subo

Dong. Outside work, I very much enjoyed spending time with Max Montenegro and family, they always made me feel like at home.

I would like to thank my family, especially my parents to whom I would like to dedicate this dissertation. Their endless love, care, guidance, and support without bounds made me get to this point and allowed me to follow my dream of doing astronomy, even though it took me so far from them. Muchas Gracias Papis. Gracias tambi´en a mis hermanos Cristian, Eugenio, Rodolfo, y a los chicos. Finally, it’s hard to describe in just a few words how lucky I feel to have met Linda and thank her

ix enough for all her encouragement, support, love, patience, and care over the last years. Thank you so much.

x VITA

November 8, 1980 ...... Born – Punta Arenas, Chile

2003 ...... B.S. Astronomy, P. Universidad Cat´olica de Chile

2003 – 2004 ...... Research Assistant, Cerro Tololo Inter-American Observatory, La Serena, Chile

2004 – 2008 ...... Graduate Research Associate, The Ohio State University

2008 – 2009 ...... Presidential Fellow, The Ohio State University

PUBLICATIONS

Research Publications

1. K. Krisciunas et al. (31 authors, incl. J. L. Prieto), “Optical and In- frared Photometry of the Type Ia Supernovae 1991T, 1991bg, 1999ek, 2001bt, 2001cn, 2001cz, and 2002bo”, AJ, 128, 3034, (2004).

2. T. Matheson et al. (31 authors, incl. J. L. Prieto), “Spectroscopy of High-Redshift Supernovae from the ESSENCE Project: The First 2 Years”, AJ, 129, 2352, (2005).

3. K. Krisciunas et al. (32 authors, incl. J. L. Prieto), “Hubble Space Tele- scope Observations of Nine High-Redshift ESSENCE Supernovae”, AJ, 130, 2453 (2005).

4. A. Rest et al. (23 authors, incl. J. L. Prieto), “Testing LMC Microlens- ing Scenarios: The Discrimination Power of the SuperMACHO Microlensing Survey”, ApJ, 634, 1103 (2005).

xi 5. A. Rest et al. (16 authors, incl. J. L. Prieto), “Light echoes from ancient supernovae in the Large Magellanic Cloud”, Nature, 438, 1132, (2005).

6. K. Krisciunas, J. L. Prieto, P. M. Garnavich, J-L G. Riley, A. Rest, C. Stubbs, and R. McMillan, “Photometry of the Type Ia Supernovae 1999cc, 1999cl, and 2000cf”, AJ, 131, 1639, (2006).

7. S. Blondin et al. (37 authors, incl. J. L. Prieto), “Using Line Proﬁles to Test the Fraternity of Type Ia Supernovae at High and Low Redshifts”, AJ, 131, 1648, (2006).

8. A. Clocchiatti et al. (29 authors, incl. J. L. Prieto), “Hubble Space Tele- scope and Ground-based Observations of Type Ia Supernovae at Redshift 0.5: Cosmological Implications”, ApJ, 642, 1, (2006).

9. M. Modjaz et al. (13 authors, incl. J. L. Prieto), “Early-Time Photome- try and Spectroscopy of the Fast Evolving SN 2006aj Associated with GRB 060218”, ApJL, 645, 21, (2006).

10. J. L. Prieto, A. Rest, and N. B. Suntzeﬀ, “A New Method to Calibrate the Magnitudes of Type Ia Supernovae at Maximum Light”, ApJ, 647, 501, (2006).

11. K. Krisciunas et al. (12 authors, incl. J. L. Prieto), “The Type Ia Su- pernova 2004S, a Clone of SN 2001el, and the Optimal Photometric Bands for Extinction Estimation”, AJ, 133, 58, (2007).

12. K. Z. Stanek et al. (12 authors, incl. J. L. Prieto), “ “Anomalous” Op- tical Gamma-Ray Burst Afterglows Are Common: Two z 4 Bursts, GRB 060206 and GRB 060210”, ApJL, 654, 21, (2007).

13. A. Garg et al. (21 authors, incl. J. L. Prieto), “Light Curves of Type Ia Supernovae from Near the Time of Explosion”, AJ, 133, 403, (2007).

14. S. A. Yost et al. (47 authors, incl. J. L. Prieto), “Exploring Broadband GRB Behavior during γ-Ray Emission”, ApJ, 657, 925, (2007).

15. M. M. Phillips et al. (44 authors, incl. J. L. Prieto), “The Peculiar SN 2005hk: Do Some Type Ia Supernovae Explode as Deﬂagrations ?”, PASP, 119, 360, (2007).

16. G. Miknaitis et al. (36 authors, incl. J. L. Prieto), “The ESSENCE Su-

xii pernova Survey: Survey Optimization, Observations, and Supernova Photometry”, ApJ, 666, 674, (2007).

17. W. M. Wood-Vasey et al. (37 authors, incl. J. L. Prieto), “Observa- tional Constraints on the Nature of Dark Energy: First Cosmological Results from the ESSENCE Supernova Survey”, ApJ, 666, 694, (2007).

18. M. Hicken, P. M. Garnavich, J. L. Prieto, S. Blondin, D. L. DePoy, R. P. Kirshner, and J. Parrent, “The Luminous and Carbon-rich Supernova 2006gz: A Double Degenerate Merger ?”, ApJL, 669, 17, (2007).

19. J. Frieman et al. (101 authors, incl. J. L. Prieto), “The Sloan Digital Sky Survey-II Supernova Survey: Technical Summary”, AJ, 135, 338, (2008).

20. M. Sako et al. (50 authors, incl. J. L. Prieto), “The Sloan Digital Sky Survey-II Supernova Survey: Search Algorithm and Follow-Up Observations”, AJ, 135, 348, (2008).

21. J. L. Prieto et al. (17 authors), “LBT Discovery of a Yellow Supergiant Eclipsing Binary in the Dwarf Galaxy Holmberg IX”, ApJL, 673, 59, (2008).

22. J. L. Prieto, K. Z. Stanek, and J. F. Beacom, “Characterizing Super- nova Progenitors via the Metallicities of their Host Galaxies, from Poor Dwarfs to Rich Spirals”, ApJ, 673, 999, (2008).

23. S. Brady, J. R. Thorstensen, M. D. Koppelman, J. L. Prieto, P. M. Gar- navich, A. Hirschauer, and M. Florack, “The Eclipsing Cataclysmic Variable Lanning 386: Dwarf Nova, SW Sextantis Star, or Both ?”, PASP, 120, 301, (2008).

24. M. Modjaz et al. (9 authors, incl. J. L. Prieto), “Measured Metallicities at the Sites of Nearby Broad-Lined Type Ic Supernovae and Implications for the Supernovae Gamma-Ray Burst Connection”, AJ, 135, 1136, (2008).

25. B. S. Gaudi et al. (22 authors, incl. J. L. Prieto), “Discovery of a Very Bright, Nearby Gravitational Microlensing Event”, ApJ, 677, 1268, (2008).

26. C. Zheng et al. (80 authors, incl. J. L. Prieto), “First-Year Spectroscopy for the Sloan Digital Sky Survey-II Supernova Survey”, AJ, 135, 1766, (2008).

27. S. F. Anderson et al. (21 authors, incl. J. L. Prieto), “Two More Can- didate AM Canum Venaticorum (am CVn) Binaries from the Sloan Digital Sky Survey”, AJ, 135, 2108, (2008).

xiii 28. A. Rest et al. (29 authors, incl. J. L. Prieto), “Spectral Identiﬁcation of an Ancient Supernova Using Light Echoes in the Large Magellanic Cloud”, ApJ, 680, 1137, (2008).

29. J. L. Prieto et al. (10 authors), “Discovery of the Dust-Enshrouded Pro- genitor of SN 2008S with Spitzer”, ApJL, 681, 9, (2008).

30. A. Rest et al. (19 authors, incl. J. L. Prieto), “Scattered-Light Echoes from the Historical Galactic Supernovae Cassiopeia A and Tycho (SN 1572)”, ApJL, 681, 81, (2008).

31. B. Dilday et al. (53 authors, incl. J. L. Prieto), “A Measurement of the Rate of Type Ia Supernovae at Redshift z 0.1 from the First Season of the SDSS-II Supernova Survey”, ApJ, 682, 262, (2008).

32. A. C. Becker et al. (39 authors, incl. J. L. Prieto), “Exploring the Outer Solar System with the ESSENCE Supernova Survey”, ApJL, 682, 53, (2008).

33. X. Dai et al. (23 authors, incl. J. L. Prieto), “Go Long, Go Deep: Finding Optical Jet Breaks for Swift-Era GRBs with the LBT”, ApJL, 682, 77, (2008).

34. C. S. Kochanek, J. F. Beacom, M. D. Kistler, J. L. Prieto, K. Z. Stanek, T. A. Thompson, and H. Yuksel, “A Survey About Nothing: Monitoring a Million Supergiants for Failed Supernovae”, ApJ, 684, 1336, (2008).

35. J. A. Holtzman et al. (32 authors, incl. J. L. Prieto), “The Sloan Digi- tal Sky Survey-II: Photometry and Supernova Ia Light Curves from the 2005 Data”, AJ, 136, 2306, (2008).

36. C. J. Grier et al. (17 authors, incl. J. L. Prieto), “The Mass of the Black Hole in the Quasar PG 2130+099”, ApJ, 688, 837, (2008).

37. J. L. Prieto, and O. Y. Gnedin, “Dynamical Evolution of Globular Clus- ters in Hierarchical Cosmology”, ApJ, 689, 919, (2008).

38. S. Blondin, J. L. Prieto, F. Patat, P. Challis, M. Hicken, R. P. Kirshner, T. Matheson, and M. Modjaz, “A Second Case of Variable Na I D Lines in a Highly Reddened Type Ia Supernova”, ApJ, 693, 207, (2009).

39. J. C. Bird, K. Z. Stanek, and J. L. Prieto, “Using Ultra Long Period

xiv Cepheids to Extend the Cosmic Distance Ladder to 100 Mpc and Beyond”, ApJ, 695, 874, (2009).

40. H. E. Bond, L. R. Bedin, A. Z. Bonanos, R. M. Humphreys, L. A. G. Berto Monard, J. L. Prieto, and F. M. Walter, “The 2008 Luminous Optical Transient in the Nearby Galaxy NGC 300”, ApJL, 695, 154, (2009).

41. M.J. Valtonen et al. (41 authors, incl. J. L. Prieto), “Tidally Induced Outbursts in OJ 287 during 2005-2008”, ApJ, 698, 781, (2009).

42. C. Badenes, J. Harris, D. Zaritsky, and J. L. Prieto, “The Stellar Ances- try of Supernovae in the Magellanic Clouds - I. the Most Recent Supernovae in the Large Magellanic Cloud”, ApJ, 700, 727, (2009).

FIELDS OF STUDY

Major Field: Astronomy

xv Table of Contents

Abstract...... ii

Dedication...... v

Acknowledgments...... vi

Vita ...... xi

ListofTables ...... xx

ListofFigures...... xxii

Chapter 1 Introduction ...... 1

1.1 MassiveStarsandSupernovae ...... 1

1.2 Probing Massive Star Evolution and Supernova Connection ...... 3

1.2.1 Indirect Studies: Host Galaxy Environments ...... 4

1.2.2 Direct Studies: Identifying Transient-Progenitor Pairs..... 5

1.3 ScopeoftheDissertation...... 7

Chapter 2 Characterizing Supernova Progenitors via the Metallicities of their Host Galaxies, from Poor Dwarfs to Rich Spirals ..... 10

2.1 Introduction...... 10

2.2 First Catalog: Supernova-Host Pairs with Known Host Metallicities (SAI ∩ SDSS-DR4)...... 15

2.2.1 Testing Supernova Trends with Metallicity ...... 20

2.2.2 Supernovae in Low-Metallicity Hosts ...... 26

xvi 2.3 Second Catalog: Supernova-Host Pairs with Unknown Host Metallicities (SAI ∩ SDSS-DR6)...... 28

2.4 Discussion and Conclusions ...... 30

Chapter 3 Discovery of the Dust-Enshrouded Progenitor of the Type IIn SN 2008S with Spitzer ...... 54

3.1 Introduction...... 54

3.2 SearchingfortheProgenitor ...... 56

3.3 BeneaththeShroud...... 60

3.4 Discussion and Conclusions ...... 62

Chapter 4 A New Class of Luminous Transients and A First Census of Their Massive Stellar Progenitors ...... 70

4.1 Introduction...... 70

4.2 TheClass ...... 75

4.2.1 SN2008S ...... 76

4.2.2 NGC300 ...... 77

4.2.3 M85 ...... 79

4.2.4 SN1999bw ...... 82

4.2.5 The Connection to Other Transients ...... 83

4.3 Rates...... 86

4.3.1 ObservedCounts ...... 87

4.3.2 Arguments for Incompleteness & Some Implications ...... 90

4.4 AFirstCensus ...... 95

4.4.1 Catalog ...... 97

4.4.2 The Color-Magnitude Diagram ...... 99

4.4.3 SpectralEnergyDistributions ...... 104

xvii 4.4.4 Variability...... 107

4.4.5 OtherGalaxies ...... 109

4.5 Discussion...... 110

4.5.1 Numbers&Rates...... 111

4.5.2 Connection to The Evolution of Massive Stars ...... 116

4.5.3 AMoreCompleteCensus ...... 124

Chapter 5 A Spitzer/IRS Spectrum of the 2008 Luminous Transient in NGC 300: Connection to Proto-Planetary Nebula ...... 137

5.1 Introduction...... 137

5.2 Spitzer Observations ...... 139

5.3 Analysis ...... 140

5.3.1 SpectralFeatures ...... 140

5.3.2 SpectralEnergyDistribution...... 144

5.4 Discussion & Conclusions ...... 149

5.4.1 Mid-IR Spectrum and SED of NGC 300-OT ...... 149

5.4.2 NGC 300-OT and SN 2008S: Connection to Proto-Planetary Nebulae ...... 152

5.4.3 The Progenitors of NGC 300-OT and SN 2008S: Massive Carbon-richAGB/post-AGBstars? ...... 155

5.4.4 Progenitors and Transients: Concluding Remarks ...... 158

Chapter 6 A Study of the Type Ia/IIn Supernova 2005gj from X-ray to the Infrared ...... 169

6.1 Introduction...... 169

6.2 Photometry ...... 173

6.2.1 SDSSandMDM ...... 173

xviii 6.2.2 CSP ...... 175

6.3 Spectroscopy ...... 177

6.4 X-rayObservation ...... 179

6.5 Results...... 182

6.5.1 Optical light curves and colors ...... 182

6.5.2 NIRlightcurves ...... 189

6.5.3 Bolometric light curve ...... 190

6.5.4 Optical spectroscopy ...... 194

6.6 Discussion...... 201

6.6.1 StructureoftheCSM...... 204

6.6.2 Rates, hosts galaxies and possible progenitors of SN 2002ic-like supernovae...... 212

Chapter 7 LBT Discovery of a Yellow Supergiant Eclipsing Binary in the Dwarf Galaxy Holmberg IX ...... 251

7.1 Introduction...... 251

7.2 Observations...... 253

7.3 LightCurve ...... 254

7.4 Discussion and Conclusions ...... 258

Appendix A A. Extreme-AGB Star Variability ...... 267

Appendix B LBV Candidate Variability ...... 270

Bibliography ...... 270

xix List of Tables

2.1 Supernovaandhostgalaxydata ...... 41

2.1 Supernovaandhostgalaxydata ...... 42

2.1 Supernovaandhostgalaxydata ...... 43

2.1 Supernovaandhostgalaxydata ...... 44

2.1 Supernovaandhostgalaxydata ...... 45

2.1 Supernovaandhostgalaxydata ...... 46

2.1 Supernovaandhostgalaxydata ...... 47

2.1 Supernovaandhostgalaxydata ...... 48

2.1 Supernovaandhostgalaxydata ...... 49

2.1 Supernovaandhostgalaxydata ...... 50

2.1 Supernovaandhostgalaxydata ...... 51

2.1 Supernovaandhostgalaxydata ...... 52

2.1 Supernovaandhostgalaxydata ...... 53

3.1 Spectral Energy Distribution of the Progenitor of SN 2008S ..... 69

4.1 MIR Catalog for 53,194 Point Sources in M33 ...... 134

4.2 Photometryforthe18EAGBsinM33 ...... 135

4.3 Photometry for the 16 LBV Candidates in M33 from Massey et al.(2007) ...... 136

5.1 Features in the Spitzer/IRS Spectrum of NGC300-OT ...... 166

5.2 Spectral Energy Distribution of the Progenitor of NGC 300-OT . . . 167

xx 5.3 Black-body Fits to the Transient and Progenitor SEDs ...... 168

6.1 SDSS ugriz and CSP u′g′r′i′ photometry of comparison stars in commonintheﬁeldofSN2005gj...... 235

6.1 SDSS ugriz and CSP u′g′r′i′ photometry of comparison stars in commonintheﬁeldofSN2005gj...... 236

6.2 SDSS and MDM ugriz photometryofSN2005gj...... 237

6.2 SDSS and MDM ugriz photometryofSN2005gj...... 238

6.3 CSP u′g′r′i′ photometryofSN2005gj...... 239

6.4 CSP YJHKs photometryofSN2005gj ...... 240

6.5 Light-curve parameters for SN 2005gj ...... 241

6.6 Spectroscopic observations of SN 2005gj ...... 242

6.6 Spectroscopic observations of SN 2005gj ...... 243

6.7 K-correctionsofSN2005gj ...... 244

6.8 Derived integrated luminosity and black-body ﬁts...... 245

6.8 Derived integrated luminosity and black-body ﬁts...... 246

6.9 LibraryofspectrausedinSNID ...... 247

6.9 LibraryofspectrausedinSNID ...... 248

6.10 Results of the Gaussian ﬁts to Hα and Hβ features ...... 249

6.10 Results of the Gaussian ﬁts to Hα and Hβ features ...... 250

7.1 Best-ﬁtBinaryModelParameters...... 266

xxi List of Figures

2.1 Metallicities of supernova host galaxies as a function of redshift and absolute B magnitudefromSDSS ...... 36

2.2 Sample of metal rich and metal poor supernova host galaxies in SDSS 37

2.3 Cumulative fraction of oxygen abundances for supernova hosts in SDSS 38

2.4 Fraction of SN Ib/c to SN II as a function of host metallicity in SDSS 39

2.5 Cumulative fraction of projected separation between supernova and host 40

3.1 Pre-explosion images of SN 2008S from LBT and Spitzer ...... 66

3.2 Light curves of the progenitor of SN 2008S from Spitzer photometry . 67

3.3 Spectral energy distribution of the progenitor of SN 2008S ...... 68

4.1 Mid-infrared color-magnitude diagram of M33 ...... 127

4.2 Same as Figure 4.1, but focused on the red and bright region of interest128

4.3 Images showing the reddest source in Figure 4.1 ...... 129

4.4 Mid-infrared color-color diagram ...... 130

4.5 Spectral energy distributions of EAGB stars in M33 and candidate LBVs131

4.6 Light curves of two selected sources ...... 132

4.7 RMS variation of sources at 4.5 µm as a function of color ...... 133

5.1 Spitzer mid-infraredspectrumofNGC300-OT ...... 161

5.2 Comparison of NGC 300-OT to type IIP supernovae ...... 162

5.3 Comparison of NGC 300-OT to massive stars ...... 163

5.4 Comparison of NGC 300-OT to proto-planetary nebula ...... 164

xxii 5.5 SpectralenergydistributionofNGC300-OT ...... 165

6.1 r-bandimageoftheﬁeldofSN2005gj ...... 219

6.2 SpectraofSN2005gj ...... 220

6.3 MorespectraofSN2005gj ...... 221

6.4 ugrizY JHKs light curves of SN 2005gj ...... 222

6.5 Diﬀerence between synthetic g − r color from spectra and photometry 223

6.6 Colors of SN 2005gj as a function of time ...... 224

6.7 Light curve of SN 2005gj in the optical compared with others. . . . . 225

6.8 LightcurveofSN2005gjinthenear-IR...... 226

6.9 Examples of black-body ﬁts to the spectral energy distribution of SN2005gj ...... 227

6.10 Bolometric light curve of SN 2005gj ...... 228

6.11 Comparison of spectra of SN 2005gj with other supernovae ...... 229

6.12 Cross-correlation comparison of SN 2005gj with other spectra . . . . 230

6.13 FitstothespectraofSN2005gj ...... 231

6.14 EvolutionofBalmerlines...... 232

6.15 Line proﬁles of Hα and Hβ ...... 233

6.16 Identiﬁcation of lines in the spectrum of SN 2005gj ...... 234

7.1 Light curve of eclipsing binary in Holmberg IX ...... 262

7.2 Color-magnitude diagrams of Holmberg IX with the binary ...... 263

7.3 LightcurveofeclipsingbinaryintheSMC ...... 264

7.4 Color-magnitude diagrams with the eclipsing binaries in Holmberg IX andSMC ...... 265

A.1 LightcurvesofEAGBsources1 ...... 268

A.2 LightcurvesofEAGBsources2 ...... 269

xxiii B.1 LightcurvesofLBVs1...... 271

B.2 LightcurvesofLBVs2...... 272

xxiv Chapter 1

Introduction

1.1. Massive Stars and Supernovae

Stars are born, they live, evolve, and die. Initially composed mainly of

Hydrogen (∼ 70%) and Helium (∼ 25%), the life cycle of a star depends most signiﬁcantly on a single parameter: mass. Stars less massive than ∼ 8 − 10 M⊙ burn

Hydrogen to Helium in the core during the main-sequence through thermonuclear reactions, then Helium to Carbon/Oxygen (M ∼> 0.5M⊙) in their post main-sequence evolution. They end their lifes expelling their outer envelope in a planetary nebula and leaving behind a Carbon/Oxygen white dwarf core of ∼< 1.4 M⊙. The lifetime of a M ∼ 1 − 8 M⊙ star expands more than two orders of magnitude, from ∼ 40 Myr

(8 M⊙) to ∼ 10 Gyr (1 M⊙).

Stars more massive than ∼ 8 − 10 M⊙ burn past C/O in the core until they form an Iron core of MChandra ≃ 1.4 M⊙. At this point the thermonuclear reactions are no longer exothermic, neutrino losses dominate, and the core cannot support itself against gravity. The core collapses in a fraction of a second, matter bounces

1 into the newly formed neutron star, and an energetic shock propagates out helped by the production of ∼ 1053 erg in neutrinos. Eventually, the shock “breaks out” through the envelope of the star and produces an energetic display of electromagnetic radiation, that can last for months, marking the catastrophic death of a massive star as a core-collapse supernova. Since supernovae have been discovered, observed, and studied in our Galaxy and external galaxies, we know that this occurs in nature at least in a fraction of massive stars, although theorists have mostly failed to produce successful supernova explosions (e.g., Thompson et al. 2003). Stars more massive than ∼ 25 M⊙ may end their life forming a black-hole instead of a neutron star without an optically bright supernova.

Mass is not the only physical parameter that regulates the evolution of a star.

Stellar evolution theory (e.g., Meynet & Maeder 2003) predicts that metallicity, rotation, and binarity are also important and can significantly affect the lives and deaths of massive stars (e.g., Heger et al. 2003; Eldridge et al. 2008). In general terms, metallicity is a source of opacity that affects the transport of photons through the star and enhances radiatively-driven winds; rotation is a source of angular momentum that changes the energetics and structure of the star; and the presence of a close binary can change the evolution of the star through interactions and mass-transfer.

Mass loss, which in turn is a complicated function of temperature, luminosity, mass, metallicity, rotation, and binarity, plays a crucial role shaping the life of

2 massive stars and their final fate as supernovae. We know this from studies of stellar populations in the Galaxy and nearby galaxies (e.g., Massey 2003), as well as from the large variety of different classes of core-collapse supernovae that have been identified (e.g., Filippenko 1997). The observational classification of core-collapse supernovae, mainly based on the presence or absence of H and He in the optical spectra obtained close to maximum light – type II-P, II-L, IIb, Ib, Ic, IIn – is primarily an imprint of the mass-loss history of the progenitor star.

1.2. Probing Massive Star Evolution and Supernova

Connection

During their life, massive stars pollute the interstellar medium (ISM) with metals through strong winds and inject energy into the ISM that aﬀects the evolution of subsequent generations of stars and galaxies (e.g., Massey 2003). At the time of their death as a supernova, they inject still more energy into the ISM (∼ 1051 erg of kinetic energy) and produce a large fraction of the elements heavier than Iron (e.g.,

Woosley et al. 2002).

Since their importance for astrophysics is large, testing the evolution of massive stars and their connection to supernovae observationally is extremely important.

This is, however, diﬃcult because massive stars are rare in any given galaxy and

3 supernovae explosions are even rarer (∼ 1 per 100 years in a galaxy like the Milky

Way).

1.2.1. Indirect Studies: Host Galaxy Environments

Statistical studies of the resolved stellar populations of massive stars in the

Milky Way and nearby galaxies have helped to constrain stellar evolution models.

For example, the ratio of the number of O-type stars to Wolf-Rayet stars in galaxies with diﬀerent metallicities and at diﬀerent locations within the galaxy can give important clues about mass-loss (e.g., Maeder & Meynet 1994; Massey 2003).

In a similar way, studying the host galaxy environment of extragalactic supernovae discovered by diﬀerent surveys oﬀers an excellent opportunity to connect the average properties of the (unresolved) massive star population (like metallicity) with stellar deaths.

This kind of statistical study needs a large number of supernovae discoveries with accurate positions, and only in the last ∼ 10 years supernova surveys undertaken by amateur and professional astronomers have been sucessfull at ﬁnding large numbers of supernovae (∼ 3300) out to cosmological distances. However, studies of host galaxies of core-collapse supernovae have been lacking mainly because of the overwhelming interest of the community in type Ia supernovae (exploding white dwarfs rather than collapsing massive stars) to measure cosmological distances,

4 with some important exceptions (e.g., van Dyk 1992; van Dyk et al. 1996; Prantzos

& Boissier 2003; James & Anderson 2006; Anderson & James 2008; Modjaz et al. 2008).

The large sky area covered by the Sloan Digital Sky Survey (SDSS), with homogeneous imaging and spectra of thousands of galaxies, provides a unique opportunity to study the host galaxies of many supernovae that overlap with the survey galaxies. Taking advantage of the results of Tremonti et al. (2004) on the

Oxygen abundances of star-forming galaxies derived from SDSS spectra, in this dissertation we present a statistical comparison of the metallicities of diﬀerent supernova types. These results can be directly connected to stellar evolution models for massive stars.

1.2.2. Direct Studies: Identifying Transient-Progenitor

Pairs

One of the most powerful tests of stellar evolution theory for massive stars is to observationally establish the causal mapping between diﬀerent populations of evolved stars that are observed in galaxies (e.g., massive AGB stars, blue-, yellow-, and red-supergiants, Wolf-Rayet stars, Luminous Blue Variables) and their ﬁnal explosive deaths.

5 This connection has been ﬁrmly proven only for a small number of objects

(< 10), most notably in the case of the core-collapse supernova SN 1987A in the

Large Magellanic Cloud (LMC). SN 1987A, which had an unusual light curve and spectral properties (e.g., Menzies et al. 1987), had a blue-supergiant progenitor star with an estimated main-sequence mass of ∼ 20 M⊙ (e.g., West et al. 1987).

The progenitor star had been identiﬁed and cataloged in pre-explosion images

(Sk-69 202). The blue-supergiant nature of the progenitor of SN 1987A came as a big surprise because stars of that mass were expected to explode as red-supergiants

(e.g., Arnett 1987).

In the last ∼ 10 years, the combination of successful supernova surveys, and the increasing volume of deep archival imaging data of nearby (∼< 30 Mpc) galaxies obtained primarily with the Hubble Space Telescope (HST), has allowed new identiﬁcations of supernova progenitors in pre-explosion images (e.g., Smartt et al. 2004; Li et al. 2007; Gal-Yam et al. 2007). Arguably, one of the most important results has been the identiﬁcation of red-supergiant stars with main-sequence masses in the range 9 ∼< M(M⊙) ∼< 17 (e.g., Smartt et al. 2009) as the progenitors of type II-P supernovae, the most common class of core-collapse supernovae (∼ 60% of all).

However, the dearth of progenitors of type II-P supernovae with initial main-sequence masses in the range ∼ 20 − 25 M⊙ (“red supergiant problem”; Smartt et al. 2009) was unexpected since red-supergiants with these masses are observed in

6 the stellar populations of our Galaxy and the Magellanic Clouds. This result might indicate that more massive stars die quietly forming a black-hole (e.g., Kochanek et al. 2008), that circumstellar dust is important at these masses hiding them from optical observations, or that the most massive red-supergiants lead to supernova types diﬀerent from type II-P (e.g., Smith et al. 2009a).

In the main part of this dissertation, we present the discovery of the progenitors of two supernova-like transients found in 2008 in the nearby galaxies NGC 6946 and NGC 300. The progenitors of these transients are unusual with respect to the progenitors of other supernovae found before. They were undetected in deep optical images, but bright in the mid-infrared, indicating the presence of circumstellar dust that absorbs the optical-UV light of the central star and re-radiates it thermally at mid-infrared wavelengths. Furthermore, the luminosities of the progenitors of

4 ≃ 5 × 10 L⊙ puts them in the range of main-sequence masses of ∼ 8 − 10 M⊙, the important transition region between low-mass and massive stars.

1.3. Scope of the Dissertation

The outline of this dissertation is as follows. In Chapter 2, we present a study of the metallicities of supernova host galaxies in the SDSS galaxy catalog and the implications of these ﬁndings for supernova progenitors. Chapter 2 has been

7 published in The Astrophysical Journal as J. L. Prieto et al. 2008, ApJ, v. 673, p. 999.

In Chapter 3, we present the discovery of the progenitor star of the low- luminosity SN 2008S in pre-explosion images obtained with the Spitzer Space

Telescope at mid-infrared wavelengths. Chapter 3 has been published in The

Astrophysical Journal Letters as J. L. Prieto et al. 2008, ApJL, v. 681, p. 9.

In Chapter 4, we present a comprehensive study of the new type of stellar transients, uncovered by SN 2008S and the transient in NGC 300, and their dusty progenitor stars. The rates of these transients and the mid-infrared properties of their progenitors compared to massive stellar populations in nearby galaxies allows us to put interesting constraints on the physical mechanism of the explosion and their importance for stellar evolution. We have submitted a paper (T. A. Thompson,

J. L. Prieto, et al. 2009) to The Astrophysical Journal with the analysis and results presented in Chapter 4.

In Chapter 5, we present a mid-infrared spectrum of the transient in NGC 300 obtained with the Spitzer Space Telescope and the analysis of the spectral energy distribution that reveal a substantial fraction of circumstellar dust that survives the explosion. The properties of this transient are consistent with the results presented in Chapter 4 on the dust-enshrouded progenitor star. Chapter 5 has been submitted for publication to The Astrophysical Journal as J. L. Prieto et al. 2009.

8 In Chapter 6, we present data and analysis of the light curves and spectra of the unusual supernova 2005gj, discovered by the SDSS-II Supernova Survey. We argue that this supernova was most likely the result of a thermonuclear explosion in a white dwarf with a relatively massive AGB star companion or a single massive

AGB star. The data and results of this chapter have been submitted for publication in The Astronomical Journal as J. L. Prieto et al. 2007.

Finally, in Chapter 7, we present the discovery of an eclipsing binary composed of two massive yellow-supergiant stars in the nearby galaxy Holmberg IX. The data for this work were obtained with the Large Binocular Telescope as part of a large project to study massive stars and supernova progenitors in a complete sample of nearby galaxies. Chapter 7 has been published in The Astrophysical Journal Letters as J. L. Prieto et al. 2008, ApJL, v. 673, p. 59.

9 Chapter 2

Characterizing Supernova Progenitors via the Metallicities of their Host Galaxies, from Poor Dwarfs to Rich Spirals

2.1. Introduction

On general grounds, it is thought that metallicity will affect the endpoints of stellar evolution, e.g., the relative outcomes in terms of different supernova types and the observed properties of each. Metals are a source of opacity that affects supernova progenitors (e.g., Kudritzki & Puls 2000) and also the supernova explosions themselves (e.g., Heger et al. 2003). However, the hypothesized metallicity effects have been rather difficult to measure directly. The number of supernova progenitors that have been identified directly from pre-explosion imaging is small and limited to core-collapse events (e.g., Hendry et al. 2006; Li et al. 2007). Previous works have either used population studies with only observational proxies for metallicity (e.g.,

Prantzos & Boissier 2003) or have considered direct metallicity measurements with only relatively small numbers of events (e.g., Hamuy et al. 2000; Gallagher et al.

2005; Stanek et al. 2006; Modjaz et al. 2008).

10 A new approach is now possible, which we employ in this paper, that takes advantage of the large sample of well-observed and typed supernovae. Due to a fortuitous match in coverage, many of these supernovae were in galaxies for which the Sloan Digital Sky Survey (SDSS) has identiﬁed the host galaxies and measured their oxygen abundances from emission lines in their spectra (Tremonti et al. 2004).

While these are central metallicities for the host galaxies, and are not measured for each supernova site, they are much more directly connected to the latter than proxies like the host luminosity. To further sharpen our tests, we compare the metallicity distributions of the host galaxies of SN Ib/c and SN Ia to those of SN II, which are taken as a control sample.

The progenitors of core-collapse supernovae (SN II and Ib/c) are massive stars, either single or in binaries, with initial main sequence masses ∼> 8 M⊙ (e.g., Heger et al. 2003). The presence of hydrogen in the spectra of SN II indicates that the massive envelopes are retained by the progenitors, of which red supergiants are probably the most common. However, SN Ib/c lack hydrogen (SN Ib) or both hydrogen and helium (SN Ic) in their spectra, and are therefore thought to have

Wolf-Rayet (WR) stars as progenitors (see Crowther 2007 for a review). The latter originate from the most massive stars, and have had their outer layers stripped oﬀ by strong winds. Thus SN Ib/c are thought to have main sequence masses ∼> 30 M⊙, which would make them ≃ (8/30)1.35 ≃ 20% of all core-collapse supernovae, assuming a Salpeter slope in the high-mass end of the initial mass function.

11 Based on theoretical considerations, the eﬀects of line-driven winds are expected to introduce a metallicity dependence in the minimum mass necessary to produce

WR stars (e.g., Heger et al. 2003; Eldridge & Tout 2004; Vink & de Koter 2005), which in turn can change the fractions of core-collapse supernovae that explode as SN II and SN Ib/c. Due to the relative frequencies, SN Ib/c will be more affected than SN II. These metallicity effects on the progenitor winds may strongly affect the rate at which radioactive 26Al is expelled into the interstellar medium before decaying (e.g., Prantzos 2004; Palacios et al. 2005), in which case the decays contribute to the observed diffuse 1.809 MeV gamma-ray line emission from the

Milky Way (e.g., Diehl et al. 2006). While 26Al appears to originate in massive stars, it is not yet known how much comes from the progenitors or the diﬀerent core-collapse supernova types (e.g., Prantzos & Diehl 1996; Higdon et al. 2004). For the most massive stars, GRB progenitors in the collapsar model (e.g., MacFadyen &

Woosley 1999, Yoon & Langer 2005), the interplay between metallicity-dependent mass loss through winds and rotation may be crucial (e.g., Hirschi et al. 2005). In all cases, binary progenitors may be more complicated (e.g., Eldridge 2007b).

Prantzos & Boissier (2003) used the absolute magnitudes of galaxies as a proxy for their average metallicities, from the luminosity-metallicity relationship, and found that the number ratio of SN Ib/c to SN II increases with metallicity; they argued that their result is consistent with stellar evolution models of massive stars with rotation (e.g., Meynet et al. 2006). If so, then one would expect a more robust

12 signature if the host metallicities were known directly. Ideally, in the latter approach, one would use the metallicities as measured from follow-up spectra obtained at the supernova sites, but this is difficult in practice. This approach of using measured as opposed to estimated metallicities was used by Stanek et al. (2006) (with compiled results from the literature) to study nearby long-duration GRBs with subsequent supernovae, finding that all of them had very low metallicity environments and that this appeared to be key to forming powerful GRB jets, and by Modjaz et al. (2008) to study nearby broad-lined SN Ic (without GRBs), finding in contrast that the metallicities of these environments were much higher. The main caveats associated with these results are the low statistics, five and twelve events, respectively. We try to combine the virtues of these two approaches, with higher statistics and mostly direct metallicity measurements.

The likely progenitors of SN Ia are white dwarfs, forming from stars with initial main-sequence masses ∼< 8 M⊙, which accrete mass from a companion

(single-degenerate model) until they reach the Chandrasekhar mass (≃ 1.4 M⊙) and produce a thermonuclear explosion that completely disrupts the star (e.g., Whelan

& Iben 1973). During the accretion process, white dwarfs could have strong winds when the accretion rate reaches a critical value (e.g., Hachisu et al. 1996), which would allow them to burn hydrogen steadily and grow in mass. At low metallicities

([Fe/H] ∼< −1), SN Ia may be inhibited through the single-degenerate channel

(Kobayashi et al. 1998), as the white dwarf wind is thought to be weak and the

13 system passes through a common envelope phase before reaching the Chandrasekhar mass. Metallicity also affects the CNO abundances of white dwarfs, which can affect the production of 56Ni in the explosion, and therefore the peak luminosities of SN Ia (e.g., Umeda et al. 1999; Höflich et al. 2000; Timmes et al. 2003; Röpke et al. 2006). Studies of the integrated metallicities of nearby SN Ia hosts (Hamuy et al. 2000; Gallagher et al. 2005) have shown that metallicity does not seem to be the main factor regulating their peak luminosities, which is consistent with some theoretical models (e.g., Podsiadlowski et al. 2006). Instead, the age of the stellar population where SN Ia progenitors originate seems to be very important: prompt

(SN Ia explode ∼ 108 yr after star formation) and delayed (SN Ia explode > 109 yr after star formation) components were suggested to explain the high rates of SN Ia in actively star-forming galaxies (late type spirals and irregulars) compared with

SN Ia in old, elliptical galaxies (e.g., Mannucci et al. 2005; Scannapieco & Bildsten

2005; Neill et al. 2006).

In this work, to our knowledge for the ﬁrst time, we compare the directly measured oxygen abundances of the hosts of SN Ib/c and SN Ia with SN II. We use the Sternberg Astronomical Institute (SAI) supernova catalog and match it with the

SDSS-DR4 catalog of oxygen abundances of a large sample of star-forming galaxies from SDSS. Using the supernova classiﬁcations presented in the literature, we can separate the sample according to diﬀerent supernova types and make statistical comparisons of the metallicity distributions of their host galaxies. We also investigate

14 some individual cases in metal-poor environments that are especially interesting and which can be used to test the strong predictions made by some theoretical models.

We create a second catalog by matching the positions of all supernovae with images from SDSS-DR6, independent of the host galaxy association. This allows us to investigate signiﬁcantly fainter SNe hosts, and we identify some even more extreme hosts for follow-up observations. To enable their further use in other studies, we make both catalogs available online, and will update them regularly.

2.2. First Catalog: Supernova-Host Pairs with Known

Host Metallicities (SAI ∩ SDSS-DR4)

We use the SAI supernova catalog1 (Tsvetkov et al. 2004) to obtain the main properties of supernovae (name, classification, RA, DEC, redshift) and their host information when available (galaxy name, RA, DEC, redshift). The SAI catalog is a compilation of information about supernova discoveries, obtained mainly from reports in the International Astronomical Union Circulars (IAUC), which include the coordinates and classification of the supernovae from the IAUCs and also basic information about the host galaxies in the cases where the galaxies can be identified in online galaxy catalogs (e.g., HyperLEDA, NED and SDSS). The version of the catalog we use contains 4,169 entries2, of which we have selected 3,050 supernovae

1 http://www.sai.msu.su/sn/sncat/ 2Version updated on June 15, 2007.

15 discovered between 1909 and 2007 classified as SN Ia, II, and Ib/c, including their sub-types. Supernovae in the catalog with no classification or only classified as

Type I are not considered for further analysis since we want to be able to distinguish between SN Ia and the core-collapse types SN Ib/c.

Tremonti et al. (2004) determined metallicities for a sample of star-forming galaxies in the SDSS Data Release 2 (SDSS-DR2; Abazajian et al. 2004) from their spectra. Here we use a larger sample of 141,317 star-forming galaxies (excluding

AGN) from the SDSS-DR4 (Adelman-McCarthy et al. 2006), with metallicities derived in the same consistent fashion, and which are available online3. The metallicities are derived by a likelihood analysis which compares multiple nebular emission lines ([O II], Hβ, [O III], Hα, [N II], [S II]) to the predictions of the hybrid stellar-population plus photoionization models of Charlot & Longhetti (2001). A particular combination of nebular emission line ratios arises from a model galaxy that is characterized by a galaxy-averaged metallicity, ionization parameter, dust-to-metal ratio, and 5500A˚ dust attenuation. For each galaxy, a likelihood distribution for metallicity is constructed by comparison to a large library of model galaxies. We use the median of the oxygen abundance distributions in this paper. The metallicities derived by Tremonti et al. (2004) are essentially on the Kewley & Dopita (2002) abundance scale (∆[12 + log (O/H)] < 0.05 dex; Ellison 2006). For further reference in this paper, we call this galaxy metallicity catalog SDSS-DR4.

3 http://www.mpa-garching.mpg.de/SDSS/DR4/

16 We restrict the initial sample of galaxies to 125,958 by applying two of the cuts that Tremonti et al. (2004) used for their ﬁnal cleaned sample: (1) the redshifts of the galaxies have to be reliable by SDSS standards; and (2) Hβ,Hα, and

[N II] λ6584 should be detected at > 5σ conﬁdence, and [S II] λλ6717,6731 and

[O III] λ5007 should at least have detections. While in our analysis we directly compare nebular oxygen abundance within the SDSS-DR4 catalog for the supernova hosts, when referring to “Solar metallicity,” we adopt the Solar oxygen abundance of 12 + log (O/H) = 8.86 (Delahaye & Pinsonneault 2006).

We cross-matched the SAI catalog with the galaxy metallicity catalog

SDSS-DR4 using a matching radius of 60′′ (∼ 48 kpc at z = 0.04). We used the coordinates of the host galaxies in the cases where they are known and identiﬁed in the SAI catalog, and the supernovae coordinates were used otherwise. We also required that the redshifts reported in the SAI catalog, which were taken from galaxy catalogs and the IAUCs, to be consistent within 20% with the redshifts of the closest galaxy from the SDSS catalog that passed the proximity cut. After selecting the supernovae that passed the proximity and redshift criteria, we visually inspected the SDSS images around the galaxies to identify the ones that were wrongly selected as hosts (e.g., close galaxy pairs). The number of supernovae that passed all these cuts is 254 in total: 95 SN Ia, 123 SN II, and 36 SN Ib/c. There were some galaxies that hosted more than one supernova: ﬁve galaxies had three supernovae each

(NGC 1084, NGC 3627, NGC 3631, NGC 3938, and NGC 5457) and 15 galaxies

17 had two supernovae (NGC 2532, NGC 2608, NGC 3627, NGC 3780, NGC 3811,

NGC 3913, NGC 4012, NGC 4568, NGC 5584, NGC 5630, NGC 6962, UGC 4132,

UGC 5695, IC 4229, and MCG +07-34-134).

In Table 2.1 we present the ﬁnal matched sample of supernovae and host galaxy metallicities from SDSS-DR4, as well as the absolute MB magnitudes of the galaxies obtained from the HyperLEDA database and SDSS. The absolute magnitudes for SDSS galaxies were calculated using Petrosian gr magnitudes transformed to

B magnitudes using the transformation of Lupton (2005), corrected by Galactic extinction (Schlegel et al. 1998) and internal extinction to a face-on geometry (Tully et al. 1998), and k-corrections (Blanton et al. 2003). To calculate the absolute

−1 −1 magnitudes, we use a ﬂat cosmology with H0 = 70 km s Mpc , ΩM = 0.3,

ΩΛ = 0.7. The typical 1σ uncertainties in the oxygen abundances are 0.05 dex at 12 + log (O/H) > 8.5, and 0.15 dex at 12 + log (O/H) < 8.5. Our estimated uncertainty in the absolute magnitudes of the hosts is ∼ 0.3 mag, calculated from a sub-sample of galaxies in the catalog with reliable absolute magnitudes from SDSS and HyperLEDA.

Our ﬁrst catalog, SAI ∩ SDSS-DR4, is available online4 and will be updated as new supernovae are discovered with host galaxy metallicities in the SDSS-DR4 catalog. It includes the information presented in Table 2.1, as well as images around the supernovae obtained from SDSS-DR6.

4 http://www.astronomy.ohio-state.edu/∼prieto/snhosts/

18 Figure 2.1 shows the distribution of metallicities as a function of redshift and MB of the supernova host galaxies, as well as the distribution of star-forming galaxies in the SDSS-DR4 catalog. The apparent “stripes” in the plots, regions with very few oxygen abundance measurements, are an effect of the grid of model parameters (metallicity, ionization parameters, attenuation, etc.) used to calculate the metallicities (see Brinchmann et al. 2004 for details). As can be seen, the redshift distribution of supernovae varies for different types, with the median redshifts of the samples at z = 0.014 (II), 0.018 (Ib/c), and 0.031 (Ia). This is a combination of several effects. First, SN Ia supernovae are, on average, ∼ 2 mag brighter at peak luminosity than core-collapse events (Richardson et al. 2002), therefore, they can be found at larger distances in magnitude-limited surveys. Second, the local rate of core-collapse supernovae in late-type galaxies is ∼ 3 times higher than the SN Ia rate

(Cappellaro et al. 1999; Mannucci et al. 2005). Finally, the great interest in SN Ia as cosmological distance indicators makes most of the supernovae searches concentrate their limited spectroscopic follow-up resources on likely Type Ia supernovae (as determined by their light curves).

As shown in Figure 2.1, the distribution of host galaxy metallicities follows the distribution of galaxies from SDSS, with a wide range spanning ∼ 1.4 dex

(7.9 < 12 + log (O/H) < 9.3). However, there appear to be significant differences present between the hosts of different supernovae types. In particular, most of the SN Ib/c hosts are concentrated in the higher metallicity/luminosity end of the

19 distribution (12 + log (O/H) ∼> 8.7), while the metallicities of SN II and SN Ia hosts are more evenly distributed and appear to be tracing each other fairly well.

Figure 2.2 shows a mosaic of SDSS-DR6 (Adelman-McCarthy et al. 2008) images5 of the host galaxies with the highest and lowest metallicities in the sample, including two supernovae of each type. This ﬁgure shows the wide range of host galaxy environments present in the sample, from big spirals (e.g., SN 2000dq,

SN 2004cc, SN 2005bc, SN 2005mb, SN 2002cg, and SN 2006iq) to small dwarfs (e.g.,

SN 1997bo, SN 2006jc, SN 2004hw, SN 1998bv, SN 2007I, and SN 2007bk), and that all types of supernovae can be found in metal-rich and metal-poor star-forming galaxies.

2.2.1. Testing Supernova Trends with Metallicity

Is the tendency of SN Ib/c hosts towards higher metallicity, compared with

SN II and SN Ia, clearly seen in Figure 2.1, a real physical eﬀect? To answer this question we identify and try to reduce some of the biases present in the sample.

The supernova sample studied in this work is far from homogeneous. The supernovae have been discovered by a variety of supernova surveys, including amateur searches that look repeatedly at bright galaxies in the local universe, professional searches that look at a number of cataloged galaxies to a certain

5 http://cas.sdss.org/dr6/en/tools/chart/chart.asp

20 magnitude limit (e.g., LOSS), and professional searches that look at all the galaxies in a given volume (e.g., SDSS-II, The Supernova Factory), among others. The host galaxies of supernovae discovered by amateur searches tend to have higher metallicities due to the luminosity-metallicity relation (see Figure 2.1), while the metallicities of galaxies observed by professional searches span a wider range.

As an example of a possible bias in the supernovae in our catalog, we note that the median metallicity decreases by ∼ 0.1 dex for the hosts of supernovae discovered between 1970 and 2007. Ideally, all the supernovae for the current study would be selected from galaxy-impartial surveys. However, the numbers of diﬀerent supernova types found by such surveys in our catalog are still small (especially core-collapse events), and do not allow a statistical comparison (see the discussion in Modjaz et al. 2008).

Another bias present in the galaxy data that we use is the so-called aperture bias (Kewley et al. 2005; Ellison 2006). The SDSS spectra are taken with a ﬁxed

fiber aperture of 3′′ (2.4 kpc at z = 0.04). Since galaxies have radial metallicity gradients (e.g., Zaritsky et al. 1994), for nearby galaxies we are, on average, only measuring the higher central metallicity, while for more distant galaxies we are covering a larger fraction of the galaxy light with the SDSS fiber. This effect also depends on galaxy luminosity, as for dwarf galaxies the fiber will cover a larger fraction of the total light than in large spirals. Kewley et al. (2005) find a mean difference of ∼ 0.1 dex, although with a large scatter (0.1 − 0.3 dex), between the

21 central and integrated metallicities of a sample of ∼ 100 galaxies of all types (S0-Im) in the redshift range z = 0.005 − 0.014.

In order to reduce these and other biases, we limit the comparison of supernova types to host galaxies in the redshift range 0.01

SN II (0.020), 19 SN Ib/c (0.021) and 38 SN Ia (0.024) hosts are consistent, while the number of galaxies in each sub-sample still allows us to make a meaningful statistical comparison. By using a small redshift slice we are, eﬀectively, reducing the aperture biases when comparing the galaxy metallicity measurements, such that they are now comparable to or smaller than the statistical errors in the metallicity determination.

We made additional checks of relative biases between supernova types in our redshift-limited sample. First, the ratios of the numbers of SN Ib/c and SN Ia to the total number of core-collapse supernovae are reasonably consistent with the ratios obtained from the local supernovae rates (e.g., Cappellaro et al. 1999; Mannucci et al. 2005). Second, the fact that the SN–host separation distributions for SN Ia and SN II agree, particularly at small radii (see below), suggests that our supernova samples are not biased (relatively, one supernova type to another) by obscuration eﬀects.

We compare the metallicity distributions of the hosts of SN Ib/c and SN Ia to

SN II, which are taken as the control sample. Given that SN II are the most common

22 type of supernovae (e.g., Mannucci et al. 2005) and that they come from massive stars from a wide range of masses that explode in all environments, presumably independent of metallicity, they are eﬀectively giving us the star-formation-rate weighting of the luminosity-metallicity (or mass-metallicity) relationship for star-forming galaxies. It would be of interest to test if indeed the SN II rates are independent of metallicity, but this is outside the scope of the current paper.

Figure 2.3 shows the cumulative distribution of metallicities for hosts of diﬀerent supernova types in the redshift ranges z < 0.04 and 0.01

• The metallicities of SN Ib/c hosts tend to be higher than those of SN II hosts,

• The SN Ia and SN II hosts have very similar metallicity distributions.

Kolmogorov-Smirnov (KS) tests between the metallicity distributions of diﬀerent supernova types in the redshift range 0.01

ﬁndings. The KS probabilities of two host metallicity samples being drawn from the same distribution are: 5% (II–Ib/c), 3% (Ia–Ib/c) and 56% (II–Ia). We obtain a similar result if we compare the mean metallicities of the host samples: 8.94±0.04

(SN II), 8.94±0.04 (SN Ia) and 9.06±0.04 (SN Ib/c), where the errors are the RMS of similar-sized samples obtained using bootstrap resampling.

23 The metallicity distribution of the SDSS-DR4 star-forming galaxies in our redshift range, weighted only by galaxy counts, is also shown in Figure 2.3. This should not be used in any comparisons, as it does not take into account the weighting with star formation rate or the supernova and galaxy selection criteria. We take all of these into account by only making relative comparisons between supernova types.

If we restrict the sample of SN Ib/c to only SN Ic and broad-lined Ic in the same redshift range, leaving out supernovae classified as SN Ib/c and SN Ib, the difference in metallicity distributions of the hosts of SN II and SN Ic+hypernovae (13 SN) becomes smaller, with a KS probability of 19%. If only the three supernova classified as SN Ib (SN 2003I, SN 2005O, and SN 2005hl) in the pseudo volume-limited sample are not considered, then the KS probability of SN II and SN Ic+hypernovae+SNIb/c being drawn from the same sample is 10%.

In Figure 2.4 we show the number ratio of SN Ib/c to SN II as a function of the metallicities of the host galaxies. This ratio is very important because the rates of core-collapse SNe are expected to change as a function of the progenitor mass and metallicity and, therefore, it can help to put constrains on massive stellar evolution models (e.g., Eldridge 2007b). We have calculated the ratio in bins of equal number of SN II+SN Ib/c, with 11 SNe per bin, to do a direct comparison with the results of Prantzos & Boissier (2003). The small statistics compared with

Prantzos & Boissier (2003), who used the absolute magnitudes of the hosts as a proxy for the average metallities through the luminosity-metallicity relationship,

24 is reﬂected in the large errors of the ratio. The large error bars do not allow us to put constraints in progenitor models, however, the general trend observed in the cumulative distribution (see Figure 2.3) is conﬁrmed with the number counts:

SN Ib/c are more common at higher metallicities compared with SN II. Our results are consistent with those of Prantzos & Boissier (2003).

Figure 2.5 shows the cumulative distributions of projected host-supernova distances for the reduced sample of 115 SNe used to compare the host metallicities

(0.01

Tsvetkov et al. 2004; Petrosian et al. 2005), which have similar distributions to each other (as also in Figure 2.3). A galactocentric concentration of SN Ib/c and their progenitors may be important for the angular distributions of diﬀuse gamma-ray line emission from the Milky Way. Besides the 1.809 MeV line from 26Al, the 0.511

MeV line from positron annihilation is poorly understood, in terms of its high flux and very strong central concentration (e.g., Casséet al. 2004; Knödlseder et al.

2005; Beacom & Y¨uksel 2006). Since the SN Ib/c are found at small separation, the central galaxy metallicities determined by the SDSS should be representative of the local environments of the supernovae. Taking into account the existence of negative metallicity gradients in increasing galactocentric radii, the local metallicities of the

SN II and SN Ia, if anything, are even lower than deduced from the SDSS central

25 metallicities. The tendency for SN Ib/c to prefer higher metallicity relative to SN II and SN Ia is probably even stronger than shown in Figure 2.3.

2.2.2. Supernovae in Low-Metallicity Hosts

Even though we have shown that there is a strong preference of SN Ib/c for high-metallicity environments, compared with SN II and SN Ia, there are four

SN Ib/c with relatively metal-poor host galaxies (12 + log (O/H) < 8.6). These events, and also some SN Ia found in low-metallicity dwarfs, made us investigate more carefully a number of individual cases. We found that among the lowest-metallicity host galaxies in the sample, there were supernovae that stood out because of their unusual properties (all shown in Figure 2.2).

SN 2006jc: Peculiar SN Ib/c supernova with strong He I lines in emission in

the spectrum (Crotts et al. 2006), thought to arise from the interaction

of the supernova ejecta with a He-rich circumstellar medium (Foley et al.

2007; Pastorello et al. 2007). Its host galaxy, UGC 4904 at z = 0.006, is a

low luminosity, blue, and relatively low-metallicity starburst (MB = −16.0,

12 + log (O/H) = 8.5). Interestingly, the host galaxy of SN 2002ao at z = 0.005

(in UGC 9299), another peculiar SN Ib/c with spectral properties very similar

to SN 2006jc (Benetti et al. 2006a) that is also present in our ﬁrst catalog, has

26 low metallicity compared with the majority of the SN Ib/c hosts, and shares

similar morphological properties with the host of SN 2006jc.

SN 2007I: Broad-lined Ic (or hypernova) with a spectrum similar to SN 1997ef

(Blondin et al. 2007a) at z = 0.022. Its host galaxy is a star-forming,

low-metallicity dwarf (MB = −16.7, 12 + log (O/H) = 8.3), unlike other

broad-lined Ic supernovae observed in higher-metallicity galaxies (Modjaz et al.

2008), and somewhat similar to the host galaxies of long GRBs associated with

supernovae (Stanek et al. 2006; Fruchter et al. 2006); however, see a detailed

discussion in Modjaz et al. (2008). The other four broad-lined Ic supernovae in

our sample that have been reported in the literature are: SN 1997ef (Iwamoto

et al. 2000), 2004ib (Sako et al. 2005), 2005ks (Frieman et al. 2005), 2006qk

(Frieman et al. 2006).

SN 2007bk: Type Ia supernova with a spectrum similar to the slow

decliner/luminous SN 1991T (Dennefeld et al. 2007) at z = 0.032. The

host galaxy is a low metallicity/luminosity dwarf, with MB = −18.2 and

12 + log (O/H) = 8.3, similar to the Large Magellanic Cloud. The supernova

was found very far from the center of its dwarf host, at a projected separation

of ∼ 9 kpc. The magnitude of the supernova at discovery (R = 16.7, Mikuz

2007) and the phase obtained from the spectrum (+50 days, Dennefeld

et al. 2007, although S. Blondin ﬁnds equally good matches with templates

at +30 days, private communication), imply that this was a very luminous

27 Type Ia event. If the reported discovery magnitude and spectral phases are

accurate, SN 2007bk was ∼ 0.5 − 1.5 mag brighter than SN 1991T at the same

phase after maximum light.

2.3. Second Catalog: Supernova-Host Pairs with

Unknown Host Metallicities (SAI ∩ SDSS-DR6)

The existence of supernovae with unusual properties among the most metal- poor, low-luminosity galaxies in the ﬁrst catalog prompted us to investigate a much larger sample of supernovae. We constructed a second catalog with images around the positions of supernovae using SDSS, matching the SAI catalog with SDSS-DR6.

We included the redshifts obtained from the SAI catalog to produce images in physical units around the supernovae. The total number of matches is 1225 for supernovae at z < 0.3. This catalog is also available online, with the ﬁrst catalog described earlier in §2.2.

This extended second catalog (SAI ∩ SDSS-DR6) does not have information about metallicities or luminosities of the hosts. It is a visual tool that can be used to explore the environments around supernovae found in the SDSS area, independent of the host galaxy association. Identiﬁcation of the supernovae hosts and their integrated properties obtained from SDSS will be added in a future study.

28 Visually inspecting the images of the second catalog, we identiﬁed a number of supernovae in what appear to be very faint galaxies, and which are likely to be low-luminosity, metal-poor galaxies not present in the ﬁrst catalog. Some examples in the redshift range 0.01 ∼< z ∼< 0.05 are (supernova types shown in parentheses):

SN 1997ab (II), SN 1997az (II), SN 2001bk (II), SN 2003cv (II), SN 2004gy (II),

SN 2004hx (II), SN 2005cg (Ia), SN 2005gi (II), SN 2005hm (Ib), SN 2005lb

(II), SN 2006L (IIn), SN 2006bx (II), SN 2006fg (II), 2007bg (Ic), 2007bu (II), and 2007ce (Ic). In this incomplete sample, which was selected by noting some especially low-luminosity galaxies, the SN Ia/SN II ratio is lower than expected.

Similarly, in our catalog of hosts with known metallicities, SN Ia may also be relatively underabundant at the lowest host luminosities and metallicities, as shown in Figure 2.1. We caution that the small statistics make these only hints, and we discuss these issues further below.

One of the most interesting supernovae in our second catalog is SN 2007bg, a recently discovered broad-lined Ic (Quimby et al. 2007; Harutyunyan et al. 2007) at z = 0.03, which has an extremely faint galaxy coincident with the position of the supernova. Using photometry and images from SDSS-DR6, we estimate the luminosity of the apparent host galaxy to be MB ≈ −12, most likely a very metal- poor dwarf (12 + log (O/H) ∼ 7.5, or ∼ 1/20 solar; see the metallicity-luminosity relationship extended to dwarf galaxies by Lee et al. 2006). Due to the extremely low luminosity of that galaxy, in fact one of the lowest-luminosity supernova hosts

29 ever seen, and also fainter than most if not all GRB hosts (see e.g., Fruchter et al.

2006), this event may represent the missing link between broad-lined SN Ic and

GRBs. This event is therefore an excellent candidate for a search for an oﬀ-axis

GRB jet in radio (Soderberg et al. 2006) and possibly other wavelengths. Follow-up spectroscopic observations and deep photometry to determine the metallicity of the host and study the supernova environment are strongly encouraged in this and other cases of very low-luminosity SN hosts.

2.4. Discussion and Conclusions

We ﬁnd that SN Ib/c tend to be in high-metallicity host galaxies, compared to

SN II, our control sample that traces the underlying star formation rates. This is the first time that such a trend has been found using the directly-measured oxygen abundances of the supernova host galaxies. This confirms and greatly strengthens an earlier result of Prantzos & Boissier (2003), which found a similar result using the absolute magnitudes of the host galaxies as an indirect estimate of their metallicities through the luminosity-metallicity relationship. This can be interpreted in relative supernova rates: the ratio of SN Ib/c to SN II increases with increasing metallicity and hence also cosmic age. We also find that SN Ib/c are consistently found towards the centers of their hosts compared with SN II and SN Ia, which had been also found in previous studies (e.g., van den Bergh 1997; Tsvetkov et al. 2004; Petrosian et al.

30 2005). This suggests that direct measurements of metallicities at the explosion sites, as opposed to the central host metallicities used here, would reveal an even stronger eﬀect, due to the radial metallicity gradients observed in spiral galaxies. The local metallicities of SN Ib/c would be less reduced with respect to the central metallicities than SN II and SN Ia, which would widen the separation seen in Figure 2.3.

The tendency towards high metallicity of SN Ib/c environments compared to those of SN II supports, in general terms, theoretical models of the eﬀects of metallicity in stellar evolution and the massive stars that are core-collapse supernova progenitors (e.g., Heger et al. 2003; Meynet et al. 2006; Eldridge 2007b; Fryer et al.

2007). Also, models of stellar evolution that include rotation, from Meynet et al.

(2006), predict that at high metallicity Wolf-Rayet stars will earlier enter the WC phase when they still are rich in helium, and that these stars would explode as

SN Ib. The fact that we do see both SN Ib and SN Ic in hosts at high metallicity should not be taken as inconsistent with these models, mainly because the number of supernovae is small and the sample has not been homogeneously selected. There is an indication, although not statistically signiﬁcant, that SN Ib may be more common in higher metallicity environments than SN Ic and broad-lined SN Ic in our sample.

The agreement between the metallicity distributions of the hosts of SN II and

SN Ia shows that their hosts are sampling a wide range of properties of star-forming galaxies, from the relatively metal-poor dwarfs to metal-rich grand design spirals.

Using models of white dwarf winds in the framework of single-degenerate progenitors

31 of SN Ia (Hachisu et al. 1996), Kobayashi et al. (1998) made a strong prediction that SN Ia would not be found in low metallicity environments, such as dwarf galaxies and the outskirts of spiral galaxies. However, we do observe SN Ia in metal poor dwarfs (e.g., SN 2004hw, SN 2006oa, and SN 2007bk, with host metallicities between ∼ 0.2 and 0.5 solar) and at large projected distances (> 10 kpc) from their star-forming hosts (e.g., SN 1988Y, SN 1992B, SN 1993I, SN 2001bg, SN 2002gf,

SN 2004ia, SN 2004ig, SN 2005ms, SN 2006ﬁ, and SN 2006gl). There are also extreme cases that have been pointed out in previous studies, like the low-luminosity dwarf (MB ≈ −12.2) host galaxy of the luminous and slow decliner SN 1999aw

(Strolger et al. 2002), which is most likely very metal-poor (12 + log (O/H) ∼ 7.5, or ∼ 1/20 solar; see Lee et al. 2006). Also, SN 2005cg was found in a dwarf with subsolar gas metallicity (Quimby et al. 2006).

We do not ﬁnd a statistically signiﬁcant low-metallicity threshold in the metallicities of SN Ia compared with SN II hosts, as predicted from theory by

Kobayashi et al. (1998) for single-degenerate progenitors of SN Ia with winds.

However, there is a preference for ﬁnding more SN II in very faint galaxies compared with SN Ia in our second catalog, which is suggestive of a luminosity or metallicity threshold for the main channel that produces SN Ia. This will have to be explored in the future with a larger sample that includes good luminosity information for the hosts and actual metallicities measured from spectra. If no metallicity threshold is found in larger samples, it means that the models and predictions of white

32 dwarf winds will have to be revisited. This would have implications for modeling and understanding of galactic chemical evolution that include the eﬀects of white dwarf winds to shut down SN Ia at low metallicities (e.g., Kobayashi et al. 2007).

Interestingly, modeling the X-ray spectra of supernova remnants from probable

SN Ia explosions in our Galaxy, the LMC and M31, Badenes et al. (2007) did not

ﬁnd the strong eﬀects of white dwarf winds predicted from theory.

On the other hand, independent of the existence (or not) of a mechanism that can shut down the production of SN Ia in low-metallicity environments, we have noted examples of SN Ia that explode in low-metallicity dwarf galaxies, like

SN 2007bk. Also, supernova remnants from probable SN Ia have been identiﬁed in the LMC (e.g., Hughes et al. 1995) and SMC (e.g., van der Heyden et al. 2004). Is this

SN Ia population dominated by a diﬀerent kind of progenitors, like double-degenerate mergers, compared to the main progenitor channel? Is the expected trend between progenitor metallicities and peak-luminosity starting to appear as we extend the sample to even lower metallicity hosts? It is suggestive that the small number of

SN Ia in low-metallicity hosts, like SN 2007bk, SN 2005cg and SN 1999aw, were all luminous events compared with normal SN Ia. Also, the very luminous SN Ia events that have spectral signatures of a strong ejecta-CSM interaction, like SN 2005gj, are mostly associated with low-luminosity, and most likely low-metallicity, hosts (Prieto et al. 2007). Is low metallicity necessary to produce this extreme class of SN Ia?

33 Detailed comparison studies of the observational properties of supernovae in these extreme environments are encouraged.

In the course of this work, we have prepared two new catalogs that should be useful for other studies. We used the SAI supernova catalog and the SDSS-DR4 sample of metallicities of star-forming galaxies from Tremonti et al. (2004) to produce a catalog of supernovae hosts with metallicities derived in a consistent fashion.

From this ﬁrst catalog, we found several interesting core-collapse (e.g., SN 2002ao,

SN 2006jc, and SN 2007I) and SN Ia events (e.g., SN 2007bk) in low-metallicity galaxies. We constructed a second catalog by matching the SAI supernova catalog with images obtained from SDSS-DR6. The second catalog does not contain information on host metallicities, but it can be used to investigate the environments of supernovae independent of the host association. In that second catalog, we found several examples of core-collapse supernovae in faint galaxies. One of most interesting cases is SN 2007bg, a broad-lined SN Ic in an extremely low-luminosity and very likely low-metallicity host. These catalogs will allow researchers to select interesting candidates for further follow-up observations. Also, as more homogeneous light curve and spectroscopic data become available for supernovae in the ﬁrst catalog, this will allow us to test possible correlations between supernova properties and the metallicities of their hosts, which may turn out to be crucial for improving our understanding of the nature of diﬀerent supernova explosions. Another possible

34 use of our catalogs is for systematically characterizing the morphologies of supernova hosts.

We stress the great importance of galaxy-impartial surveys for ﬁnding and studying the properties of all supernovae types. Some very interesting and potentially informative supernovae have been found in very low-luminosity, low-metallicity galaxies, hosts which are not included in supernova surveys based on catalogs of normal galaxies. These unusual supernovae and hosts may help probe the relationship between the SN Ib/c and SN II core-collapse supernova types, the progenitors of SN Ia as well as the possible correlations between observed SN Ia properties and host metallicities, the supernova-GRB connection (e.g., Stanek et al.

2003) and its possible metallicity dependence (e.g., Stanek et al. 2006; Modjaz et al.

2008), and also to test the consistency between the cosmic stellar birth and death rates (e.g., Hopkins & Beacom 2006). As we pointed out in §2.2.1, presently the comparison of host metallicities using supernovae discovered by galaxy-impartial surveys is limited by their small numbers, especially for core-collapse events, since

SN Ia receive much more attention when decisions about spectroscopic follow-up are made. This is also true for the study of their observational properties (e.g., light curves and spectra). However, in order to better understand all types of cosmic explosions and put further constraints on the predictions of stellar evolution theory, a larger eﬀort on other supernovae types is greatly needed.

35 Fig. 2.1.— Metallicities of supernova host galaxies from SDSS-DR4 as a function of redshift and absolute B magnitude. The symbols distinguish diﬀerent supernova types: SN II (triangles), SN Ib/c (squares) and SN Ia (circles). The dots in the left panel are 125,958 star forming galaxies in SDSS-DR4 with reliable metallicity and redshift measurements. The dots in the right panel are a subsample of 86,914 star forming galaxies (z > 0.005) selected from the main SDSS-DR4 galaxy sample.

36 II Ib/c Ia SDSSJ020820.18+011103.5 NGC4568 NGC5698 SN2000dq (II) SN2004cc (Ic) SN2005bc (Ia)

NGC4963 UGC10415 SDSSJ213933.98+102900 SN2005mb (II) SN2002cg (Ic) SN2006iq (Ia)

SDSSJ121952.54+074352.3 UGC04904 SDSSJ235117.28+011339.7 SN1997bo (II) SN2006jc (Ib/c) SN2004hw (Ia)

SDSSJ103825.73+474236.7 SDSSJ115913.13−013616.2 SDSSJ152845.00+585200.1 SN1998bv (II) SN2007I (Ic) SN2007bk (Ia)

Fig. 2.2.— SDSS color images of the most metal rich (top six images) and most metal poor (bottom six images) host galaxies in our sample. We show two galaxies of each supernova type. The images are centered on the position of the supernova explosion, marked with a cross, with North-up and East-left. They have the same physical size of 30 kpc at the distance of each galaxy.

37 Fig. 2.3.— Cumulative fraction with the oxygen abundance of the supernova host galaxies. The number of host galaxies of each supernova type is indicated in each panel and the lines correspond to: SN II (solid), SN Ib/c (dot-dashed) and SN Ia (dotted). The left panel includes host galaxies with redshifts z < 0.04, while the right panel includes host galaxies with redshifts 0.01

38 Fig. 2.4.— Number ratio of SN Ib/c to SN II as a function of metallicity of the host galaxies. The open circles are the values obtained with our sample from directly measured central metallicities, and the ﬁlled squares are the results from Prantzos & Boissier (2003) using absolute magnitudes as a proxy to host metallicities. The error bars are obtained from Poisson statistics. The solid line shows the predicted ratio from the binary models of Eldridge (2007b); the dashed line is from the models of single stars with rotation of Maeder & Meynet (2004); and the long-dashed line is from the single star models of Eldridge (2007b).

39 Fig. 2.5.— Cumulative fraction plot of the projected separation between the supernova and its host for the reduced sample in the redshift range 0.01

40 a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

1909A II: 210.5129 54.4661 NGC5457 . . . 0.00082 –21.03 9.12 1920A II: 128.8156 28.4754 NGC2608 3.6 0.00727 –20.21 9.12 1936A IIL 184.9830 5.3522 NGC4273 6.0 0.00783 –20.57 9.14 1951H II: 210.9804 54.3614 NGC5457 . . . 0.00082 –21.03 9.12 1961U IIL 178.2366 44.1504 NGC3938 . . . 0.00263 –20.01 9.20 1963J Ia 177.6596 55.3504 NGC3913 . . . 0.00318 –18.00 9.01 1963P Ia 41.5088 –7.5807 NGC1084 . . . 0.00436 –20.58 8.89 1964A IIPec 170.2256 53.1945 NGC3631 . . . 0.00387 –21.00 9.20 1964L Ib 178.2046 44.1293 NGC3938 . . . 0.00263 –20.01 9.20

41 1965L II 170.2423 53.1851 NGC3631 . . . 0.00387 –21.00 9.20 1966B IIL 191.9371 4.3256 NGC4688 . . . 0.00325 –17.64 8.77 1966E II: 183.4414 13.4149 NGC4189 6.8 0.00702 –20.20 9.14 1967C Ia 162.1044 12.5453 NGC3373 . . . 0.00433 –19.67 9.02 1969C Ia 175.3243 47.6925 NGC3811 3.1 0.01057 –20.76 9.06 1970G IIL 210.7535 54.2424 NGC5457 . . . 0.00082 –21.03 9.12 1971G Ia 183.0500 13.2378 NGC4165 4.8 0.00624 –18.33 8.94 1971I Ia 198.9567 41.9875 NGC5055 . . . 0.00131 –21.20 9.06 1971K II: 175.3082 47.6861 NGC3811 7.3 0.01057 –20.76 9.06 1971S II: 20.5505 0.9557 NGC0493 8.5 0.00789 –20.65 8.83 1973R IIP 170.0481 12.9987 NGC3627 . . . 0.00193 –21.17 9.16

(cont’d) Table 2.1. Supernova and host galaxy data Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

1976B Ib: 186.5508 13.1115 NGC4402 . . . 0.00079 –18.73 9.05 1978H II 174.8486 56.2669 NGC3780 3.2 0.00841 –20.97 9.16 1979B Ia 177.6808 55.3592 NGC3913 . . . 0.00318 –18.00 9.01 1983W Ia 170.1356 57.7820 NGC3625 1.4 0.00647 –19.60 8.97 1985G IIP 187.1683 9.2611 NGC4451 . . . 0.00258 –17.69 9.06 1985H II 161.6339 63.2304 NGC3359 . . . 0.00339 –20.53 8.74 1987C IIn 127.5056 52.6926 UGC04438 6.2 0.01415 –20.47 9.24

42 1988C Ia 114.4125 41.9499 UGC03933 9.4 0.01970 –21.20 9.06 1988L Ib 211.5913 50.7298 NGC5480 2.4 0.00637 –19.91 9.18 1988M II 187.9206 3.9231 NGC4496B 3.9 0.01511 –19.71 9.16 1988Q II 248.3362 34.8058 SDSSJ163320.66+344825.9 3.5 0.03511 –19.61 8.74 1988R Ia 207.0593 54.7982 MCG+09–23–009 4.5 0.02548 –20.48 8.97 1988Y Ia 41.2288 –8.4079 SDSSJ024456.18–082411.9 14.5 0.02948 –19.61 8.92 1989B Ia 170.0581 13.0054 NGC3627 . . . 0.00193 –21.17 9.16 1989C IIn 146.9395 2.6267 UGC05249 0.4 0.00625 –19.10 8.72 1989U II 148.2191 42.8622 UGC05295 18.7 0.01603 –20.23 9.17 1990B Ib 189.1410 11.2416 NGC4568 2.1 0.00744 –21.66 9.24 1990E IIP 39.8713 –8.1371 NGC1035 . . . 0.00429 –19.23 9.02 1991C II 173.1695 5.3421 SDSSJ113240.27+052035.9 4.4 0.02814 –18.05 8.73 1991G IIP 181.3702 50.5366 NGC4088 . . . 0.00262 –20.28 9.11

(cont’d) Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

1991J II 198.1567 12.5843 NGC5020 15.9 0.01097 –20.64 9.16 1991L Ib/c 250.3127 39.2924 MCG+07–34–134 5.1 0.03054 –19.86 8.97 1991T IaPec: 188.5425 2.6657 NGC4527 6.7 0.00529 –21.53 9.12 1991aj Ia 247.4371 41.7786 MCG+07–34–084 1.8 0.03139 –18.61 8.88 1992B Ia 169.0415 55.4919 SDSSJ111608.45+552925.1 15.7 0.05760 –20.30 9.10 1992bt IIP 174.8323 56.2711 NGC3780 4.0 0.00841 –20.97 9.16 1993E II 145.9340 49.2989 SDSSJ094345.02+491748.1 6.0 0.02510 –18.41 8.77

43 1993I Ia 188.6779 9.0022 MCG+02–32–144 13.6 0.04295 –21.13 9.14 1993N IIPec 157.4431 13.0208 UGC05695 2.5 0.00979 –19.00 8.96 1994J Ia 150.4032 54.5824 SDSSJ100136.65+543458.8 2.7 0.05632 –20.12 9.02 1994N II 157.4455 13.0209 UGC05695 2.3 0.00979 –19.00 8.96 1994P II 179.8378 52.7164 UGC06983 . . . 0.00361 –18.37 8.73 1994Q Ia 252.4631 40.4322 SDSSJ164951.17+402600.2 2.6 0.02946 –19.70 8.98 1994ak IIn 138.5061 40.1060 NGC2782 9.5 0.00853 –20.90 8.94 1995I II 200.2494 3.5989 SDSSJ132059.90+033550.6 8.2 0.07565 –19.92 8.76 1995J II 188.1598 63.8859 UGC07700 8.8 0.00989 –19.00 8.63 1995R Ia 208.3858 –1.1917 UGC08801 6.9 0.02374 –21.28 9.06 1995T Ia 336.8017 –9.4957 SDSSJ222712.65–092942.0 4.6 0.05640 –20.28 8.95 1995V II 41.6115 –0.4988 NGC1087 1.9 0.00507 –20.58 9.10 1995ah II 4.7955 15.1073 SDSSJ001911.01+150622.7 1.1 0.01456 –16.67 8.27

(cont’d) Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

1996ak II 198.0184 46.1940 NGC5021 12.2 0.02864 –21.86 8.85 1996an IIP 41.5035 –7.5723 NGC1084 . . . 0.00436 –20.58 8.89 1996aq Ic 215.5947 –0.3900 NGC5584 2.3 0.00553 –19.83 9.03 1996bg II 13.2182 –0.0753 SDSSJ005252.42–000427.8 6.8 0.11571 –21.89 9.06 1996bu IIPec 170.2471 53.2023 NGC3631 . . . 0.00387 –21.00 9.20 1997X Ib/c 192.0596 –3.3331 NGC4691 . . . 0.00372 –19.58 9.06 1997bo II 184.9694 7.7305 SDSSJ121952.54+074352.3 0.9 0.01257 –16.91 7.94

44 1997bs IIn 170.0594 12.9721 NGC3627 . . . 0.00193 –21.17 9.16 1997bz Ia 170.6061 1.1893 SDSSJ112225.46+011122.2 0.4 0.02980 –18.48 8.94 1997cs IIPec 228.4158 2.8957 SDSSJ151338.52+025342.4 14.2 0.03690 –19.95 8.88 1997ef IcPec 119.2618 49.5612 UGC04107 6.0 0.01168 –20.13 9.14 1997ei Ic 178.7499 58.4907 NGC3963 3.3 0.01063 –21.07 9.19 1998S IIn 176.5257 47.4821 NGC3877 . . . 0.00303 –20.37 9.18 1998ab IaPec 192.1968 41.9245 NGC4704 8.4 0.02725 –21.05 9.16 1998aq Ia 179.1075 55.1282 NGC3982 . . . 0.00395 –19.83 9.13 1998bv IIPec 159.6058 47.7092 HS1035+4758 0.6 0.00526 –15.25 8.11 1998dk Ia 3.6340 –0.7364 UGC00139 1.6 0.01319 –20.19 9.09 1998dl II 41.5061 –7.5736 NGC1084 . . . 0.00436 –20.58 8.89 1999ab II 156.2527 53.8769 SDSSJ102500.07+535230.0 5.6 0.03215 –19.72 8.65 1999ap II 127.4477 4.6274 SDSSJ082947.33+043735.8 2.7 0.03901 –19.03 8.41

(cont’d) Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

1999ay II: 221.1827 58.9284 SDSSJ144444.23+585544.4 3.1 0.04379 –18.56 8.69 1999bg II: 181.0304 62.5003 IC0758 . . . 0.00426 –18.13 8.81 1999br IIPec 195.1742 2.4961 NGC4900 . . . 0.00322 –19.04 9.18 1999bw Ib: 154.9450 45.5264 NGC3198 . . . 0.00222 –20.48 9.10 1999cb Ia 246.4517 40.3423 SDSSJ162549.27+402042.7 8.3 0.02904 –20.45 8.85 1999ce Ia 172.2149 57.1344 SDSSJ112851.62+570803.1 1.5 0.07748 –20.07 8.84 1999eh Ib 137.3859 33.1213 NGC2770 2.4 0.00645 –20.73 9.00

45 1999gb IIn 122.5571 33.9583 NGC2532 7.1 0.01751 –21.77 9.13 1999gk II 190.9663 –0.5487 NGC4653 9.6 0.00875 –20.29 9.04 2000I II 145.1731 11.8857 NGC2958 4.5 0.02175 –21.54 9.18 2000L II 153.7576 65.1343 UGC05520 5.1 0.01101 –19.83 8.94 2000cv Ia 183.9259 61.8900 SDSSJ121540.80+615323.6 4.2 0.02026 –19.68 9.19 2000db II 178.4156 47.8629 NGC3949 . . . 0.00285 –19.93 8.77 2000de Ib 186.3007 54.5081 NGC4384 1.8 0.00833 –19.66 8.89 2000dq II 32.0874 1.1836 SDSSJ020820.18+011103.5 10.0 0.04232 –21.14 9.27 2001D II 176.2124 –1.6051 IC0728 9.3 0.02840 –21.65 9.20 2001K II 168.4838 12.3017 IC0677 1.5 0.01080 –20.55 9.08 2001W II 250.3129 39.2957 MCG+07–34–134 12.3 0.03054 –19.86 8.97 2001ad IIb 261.0101 58.9978 NGC6373 10.0 0.01103 –19.23 8.83 2001ae II 200.6090 –2.4236 IC4229 9.4 0.02320 –20.93 9.19

(cont’d) Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

2001ag II 143.6292 46.4607 MCG+08–18–009 3.5 0.02649 –19.82 9.10 2001ai Ic 205.4140 55.6681 NGC5278 5.1 0.02597 –21.02 9.10 2001bg Ia 128.8286 28.4683 NGC2608 4.5 0.00727 –20.21 9.12 2001bp Ia 240.5385 36.7189 SDSSJ160208.90+364313.8 12.6 0.09484 –20.53 9.10 2001dr II 195.6629 50.4443 NGC4932 12.2 0.02350 –21.04 8.77 2001dy II 256.2476 23.1684 SDSSJ170459.90+231008.8 4.3 0.03007 –20.21 9.17 2001em Ib/c 325.5986 12.4975 NGC7112 4.6 0.01951 –20.63 9.16

46 2001er Ia 148.3361 42.8458 UGC05301 2.0 0.01615 –19.81 8.70 2001fb II 2.5256 –0.4384 SDSSJ001006.62–002609.6 6.9 0.03215 –20.37 8.81 2001fg Ia 318.1885 –0.8767 SDSSJ211245.45–005232.3 3.1 0.03164 –18.27 8.75 2001im IIb 31.0700 0.6531 SDSSJ020416.94+003909.7 3.5 0.07501 –20.05 8.85 2001km Ia 137.9099 –0.7150 SDSSJ091138.38–004254.0 0.1 0.07022 –19.73 8.95 2001kt Ia 244.3058 48.4744 SDSSJ161713.39+482827.8 0.3 0.10373 –20.50 9.15 2002I Ia 200.6118 –2.4178 IC4229 5.2 0.02320 –20.93 9.19 2002ao Ic: 217.3989 –0.0155 UGC09299 2.4 0.00517 –18.31 8.57 2002at II: 173.0915 0.8024 NGC3720 3.3 0.01997 –21.09 9.20 2002ca II 203.1313 1.8450 UGC08521 3.6 0.01090 –19.50 9.06 2002cb IIn 196.0998 47.5981 MCG+08–24–034 7.1 0.02956 –20.32 9.07 2002ce II 128.3421 29.5336 NGC2604A 3.7 0.00694 –19.04 8.98 2002cg Ic 247.2520 41.2834 UGC10415 2.0 0.03189 –21.58 9.29

(cont’d) Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

2002ck Ia 236.7531 –0.9904 UGC10030 12.4 0.02975 –21.41 9.23 2002ea IIn 32.1045 14.3480 NGC0820 1.7 0.01477 –21.19 9.15 2002ew II 313.6271 –0.1406 SDSSJ205430.46–000820.8 3.1 0.02992 –19.28 8.74 2002fs Ia 343.5037 14.6568 SDSSJ225400.65+143924.5 2.5 0.03828 –18.41 8.56 2002gb Ia 340.8388 –0.1147 SDSSJ224321.26–000654.7 2.6 0.07392 –18.62 8.61 2002gf Ia 313.8245 –0.0726 SDSSJ205518.46–000424.4 14.7 0.08622 –21.05 9.14 2002gr Ia 36.8051 0.8994 SDSSJ022713.44+005356.0 6.3 0.09102 –19.19 8.72

47 2002ha Ia 311.8274 0.3127 NGC6962 8.3 0.01447 –21.58 8.94 2002hg II 159.2942 12.6536 NGC3306 1.5 0.00994 –19.91 8.86 2002hn Ic 122.5623 33.9554 NGC2532 2.0 0.01751 –21.77 9.13 2002ii Ia 330.4684 –1.1734 SDSSJ220152.75–011018.8 13.2 0.10059 –21.30 9.16 2002in II 326.0724 –0.2421 SDSSJ214417.38–001431.5 0.7 0.07553 –18.01 8.38 2002iq II 338.0294 0.9156 SDSSJ223207.06+005456.0 1.6 0.05575 –18.44 8.29 2002jo Ia 219.5684 40.4539 NGC5708 2.0 0.00904 –20.33 9.00 2002ln II: 249.8539 41.7914 SDSSJ163925.01+414737.0 20.0 0.14199 –21.01 9.09 2003I Ib 141.8728 3.9293 IC2481 3.9 0.01778 –20.59 9.06 2003L Ic 165.8014 11.0772 NGC3506 4.2 0.02119 –21.93 9.12 2003S Ia 205.3312 55.6768 SDSSJ134119.63+554040.1 3.0 0.03982 –20.63 8.94 2003ci II 167.5993 4.8266 UGC06212 10.3 0.03034 –21.69 9.18 2003cn II 196.9044 –0.9472 IC0849 11.7 0.01808 –20.56 9.11

(cont’d) Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

2003da II 140.9482 42.1802 UGC04992 3.5 0.01385 –19.88 8.87 2003dg Ib/cPe 179.3832 –1.2538 UGC06934 1.5 0.01840 –20.73 8.98 2003ds Ic 154.7431 46.4543 MCG+08–19–017 2.4 0.02993 –19.76 8.95 2003dt Ia 311.8232 0.3119 NGC6962 10.8 0.01447 –21.58 8.94 2003du Ia 218.6492 59.3344 UGC09391 2.2 0.00635 –17.83 8.65 2003eh Ia 167.1014 3.4964 MCG+01–29–003 4.4 0.02539 –21.36 9.18 2003ej II 189.7962 0.7251 UGC07820 10.2 0.01697 –20.34 9.01

48 2003ez Ia 190.7995 –2.5011 SDSSJ124312.22–023003.2 5.1 0.04800 –21.04 9.19 2003ke IIn 146.4707 34.6837 MCG+06–22–009 4.8 0.02043 –20.90 9.10 2003la II 157.5842 61.2635 MCG+10–15–089 3.2 0.03071 –20.39 9.12 2003ld II 3.9660 16.0893 UGC00148 2.1 0.01386 –19.56 8.78 2004C Ic 171.8738 56.8801 NGC3683 2.7 0.00570 –20.04 9.06 2004F IInPec 49.4742 –7.2953 NGC1285 3.7 0.01752 –21.49 9.24 2004G II 218.3392 4.4471 NGC5668 5.7 0.00527 –19.94 8.72 2004I II 40.8801 0.3085 NGC1072 3.2 0.02600 –21.75 9.06 2004ak II 127.4482 48.7730 UGC04436 16.8 0.02402 –20.82 9.11 2004aq II 179.6135 10.0179 NGC4012 4.2 0.01388 –20.50 8.97 2004at Ia 164.6883 59.4867 MCG+10–16–037 5.9 0.02203 –19.00 8.68 2004bk Ia 204.3697 4.1031 NGC5246 5.3 0.02306 –20.73 9.20 2004bl Ia 183.8003 –3.4384 MCG+00–31–042 4.9 0.01739 –20.66 8.60

(cont’d) Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

2004bn II 163.1294 7.2281 NGC3441 5.4 0.02173 –21.61 9.18 2004cc Ic 189.1433 11.2424 NGC4568 2.3 0.00744 –21.66 9.24 2004cj Ia 155.7088 11.7030 SDSSJ102250.11+114210.8 0.2 0.10191 –20.86 9.16 2004cm II 211.8540 55.1031 NGC5486 . . . 0.00453 –18.27 8.41 2004dt Ia 30.5532 –0.0976 NGC0799 4.7 0.01944 –20.74 9.08 2004eb II 262.1012 57.5460 NGC6387 2.5 0.02853 –20.58 8.82 2004el II 224.9659 54.6183 MCG+09–25–004 5.1 0.02627 –19.77 8.93

49 2004ey Ia 327.2825 0.4442 UGC11816 4.0 0.01583 –19.96 9.10 2004fc II 27.7660 –9.7019 NGC0701 0.5 0.00622 –19.79 9.01 2004hw Ia 357.8223 1.2276 SDSSJ235117.28+011339.7 1.2 0.05993 –17.53 8.23 2004hy II 316.5177 1.2158 SDSSJ210604.63+011258.0 6.4 0.05806 –19.26 8.69 2004ia Ia 34.5096 –0.5590 SDSSJ021802.06–003334.9 10.9 0.14370 –21.77 9.06 2004ib Ic 40.2350 –0.1801 SDSSJ024056.35–001045.1 3.5 0.05621 –19.33 8.50 2004ie Ia 330.4441 1.2367 SDSSJ220147.09+011412.8 7.5 0.05133 –19.13 8.86 2004ig Ia 1.4658 –0.9959 SDSSJ000551.98–005946.7 10.1 0.18300 –20.93 9.11 2005H II 32.4105 –10.1454 NGC0838 0.8 0.01284 –20.27 9.12 2005J II 179.6189 10.0192 NGC4012 5.2 0.01388 –20.50 8.97 2005O Ib 160.5763 –0.3781 NGC3340 2.6 0.01814 –21.25 9.12 2005Y II 24.4054 0.0416 UGC01159 1.8 0.01640 –19.19 8.84 2005ab II 190.2716 50.3823 NGC4617 12.9 0.01490 –21.47 9.11 2005ad II 37.1227 –1.1389 NGC0941 4.8 0.00534 –19.06 8.75

(cont’d) Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

2005ay II 178.2003 44.1051 NGC3938 . . . 0.00263 –20.01 9.20 2005bb II 194.2995 –1.7048 UGC08067 1.6 0.00913 –19.44 8.88 2005bc Ia 219.3127 38.4564 NGC5698 2.3 0.01210 –20.60 9.31 2005bk Ic 240.5710 42.9154 MCG+07–33–027 3.5 0.02445 –20.64 9.09 2005ci II 218.6870 48.6722 NGC5682 1.1 0.00759 –18.75 8.69 2005cr Ia 185.5717 12.3970 SDSSJ122216.99+122348.8 1.5 0.02139 –18.62 8.90 2005dp II 216.9026 41.2542 NGC5630 2.5 0.00891 –20.37 8.74

50 2005eh Ia: 327.4350 0.6576 SDSSJ214944.35+003924.8 5.9 0.12488 –20.05 8.79 2005em IIb: 48.4488 –0.2436 SDSSJ031347.86–001435.9 1.2 0.02517 –16.04 8.27 2005en II 119.8026 32.9166 UGC04132 3.2 0.01730 –22.21 9.23 2005eo Ic 119.8083 32.9218 UGC04132 10.2 0.01730 –22.21 9.23 2005gp Ia 55.4970 –0.7827 SDSSJ034159.35–004658.4 2.8 0.12660 –19.63 9.00 2005hk IaPec 6.9620 –1.1979 UGC00272 4.5 0.01301 –19.47 8.70 2005hl Ib 313.8325 0.5430 SDSSJ205519.27+003226.2 5.2 0.02319 –20.52 9.18 2005ho Ia 14.8504 0.0026 SDSSJ005924.12+000009.7 0.5 0.06283 –19.57 8.69 2005if Ia 52.5536 –0.9746 SDSSJ033012.89–005828.2 0.5 0.06706 –19.89 9.01 2005ij Ia 46.0886 –1.0629 SDSSJ030421.26–010347.2 1.2 0.12446 –21.17 8.87 2005ip II 143.0267 8.4457 NGC2906 2.3 0.00678 –20.29 9.13 2005ir Ia 19.1824 0.7946 SDSSJ011643.87+004737.0 5.4 0.07635 –20.26 9.09 2005kb II 12.7112 0.8536 SDSSJ005051.62+005104.3 4.8 0.01526 –18.76 8.38 2005ks Ib/c 324.4857 –0.0325 SDSSJ213756.52–000157.7 1.8 0.09866 –19.58 8.91

(cont’d) Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

2005ku Ia 344.9275 –0.0137 SDSSJ225942.67–000049.0 0.9 0.04541 –20.07 8.88 2005lj Ia 29.4293 –0.1794 SDSSJ015743.10–001045.8 1.6 0.07774 –19.78 8.88 2005lm Ia 3.7705 0.3551 SDSSJ001504.87+002118.4 0.9 0.08466 –18.41 8.70 2005mb II 196.4686 41.7164 NGC4963 10.2 0.02384 –21.19 9.25 2005mf Ic 137.1764 44.8143 UGC04798 7.3 0.02675 –20.45 9.05 2005mn Ib 57.3268 –0.6921 SDSSJ034918.34–004129.4 2.3 0.04737 –20.17 8.73 2005ms Ia 132.3098 36.1300 UGC04614 23.2 0.02523 –20.93 9.06

51 2006E Ia 208.3688 5.2062 NGC5338 . . . 0.00270 –16.76 8.94 2006am IIn 216.9052 41.2598 NGC5630 2.0 0.00891 –20.37 8.74 2006ar Ia 159.3781 65.0161 SDSSJ103731.99+650105.9 5.1 0.02249 –20.33 9.09 2006ck Ic 197.4184 –1.0492 UGC08238 5.4 0.02443 –20.91 8.99 2006ct Ia 182.4876 47.0955 SDSSJ120956.71+470545.6 2.4 0.03145 –19.57 8.71 2006db IIn 178.9112 44.3837 SDSSJ115538.31+442301.8 2.0 0.02308 –17.54 8.33 2006ed II 225.9610 42.1135 UGC09684 2.5 0.01687 –20.21 9.02 2006fe Ia 313.0382 –0.5111 SDSSJ205209.10–003039.2 1.6 0.07049 –20.48 9.10 2006ﬁ Ia 334.9594 0.0244 SDSSJ221950.56+000125.2 19.2 0.23062 –21.98 8.96 2006fo Ic 38.1620 0.6175 UGC02019 2.4 0.02074 –20.41 9.02 2006fq IIP 5.0034 –0.6250 SDSSJ002000.77–003731.4 2.1 0.06790 –20.10 8.95 2006fs Ia 317.4958 0.4088 SDSSJ210958.97+002431.0 1.4 0.09914 –21.29 8.95 2006fu Ia: 357.7850 –0.7464 SDSSJ235108.38–004447.8 3.3 0.19852 –20.50 8.87

(cont’d) Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

2006fy Ia 351.6675 –0.8403 SDSSJ232640.14–005026.2 2.1 0.08265 –19.90 8.72 2006gb Ia 359.8188 –1.2504 SDSSJ235916.46–011502.5 5.4 0.26607 –21.58 9.13 2006gl Ia 16.4588 0.1448 SDSSJ010549.99+000843.4 11.3 0.26535 –20.78 9.02 2006gx Ia 42.0587 –0.3470 SDSSJ024814.09–002048.5 2.5 0.18063 –19.77 8.84 2006ha Ia 344.6429 15.1738 IC1461 2.2 0.03084 –20.76 9.20 2006iq Ia 324.8906 10.4849 SDSSJ213933.98+102900.6 8.8 0.07885 –20.83 9.23 2006iv IIb 177.0515 54.9874 UGC06774 3.2 0.00810 –18.82 8.53

52 2006iw II 350.3312 0.2597 SDSSJ232119.13+001532.8 3.4 0.03078 –18.73 8.86 2006ix II 359.8085 –0.3110 SDSSJ235913.99–001839.2 1.0 0.07565 –19.51 8.70 2006jc Ib/cPe 139.3367 41.9091 UGC04904 1.6 0.00555 –15.98 8.48 2006ju Ia: 351.1625 –0.7183 SDSSJ232438.93–004304.4 5.1 0.14864 –21.54 9.15 2006kh II 27.2994 –0.6052 SDSSJ014911.88–003618.4 0.8 0.05964 –18.18 9.03 2006kn II 314.5217 0.9005 SDSSJ205805.21+005402.2 1.1 0.12020 –20.10 8.43 2006kq Ia 318.9024 –0.3214 SDSSJ211536.50–001918.1 5.0 0.19827 –21.36 9.06 2006lb Ia 49.8675 –0.3180 SDSSJ031928.18–001904.8 0.7 0.18191 –20.30 8.98 2006nd Ia 341.2461 –1.0066 2MASXJ22445879–0100229 8.2 0.12876 –21.08 9.08 2006ns II 323.3843 0.7848 SDSSJ213332.23+004705.3 0.5 0.11991 –20.53 8.63 2006oa Ia 320.9289 –0.8435 SDSSJ212342.96–005035.2 1.6 0.06256 –17.95 8.45 2006pc II 26.0441 –0.1559 SDSSJ014411.09–000916.1 9.7 0.05532 –19.17 8.72 2006pt Ia 36.8174 –0.3935 SDSSJ022716.08–002335.5 7.4 0.29883 –20.90 9.07 2006qk Ic: 336.3849 0.1542 SDSSJ222532.38+000914.8 0.4 0.05831 –17.89 8.95

(cont’d) Table 2.1—Continued

a b c SN Name Type RA Dec Host Name Distance z MB 12 + log(O/H) (deg) (deg) (kpc) (mag) (dex)

2006rz Ia 56.5282 0.3897 SDSSJ034606.86+002325.1 1.4 0.03089 –18.35 8.55 2006ss IIb 215.1144 35.1952 NGC5579 6.7 0.01200 –19.62 8.95 2006te Ia 122.9291 41.5547 SDSSJ081143.49+413318.0 3.7 0.03155 –20.43 9.10 2007F Ia 195.8128 50.6188 UGC08162 5.9 0.02362 –20.07 9.05 2007I Ic 179.8048 –1.6053 SDSSJ115913.13–013616.2 1.3 0.02156 –16.73 8.34 2007S Ia 150.1302 4.4073 UGC05378 3.4 0.01385 –20.05 9.07

53 2007aa II 180.1154 –1.0810 NGC4030 . . . 0.00495 –20.76 9.24 2007af Ia 215.5876 –0.3938 NGC5584 6.1 0.00553 –19.83 9.03 2007av II 158.6798 11.1938 NGC3279 . . . 0.00464 –19.35 9.02 2007be IIP: 189.5277 –0.0309 UGC07800 6.7 0.01251 –19.76 9.05 2007bk Ia 232.1899 58.8702 SDSSJ152845.00+585200.1 8.7 0.03214 –18.17 8.27

Note. — Complete table is available online. aSupernova classiﬁcation in the SAI Catalog. The types followed by a colon indicate a provisional classiﬁcation in the SAI Catalog. bProjected SN-Host distance. cRedshift of the host galaxy from SDSS-DR4. Chapter 3

Discovery of the Dust-Enshrouded Progenitor of the Type IIn SN 2008S with Spitzer

3.1. Introduction

Over the last ∼ 20 years, several signiﬁcant milestones have been reached in the pre-explosion detection of core-collapse supernova progenitors. These began with the “peculiar” type II-P supernova 1987A in the Large Magellanic Cloud

(e.g., Menzies et al. 1987), where a cataloged ∼ 20 M⊙ blue supergiant star was identiﬁed as the progenitor (Sk −69 202; e.g., West et al. 1987). Next came the transition type IIb 1993J in M81, with a progenitor identiﬁed as a red supergiant in a binary system (e.g., Podsiadlowski et al. 1993; Maund et al. 2004). During the last decade, analyses of pre-explosion archival optical imaging of nearby galaxies obtained (mainly) with the Hubble Space Telescope have convincingly shown red supergiants with masses 8 M⊙ ≤ M ≤ 20 M⊙ to be the typical progenitors of type II-P supernovae (e.g., Smartt et al. 2004; Li et al. 2007), the most common core-collapse supernovae. Curiously, the progenitors of nearby type Ib/c supernovae, thought to result from very massive (∼> 20 M⊙) stars with strong winds that end

54 their lives as Wolf-Rayet stars, have evaded optical detection (e.g., Crockett et al.

2008).

The rarest and most diverse class of core-collapse supernovae are the type IIn

(Schlegel 1990), which represent ∼ 2 − 5% of all type II supernovae (e.g., Capellaro et al. 1997). Their optical spectra, dominated by Balmer lines in emission, and slowly declining light curves show clear signatures of interactions between the supernova ejecta and a dense, hydrogen-rich circumstellar medium (e.g., Filippenko

1997). Mainly due to their low frequencies, high mass loss rates, and the massive circumstellar envelopes generally required to explain the observations, some luminous type IIn supernovae have been associated with the deaths of the most massive stars (e.g., Gal-Yam et al. 2007; Smith 2008 and references therein). Recently, evidence for this association has increased with the report of a very luminous source in pre-explosion images of the type IIn SN 2005gl (Gal-Yam et al. 2007) and the discovery of an LBV eruption two years before the explosion of SN 2006jc (Pastorello et al. 2007). On the other hand, some low luminosity type IIn have been associated with the super-outbursts of LBVs like η Carinae (e.g., van Dyk et al. 2000; van Dyk et al. 2006).

The appearance of the type IIn SN 2008S in the nearby galaxy NGC 6946

(d ≃ 5.6 Mpc; Sahu et al. 2006) was fortuitous, since a massive stellar progenitor would be relatively easy to ﬁnd. However, pre-explosion images serendipitously obtained from the Large Binocular Telescope revealed nothing at the position of

55 SN 2008S, allowing us to put stringent limits on the optical emission. In this paper, we report the discovery of an infrared point source coincident with the site of

SN 2008S using archival Spitzer Space Telescope data. The Spitzer mid-IR detection, and deep optical non-detections, of the progenitor are the tell-tale signs of a ∼ 10 M⊙ star obscured by dust. We describe the available data in §3.2, our analysis in §3.3, and our conclusions in §3.4.

3.2. Searching for the Progenitor

NGC 6946 is quite a remarkable galaxy, giving birth to (at least) nine SNe in the last century. The latest event discovered in NGC 6946 is SN 2008S, found on February 1.79 UT at ∼ 17.6 mag (Arbour & Boles 2008) and located 52′′ West and 196′′ South of the nucleus of NGC 6946. It was spectroscopically classiﬁed as a likely young type IIn supernova from the presence of narrow Balmer lines in emission, highly reddened by internal extinction with a measured Na D absorption equivalent width of 5 A˚ (Stanishev et al. 2008). Steele et al. (2008) later reported that it had a peculiar spectrum due to the presence of narrow emission lines from the [Ca II] 730 nm doublet, Ca II infrared triplet, and many weak Fe II features.

The spectral properties and low peak luminosity led Steele et al. (2008) propose that SN 2008S was a supernova impostor such as SN 1997bs (van Dyk et al. 2000).

56 Accurate coordinates are needed in order to search for the progenitor in pre-explosion images. Fortunately, Swift started monitoring SN 2008S with UVOT and XRT shortly after the discovery. We retrieved the UVOT ubv optical images obtained on Feb. 4.8, 6.0, and 10.5 (UT) from the Swift archive. We used WCSTools v3.6.7 (Mink 1999) and the USNO-B astrometric catalog (Monet et al. 2003) to obtain astrometric solutions for the images. The mean coordinates of SN 2008S are

h m s ◦ ′ ′′ ′′ α = 20 34 45.37, δ = 60 05 58.3 (J2000.0), with rms uncertainties of σα = 0.5 and

′′ σδ = 0.3.

The Large Binocular Telescope (Hill et al. 2006) obtained deep optical images of NGC 6946 on 19−21 May 2007, 225 days before discovery, during Science

Demonstration Time using the LBC/Blue camera (Ragazzoni et al. 2006; Giallongo et al. 2008). We combined the 12 × 300 sec images obtained using the U ﬁlter (seeing

1′′.0), and the 4 × 300 sec images obtained using the B and V ﬁlters (seeing 1′′.5).

After ﬁnding an astrometric solution for the combined images using the USNO-B

′′ catalog (σα ≃ σδ = 0.2), we do not detect a source at the position of SN 2008S (see

Fig. 3.1). After calibrating the images using ancillary optical data obtained by the

Spitzer Infrared Nearby Galaxies Survey (SINGS; Kennicutt et al. 2003) and Swift, we obtain 3σ upper limits on the progenitor magnitudes of U > 25.8, B > 25.9 and

V > 26.0, which correspond to absolute magnitudes MU > −4.8, MB > −4.3, and

MV > −3.8, correcting for AV = 1.1 mag of Galactic extinction (Schlegel et al.

1998). The upper limits are calculated using aperture photometry from the standard

57 deviation of the sky at the SN position using a 10 pixel (2′′.2) diameter aperture.

We correct these values with aperture corrections estimated using bright stars.

The ∼ 0.2 mag uncertainties in the 3σ upper limits are due to the uncertainties in the aperture corrections and the standard deviation of the sky (which is estimated from the rms variations in the standard deviation measured in equal-sized apertures placed in the background around the SN position). Welch et al. (2008) reported 3σ upper limits from pre-explosion Gemini/GMOS observations of V > 24.0, R > 24.5 and I > 22.9. These correspond to absolute magnitudes MV > −6.8, MR > −5.2, and MI > −6.5, correcting for Galactic extinction.

Such a deep non-detection led us to investigate IRAC (3.6 − 8.0 µm; Fazio et al. 2004) and MIPS (24 − 160 µm; Rieke et al. 2004) images obtained by the

SINGS Legacy Survey in 2004. We astrometrically calibrated the images in the same way as the optical images from Swift and LBT. We detect a point source at α = 20h34m45s.35, δ = 60◦05′58.0′′ in the 4.5, 5.8, and 8.0 µm IRAC bands

′′ ′′ (see Fig 3.1), with rms uncertainties σα = 0.5, σδ = 0.2. This is consistent with the position of SN 2008S given the estimated uncertainties, and thus likely to be the progenitor. The source is not detected at 3.6, 24, or 70 µm. We estimate a probability of random coincidence given the uncertainty in the SN position (0′′.5) of

0.8% (0.02%) from the density of 4.5 micron sources (with [3.6]−[4.5] > 1.5 mag) detected within a 1′ radius of the SN position.

58 We searched the Spitzer archive for all the programs that have observed

NGC 6946. Observations by the SINGS survey (PID: 159), and two programs

(PIs: Meikle, Sugerman, Barlow) monitoring the type II-P SNe 2002hh and 2004et

(PID: 230, 20256, 30292, 30494) provide a 2.5-year baseline (June 2004 − January

2007) of IRAC and MIPS observations prior to the discovery of SN 2008S. We used aperture photometry (a 2 pixel extraction radius with aperture corrections) in the

ﬂux-calibrated images provided by the Spitzer Science Center to derive light curves for the progenitor. Fig. 3.2 shows the ﬂux density as a function of time in the 4.5,

5.8, and 8.0 µm bands starting from June 2004. There is no sign of variability at the

∼ 10% level. The non-detection at 3.6, 24, and 70 µm in single and stacked images allows us to place useful upper limits on these ﬂuxes.

Finally, we searched the Chandra archive to determine if the progenitor was an X-ray source. All ﬁve ACIS-S observations of NGC 6946 include the location of

SN 2008S. These observations include a 60 ks exposure in 2001, a 30 ks exposure in 2002, and 3 × 30 ks exposures in 2004. No source is detected at the supernova position in any of these images. We set a 3σ upper limit on the ﬂux of the progenitor

−15 −2 −1 37 −1 of fX < 3 × 10 erg cm s (LX < 10 erg s ) in the broad X-ray band

(0.5−8 keV), which rules out a bright X-ray binary as the progenitor. This ﬂux limit corresponds to 20 counts in the longest exposure. Table 3.1 summarizes the detections and 3σ upper limits on the progenitor ﬂuxes.

59 3.3. Beneath the Shroud

The measured ﬂuxes and upper limits in the mid-IR bands are shown in Fig. 3.3.

The shape of the spectral energy distribution (SED) suggests thermally-radiating dust as the source of the emission. We derive a best-ﬁt single-temperature blackbody

4 of T ≃ 440 K, with a luminosity of Lbol ≃ 3.5 × 10 L⊙ (d = 5.6 Mpc; Sahu et al.

1 2006), which implies a blackbody radius RBB ≃ 150 AU. This luminosity points to a ∼ 10 M⊙ star at the end of its life (e.g., Meynet & Maeder 2003). The 3σ upper limit at 70 µm further limits the total luminosity of the dust-enshrouded source and the geometry of obscuring dust distribution.

As shown in Fig. 3.3, a blackbody yields a relatively poor ﬁt to the data

(χ2 ≈ 4.9 per d.o.f.). The inability of a single-temperature blackbody to accommodate the data follows primarily from the rapid change in the SED implied by the 3.6 µm upper limit and the 4.5 µm detection. Radiation transport calculations using DUSTY (Ivezic & Elitzur 1997) were performed as a sanity check. Using a central incident blackbody with T = 3000 − 20000 K we calculated the emergent spectrum from a spherical dusty shell extended over approximately one decade in radius. As expected, the best correspondence with the data is obtained for a total optical depth at 8.0 µm of order unity, although the precise value depends

1 4 These values would change to Lbol ≃ 8 × 10 L⊙ and RBB ≃ 230 AU if we assume an extreme distance to NGC 6946 of 8.5 Mpc, which is the 3σ upper limit of the distance used by Li et al.

(2005; 5.5±1.0 Mpc).

60 on the assumed radial gradient of the density, the radial extent of the obscuring medium, and the mixture of grain types. Although a detailed investigation of the dust properties is beyond the scope of this paper, we note that the strong evolution in the SED between 3.6 and 4.5 µm may signal the need for relatively large grains

(e.g., Ivezic & Elitzur 1996).

We can estimate the mass of obscuring gas and dust by assuming that the medium is marginally optically thick at 8.0 µm. Setting τλ ≈ 1 ≈ κλρRBB,

3 −3 −1 and assuming ρ = M/(4πRBB/3), we ﬁnd that M ∼ 10 κλ, 10 M⊙, where

2 −1 κλ, 10 = κλ/10 cm g is a typical value for the Rosseland-mean dust opacity for gas at ∼ 440 K (e.g., Semenov et al. 2003). This suggests a gas density on the scale

7 −3 RBB of n ∼ 3 × 10 cm . We also estimate a minimum mass loss rate from the

˙ 2 −5 −1 −1 progenitor of Mmin = 4πRBBρcg ∼ 10 M⊙ yr , where cg ∼ 2 km s is the gas sound speed in the medium on the scale of RBB.

The lack of variability in the mid-IR ﬂuxes (see Fig. 3.2) limits the expansion velocity of the photosphere. Given our estimated temperature and luminosity, keeping the mid-IR ﬂuxes constant to within ∼ 10% over the ∼ 103 days covered by the observations means that the dust photosphere cannot be expanding by more

−1 −1 than ∼ 10 km s , which is below the escape velocity of 13 km s for a 10 M⊙ star at the estimated photospheric radius of 150 AU. This is further evidence that the dust is part of a relatively steady, massive wind rather than an explosively-expelled dust shell.

61 3.4. Discussion and Conclusions

Our pre-explosion detection of the progenitor of the type IIn SN 2008S is, to the best of our knowledge, the ﬁrst in the mid-IR. The Spitzer observations suggest

−3 an enshrouded star with a mass of ∼ 10 M⊙, buried in ∼ 10 M⊙ of gas and dust.

If SN 2008S was a real supernova explosion, this is direct evidence that relatively low-mass stars can end their lives as type IIn SNe when they have a suﬃciently dense CSM from a massive wind, as proposed by Chugai (1997b). If this event was the luminous outburst of an LBV, it presents evidence for low-luminosity, low-mass

LBVs that have not been observed before2. These conclusions about the relatively low mass hold even if the identiﬁcation of the progenitor with the Spitzer source is incorrect. In this case, we know the total extinction from the colours of the SN (see below). As shown in Fig. 3.3, our optical limits with this extinction correspond to

3 mass limits of ∼< 12 M⊙ for red supergiants and ∼< 20 M⊙ for blue supergiants .

Interestingly, we see luminous dust-enshrouded stars in the Milky Way and the

LMC whose physical properties match well the observed properties of the progenitor of SN 2008S. van Loon et al. (2005, and references therein) studied the properties

2 5 The lowest-mass LBVs known have initial masses of 20 − 25 M⊙ and luminosities > 10 L⊙

(e.g., Smith et al. 2004; Smith 2007). 3 We obtain an upper limit in the absolute optical magnitude of the progenitor of MV ∼> −7.1 if we assume an upper limit on the extinction estimate from the SN color (AV ≃ 3.5 mag; bluest black-body possible) and an extreme distance to NGC 6946 of 8.5 Mpc.

62 (T⋆, Tdust, Lbol, M˙ ) of dust-enshrouded AGB stars and red supergiants in the LMC using mid-IR observations. These are M-type stars with eﬀective temperatures

∼ 2500 − 3800 K, which have strong winds with high (gas + dust) mass loss rates

−6 −3 −1 (M˙ ∼ 10 − 10 M⊙ yr ), and warm dust emission from their dusty envelopes

(200 K < Tdust < 1300 K). Due to these similarities, we conclude that the progenitor of SN 2008S was likely a dust-enshrouded AGB (core-collapse produced from electron capture in the O-Ne-Mg core, e.g., Eldridge et al. 2007; Poelarends et al. 2008) or red supergiant like the ones observed in the LMC.

Although the detection and physical properties of the progenitor are the main results of this study, we can also try to understand something about the progenitor and explosion mechanism from the supernova itself. The classiﬁcation spectrum of SN 2008S is similar to the published spectrum of SN 1997bs (van Dyk et al.

2000), which showed narrow Balmer lines in emission and many weaker Fe II lines

(V. Stanishev, priv. comm.; Steele et al. 2008). SN 2003gm had photometric and spectroscopic characteristics similar to SN 1997bs (Maund et al. 2006). Since both of these were faint (MV ∼ −14 mag) compared with the typical absolute magnitudes at maximum of type II SNe (MV ∼ −16 to −18 mag), it is still debated whether they were intrinsically faint explosions or super-outbursts of LBVs.

The early optical photometry obtained with Swift also indicates that SN 2008S was a low-luminosity object, with MV ∼ −14 mag after correcting for the total extinction along the line of sight. We estimate the total extinction for RV = 3.1

63 to be AV ≈ 2.5 mag from the observed color B − V ≃ 0.8 mag and assuming a typical intrinsic temperature of ∼ 10000 K at this early phase of the evolution.

This value is roughly consistent with the estimated reddening obtained from the reported equivalent width of the Na D absorption feature (2.5 < AV < 7.8; based on Turatto et al. 2002). This implies the presence of significant internal extinction with AV ≃ 1.4 mag after correcting for AV (Gal) = 1.1 mag. Although the light from the supernova likely destroyed the dust that obscured the progenitor to significantly beyond the blackbody scale of ∼ 150 AU, the existence of internal extinction in the supernova light curve implies a more tenuous dusty obscuring medium on larger scales. In fact, the rare detection of the [Ca II] 730 nm doublet in emission by Steele et al. (2008) may provide direct, and independent, evidence for a significant amount of dust in the CSM that was destroyed by the UV-optical flash (e.g., Shields et al.

1999). The future spectra and light curves of SN 2008S, optical as well as radio and

X-ray, should further probe the environment as they show signs of interactions with the progenitors’s wind.

The ﬁeld of supernova forensics has advanced rapidly in recent years, with

∼ 10 SN progenitors now known (e.g., Smartt et al. 2004; Li et al. 2007). Moving forward, several groups are obtaining the data required to more fully characterize the progenitors of future nearby SNe (e.g., Kochanek et al. 2008). We note that the discovery of the progenitor of SN 2008S itself would not have been possible only

64 few years ago without Spitzer. Future multi-wavelength surveys of the local universe are thus encouraged in order to catch other unexpected stellar phenomena, potentially even before they occur.

65 LBT V IRAC 3.6 µmIRAC 4.5 µ m

IRAC 5.8µm IRAC 8.0 µm MIPS 24µm

Fig. 3.1.— Pre-supernova images (30′′ × 30′′) of the site of SN 2008S. We show the LBT/LBC optical non-detection of the progenitor and the images obtained with Spitzer by the SINGS project at 3.6, 4.5, 5.8, 8.0, and 24 µm. The progenitor is clearly detected at 4.5, 5.8, and 8.0 µm. The circle in each panel has a radius of 2′′ and is centered on the position of the supernova, corresponding to 4 times the astrometric uncertainty of 0′′.5. The dark line in the LBT image is bleeding from a saturated star.

66 Fig. 3.2.— Flux densities at 4.5, 5.8, and 8.0 µm as a function of time (in days before the discovery) for the progenitor of SN 2008S. The solid line in each panel shows the mean for each band and the dashed lines show the rms deviations of ±3.3, 12.2, and 13.0 µJy, respectively.

67 Fig. 3.3.— The spectral energy distribution of the progenitor of SN 2008S from Spitzer observations. Detections are shown as open squares at 4.5, 5.8, and 8.0 µm. Upper limits (3σ) from the combined images at 3.6 and 24 µm are also indicated. The solid line is the best-fit blackbody with T = 440 K. We also show the UBV upper limits (3σ) from LBT and the RI limits from Welch et al. (2008). The measured fluxes are not extinction corrected. The dotted line shows a reddened blackbody with the 4.5 3.5 luminosity (10 L⊙) and effective temperature (10 K) of a 12 M⊙ red supergiant. The dashed line shows a reddened blackbody with the approximate temperature 4.2 5 (10 K) and luminosity (10 L⊙) of the blue supergiant progenitor of SN 1987A, which has similar properties to the lowest luminosity LBVs observed (e.g., Smith 2007). The models were reddened with AV = 2.5 mag, the total extinction estimated from the colors of SN 2008S.

68 λ λFλ Source (10−17 W m−2)

0.3-8 keV < 0.3 Chandra/ACIS-S 0.36 µm < 0.07 LBT/LBC-Blue 0.44 µm < 0.11 LBT/LBC-Blue 0.55 µm < 0.08 LBT/LBC-Blue 0.64 µm < 0.22 Welch et al. (2008) 0.80 µm < 0.63 Welch et al. (2008) 3.6 µm < 0.45 Spitzer/IRAC 4.5 µm 1.47 ± 0.22 Spitzer/IRAC 5.8 µm 2.54 ± 0.64 Spitzer/IRAC 8.0 µm 2.48 ± 0.50 Spitzer/IRAC 24 µm < 1.20 Spitzer/MIPS 70 µm < 40 Spitzer/MIPS

Table 3.1. Spectral Energy Distribution of the Progenitor of SN 2008S

69 Chapter 4

A New Class of Luminous Transients and A First Census of Their Massive Stellar Progenitors

4.1. Introduction

Identifying the progenitors of core-collapse supernovae, the outbursts of

Luminous Blue Variables (LBVs), and other massive-star transients is essential for understanding the physics, demographics, variability, evolution, and end-states of massive stars. The problem of identifying the progenitors of bright transients from massive stars is diﬃcult, and traditionally limited to serendipitous archival imaging of nearby galaxies in the optical and near-infrared (e.g., van Dyk et al. 2003; Smartt et al. 2004; Li et al. 2007; see the extensive summary in Smartt et al. 2009). Progenitor searches are complemented by statistical studies of supernova environments within their host galaxies, which provide indirect evidence for associations between certain types of supernovae and broad classes of progenitors (e.g., James & Anderson 2006;

Kelly et al. 2008; Prieto et al. 2008b; Anderson & James 2008).

70 A much more direct method for understanding the relation between types of massive stars and their transients is to catalog all of the massive stars in the local universe (D ∼< 10 Mpc) before explosion. While surveys for bright optical transients in the local universe are well-developed (e.g., Li et al. 2001), a fairly complete census of the massive stars in nearby galaxies has only recently been proposed and undertaken (Massey et al. 2006; Kochanek et al. 2008). Despite the technical challenges required by the depth, area, and cadence of the observations, these surveys are critical for our understanding of the one-to-one correspondence between massive stars and their end-states, whether they are successful or failed explosions (e.g., Kochanek et al. 2008). The long-term promise of these surveys is to produce a catalog within which the characteristics of progenitors of future supernovae are listed pre-explosion, as in the case of SN 1987A (West et al. 1987;

Menzies et al. 1987). They will provide an essential mechanism for understanding the direct causal mapping between individual progenitor types and their transients

(see, e.g., Gal-Yam et al. 2007).

Here, we describe a new link in this causal mapping. The discovery by

Prieto et al. (2008c) and Prieto (2008d) of the dust-obscured progenitors of the luminous outbursts in NGC 6946 (SN 2008S; Arbour & Boles 2008) and in NGC 300

(Monard 2008) with Spitzer, opens up qualitatively new possibilities in the study of the connection between massive stars and their explosions. We show that these discoveries allow us to make a strong — and perhaps unprecedented — connection

71 between a dust-enshrouded sub-population of massive stars and a new relatively common class of bright transients.

The argument presented in this paper can be summarized in four points:

1. The transients SN 2008S and NGC 3001 constitute a class. Both have peak absolute V -band magnitude MV ≈ −14 ± 1(≈ 2 − 3 magnitudes fainter than normal core-collapse supernovae; e.g., Richardson et al. 2002), strong evidence for internal extinction in their spectra, narrow emission lines (similar to low-luminosity Type IIn supernovae; e.g., Filippenko 1997), and progenitors that are optically obscured and deeply dust-embedded (dust-reprocessed emission giving blackbody temperatures

4 ∼< 500 K), and with bolometric luminosities of ∼ 5 × 10 L⊙. In addition, they show little infrared variability on a few-year baseline. The details of this unique class of progenitor-transient pairs and its members, as well as a comparison with other classes of optical transients are presented in §4.2. An in-depth search for analogs to the progenitors of SN 2008S and NGC 300 in M33 is presented in §4.4.

2. Transients of this type are relatively common with respect to core-collapse supernovae. A total of ≈ 22 core-collapse supernovae or supernova-like transients

1Throughout this paper we denote “the transient in NGC 300” as “NGC 300” (e.g., “the progenitor of NGC 300”) unless we speciﬁcally refer to the host galaxy.

72 have been discovered within ≈ 10 Mpc since 1999. Sixteen were conﬁrmed supernovae, two were LBV eruptions, one was a Type IIn supernova (SN 2002bu in NGC 4242, D ≈ 8 Mpc; Puckett & Gauthier 2002), whose relatively low peak magnitude (MV ≈ −15; Hornoch 2002) suggests a close similarity with the remaining two bright transients (see §4.2.5), which are of primary interest in this paper: SN

2008S (D ≈ 5.6Mpc) and NGC 300 (D ≈ 1.9 Mpc), whose physical nature is uncertain. They may be either true (but optically sub-luminous) supernovae, or a new class of massive star eruptions. Taken at face value, these numbers imply that the rate of SN 2008S-like transients is of order ∼ 10% of the supernova rate.2

Because of incompleteness, the true rate is likely higher. We discuss the frequency of these events in detail in §4.3.

3. The progenitors of this class are extremely rare among massive stars at any moment, in any star-forming galaxy. Although the bolometric luminosities of the progenitors of SN 2008S and NGC 300 are unremarkable for massive stars

4 −4 (∼ 5 × 10 L⊙), their colors put them in a class consisting of less than 10 of all massive stars ([3.6] − [4.5]µm color ∼> 2.0 and ≈ 2.7, respectively). In a mid-infrared

(MIR) color-magnitude diagram (see Figs. 4.1 & 4.2), these progenitors lie at the extremum of the AGB sequence in both luminosity and MIR color. We refer to them as “extreme AGB” (EAGB) stars throughout this work. Because of their relatively

2Throughout this paper we use “supernova rate” to mean the core-collapse supernova rate unless we speciﬁcally mention the contribution from Type Ia supernovae.

73 low bolometric luminosities, they are not η Carinae, cool hypergiant, or classical

LBV analogs (see, e.g., Smith 2008). They are thus distinct from the “supernova impostors,” produced by bright outbursts of optically-luminous LBVs (e.g., SN

1997bs; van Dyk et al. 2000, 2003). In §4.4 we present results from a comprehensive survey of M33 for massive stars with properties similar (in bolometric luminosity, obscuration, and variability) to the progenitors of SN 2008S and NGC 300. We ﬁnd remarkably few. We compare with MIR surveys of the LMC, SMC, and NGC 300.

4. Conclusion: A large fraction of all massive stars undergo a dust-enshrouded phase

4 within ∼< 10 yr of explosion. This is the most natural explanation for the facts of points (2) and (3) above. If these transients have a rate comparable to the supernova rate (∼ 20%; point 2), then the timescale for the obscured phase is determined by the ratio of the number of dust-obscured massive stars relative to the entire population

−4 7 (∼< 10 ; see §4.4, §4.5) times the average lifetime of a massive star (∼ 10 yr).

Importantly, from the rarity of SN 2008S-like progenitors alone (∼ 10−4 of massive stars; point 3), one would naively expect a comparable fraction of supernovae to have progenitors of this type, if the dust-obscured phase occurs at a random time in the life of a massive star. However, the relative frequency of these explosions (point

4 2) shows that this phase must come in the last ∼< 10 yr, just before explosion. Thus, there must be a causal relation between the occurrence of the highly dust-enshrouded phase and eruption. These points, together with a discussion of the theory of

74 the evolution of massive stars, the potential connection with electron-capture supernovae, white dwarf birth, and other hypotheses for the physical mechanism of

2008S-like explosions, as well as a call for a more comprehensive Spitzer survey for analogous sources within D ≈ 10 Mpc, are presented in §4.5.

4.2. The Class

We start by listing the objects we view as likely to represent this new class of

SN 2008S-like transients and progenitors. The two objects that deﬁne the class —

SN 2008S (§4.2.1) and NGC 300 (§4.2.2) — are unique among transient-progenitor pairs. The progenitors are relatively low-luminosity, have low variability, and are deeply dust-embedded on ∼ 100 AU scales. The transients are low-luminosity, with spectra exhibiting both narrow Balmer lines (similar to low-luminosity IIn’s and impostors), and [Ca II] in emission, and have rapidly decaying lightcurves compared to IIP supernovae. The transient in M85 (§4.2.3) and SN 1999bw (§4.2.4) may also be members of the same class, but we cannot conﬁrm the existence of a dust-obscured progenitor similar to SN 2008S/NGC 300. In §4.2.5, we contrast this class with other peculiar outbursts, such as the supernova impostors and low-luminosity Type

IIP supernovae, and we note a number of other transients that are not excluded as members of the SN 2008S-like class.

75 4.2.1. SN 2008S

SN 2008S in NGC 6946 (D ≈ 5.6 Mpc; Sahu et al. 2006) was discovered

February 1.79 UT (Arbour & Boles 2008). Because of the presence of narrow

Balmer lines (FWHM ≈ 1000 km s−1) it was initially classiﬁed as a young Type IIn supernova. Stanishev et al. (2008) reported strong Na D absorption with equivalent width of 5 A,˚ indicating a high degree of internal extinction. On the basis of its relatively low luminosity (MV ≃ −14) and peculiar spectrum (including the presence of strong and narrow [Ca II] 730 nm doublet in emission), Steele et al. (2008) proposed that SN 2008S was a supernova impostor such as SN 1997bs (van Dyk et al. 2000). Botticella et al. (2009) present the late-time lightcurve of SN 2008S, which shows evidence for a power-law time dependence with a slope indicative of being powered by the radioactive decay 56Co. Following the suggestion by Prieto et al. (2008c) and the discussion presented in §4.5, Botticella et al. argue that SN

2008S may have been an electron-capture supernova.

Prieto et al. (2008c) used a deep pre-explosion archival Large Binocular

Telescope (LBT) image of NGC 6946 to put stringent upper limits on the optical emission from the site of the supernova (3σ limits of MU > −4.8, MB > −4.3, and MV > −3.8). The failure to detect a progenitor in the optical led Prieto et al. (2008c) to examine archival Spitzer IRAC imaging of NGC 6946 from the SINGS

Legacy Survey (Kennicutt et al. 2003) for the progenitor. A point source at the

76 location of SN 2008S was detected at 4.5, 5.8, and 8µm, but undetected at 3.6µm and 24µm, leading to a lower limit on the [3.6] − [4.5] color of ∼> 2. The best-ﬁtting blackbody temperature to the SED was ≈ 440 K.3 The integrated luminosity was

4 ≈ 3.5 × 10 L⊙, consistent with a relatively low-mass massive-star progenitor with a zero-age main sequence (ZAMS) mass of M ≈ 10 M⊙. The progenitor SED is shown in the left panel of Figure 4.5. Simple arguments suggest that the obscuration was circumstellar, with an optical depth at visual wavelengths larger than a few on a physical scale of order 150 AU (Prieto et al. 2008c). The explosion itself was also serendipitously observed by Spitzer 5 days after discovery (Wesson et al. 2008), with an infrared luminosity ∼50 times larger than the progenitor, suggesting a substantial amount of dust-reprocessing (see Fig. 2 of Prieto et al. 2008e). The lack of variability of the progenitor on a ∼ 3 yr timescale argues that the obscuring medium was either a continuous wind with a steady photosphere or a (implausibly?) slowly (∼< 10 km s−1) expanding shell.

4.2.2. NGC 300

A luminous optical transient (MV ≃ −13) in the nearby galaxy NGC 300

(D ≈ 1.9 Mpc; Gieren et al. 2005) was discovered by Monard (2008), and reported by Berger & Soderberg (2008). The latter used archival imaging with ACS/WFC onboard the Hubble Space Telescope (HST) to put very tight limits on the optical

3A blackbody provides a rather poor ﬁt to the SED, perhaps indicating an interesting grain size distribution in the obscuring material (Prieto et al. 2008c).

77 emission (see Fig. 4.5). These limits led them to suggest that the NGC 300 transient was analogous to the outburst V838 Monocerotis (Bond et al. 2003).

Similar to the case of SN 2008S, Prieto (2008d) discovered a deeply-embedded source in archival Spitzer imaging (PI: R. Kennicutt) at the location of the transient.

The source was detected in all IRAC bands, as well as in MIPS 24µm, and had a [3.6] − [4.5] color of ≈ 2.7. The SED implies a ≈ 330 K blackbody (as with SN

2008S, a blackbody is a fairly poor ﬁt to the SED) with a bolometric luminosity

4 of ≈ 5.6 × 10 L⊙. This ﬁnding conﬁrms the massive stellar origin of the NGC 300 transient, and is consistent with a relatively low-mass massive star (see Fig. 4.5).

Importantly, depending on the details of stellar models for ZAMS masses in the range of ∼ 10 M⊙, at ﬁxed ﬁnal bolometric luminosity, the inferred initial progenitor

4 mass may be multiply-valued and a luminosity of ≈ 5.6 × 10 L⊙ can imply a ∼ 5,

4 ∼ 8 or ∼ 11 M⊙ progenitor (see Smartt et al. 2009, their Fig. 2; §4.5.2).

The luminosity and blackbody temperature of the progenitor of NGC 300 suggest an obscuring medium with physical scale of order 300 AU. The deep limits on the optical emission from the progenitor with HST suggest an optical depth at V considerably larger than unity (∼ 8 − 10).5

4Prieto (2008d) assumed standard singly-valued masses, and given the measured luminosity of the 5 progenitor of NGC 300 quoted (∼ 10 L⊙) they inferred an initial progenitor mass of ∼ 15 − 20 M⊙. 5Assuming a spherical homogeneous medium and a Galactic dust-to-gas ratio, this optical depth −3 at V requires a total mass of obscuring material of roughly ∼ 2 × 10 M⊙ on a scale of 300 AU.

78 The fact that the transient in NGC 300 and SN 2008S were similar, both in their luminosities (both relatively faint with respect to typical supernovae with

MV ≈ −14) and spectra (with narrow Balmer lines and strong [Ca II] 730 nm doublet in emission; Bond et al. 2008), and that their progenitors were similar

(highly dust-obscured, relatively modest bolometric luminosities) led Prieto (2008d) to propose that NGC 300 and SN 2008S share a common origin: the explosion — whether supernova or outburst — of a massive star enshrouded in its own dust.

As we show below, the fact that their progenitors are so rare among massive stars implies that just two events (although, see §4.2.3 & 4.2.4) are suﬃcient to deﬁne a class.

4.2.3. M85

The optical transient in the Virgo galaxy M85 (NGC 4382) was discovered in early 2006 by the Lick Observatory Supernova Search Team (KAIT; Li et al. 2001) and discussed in Kulkarni et al. (2007) and Pastorello et al. (2007) and may also be a member of the class deﬁned by SN 2008S and NGC 300. The transient had peak R-band absolute magnitude of ≈ −12 with a ∼ 80 − 100 day plateau, similar to low-luminosity Type IIP supernovae. Optical limits constrain the progenitor to be less than ≈ 7M⊙ or highly obscured (Ofek et al. 2008; see also Pastorello et al. 2007).6 The low optical luminosity, plateau, and redward spectral evolution of the

6Ofek et al. (2008) quote a limiting absolute magnitude at F850LP(z) with HST of > −6.2.

79 transient led Kulkarni et al. (2007) to propose that it was analogous to the outburst

V838 Mon. In contrast, Pastorello et al. (2007) argued that it was a low-luminosity

Type IIP supernova (see §4.2.5).

Prieto et al. (2008e) again searched archival Spitzer imaging and discovered an infrared source at the site of the optical transient, but, by chance, taken 8.8 days before the discovery in the optical by KAIT. The source was detected in all

IRAC bands and was associated with the transient itself, and not the progenitor, since archival Spitzer imaging from 2004 (PID 3649; PI P. Cˆot´e) does not reveal a point source at the location of the M85 transient. Using these images, we derive

5 5 3σ upper limits of 4 × 10 L⊙ and 2 × 10 L⊙ at 3.6 µm and 4.5 µm, respectively, for the progenitor. The transient was also detected in the mid-IR seven months after the initial discovery by Rau et al. (2007), the ﬂuxes having decreased by a factor of ≈ 5 over that time. The bright infrared transient discovered by Prieto et al. (2008e) is adequately ﬁt by a blackbody with temperature of ≈ 800 K with

6 luminosity ≈ 2 × 10 L⊙. The optical and NIR photometry of Kulkarni et al. (2007) indicate a second component to the SED with a blackbody temperature of ≈ 3900 K

6 and with a luminosity of ≈ 5 × 10 L⊙.

The cooler re-radiated dust emission arises from a region of order 300 − 400 AU in physical scale. Assuming that the optical emission did not vary in the ≈8.8 days between the IR discovery and the optical discovery, the ratio of the power in these two blackbody components implies that the optical depth at V (τV ) is less

80 than unity. The physical scale of the obscuring medium indicates that it is likely circumstellar, and the result of a mass-loaded wind. In addition, the luminosity of the transient suggests that any pre-existing dust within ∼ 100 − 200 AU would have been destroyed during the explosion. Given the fact that the optical depth to the

−1 source scales as r in a freely-expanding wind, it is not implausible that τV to the progenitor was a factor of ∼ 10 − 20 larger before explosion. These estimates suggest a pre-explosion obscuring medium similar in its gross properties to the SN 2008S and NGC 300 progenitors.

The M85 transient also showed narrow Balmer lines in emission, as well as some Fe II lines, similar to SN 2008S and NGC 300. Because of the strong evidence for obscuration of the progenitor, as evidenced by the bright IR transient, and the similarity of the spectra, Prieto et al. (2008e) proposed that these outbursts share a common origin and that their obscured progenitors may give rise a new class of

2008S-like transients.

We emphasize that because the character of the progenitor is not known (except for the optical and IR limits), the connection to SN 2008S- and NGC 300-like events is plausible rather than certain. Nevertheless, if the M85 transient was associated with an embedded massive star, the IR limits we derive are consistent with the luminosities derived for the 2008S and NGC 300 progenitors.

81 4.2.4. SN 1999bw

The Lick Observatory Supernova Search reported in April 1999 the discovery of a possible supernova in the galaxy NGC 3198 (Li 1999). The optical spectrum of the transient, dominated by narrow Balmer lines in emission (Garnavich et al. 1999;

Filippenko et al. 1999), and its low V -band absolute magnitude at maximum of

≃ −13 (D ≈ 13.7 Mpc; Freedman et al. 2001) led Li et al. (2002) to propose that this transient was an LBV-like outburst. Like SN 2008S and NGC 300, its spectrum showed [Ca II] in emission. Additionally, an infrared source coincident with the optical position of the transient was detected in archival Spitzer imaging obtained with IRAC by the SINGS Legacy Survey ﬁve years after the discovery of SN 1999bw

(Sugerman et al. 2004). The source was detected in all IRAC bands and the SED

5 was well-ﬁt by a 450 K blackbody with an integrated luminosity of ≈ 1.4 × 10 L⊙, which translates into a blackbody scale of ∼ 300 AU.7 We have checked archival

IRAC images obtained in December 2005 (PID 20320; PI B. Sugerman), 1.5 yr after the detection in the SINGS images, and we conﬁrm that the MIR source is indeed the transient, since the ﬂuxes have declined by a factor of more than 3 during this time.

The combination of a low optical luminosity at maximum, an optical spectrum dominated by narrow Balmer lines in emission, the presence of [Ca II] emission, and a luminous infrared emission detected with Spitzer, make SN 1999bw similar to SN

7The luminosity and blackbody scale have been adjusted to the distance employed here.

82 2008S, NGC 300, and the transient in M85. However, as in the case of M85, we emphasize that because there is no information on the progenitor we cannot be sure that SN 1999bw was of the same class as SN 2008S and NGC 300.

4.2.5. The Connection to Other Transients

As we discuss in detail in §4.4, perhaps the primary distinguishing characteristic of this class of transients is their deeply embedded progenitors. Since we are unable to conﬁrm the presence of such progenitors for the M85 transient or SN 1999bw, we are unable to make a direct analogy with SN 2008S and NGC 300, and instead rely on the fact that the transients themselves provides strong evidence for obscuration on few-hundred AU scales.

In our eﬀort to understand which cataloged transients might belong to the

SN 2008S/NGC 300-like class we have examined archival imaging of many recent supernovae, as well as archetypal peculiar supernovae, including supernova impostors,

LBV outbursts, and low-luminosity Type IIP supernovae. Here, we provide a brief discussion in an eﬀort to orient the reader.

As we mentioned in §4.1 (point 2), the low-luminosity Type IIn SN 2002bu is an interesting transient that may be a member of the class deﬁned by SN 2008S and NGC 300. We checked archival Spitzer data of the host galaxy NGC 4242 (PID

69; PI G. Fazio) taken 2 years after discovery and we ﬁnd a bright infrared point

83 source detected in all IRAC bands within 0′′.6 of the supernova position. Also, two epochs of MIPS data (PID 40204; PI R. Kennicutt), obtained in 2008, 6 years after explosion, and separated by just 6 days, reveal a 24 µm source at the position of the supernova. This is qualitatively similar to the case of M85 and SN 1999bw, but because only a single post-explosion IRAC epoch exists, and because the two MIPS epochs are separated by such a short time, we are unable to deﬁnitively conﬁrm that the MIR source is associated with SN 2002bu.

SN 1997bs is an intriguing example of an object that does not fit into this class, although its peak absolute magnitude, lightcurve, color, and some spectral features are comparable to SN 2008S and NGC 300. In this case, a luminous un-obscured progenitor has been identified in the optical (MV ≈ −7) and the transient itself has been argued to be the outburst of an LBV (van Dyk et al. 2003). Of interest is the fact that no object has been subsequently identified in the optical at the site of the transient (van Dyk 2005). We have checked archival Spitzer data of the host galaxy

(NGC 3627) obtained in 2004 (∼ 7 years after discovery) by SINGS and we do not detect a bright MIR source at the site of SN 1997bs, in contrast with SN 2008S,

M85, and SN 1999bw. The event SN 2003gm is also interesting in this context since it had photometric and spectroscopic evolution similar to SN 1997bs, and also showed an optically luminous progenitor (MV ≈ −7.5; Maund et al. 2006). The fact that both SN 1997bs and 2003gm had bright unobscured progenitors is our primary reason for excluding them from the class deﬁned by SN 2008S and NGC 300.

84 Historical LBV eruptions in nearby galaxies that have been initially classiﬁed as supernovae are also worth mentioning here. These include SN 1954J (e.g., Smith et al. 2000) and SN 2002kg (e.g., Maund et al. 2006; van Dyk et al. 2006) in NGC

2403, SN 1961V (e.g., Humphreys 2005) in NGC 1058, and SN 2000ch in NGC 3432

(Wagner et al. 2004). As in the cases of SN 1997bs and SN 2003gm, a very important common diﬀerence between these objects and SN 2008S or NGC 300 is that they all had optically luminous progenitors with absolute magnitudes ∼< −6, consistent with originating from very massive stars. These transients also have other properties that are not consistent with SN 2008S-like explosions: (1) their peak absolute magnitudes range between −18 ∼> MV ∼> −9, (2) the transient timescales vary widely from a few days to years, and (3) they are not luminous MIR sources in archival Spitzer data.

Other optical transients classiﬁed as sub-luminous Type IIP supernovae that might potentially fall into the class deﬁned by SN 2008S and NGC 300 include SN

1994N, SN 1997D, SN 1999eu, 1999br, 2001dc, 2003Z, and 2005cs (e.g., Pastorello et al. 2004, 2006). We note, however, that there are fundamental diﬀerences in the spectra of low-luminosity Type IIP SNe compared with SN 2008S-like transients.

In particular, sub-luminous Type IIP SNe show Balmer lines with strong P-Cygni absorption proﬁles and velocities of a few thousand km s−1, as is observed in more luminous Type IIP SNe. This stands in sharp contrast with the Balmer lines in

SN 2008S-like transients, which are fairly narrow (FWHM ∼ 1000 km s−1) and which do not show strong P-Cygni absorption features. In this way, SN 2008S-like

85 transients most closely resemble the spectra of low-luminosity Type IIn SNe and

LBV outbursts.

In addition to the very interesting case of 2002bu, there were five other low-luminosity transients classified as “impostors” or “unknown” that could have been LBV outbursts, but for which no progenitor has been identified, and which might be 2008S-like: NGC 4656, SN 2001ac, 2006bv, 2006fp, and 2007sv (see §4.3).

However, as we have emphasized and as the sources discussed above imply, the properties of the progenitor cannot be deduced from the character of the optical outburst alone (e.g., contrast SN 1997bs and SN 2003gm with SN 2008S). Thus, in order to understand the causal mapping between progenitor and explosion, a census of the progenitors must ﬁrst be completed. What is clearly needed is a comprehensive survey for bright MIR sources in all nearby galaxies (∼< 10 Mpc) with

(warm) Spitzer, analogous to the survey proposed by Kochanek et al. (2008) in the optical. In the next section we discuss our search for deeply-embedded progenitors in M33. We discuss a more complete census in §4.5.

4.3. Rates

The absolute rate of transients analogous to NGC 300 and SN 2008S is uncertain. Current samples of supernovae over the last decade in the local volume within 10, 20, and 30 Mpc allow us to make only a rough estimate of the true rate. A systematic transient search in the local volume is crucial to solidify these

86 numbers. Nevertheless, we estimate that 2008S-like transients occur with a frequency equivalent to ∼ 20% of the Type II supernova rate. We discuss the observed rates in

§4.3.1 below, and then we enumerate several arguments suggesting that the sample of SN 2008S-like transients may be highly incomplete in the local universe (§4.3.2).

4.3.1. Observed Counts

D ∼< 10 Mpc

In addition to SN 2008S and NGC 300, ≈ 20 other core-collapse supernovae or supernova-like transients have been discovered within ≈ 10 Mpc since 1999.8

Sixteen were conﬁrmed supernovae; they are SN 1999em, 1999ev, 1999gi, 1999gq,

2002hh, 2002ap, 2003gd, 2004am, 2004dj, 2004et, 2005af, 2005at, 2005cs, 2007gr,

2008bk, and 2008ax. Two were bona ﬁde LBV eruptions (SN 2000ch, Wagner et al. 2004; 2002kg, Schwartz et al. 2003; Weis & Bomans 2005; Maund et al. 2006; van

Dyk et al. 2006). One was a Type IIn supernova potentially of the SN 2008S class

(SN 2002bu in NGC 4242, D ≈ 8 Mpc; Puckett & Gauthier 2002; Hornoch 2002).

Finally, the transient in NGC 4656 also had some of the spectral characteristics of low-luminosity IIn supernovae (e.g., narrow Hα in emission), but only reached an absolute magnitude of ≈ −11.5, (Rich et al. 2005; Elias-Rosa et al. 2005).

Taken at face value, with no correction for incompleteness, these numbers imply that 2/22 ≈ 9% or 3/22 ≈ 14% (including SN 2002bu) of all optically bright

8Throughout the discussion here we exclude Type Ia supernovae.

87 transients are SN 2008S-like. Removing the two bona ﬁde LBV outbursts (2000ch and 2002kg) for comparison with the supernova sample proper, if SN 2008S and

NGC 300 are supernovae, they represent ≈ 10% and ≈15% (again, with SN 2002bu) of the sample.

D ∼< 20 Mpc

A similar exercise can be carried out within the larger volume of ∼ 20 Mpc.

With a peak absolute magnitude of ∼ −14, SN 2008S-like transients would have an apparent magnitude of 17.5 at D = 20 Mpc, without including a correction for extinction intrinsic to the transient. In fact, SN 2008S and NGC 300 had AV ≈ 1.2

(Prieto et al. 2008c) and AV ∼ 0.3 − 1.2 (Bond et al. 2008), respectively.

Using the Smartt et al. (2009) compilation, we ﬁnd 29 IIP, 4 IIb, 15 Ib/c, 2 IIn

(1998S & 2002bu), and 2 IIL supernovae in the last decade. There are 6 classiﬁed as “LBV eruptions/impostors” (1999bw, 2000ch, 2002kg, 2003gm, 2007sv, and NGC

46569), but only 2000ch and 2002kg have strong evidence for an LBV progenitor.

Whether the yellow supergiant progenitor of 2003gm survived the explosion, and hence whether 2003gm was in fact a supernova, has not yet been deﬁnitively established (Maund et al. 2006). Similarly, although SN 2007sv, which reached an absolute magnitude of ≈ −14, bears some similarity to 1997bs, the nature of its progenitor has not been established (Duszanowicz et al. 2007; Harutyunyan et

9Note that NGC 4656 is not included in Smartt et al. (2009).

88 al. 2007). Finally, there are 3 events whose nature is classiﬁed in Smartt et al. (2009) as unknown: M85, NGC 300, and SN 2008S.

With no correction for incompleteness, and taking only NGC 300 and SN 2008S, this compilation implies an overall rate of 2/61 ≈ 3.3% within 20 Mpc. Including

M85 and 2002bu in the sample of 2008S analogs doubles the rate. Including 1999bw,

2007sv, and NGC 4656 brings the overall observed rate of SN 2008S-like transients to ∼ 10% within D ∼< 20 Mpc.

D ∼< 30 Mpc

With a limiting magnitude of ≈ 18 − 19, amateur and professional supernova surveys could in principle ﬁnd SN 2008S-like transients with MV ≈ −14 to a luminosity distance of ∼ 25 − 40 Mpc (again assuming no extinction). Objects of interest discovered over the last ∼ 10 years with faint absolute magnitudes in this distance range, and with the spectral characteristics of IIn supernovae analogous to

SN 2008S include 2001ac, 2006bv, and 2006fp.

A total of 92 core-collapse supernovae and 7 LBV eruptions appear in the recent compilation of Smartt et al. (2009). Three events — M85, NGC 300, and

SN 2008S — are classiﬁed as “unknown.” Of the 7 “LBV eruptions,” only 2000ch,

2002kg, and 2003gm have LBV-like progenitors. The remainder — 1999bw, 2001ac,

2006fp, and 2007sv — have little or no progenitor information, spectra that resemble

IIn’s, and may be SN 2008S analogs. Given the number of conﬁrmed NGC 300

89 and SN 2008S-like analogs (just 2) and those suspected of belonging to this class

(M85, 1999bw, 2002bu), as well as those tradiationally labelled “LBV outbursts,” but with no strong conﬁrmation (4656, 2001ac, 2006bv, 2006fp, 2007sv) the overall rate ranges from 2/102 ≈ 2% to 10/102 ≈ 10% when measured within 30 Mpc.

4.3.2. Arguments for Incompleteness & Some Implications

To summarize the previous subsections, conservatively taking NGC 300 and SN

2008S as the only examples of their type ever observed (that is, excluding all other low-luminosity transients), the observed rate is ∼ 9%, ∼ 3%, and ∼ 2% within 10,

20, and 30 Mpc volumes, respectively, with respect to all bright optical transients when averaged over the last 10 years.

It is difficult to estimate the degree of uncertainty in these numbers since the surveys that find local supernovae are a combination of professional and amateur, with complicated and unquantitified selection functions for transient identification.

Most surveys responsible for transient discoveries in the local universe do not have detailed descriptions of completeness in the literature. On the contrary, the large majority of the transients in the local volume are discovered by amateurs.10

Nevertheless, it is possible to make an estimate of completeness that gives a sense of how large the correction to the rate of SN 2008S-like transients may be.

10 Of the 13, 17 and 13 local (D ∼< 30 Mpc) supernovae (including Type Ia’s) discovered in 2006, 2007, and 2008, only 1, 4 and 4 were discovered by professional searches.

90 As an example, B. Monard typically quotes ±0.2 magnitude (unﬁltered) photometric errors on discovery observations of SNe detected at ≈ 16−17 magnitude

(e.g., Monard 2006). This photometric error translates to a signal-to-noise ratio of

S/N ≈ 5. In order to estimate a lower limit on the incompleteness, we can compare this value for S/N with the mean detection eﬃciency for Type Ia supernovae in the

SDSS-II Supernova Survey, which employs a well-tested photometric pipeline that uses diﬀerence imaging to subtract oﬀ the host galaxy (see Dilday et al. 2008). Their

Figure 7 shows that for S/N ≈ 5, the detection eﬃciency is ≈ 0.5 in Sloan gri. The detection eﬃciency drops to ∼ 0.1 − 0.2 for S/N ≈ 2, corresponding to a photometric error of ±0.4 magnitudes.

Momentarily ignoring the diﬀerence in the shape of Type Ia lightcurves with respect to those of 2008S and NGC 300, the results of Dilday et al. (2008) imply that surveys achieving limiting magnitudes of ≈ 18 − 19 are of order ∼ 10% complete for

SN 2008S-like transients with MV ≈ −14 at 25 − 40Mpc (S/N ∼ 1 − 2). Thus, a factor of ∼ 10 correction should be applied to the 30 Mpc sample in §4.3.1 from the

Smartt et al. (2009) catalog for the true rate of low-luminosity IIn supernovae like

SN 2008S. Within 20 Mpc the correction for incompletness is likely a factor of ∼ 5, and within 10 Mpc the incompleteness is probably closer to a factor of ∼ 2. Similar corrections should be applied to the observed rate of true LBV eruptions.

There is another argument for incompleteness at the factor of ∼ 2 level within

10 Mpc for SN 2008S-like transients. As summarized by Horiuchi et al. (2008) (their

91 Section II.B), the observed rate of supernovae of all types within 30 Mpc yields a ratio of Type Ia supernovae to core-collapse supernovae that is signiﬁcantly in excess of the cosmic ratio measured at high redshift (0.5 ∼< z ∼< 1; Dahlen et al. 2008). Indeed, the ratio within 30 Mpc is large enough that we would expect to have seen several Type

Ia supernovae within 10 Mpc in the last 10 years. Yet, none have been found. This fact implies that the ratio of Ia to core-collapse supernovae has been over-estimated within 30 Mpc because normal core-collapse supernovae are intrinsically fainter than

Type Ia’s. Thus, the sample of normal core-collapse supernovae is incomplete within

30 Mpc at the factor of ∼ 1.5 − 2 level, even though these objects typically have peak absolute visual magnitudes of MV ≈ −16 to −18. Naively, analogs to SN

2008S with MV ≈ −14 would be ∼ 6 times more incomplete than normal supernovae at D ≈ 30 Mpc. Of course, the actual incompleteness correction depends on the overall extinction correction for the transient population and on the cadence of the observations since the lightcurve declines much faster for 2008S-like transients than for supernova of type IIP. This implied nearly order-of-magnitude incompleteness correction at D ≈ 30 Mpc for 2008S-like events strongly indicates that the sample is incomplete at order unity within D ≈ 10 Mpc.

Additionally, a plot of the discovery rate of all supernovae (Ia’s included) within

30 Mpc over the last 10 years shows an increasing trend, super-Poisson variance, and strong dependence on the results and observing strategy of a single survey (LOSS;

Li et al. 2000). There is also an asymmetry between the rate of discovery in the

92 northern and the southern sky in excess of a simple extrapolation of star formation from the catalog of Karachentsev et al. (2004). Finally, there is an unquantiﬁed bias against small star-forming galaxies in the local universe. These points further solidify the case that the normal core-collapse supernova rate is incomplete, which implies that we are missing 2008S analogs in abundance in the local universe.

Taking yet another angle on the question of overall rate, we may consider the a posteriori statistics of the events SN 2008S and NGC 300 themselves. Taking the overall supernova rate as ∼ 1 − 2 yr−1 within 10 Mpc implies a probability of

∼ 4 − 8 × 10−4 of seeing two events in a single year if the overall rate of 2008S-like transients is 2% of the supernova rate. We consider this uncomfortably small.

Similarly, if we were to take the true SN 2008S-like transient rate to be ∼ 100% of local supernova rate, we would be forced to explain the fact that only ∼ 2 such events have been seen within 10 Mpc in the last 10 years. Given the discussion of incompleteness above, an overall rate of ∼ 20% with respect to supernovae gives a reasonable chance of seeing two in one year and of seeing only a handful on a 10 year baseline.

Taken together, these arguments imply that the sample of transients in the local universe when averaged over the last 10 years is highly incomplete. We suggest that the incompleteness correction is a factor of ∼ 2, ∼ 5, and ∼ 10 for MV ≈ −14 transients at 10 (≈ 16.0 mag), 20 (≈ 17.5 mag), and 30Mpc (≈ 18.4 mag), even before accounting for the potentially higher average extinction of these transients

93 relative to normal supernovae. These estimates are consistent with the Richardson et al. (2002), who conclude that low-luminosity supernovae with MB ∼> −15 may constitute more than 20% of the overall supernova population (for related arguments, see Schaefer 1996; Hatano, Fisher, & Branch 1997; Pastorello et al. 2004). There are several immediate implications:

1. The true rate of SN 2008S-like transients is ∼ 20% of the core-collapse

supernova rate. However, we emphasize that lower and higher values at the

factor of ∼ 2 level are not excluded until a more thorough census has been

made.

2. The true rate of massive star eruptions (LBV-like and otherwise) is similarly

incomplete. The observed rate of “LBV eruptions” within 10 and 20Mpc in

§4.3.1 and §4.3.1 implies that they are ∼ 1 − 3 times more common than SN

2008S-like transients. Thus, the true rate of massive LBV eruptions is likely

∼ 20 − 60% of the core-collapse supernova rate.11

3. The observed rate of core-collapse supernovae is incomplete at the factor of

∼ 2 level for D ∼< 30 Mpc.

11We thank the anonymous referee for pointing this out.

94 4.4. A First Census

Because of the implied frequency of events similar to SN 2008S and NGC 300

(§4.1, point 2) and the interesting character of their progenitors, we searched for analogous sources in archival Spitzer imaging of nearby galaxies. Our goal was to identify the underlying sub-population of massive stars from which these progenitors emerge, to characterize their properties and frequency, and to catalog them for future study.

The key characteristics of the progenitors of SN 2008S and NGC 300 are that they are optically-obscured and deeply embedded, with very red MIR colors, that their bolometric luminosities are indicative of relatively low-mass massive stars

4 4 (L ≈ 4 × 10 and ≈ 5.6 × 10 L⊙, respectively), and that the several epochs on

NGC 6946 revealed that the progenitor of SN 2008S was not highly variable in the

≈ 3 years before explosion (Prieto et al. 2008c). We discuss the variability of the progenitor of NGC 300 in §4.4.4 below based on just two pre-explosion epochs.

Although we are only able to derive a lower limit to its RMS variation at 4.5 µm, like the progenitor of SN 2008S, we ﬁnd that it too is consistent with a low level of variability.

For a ﬁrst census, we searched for the deepest archival Spitzer observations of a nearby relatively massive bright star-forming galaxy, with already extant optical catalogs. The Triangulum galaxy M33 is a perfect test case. It has an absolute

95 B-band magnitude of MB ≈ −19.2, a distance of ≈ 0.96Mpc (distance modulus

µ = 24.92; Bonanos et al. 2006), and it has extensive optical (e.g., Hartman et al. 2006; Massey et al. 2006), Hα (Massey et al. 2007), and MIR and FIR imaging

(McQuinn et al. 2007). A similarly rich dataset exists for several other local galaxies, including the Magellanic clouds (e.g., Blum et al. 2006; Bolatto et al. 2007) and M31

(e.g., Mould et al. 2008). An analysis similar to that described below will be the subject of future work (but, see §4.4.5).

In this Section, we present the MIR color-magnitude and color-color diagrams for all cataloged point sources in M33 obtained from multi-epoch archival

Spitzer/IRAC (Fazio et al. 2004) observations (PI R. Gehrz; PID 5). This dataset allows us to search for and identify stars analogous to the progenitors of SN 2008S and NGC 300. We also present a variability study for both the reddest sources we

ﬁnd in M33 (all extreme-AGB “EAGB” stars) and for the optically-selected LBV candidates from Massey et al. (2007), which are detected in the MIR imaging. We

ﬁnd that the population of the reddest sources — those most likely to be true analogs of the progenitors of SN 2008S and NGC 300 — is completely distinct from the population of LBV candidates in the primary metrics of color, magnitude, and variability. Indeed, we ﬁnd very few sources with the properties of the SN 2008S and

NGC 300 progenitors.

Section 4.4.1 describes our procedure for extracting point sources and the resulting catalog. In §4.4.2 we present the color-magnitude diagram for M33 and

96 we discuss the reddest sources vis ´avis the optically-selected LBV candidates from

Massey et al. (2007). In §4.4.3 and §4.4.4 we discuss their SEDs and variability, respectively. Finally, in §4.4.5 we discuss a preliminary search for similar sources in

NGC 300, the LMC, and the SMC.

4.4.1. Catalog

We coadded six epochs of mid-IR imaging of M33 obtained between Jan. 9,

2004 and Feb. 4, 2006 with IRAC (3.6-8.0µm; see McQuinn et al. 2007 for details of the observing program). We produced the coadds in all four IRAC channels from the ﬂux calibrated mosaics provided by the Spitzer Science Center (post-BCD data). Our ﬁnal mosaics cover an area of ∼ 33′ × 33′ (1650 × 1650 pixels, with

′′ ′ 1.2/pix) centered on M33, this is approximately within R25 (≃ 35 ; de Vaucouleurs et al. 1991). We performed source detection and PSF ﬁtting photometry on the coadds using the DAOPHOT/ALLSTAR package (Stetson 1992). The PSF magnitudes obtained with ALLSTAR were transformed to Vega-calibrated magnitudes using simple zero point shifts derived from aperture photometry (using a 12′′ radius), performed in the original images, of 10 − 20 bright and isolated stars in each band. We estimate errors in the photometric transformations to the Vega system of

0.04 mag in 3.6 µm, 0.05 mag in 4.5 µm, 0.07 mag in 5.8 µm, and 0.07 mag in 8.0 µm channel. For the 3.6 and 4.5 µm bands, the detection limits (3σ) in the coadds are

[3.6] ≃ 18.9 mag and [4.5] = 18.2 mag, respectively. The sample becomes incomplete

97 at [3.6] ≃ [4.5] > 17.1, approximately 0.5 magnitudes deeper than McQuinn et al. (2007). A total of ≈ 80, 000 sources are detected in either 3.6 or 4.5 µm. We cross-matched the two catalogs using a 0.5 pixel (0′′.6) matching radius to obtain a

ﬁnal catalog with ≈ 53, 200 individual sources detected at both 3.6 and 4.5 µm (see

Table 4.1).

There are several reasons for producing new point source and variability (see

§4.4.4) catalogs, given the already existing catalog from McQuinn et al. (2007): (1) because we are speciﬁcally looking for objects with extreme colors, we wanted to be able to relax the criterion for point source detection in all IRAC bands (in the

Figures that follow, all sources of interest are detected with 3σ conﬁdence); (2) we wanted to be able to derive our own upper limits in each band for the same reason;

(3) we wanted full IRAC SEDs, whereas the catalog of McQuinn et al. (2007) does not provide data at 5.8 µm; (4) we wanted to combine the images used by McQuinn et al. (2007) with a sixth archival epoch (PI R. Gehrz; PID 5); (5) for the reddest stars, we wanted to derive full six-epoch lightcurves for a more complete measure of variability. The resulting MIR color-magnitude diagram looks diﬀerent from that in

McQuinn et al. (2007). Of the 18 sources we discuss extensively below, only 2 were detected at both 3.6 and 4.5 µm in the catalog of McQuinn et al. (2007).

98 4.4.2. The Color-Magnitude Diagram

The primary result of this eﬀort on M33 was the production of the MIR color-magnitude diagram (CMD), shown in Figure 4.1, which shows the [3.6] − [4.5] color for all the sources detected at both 3.6 and 4.5 µm as a function of absolute magnitude at 4.5 µm, M4.5. The dashed line marks the 3σ completeness limit in this plane. The Spitzer colors and magnitudes of the progenitors of SN 2008S and NGC

300 are shown for comparison (ﬁlled triangle and ﬁlled square, respectively; Prieto et al. 2008c, Prieto 2008d). Note that there are remarkably few objects inhabiting the bright and (very) red region of the CMD. Among the ∼ 5 × 104 massive stars in M33

(see §4.5), ≈ 18, 186, and 567 have both M4.5 < −10 and [3.6] − [4.5] color larger than 1.5, 1.0, and 0.7, respectively, which correspond to blackbody temperatures of

≈ 500, 700, and 1000 K, respectively. A total of 2264 point sources are detected with

[3.6] − [4.5] ≤ 0.7 and M4.5 < −10.

Figure 4.2 shows an expanded view of the brightest MIR sources. In addition to the progenitors of SN 2008S and NGC 300, we include several well-studied LBVs

(e.g., η-Carina), and a number of cool hyper-giants (VY CMa, NML Cyg, and

Var A in M33) for comparison. The MIR magnitudes and colors for η-Carina,

VY CMa, and NML Cyg were synthesized from ISO spectra (Sloan et al. 2003).

The magnitudes of M33 Var A were obtained from Humphreys et al. (2006). Also included are the 16 sources matched between the catalog of MIR sources presented

99 here and the LBV sample of Massey et al. (2007), obtained from narrow-band Hα imaging, using a 0′′.6 matching radius. The larger circles within the dotted lines show the EAGB stars, which we discuss in detail below.

The primary point of Figures 4.1 and 4.2 is to show that there are very few massive stars with the colors and MIR luminosities of the progenitors of SN 2008S and NGC 300. Although it is diﬃcult to identify a quantitative criterion for inclusion in the class deﬁned by the progenitors of SN 2008S and NGC 300, it is clear from

Figure 4.1 that the number of analogs in color and magnitude is very small with respect to the total number of massive stars in M33. For example, if to be included as an analog to the SN 2008S and NGC 300 progenitors we require that M4.5 be brighter than or equal to the progenitor of NGC 300 and we require that the color be redder than the lower limit on the progenitor of SN 2008S, we find a single source. If we require M4.5 < −10 and color redder than SN 2008S, we find just two sources. Casting the net more widely, for the purpose of having a sample larger than one or two objects and in an effort to be conservative, we use M4.5 < −10 and

[3.6] − [4.5] > 1.5 to identify a sample of 18 EAGB stars (large open circles). We discuss the spectra and variability properties of these sources in §4.4.3 and §4.4.4.

Our choice of the cuts M4.5 < −10 and [3.6] − [4.5] > 1.5 to identify objects of interest is somewhat arbitrary. Because our argument in this paper relies on the fact that analogs to the progenitors of SN 2008S and NGC 300 are intrinsically rare

(§4.1 & §4.5), this issue deserves discussion. The magnitude limit is straightforward:

100 it is meant to select objects that have bolometric luminosities indicative of massive stars (∼> 8M⊙). In §4.4.3 below, we show that M4.5 < −10 is conservative; only half of the 18 sources selected have bolometric luminosities large enough to be massive stars. Note that had we required M4.5 to be brighter than or equal to the NGC 300 progenitor, we would exclude 8 of our 18 sources (see Fig. 4.5). Our goal was to not miss any deeply-embedded massive stars and the criterion M4.5 < −10 accomplishes that goal.

The cut on color is more complicated. We were motivated by several factors.

First, we wanted to avoid the AGB sequence blueward of 1.5, where the density of points increases dramatically and where the sample would consist largely of Carbon stars (see Figs. 4.1 & 4.2). Second, this cut essentially eliminates contamination from background active galaxies (Stern et al. 2005). Third, the [3.6] − [4.5] > 1.5 cut gives us a reasonable number of objects to assess individually — it is neither too many, nor too few (again, had we taken [3.6] − [4.5] > 2.0 the sample would consist of just one or two objects).

These considerations leave us open to the potential criticism that optically- obscured massive stars may exist in the region M4.5 < −11.5 (again, avoiding the

AGB feature in the CMD) and with 0.5 < [3.6]−[4.5] < 1.5. To address this, we have examined the 45 sources that occupy this region of the CMD. Four are identiﬁed with the LBV candidate catalog from Massey et al. (2007), which we discuss more below. Sixteen of the remaining 41 sources have bright optical counterparts from

101 the catalog of Massey et al. (2006). Of the 25 sources that do not appear in Massey et al. (2006), 16 are optically-detected, but at ﬂux levels where the catalog is highly incomplete. That leaves 9 optically-obscured sources that have MIR luminosities indicative of massive stars. Like the very brightest of the 18 sources within the dotted lines in Figure 4.2, about half of these have very large MIR luminosities, considerably larger than the progenitors of SN 2008S and NGC 300. We conclude that in this region there is just a handful of sources that might be true analogs to these progenitors — and we note that these have colors ∼ 0.5 − 1.0 magnitudes bluer than the lower limit on the progenitor of SN 2008S. Indeed, self-obscuration to the extent of the NGC 300 and SN 2008S progenitors is exceedingly rare for the most luminous stars, as evidenced by the lack of objects in the upper right corner of

Figures 4.1 and 4.2.

Finally, we note that we have searched for 4.5 µm sources without 3.6 µm detections that would lie within the dotted lines in Figure 4.2 and we ﬁnd just one source. Close inspection of the images reveals a marginal 3.6 µm detection and

[3.6] − [4.5] ≈ 1.5.

Despite this long discussion of color and magnitude selection, the primary point of Figures 4.1 and 4.2 still stands: there are remarkably few massive stars in M33 that have the color and luminosity of the progenitors of SN 2008S and NGC 300. As we discuss in §4.4.3, all are consistent with relatively low-mass massive stars.

102 The second point to note from Figures 4.1 and 4.2 is that the EAGB stars we have selected are not optically-luminous LBVs, η-Carina analogs, or cool hyper- giants. Indeed, all of the LBV candidates (open triangles) have [3.6] − [4.5] ∼< 0.8, and about half have colors ∼< 0.3. The cool hyper-giants and η-Carina-like objects are also bluer than the EAGB population, and considerably brighter than the progenitors of SN 2008S and NGC 300. Indeed, the latter are most naturally associated in this diagram with the luminous red extremum of the AGB population, hence our use of “extreme-AGB” (EAGB) stars.

The color-color diagram for all the sources detected in the four IRAC bands is shown in Figure 4.4. The symbols are the same as in Figure 4.2. The small points with extremely red [5.8] − [8.0] colors are relatively dim, with M4.5 > −10 and are likely young stellar objects (YSOs; e.g., Bolatto et al. 2007). The strong deviation of the SN 2008S and NGC 300 progenitors from the blackbody curve

(solid line) reinforces the fact that the SEDs of these sources are not well-ﬁt by a simple blackbody (Prieto et al. 2008c). Despite this, the SN 2008S and NGC 300 progenitors, EAGBs, LBV candidates, and cool-hypergiants do not stand out as separate populations in [5.8] − [8.0] color. This is important because it means that

[5.8] − [8.0] color alone cannot be used as a metric for inclusion or exclusion from the class of SN 2008S/NGC300-like progenitors.

103 4.4.3. Spectral Energy Distributions

The limits on the optical emission from the progenitors of NGC 300 and SN

2008S are tight, and eﬀectively rule out an optically unobscured massive star (Berger

& Soderberg 2008; Prieto et al. 2008c; Fig. 4.5). For the purposes of ﬁnding analogs to these sources, it is critical to derive the optical luminosities and/or upper limits for the 18 EAGBs identiﬁed in §4.4.2. Here, we present these results and we compare the derived optical-to-MIR SEDs of the EAGBs to the optically-bright (narrow band

Hα-selected) LBV candidates from Massey et al. (2007).

As part of the Survey of Local Group Galaxies Currently Forming Stars

(SLGG), Massey et al. (2006) presented a catalog of ∼ 150, 000 point sources in M33 with well-calibrated UBVRI photometry obtained from observations at the KPNO

4m telescope with the Mosaic imager. We use the published photometric catalog and images from Massey et al. (2006) to complement the Spitzer MIR photometry described in §4.4.2.

We ﬁrst cross-matched the positions of the EAGB stars with the Massey et al. (2006) photometric catalog. Importantly, we do not ﬁnd any optical counterpart to the EAGB sources using a matching radius of 0′′.5. Since the completeness of the catalog starts to decline rapidly at V ≃ 22 mag, we analyze the images independently to look for faint optical counterparts to the EAGB stars. We use SEXtractor (Bertin

& Arnouts 1996) with a low detection threshold (2σ above the local background)

104 to detect and measure aperture photometry (using a small aperture of 3 pixels radius) of all the sources detected in the KPNO/Mosaic UBV RI images of M33 from Massey et al. (2006). We calibrate the photometry relative to the magnitudes in the catalog of Massey et al. (2006). Using a radius of 0′′.5 to cross-match the MIR positions of the EAGB stars with our multi-band catalogs of faint optical sources, we only detect two EAGB stars in the BVR bands (again, 2σ). The remaining

17 sources do not have optical counterparts. We estimate 3σ upper limits on the

UBV RI magnitudes using the local background RMS at the positions of the EAGB stars. The median 3σ upper limits are: U = 24.0, B = 24.0, V = 23.5, R = 23.0,

I = 22.5. The data for each of the 18 EAGB stars is listed in Table 4.2.

In order to convert MIR and optical magnitudes to ﬂuxes, we used zeropoints in Reach et al. (2005) and Cohen et al. (2003) for the EAGBs and LBV candidates, respectively. The luminosities of all sources were calculated assuming a constant reddening of E(B −V ) = 0.15 mag and a distance of µ = 24.92 (Bonanos et al. 2006).

The reddening correction is motivated by the uncorrected B − V color-magnitude diagram of Massey et al. (2006), which shows that the bluest sources only reach

B − V ≈ −0.2, instead of ≈ −0.33, as would be expected from an un-reddened massive star. In addition, Bonanos et al. (2006) also quote an average reddening correction of 0.1 mag to massive stars for M33. A larger adopted reddening correction increases our upper-limits for the optical ﬂuxes of the EAGB stars.

105 The primary result of this procedure is Figure 4.5, which shows the SEDs of the EAGB stars (left panel) and LBV candidates (right panel). The dotted lines in the left panel show the range of upper limits (and one BVR detection, as described above) for the 18 EAGB stars at UBV RI (see Table 4.2). The ﬁlled triangles show the optical and 3.6 µm upper limits for the progenitor of SN 2008S. The solid squares show the MIR detections and optical upper limits for the NGC 300 progenitor.

These sources should be contrasted with the optically-luminous LBV candidates from Massey et al. 2007 (right panel). Note that 3 of the LBVs do not have 5.6 and

8.0 µm detections. The fact that one of the non-detections would appear to have a

5.6 and 8.0 µm flux larger than some of the other detections is a consequence of the locally higher MIR diffuse flux near that particular object.

There are a number of points to take away from the two panels of Figure 4.5.

First, 9 of the 18 EAGB stars we identiﬁed in Figure 4.1 do not have bolometric

4 luminosities indicative of massive stars; they have Lbol ∼< 2 × 10 L⊙. Thus, these are not likely to be true analogs to the SN 2008S and NGC 300 progenitors. Second, all

−2 of the sources are highly optically-obscured, with λLλ[V ]/λLλ[4.5 µm] ∼ 10 . Third, these sources are qualitatively diﬀerent from the more bolometrically luminous LBV candidates (right panel). The LBVs are interesting in their own right, dividing approximately into two classes: (1) relatively optically-dim with a strong MIR excess and (2) optically-bright with little MIR excess, if at all. This division is also

106 evidenced by their positions in the CMD (Fig. 4.1), which indicates a bimodality in

MIR color. For a possible analog to LBVs with a MIR excess, see Smith (2007).

4.4.4. Variability

Because four epochs of archival Spitzer data were available for NGC 6946 in the three years prior to the discovery of SN 2008S, Prieto et al. (2008c) investigated potential variability of the progenitor. They found that there was remarkably little and showed that this fact could be used to constrain the motion of the obscuring medium, under the assumption of a geometrically-thin, but optically-thick shell (see

§4.2; Prieto et al. 2008c).

Motivated by this result, and by the fact that six epochs of archival data over two years exist for M33, we investigated the MIR variability of all the sources detected at 3.6 and 4.5 µm. To generate lightcurves, we used the diﬀerence imaging analysis package ISIS, based on the techniques of Alard & Lupton (1998), Alard

(2000).12 For a discussion, see Hartman et al. (2004).

There is a striking diﬀerence between the 4.5 µm light curves of the 18 EAGB stars and the LBV candidates from Massey et al. (2007). Most of the EAGB stars are highly variable, both in magnitude (0.1 ∼< RMS(mag) ∼< 0.8) and in [3.6] − [4.5] color. In contrast, the majority of the LBVs are not variable (RMS ∼< 0.1 mag); only two sources show clear variability (both are blue in [3.6] − [4.5]). To illustrate these

12 See http://www2.iap.fr/users/alard/package.html.

107 diﬀerences, we show in Figure 4.6 the light curve at 4.5 µm and color variations of the reddest source in our EAGB sample S1 (left panel) and an LBV candidate

(right panel). For completeness, we present all of the 4.5 µm light curves and color variations of the 18 EAGB stars in Appendix A (Figures A.1 & A.2), and of the

LBVs in Appendix B (Figures B.1 & B.2).

Figure 4.7 summarizes these ﬁndings. It shows the measured RMS at 4.5 µm as a function of [3.6] − [4.5] color for all the bright sources with M4.5 < −10. The symbols are the same as in Figures 4.2 and 4.4. There is a clear correlation evident between the RMS (or amplitude) and color for the AGB stars (see also McQuinn et al. 2007).

For comparison, the RMS variation of the progenitor of SN 2008S derived from its 3-year light curve is also shown. In the case of the NGC 300 progenitor we can only put a lower limit on its RMS variation, because only two epochs of archival

Spitzer imaging exist. Nevertheless, it is striking that both the SN 2008S and the NGC 300 progenitors are consistent with very little variation in the few years preceding their explosions. In particular, SN 2008S is inconsistent with the clear trend among the AGB stars to become more variable as they become redder. Only a handful of the EAGB stars vary so little, which suggests that a lack of variability among an otherwise variable EAGB star population may be used as a selection criterion for analogs to the SN 2008S and NGC 300 progenitors. As an example, requiring the RMS to be ∼< 0.3 magnitudes, we ﬁnd 5 sources. They are S12, S7,

108 S9, S17, and S8, in order of increasing RMS (see Table 4.2). Sources S8 and S9 are among the lower luminosity sources in the left panel of Figure 4.5, and are therefore not likely true SN 2008S analogs. In contrast, S7 is the brightest of our EAGB stars.

Finally, the least variable source (S12) has MV ≈ −11.6, which is quite close to the

SN 2008S progenitor, even though it is ∼ 0.5 magnitudes bluer.

Although we have identified a few rare and interesting sources, the sparsity of data in Figures 4.2 and 4.7 with [3.6] − [4.5] > 1.5 makes it difficult to construct a strict quantitative joint criterion in the space of luminosity, color, and variability for inclusion in the class of SN 2008S-like progenitors. Based on the discussion above, as in §4.4.2 and §4.4.3, we expect a total of ∼ 1 − 10 in M33. A more complete multi-epoch survey of EAGB stars in the local universe may fill in the region in

Figure 4.7 between the AGB locus and the SN 2008S and NGC 300 progenitors.

Most importantly, it might make clear a quantitative criterion for “SN 2008S-like” in the RMS-color plane.

4.4.5. Other Galaxies

Blum et al. (2006) and Bolatto et al. (2007) present Spitzer point source catalogs for the LMC and SMC, respectively. We have searched these catalogs for sources that satisfy the selection criteria M4.5 < −10 and [3.6] − [4.5] > 1.5 used to identify the 18 EAGB stars discussed throughout this section. In the catalog of

Blum et al. (2006), we ﬁnd 9 sources. Three are coincident with 2MASS sources and

109 5 appear in the IRAS catalog. Although more careful follow-up is clearly required, a subset of these 9 sources may be EAGB stars. In the catalog of Bolatto et al. (2007) for the SMC, we ﬁnd a single source that satisﬁes M4.5 < −10 and [3.6] − [4.5] > 1.5.

Finally, we have also completed a cursory search for EAGB stars in archival imaging of NGC 300 (PI R. Kennicutt; ID 40204), and we ﬁnd just 4 potential sources.

In sum, even the relatively conservative criteria M4.5 < −10 and [3.6]−[4.5] > 1.5 pick out remarkably few stars in any galaxy — M33 is not peculiar in this regard.

Given the fact that we expect only a fraction of the 18 sources in M33 to be bona

ﬁde analogs to the progenitors of SN 2008S and the transient in NGC 300 (based on luminosity, color, and variability; see §4.4.3, §4.4.4), a more careful look at the sources of interest in the LMC (≈ 9), SMC (≈ 1), and NGC 300 (≈ 4) is likely to further decrease the total number of sources of interest in these systems. As emphasized in §4.1 and in §4.5 below, the scarcity of SN 2008S-like progenitors with respect to the total massive star population is remarkable in light of the fact that SN

2008S-like transients are likely to be relatively common with respect to the overall supernova rate.

4.5. Discussion

We have shown that in the primary metrics of color, luminosity, and variability, stars analogous to the progenitors of SN 2008S and NGC 300 are exceedingly rare in star-forming galaxies. They have luminosities characteristic of low-mass massive

110 stars, are deeply dust-obscured with extremely red MIR colors, and show little MIR variability (see Figs. 4.1, 4.2, 4.5, and 4.7). In luminosity and color they are distinct from the population of optically-luminous LBV candidates selected from Massey et al. (2007). Although many of the reddest objects selected as EAGB stars in Figure

4.2 are highly variable, the few (∼ 1 − 5) least-variable sources most closely resemble the SN 2008S and NGC 300 progenitors. In this way (but in only this way), they are similar to the LBV candidates.

In this section we discuss the implications of our ﬁnding that stars with characteristics analogous to the progenitors of SN 2008S and NGC 300 are rare. In

§4.5.1 we estimate the overall fraction of massive stars that are deeply dust-embedded and the lifetime of stars in that state. In §4.5.2, we connect with the evolution of massive stars, including the possibility that SN 2008S-like transients are the result of electron-capture supernovae or massive white dwarf birth. Section 4.5.3 discusses how many EAGB stars can be found within the local universe (D ∼< 10 Mpc) using

Spitzer.

4.5.1. Numbers & Rates

The total number of analogs to the progenitors of SN 2008S and NGC 300 in M33 is uncertain. This uncertainty comes primarily from the fact that we are unable to identify an absolute quantitative criterion for inclusion into this progenitor class based on our three primary metrics: color, luminosity, and variability. To

111 be conservative, we have identiﬁed 18 sources in the region M4.5 < −10 and

[3.6] − [4.5] > 1.5 of the CMD that satisfy the minimal criteria of being bright and extremely red (larger open circles in Fig. 4.2). However, a very strict cut in color and magnitude (i.e., all sources redder than the lower limit to SN 2008S and brighter than NGC 300) yields just two sources. Among the 18 selected sources, we have shown that roughly half do not have bolometric luminosities indicative of massive stars. That is, they do not have luminosities as large as one would expect for stars

4 who are traditionally thought to end their lives as supernovae (Lbol ∼> 4 × 10 L⊙).

It is important to note, however, that at ﬁxed ﬁnal luminosity of a massive star its initial ZAMS stellar mass may be multi-valued, implying progenitors with either

4 ∼ 5 − 7, ∼ 8 − 9, and ∼ 11 − 14 M⊙ for L ≈ 5 − 10 × 10 L⊙. (see Fig. 2 of Smartt et al. 2009; §4.5.2; see footnote 4). Of course, the explosions SN 2008S and NGC

300 may not be true supernovae, but rather a new class of bright eruptions from obscured massive stars (see §4.5.2). Nevertheless, comparing these 18 sources to the progenitors of SN 2008S and NGC 300 in Figure 4.5 we would argue that only roughly half belong to this progenitor class based on SED alone. Finally, Figure

4.7 shows that only ≈ 1 − 5 of the 18 sources vary as little as the progenitors of

SN 2008S and (potentially) NGC 300. Importantly, 16 of the 18 sources satisfy the criterion of being highly optically-obscured (see Table 4.2).

In summary, very few massive stars have the color, luminosity, and variability of the SN 2008S and NGC 300 progenitors. Our best guess is that the number of

112 true analogs may be as few as zero and as large as ∼ 10 − 20. We denote this number in M33 — the number of true analogs — as NEAGB. A larger sample of stars, culled from a larger multi-epoch study of local star-forming galaxies (see §4.5.3) is clearly needed to ﬁll in the parameter space in the extreme red and bright side of the CMD.

This is the most robust way to understand NEAGB and its uncertainty.

In order to evaluate the fraction of stars in M33 that might be analogs to the progenitors of SN 2008S and NGC 300, and so constrain the rate of production of such objects, we must ﬁrst estimate the total number of massive stars in M33

(N⋆; i.e., with MZAMS ∼> 8M⊙). This number can be estimated in several ways: (1) extinction-corrected Hα luminosity (e.g., Hoopes et al. 2001; Hoopes & Walterbos

2000; Greenawalt 1998), (2) dust-reddening corrected UV continuum luminosity (for

GALEX observations, see Thilker et al. 2005), (3) total number of main-sequence optical point sources detected with MV ∼< −2 (appropriate for stars with ZAMS masses above ≈ 9 − 10 M⊙; Lejeune & Schaerer 2001), or (4) total number of red supergiants (RSGs) (MV ∼< −3.5) times the ratio of the lifetime of a massive star to the time spent as an RSG (t⋆/tRSG ≈ 10; e.g., Schaller et al. 1992). Using the latter method, and selecting RSGs with V − R > 0.5 and MV < −3.5 from the

4 catalog of Massey et al. (2006), we ﬁnd ≈ 5400 sources, implying N⋆ ≈ 5.4 × 10 .

Taking a more conservative color cut of V − R > 0.7 and MV < −3.5, we ﬁnd that

4 4 N⋆ ≈ 3.5 × 10 . Similar estimates in the range of N⋆ ≈ 3 − 6 × 10 are obtained using method (3) with the Massey et al. (2006) catalog, although this estimate suﬀers

113 4 signiﬁcantly from incompleteness. We take N⋆ = 5 × 10 as a ﬁducial number and include it in our scalings below. Note that estimates of the total star formation rate

−1 in M33 range from ∼ 0.3 to ∼ 0.7 M⊙ yr , consistent with the UV, Hα, and FIR luminosities (e.g., Gardan et al. 2007 and references therein), implying a supernova rate for the galaxy of ∼ 0.005 yr−1 (e.g., Gordon et al. 1998).

4 Taking NEAGB ∼ 5 and N⋆ ≈ 5 × 10 , we ﬁnd that a fraction

4 NEAGB −4 NEAGB 5 × 10 fEAGB = ∼ 1 × 10 (4.1) N⋆ 5 " N⋆ # of the massive stars in M33 may be analogs to the progenitors of SN 2008S and NGC

300.

As noted in §4.1 (point 2) and §4.3, only a fraction of all massive stars go through this highly dust-enshrouded phase, and produce transients like SN 2008S and NGC 300. Since, by assumption, roughly all of the massive stars in any galaxy become normal core-collapse supernovae (but, see Kochanek et al. 2008), the rate of

SN 2008S-like explosions can be characterized by their fractional rate with respect to the overall supernova rate. This fraction is determined by dividing the observed rate of SN 2008S-like transients by the total number of supernovae within some volume, times an incompleteness correction that accounts for the fact that SN 2008S-like transients are intrinsically less optically luminous. Based on the numbers presented in §4.3, we estimate that fSN ≈ 0.2, although higher and lower values are not

114 excluded. For example, it is possible that SN 2008S-like transients are intrinsically rare and the fact that NGC 300 and SN 2008S occurred in the same year was simply chance. Although we cannot exclude this possibility, we note that such an explanation appears improbable in the face of what is known about the rarity of their progenitors (fEAGB; eq. 4.1). Conversely, it is possible that the incompleteness correction exceeds the factor of ∼ 2 advocated in §4.3 within 10 Mpc and that such transients are indeed common with respect to supernovae. However, it then becomes increasingly diﬃcult to explain why no more SN 2008S-like transients were observed within 10 Mpc in the last 10 years. There is no way to circumvent these uncertainties without a more complete census of progenitors and outbursts.

As stated in §4.1, the simplest explanation for the fact that SN 2008S- and

NGC 300-like transients are simultaneously common with respect to supernovae

(fSN ∼ 0.2) and that their progenitors are very rare by number at any moment,

−4 in any star-forming galaxy (fEAGB ∼ 10 ) with respect to massive stars, is that a signiﬁcant fraction of all massive stars (∼ 0.2) go through a brief evolutionary epoch in which they are highly dust-obscured, just before explosion. Taking the average lifetime of massive stars with ZAMS masses in the range of 9 − 10 M⊙ to

7 be t⋆ ≈ 3 × 10 yr (e.g., Schaller et al. 1992), we ﬁnd that the duration of this dust-obscured phase is

t N 5 × 104 0.20 ∼ × 4 ⋆ EAGB tEAGB 1 10 7.5 yr. (4.2) "10 yr# 5 " N⋆ #" fSN # 115 We consider the uncertainty in fSN to be at the factor of two level and the uncertainty in N⋆ to be at the level of a factor of 1.5. However, as we have stressed, NEAGB may

3 be as much as a factor of 5 or more lower (tEAGB ∼< 10 yr), or a factor of ∼ 2 − 4

4 higher (tEAGB ∼ 6 × 10 yr). To improve these numbers signiﬁcantly, a careful monitoring program for optical transients like SN 2008S within the local universe

(D ∼< 10 Mpc), coupled with a survey of all local galaxies for bright MIR point sources with (warm) Spitzer (see §4.5.3) should be undertaken. The combination of watching for more transients of this type and associating them with individual progenitors whose luminosities and variability have been cataloged will signiﬁcantly decrease the uncertainty in both fSN and NEAGB, and signiﬁcantly increase our understanding of the causal mapping between progenitors and their outbursts.

4.5.2. Connection to The Evolution of Massive Stars

The relation between ﬁnal luminosity and initial stellar mass may be triply-valued (Smartt et al. 2009) at ∼ 5 − 7, ∼ 8 − 9, and ∼ 11 − 14 M⊙ for

4 L ≈ 5 − 10 × 10 L⊙. This relation is of course uncertain, particularly in the mass range singled out by the bolometric luminosity of the 2008S and NGC 300 progenitors near ∼ 10 M⊙. It is likely further complicated by binarity, and by the mass-loss history, metallicity, and rotation of massive stars. Because the absolute rate of these outbursts as well as whether or not they should be associated with

116 the death of the progenitor are still uncertain, we consider a number of potential scenarios below. We list a subset of the possibilties in order of increasing progenitor mass.

Massive White Dwarf Birth: M ≈ 6 − 8 M⊙

We have referred to the progenitors of SN 2008S and NGC 300 throughout this work as extreme (“E”) AGB stars because they lie at the red extremum of the AGB sequence in the MIR color-magnitude diagram (Figs. 4.1 & 4.2). Taken literally, these stars may indeed be the progenitors of the most massive O-Ne-Mg white dwarfs, undergoing explosive core-envelope separation as they transition to proto-planetary nebulae (e.g., Riera et al. 1995; Garc´ıa-Hern´andez et al. 2007).

Perhaps the 2008S and NGC 300 progenitors were then akin to the most massive highly-evolved carbon- or oxygen-rich AGB stars (Kwok 1993).

Based on analogy with local proto-planetary nebulae, we would expect bi-polar explosion morphology and eventually the emergence of a hot ionizing continuum source as the newly-born white dwarf begins its cooling phase (perhaps similar to Hen 3-1475/IRAS 17423-1755; Riera et al. 2003). The initial luminosity of

4 the central source would be of order ∼ 5 × 10 L⊙ for a white dwarf near the

Chandrasekhar mass and it should cool on a timescale comparable to ∼ 105 yr.

Thus, the bolometric luminosity of the transient should eventually decrease back to approximately pre-outburst levels. The primary distinguishing characteristic of this

117 particular scenario is the (eventual) emergent hot continuum source and emission lines, bi-polar morphology, and the fact that the bolometric luminosity should not decrease to pre-outburst levels in the next decades (e.g., Kwok 1993).

Electon-Capture Supernova: M ≈ 9 M⊙

The timescale estimated in equation (4.2) is of the right order of magnitude to be associated with the onset of carbon burning in relatively low-mass massive stars.

This is traditionally a very diﬃcult phase to model (see the summary in Woosley et al. 2002; Siess 2006; Poelarends et al. 2008).

One of the most intriguing explanations for the physics of SN 2008S-like transients is that they result from electron-capture SNe (ecSNe) of O-Ne-Mg cores of relatively low-mass massive stars (Miyaji et al. 1980). While speculative, this explanation accounts for many of the observed characteristics of both the transients and their progenitors. In particular, it accounts for the fact that the progenitors of

NGC 300 and SN 2008S were relatively low luminosity and deeply embedded. Here we follow the scenario detailed by Poelarends et al. (2008) (see also Nomoto 1984a,

1987; Ritossa et al. 1996, Seiss 2006, 2007; as well as Chugai 1997b; Wheeler et al. 1998; Woosley et al. 2002; Eldridge et al. 2007; Wanajo et al. 2009).

We know from the properties of the observed progenitors and our analysis of the luminous stars in M33 that the progenitors are extreme AGB stars. In these systems, the combination of thermal pulses due to He shell burning and dredge-up produces a

118 massive, dusty wind (for lower luminosity and less enshrouded analogs, see the work on carbon stars in the Magellanic clouds by Groenewegen et al. 2007, as well as van

Loon et al. 2005, 2006). In the Poelarends et al. (2008) models, the mass loss peaks

−4 −1 5 at nearly 10 M⊙ yr for stars of mass M ≃ 9M⊙ and luminosity L ≃ 10 L⊙, and then drops precipitously for slightly more massive stars which can support core Neon burning, and which eventually become normal iron-core core-collapse supernovae

(ccSNe) (see Fig. 13 from Poelarends et al. 2008). The thermal pulses driving the mass loss occur at very high rates for the EAGB stars (on timescales of years), suggesting that mass loss may appear as a steady wind, as seems to be required for the small variability in the lightcurves of the SN 2008S and NGC 300 progenitors

(see Fig. 4.7; discussion in Prieto et al. 2008c), rather than as impulsive ejections of optically-thick shells expected for normal carbon stars. The low degree of variability seen in SN 2008S (particularly when contrasted with the EAGB stars in Fig. 4.7) might also be explained by the onset of core carbon burning as the ﬁnal phases of stellar evolution commence. Note that the work of Nomoto (1984a, 1987) implies that the EAGB envelope and mass-loss would be carbon-enhanced.

For a narrow mass range near 9 M⊙, the balance between mass loss and the growth of the core allows the core to become unstable to collapse without igniting core

Neon burning, leading to an ecSNe. This occurs only for a small model-dependent mass range near 9 M⊙ (see also, e.g., Nomoto 1984a; Seiss 2006, 2007). Poelarends et al. (2008) estimated a mass range of Mmin ≃ 9M⊙ to Mmax ≃ 9.25 M⊙. For

119 a standard Salpeter IMF, and assuming that all stars from Mmax to 20M⊙ form ccSNe, then the ecSN fraction is just ≈ 6% (9.0 ≤ M ≤ 9.25 M⊙). This is below our (albeit, uncertain) ﬁducial estimate for the rate of SN 2008S-like transients relative to the normal ccSN rate, fSN ∼ 0.2. However, other studies have found somewhat broader mass ranges for ecSNe. For example, Seiss (2007) ﬁnd that

Mmax − Mmin ≈ 1 − 1.5M⊙, which implies a fractional rate for ecSNe more in accord with our nominal estimate for fSN.

While the observed rate of SN 2008S-like transients is uncertain, we have argued that they represent a modest fraction of the normal ccSN rate, consistent with a limited progenitor mass range. The relatively low luminosity of their progenitors implies that they are low-mass massive stars, potentially near the boundary between electron-capture and normal core-collapse supernovae. In addition, Kitaura et al. (2006) argue that ecSNe should be sub-luminous compared to normal ccSNe, because of their low Ni yields, potentially explaining the low luminosity of SN

2008S-like transients. Finally, for the ﬁducial ecSN model of Poelarends et al. (2008)

(see also Nomoto 1984a), the AGB phase lasts for ≈ 4 × 104 yr, which although short with respect to the lifetime of the star itself, is of order tEAGB in equation (4.2) based on the number of analogs to the SN 2008S and NGC 300 progenitors in M33.

Of course, these timescales need not be identical, since the ﬁducial ecSN progenitors of Poelarends et al. may evolve signiﬁcantly in color as a result of mass-loss during

120 their super-(“extreme”) AGB phase, becoming increasingly like the SN 2008S and

NGC 300 progenitors as they approach the end of their lives.

Fortunately, this speculative explanation has at least one simple and testable prediction: there should be no surviving progenitor once the transient fades. This may be testable in the optical, since most of the dust enshrouding the progenitor was likely destroyed by the explosion, but observations in the MIR will be required to be certain that a new shroud has not formed. A second test is to ﬁnd strong evidence in the late-time lightcurve for synthesized 56Ni. This may be diﬃcult both because ecSN may produce little 56Ni (Kitaura et al. 2006), and because dust, whether that remaining from the EAGB phase or dust formed in the ejecta (e.g.,

Prieto et al. 2008e for M85), may make it diﬃcult to correctly measure the late-time decay rate. Spitzer may again be key in constraining the nature of these events because of this obscuration. A third test is to ensure that Spitzer has surveyed all the nearby galaxies so that future examples of these transients can be causally connected to their deeply obscured progenitors. Finally, as with ccSNe, very nearby ecSNe (D ∼< 300 kpc) should produce neutrino signatures characteristic of neutron star formation, which detectors such as SuperKamiokande and its successors would observe (see, e.g., Thompson et al. 2003; Kistler et al. 2008).

The recent paper by Botticella et al. (2009) in part corroborates the interpretation discussed in Prieto et al. (2008a) and proposed here that SN 2008S may be an electron-caputre supernova. They present the late-time quasi-bolometric

121 lightcurve of SN 2008S, which shows evidence for a power-law time dependence with a slope indicative of being powered by the radioactive decay 56Co. Although this need not uniquely signal ecSN as the physical mechanism (see §4.5.2), it provides some evidence for core-collapse, and something perhaps akin to normal neutron star formation (Kitaura et al. 2006). A lightcurve with similar cadence and photometric coverage has recently been published in Bond et al. (2009) for NGC 300.

Intrinsically Low-Luminosity Iron Core-Collapse Supernova:

M ∼ 10 − 12 M⊙

Heger et al. (1997) discuss a mechanism for generating a potentially obscuring

“superwind” via pulsational mass-loss in red supergiants between 10 and 20 M⊙ during the last 104 yr before explosion. The prediction of enhanced AGB-like obscuration (they compare directly with luminous OH/IR stars), the evolutionary timescale (compare their 104 yr with our eq. 4.2), and the secular increase in the fundamental mode pulsation period as the star approaches death (see their Fig. 7) are all in good agreement with the requirements on the 2008S and NGC 300 progenitors we discuss in this paper.

The physical mechanism of iron core-collapse supernovae is unknown (Rampp

& Janka 2000; Liebend¨orfer et al. 2001; Thompson et al. 2003; Buras et al. 2003;

Burrows et al. 2006). Recent observations hint that low-luminosity Type IIP supernovae may be more common than previously thought (e.g., Chugai &

122 Utrobin 2000; Pastorello et al. 2004, 2007), particularly when one accounts for the incompleteness corrections discussed in §4.3. Because the mechanism of supernovae has yet to be conclusively identiﬁed, it is diﬃcult to interpret the diversity in inferred 56Ni yield physically. In fact, that diversity may be larger than previously thought, and we are only now appreciating the existence of a very low-luminosity tail to the Type IIP luminosity function. If so, it is natural to imagine that these low-luminosity core-collapse events might have analogs that occur in the very dusty circumstellar medium of their massive stellar progenitors, as in Heger et al. (1997), and thus may give rise to events like SN 2008S and NGC 300.

This scenario yields many of the predictions of the ecSN scenario discussed in

§4.5.2. Indeed, even with a complete sampling of “ec-” and “cc-” supernovae, it may be diﬃcult to disentangle the two populations since many of the predictions

— radioactive decay powered lightcurves, potentially embedded progenitor, no

“postgenitor” — are the same in both.

Massive Star Outburst: M ≈ 10 − 15 M⊙

On the basis of the relatively low luminosity of their progenitors, we view the ecSN and massive white dwarf birth scenarios discussed above as the most probable explanation SN 2008S and NGC 300. Nevertheless, there is of course the possibility that they are instead a new class of outbursts from relatively low-mass massive stars, potentially analogous to the pulsational instabilities discussed in Poelarends

123 et al. (2008) or Heger et al. (1997). The majority of the true “LBV” eruptions with documented progenitors (e.g., 1997bs, 2002kg) came from optically bright massive stars signiﬁcantly more bolometrically luminous than the progenitors of

SN 2008S and NGC 300. As we have shown in Figures 4.2 and 4.5, the EAGB population is separate from the sources traditionally classiﬁed as LBVs: they are less bolometrically luminous and much more dust-obscured. These facts suggest that if these transients were the outbursts of massive stars then they are distinct from from the classical supernova impostors. If these events are not supernovae, but merely outbursts, then their existence is likely connected to the physics of the transition between stars that become ecSNe and/or ccSNe, and those that do not.

The degree of dust-obscuration at outburst is a crucial clue to their evolution. A simple prediction of this possibility is that the progenitors of SN 2008S and NGC 300 should eventually be re-discovered in the optical and/or infrared after the outburst emission has faded. For further arguments on the nature of SN 2008S and NGC 300 related to this discussion, see Bond et al. (2008), Smith et al. (2009b), and Berger et al. (2008).

4.5.3. A More Complete Census

Equation (4.1) implies that a fraction ∼ 1 × 10−4 of the massive star population in any given galaxy appears to be in the evolutionary state that led to the explosions observed as SN 2008S and NGC 300. The simplest explanation, adopted throughout

124 4 this work, is that the deeply dust-enshrouded phase marks the last tEAGB ∼< 10 yr

(eq. 4.2) in the life of a fraction fSN ≈ 0.2 of the massive star population.

Compilations of star formation and supernova rates in the local universe (e.g.,

Ando et al. 2005) suggest that the latter is ≈ 2 yr−1 within 10 Mpc (see §4.1, point

6 3 2), implying that there are ∼ 5 × 10 massive stars and ∼< 10 EAGB stars within this volume. If the lifetime in the pre-explosion, highly dust-obscured phase is tEAGB

(eq. 4.2), we would expect to see one SN 2008S-like transient every few years, in accord with our estimate for fSN.

A multi-epoch survey of all the local star-forming galaxies within 10 Mpc with Spitzer would allow for a comprehensive census of EAGB stars. It would signiﬁcantly increase our knowledge of the variability properties and SED evolution of these objects, and it might allow us to deﬁne more strict criteria for inclusion in the class of SN 2008S/NGC 300-like progenitors. It would therefore decrease the considerable uncertainty in NEAGB in equations (4.1) and (4.2). Coupled with the supernova surveys in the local volume, such a study would improve our knowledge of the fraction fSN of stars that eventually go through the deeply embedded phase just before explosion. Of course, the most intriguing possibility is that the number of true analogs to the progenitors of SN 2008S and NGC 300 is in fact NEAGB ∼ 0 − 1

3 13 in M33 and that tEAGB ∼< few − 10 yr. If so, the ﬁnal catalog of EAGB stars that

13The lower limit here comes from the fact that pre-explosion imaging of the SN 2008S and NGC 300 progenitors establishes a few year baseline.

125 would be produced by a Spitzer survey would have just ∼ 50 − 100 members. These could be followed up repeatedly, since, given these numbers one would expect to wait just ∼ 10 years before one of these individual sources exploded. This would give a direct observational link in the causal mapping between a sub-population of massive stellar progenitors and their explosions, connecting them with a short timescale.

Indeed, the ability to identify an individual star as marked for imminent death (or eruption) would be an astonishing consequence of this work.

126 Fig. 4.1.— Mid-infrared color-magnitude diagram for M33. Absolute magnitude at 4.5µm is plotted versus [3.6] − [4.5] color for all detected sources (3σ limits denoted by dashed lines; see §4.4.1; Table 4.1). For comparison, the positions of the progenitors of NGC 300 (filled square) and SN 2008S (filled triangle; lower limit) are also plotted (Prieto et al. 2008c; Prieto 2008d). Stars analogous to these progenitors are exceedingly rare. Compare with Figure 4.2. The main sequence, AGB, and EAGB stars are clearly visible. Note the “spur” in the data extending to fainter M4.5 and redder color at [3.6] − [4.5] ≈ 1 and M4.5 ≈ −10.5, originating at the red extremum of the AGB population. To our knowledge, this is the first time that such a feature has been identified in a MIR CMD.

127 Fig. 4.2.— Same as Figure 4.1, but focused on the red and bright region of interest. Here, the [3.6] − [4.5] > 1.5 and M4.5 < −10 selection for extremely red and bright objects is shown explicitly by the dotted lines, as are the 18 EAGB sources in M33 that meet these criteria (large open circles; see §4.4.2). We denote them “S1-S18,” ordered by color. The reddest source, S1, is shown in Figure 4.3. Most of these sources are highly variable (see Figs. 4.7, A.1, & A.2). SEDs are shown in Figure 4.5 (see Table 4.2). The EAGB population is clearly distinct from the optically-luminous LBV candidates from the catalog of Massey et al. 2007, shown by the open triangles (see §4.4; Table 4.3). Most of the LBV candidates are considerably more bolometrically luminous and much less variable at 4.5µm than the EAGB sample (see Figs. 4.5, 4.7, B.1, & B.2). A number of cool hyper-giants such as VY CMa, NML Cyg, and M33 Var A, as well as η-Carina, are shown for comparison. The brightest LBV candidate is M33 Var C. The LBV candidates appear to be bimodal in MIR color (right panel of Fig. 4.5).

128 3.6 4.5

Fig. 4.3.— Image showing the point source S1 at 3.6 µm, 4.5 µm, as well as V and I band (Massey et al. 2006). This is the reddest of the 18 sources in [3.6] − [4.5] color selected in Figure 4.1. It is optically obscured (see Fig. 4.5). S1 is also highly variable in color and 4.5 µm magnitude (see Figs. 4.6 & 4.7).

129 Fig. 4.4.— Color-color diagram showing [5.8] − [8.0] versus [3.6] − [4.5] colors for all (≈ 1800) the sources detected in all four IRAC bands. Symbols are the same as in Figure 4.1. The solid line and open squares show the expectation for a blackbody of temperature TBB = 5000, 1000, 400, and 280 K. The small filled points with [5.8] − [8.0] ≈ 1.6 − 1.8 and [3.6] − [4.5] ∼> 2 are not sufficiently bright at 4.5µm (M4.5 < −10) to be included in the sample defined by the dotted lines in Figure 4.2.

130 Fig. 4.5.— Left Panel: Spectral energy distributions of the 18 EAGB stars in M33 with [3.6] − [4.5] > 1.5 and M4.5 < −10 (see Fig. 4.2; Table 4.2). The lower dotted line shows the best lower limits obtained in the optical, whereas the upper dotted line shows the worst lower limits at U and I and the two BVR detections (see discussion §4.4.3; Table 4.2). The SEDs of the SN 2008S (ﬁlled triangles) and NGC 300 (ﬁlled squares) progenitors are also shown. For all the sources in M33 we assume a total extinction of E(B − V ) = 0.1 mag. Right Panel: Spectral energy distributions of the 16 LBVs detected in MIR from the Hα selected catalog of Massey et al. 2007 (see Fig. 4.2; Table 4.3). The relative increase in λLλ in the R-band (0.6 µm) is due to the presence of strong Hα emission.

131 Fig. 4.6.— Left panel: Lightcurve for S1 (see Fig. 4.3), the reddest of the 18 EAGB stars selected in Figure 4.2. Note the high degree of variability, which for this source, is inconsistent with the progenitor of SN 2008S and (potentially) NGC 300 (see Fig. 4.7). Lightcurves for all of the EAGB stars are shown in Figures A.1 and A.2. Although not strong in this example, most of the sources exhibit correlated color-magnitude variations. Right panel: 4.5 µm lightcurve for one of the LBV candidates from Massey et al. 2007 (see Figs. B.1 & B.2 for the complete set). The large majority of the 16 LBV candidates are not highly variable at 4.5 µm, although there are two exceptions (Fig. 4.7).

132 Fig. 4.7.— RMS variation in 4.5µm magnitude as a function of [3.6] − [4.5] color for all of the sources detected in 3.6µm and 4.5µm with M4.5 < −10 (points) and the 16 LBV candidates (open triangles). As in Figure 4.2, the 18 EAGB stars are denoted with larger open circles. As noted in Figures A.1 & A.2, as well as B.1 & B.2, the extreme-AGB stars are highly variable, whereas all but two of the LBV candidates are not. Variability of the progenitors of SN 2008S and NGC 300 is also shown. For the former, the data are taken from Prieto et al. 2008c (their Fig. 2). For NGC 300 only two epochs are available and hence the value of the RMS (≈ 0.05) is a lower limit on the variability of the progenitor.

133 RA Dec [3.6] σ3.6 [4.5] σ4.5 (deg) (deg) (mag) (mag)

23.03815 30.82294 18.70 0.21 18.19 0.12 23.04489 30.82397 17.67 0.07 18.24 0.10 23.04532 30.81396 18.15 0.14 18.59 0.15 23.04906 30.81863 15.69 0.05 16.19 0.07 23.04955 30.82077 17.52 0.08 17.82 0.13 23.05103 30.82549 16.79 0.06 17.13 0.07 23.05131 30.81658 17.96 0.09 18.33 0.12 23.05277 30.79529 16.50 0.07 17.13 0.07 23.05287 30.80321 17.45 0.06 17.94 0.09 23.05303 30.82159 16.96 0.05 17.38 0.07 23.05314 30.78955 18.73 0.11 18.41 0.19 23.05333 30.81954 18.14 0.07 18.13 0.11 23.05470 30.79727 17.57 0.06 18.04 0.12 23.05473 30.81597 17.84 0.07 18.29 0.10 23.05481 30.80367 17.86 0.06 18.14 0.10 ......

Table 4.1. MIR Catalog for 53,194 Point Sources in M33

134 Name RA Dec U a Ba V a Ra Ia [3.6] [4.5] [5.8] [8.0] (deg) (deg) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag)

S1 23.45485 30.85704 23.47 23.76 23.37 22.90 22.47 16.35 14.21 12.55 11.37 S2 23.44397 30.79731 23.96 24.15 23.55 23.15 22.47 16.84 14.85 13.52 12.42 S3 23.56813 30.87755 23.91 23.97 23.48 23.10 22.50 16.56 14.73 13.42 12.55 S4 23.43452 30.57106 23.87 23.74 23.21 22.71 22.02 15.87 14.15 13.23 12.07 S5 23.40436 30.51738 22.99 23.10 22.97 22.71 22.10 15.08 13.38 12.12 11.32 S6 23.49194 30.82791 24.10 23.84 23.42 22.89 22.09 16.51 14.84 13.46 12.32 S7 23.55640 30.55211 24.52 24.50 23.86 23.31 22.75 14.29 12.63 11.18 10.17 S8 23.46408 30.64138 22.79b 23.11b 22.32b 21.75b 20.96b 16.17 14.52 14.14 12.87 S9 23.53817 30.73269 21.75b 21.93b 22.14b 22.10b 21.89b 15.67 14.08 12.69 11.51

135 S10 23.29877 30.59901 24.31 23.97 23.67 23.30 22.75 16.35 14.78 13.55 12.37 S11 23.29777 30.50744 23.82 23.64 23.53 23.05 22.83 15.30 13.72 12.47 11.26 S12 23.55236 30.90564 24.20 23.96 23.50 23.21 22.62 14.90 13.33 12.01 11.08 S13 23.37907 30.70096 24.05 24.14 23.60 23.03 22.41 16.38 14.82 13.44 12.36 S14 23.26238 30.34469 23.96 23.95 23.48 23.03 22.47 16.33 14.79 13.61 12.43 S15 23.39709 30.67737 24.13 24.02 23.29 22.80 22.23 15.25 13.71 12.22 10.97 S16 23.43722 30.64242 23.75 23.39 22.79 21.99 21.39 16.30 14.76 14.15 12.76 S17 23.47176 30.67430 23.81 23.67 23.03 22.51 21.60 15.27 13.74 13.15 11.56 S18 23.34248 30.64602 24.21 23.95 23.65 23.22 22.63 14.86 13.32 12.03 10.88

aExcept where otherwise noted, all UBV RI data in this table are upper limits.

bSource detections. Magnitudes from Massey et al. (2006).

Table 4.2. Photometry for the 18 EAGBs in M33 Identiﬁer U B V R I [3.6] [4.5] [5.8] [8.0] (from Massey et al. 2007) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag)

J013248.26+303950.4 16.13 17.32 17.25 17.01 16.89 15.70 15.36 15.02 14.57 J013324.62+302328.4 18.66 19.49 19.58 19.27 19.53 14.26 13.55 13.04 12.27 J013333.22+303343.4 18.30 19.30 19.40 18.44 18.89 13.09 12.50 12.20 11.65 J013335.14+303600.4 15.54 16.53 16.43 16.30 16.14 14.45 14.35 14.48 14.15 J013341.28+302237.2 15.18 16.24 16.29 16.28 16.32 15.08 14.96 15.07 14.45 J013350.12+304126.6 15.76 16.85 16.82 16.43 16.30 12.22 11.65 11.06 10.64 J013406.63+304147.8 15.12 16.26 16.08 15.86 15.76 14.41 13.95 13.83 ··· a J013410.93+303437.6 15.13 16.13 16.03 15.87 15.70 14.84 14.65 14.69 13.97 136 J013416.10+303344.9 16.32 17.17 17.12 16.96 16.85 15.69 15.48 15.01 ··· a J013422.91+304411.0 16.36 17.28 17.21 17.14 17.07 16.24 16.35 ··· a ··· a J013424.78+303306.6 16.21 16.97 16.84 16.72 16.54 15.24 15.00 14.74 13.97 J013426.11+303424.7 18.38 19.23 18.97 18.59 18.27 14.57 13.79 13.09 12.24 J013429.64+303732.1 16.24 17.12 17.11 17.05 16.93 16.55 16.64 ··· a ··· a J013442.14+303216.0 18.35 18.20 17.34 16.89 16.44 13.56 13.02 12.39 11.62 J013459.47+303701.9 17.35 18.59 18.37 17.88 17.69 14.42 13.73 13.16 12.31 J013500.30+304150.9 18.37 19.22 19.30 18.60 19.11 13.93 13.24 12.54 11.93

aNon-detection.

Table 4.3. Photometry for the 16 LBV Candidates in M33 from Massey et al. (2007) Chapter 5

A Spitzer/IRS Spectrum of the 2008 Luminous Transient in NGC 300: Connection to Proto-Planetary Nebula

5.1. Introduction

An intriguing luminous optical transient was discovered in the nearby galaxy

NGC 300 (hereafter NGC 300-OT) by the amateur astronomer B. Monard on May

14.1, 2008 (Monard 2008). The transient was faint compared to normal core-collapse supernovae, with an absolute magnitude at discovery of MV ≃ −13 (Bond et al. 2008). The optical spectrum obtained by Bond et al. (2008) close to discovery was dominated by relatively narrow Hydrogen Balmer and Ca II lines (infrared triplet and forbidden doublet) in emission, as well as strong Ca II H&K in absorption.

Shortly after discovery, Berger & Soderberg (2008) reported strong upper limits on the optical luminosity of the progenitor star obtained from deep archival HST data, which led them to suggest that the progenitor was a low-mass main sequence star and the transient was a stellar merger, similar to the red Galactic nova V838

Monocerotis (e.g., Bond et al. 2003).

137 However, Prieto (2008d) reported the discovery of a luminous mid-infrared

(mid-IR) progenitor to the transient in archival Spitzer images. The progenitor was a luminous dust-enshrouded star, whose spectral energy distribution was consistent

4 with a black-body of R ≃ 300 AU radiating at T ≃ 300 K, with Lbol ≃ 6 × 10 L⊙.

This discovery showed that NGC 300-OT was connected to an energetic explosion in a relatively low mass ∼ 10 M⊙ star. The relatively low luminosity of the transient compared to normal core-collapse supernovae, spectral properties, and dust-enshrouded nature of the progenitor star, made NGC 300-OT a “twin” of

SN 2008S (Prieto et al. 2008c; Prieto 2008d), which was discovered earlier in 2008 in the galaxy NGC 6946 (Arbour & Boles 2008; Stanishev et al. 2008; Chandra &

Soderberg 2008; Steele et al. 2008; Yee et al. 2008; Wesson et al. 2008).

There have been a number of studies of NGC 300-OT and SN 2008S, and the nature of these transients is still under debate (Prieto et al. 2008c; Thompson et al. 2008; Smith et al. 2009b; Bond et al. 2009; Berger et al. 2009; Boticella et al. 2009; Wesson et al. 2009). Thompson et al. (2008) present and discuss various possible physical mechanisms that can explain these transients and the likely range of main-sequence masses of the dusty progenitor stars that are consistent with the observations of NGC 300-OT and SN 2008S: (1) massive white-dwarf birth (MZAMS ≈ 6 − 8 M⊙); (2) electron-capture supernova (MZAMS ≈ 9 M⊙); (3) intrinsically low-luminosity iron core-collapse supernova (MZAMS ≈ 10 − 12 M⊙); and (4) massive star outburst (MZAMS ≈ 10 − 15 M⊙). Any of these potential

138 scenarios suggests that these transients are very important for our understanding of the evolution of stars at the dividing line between “high” and “low” mass (i.e.,

8 − 10 M⊙).

Here we report on a low-resolution mid-IR spectrum of NGC 300-OT obtained with Spitzer on August 14, 2008, 93 days after the discovery and 113 days after the

ﬁrst detection (Monard 2008). The transient is luminous in the mid-IR spectral range and shows broad emission features that we interpret as signs of carbon-rich dust, similar to the spectra of carbon-rich proto-planetary nebulae in the Galaxy.

The paper is organized as follows. In §5.2, we discuss the observations and data reduction. In §5.3, we present the analysis of the spectrum and spectral energy distribution of the transient. In §5.4, we discuss the implications of our ﬁndings.

Hereafter we adopt a distance of 1.88 Mpc to NGC 300 (Gieren et al. 2005).

5.2. Spitzer Observations

We observed NGC 300-OT with the Short-Low (SL; 5.2 − 14 µm,

R = ∆λ/λ = 60 − 120) module of the Infrared Spectrograph (IRS; Houck et al. 2004) on August 14.4, 2008 (UT). The observations were obtained in staring mode as part of a Spitzer Director’s Discretionary Time (DDT) proposal request

(PID 487; AOR key 28139008). The ramp time was set to 60 sec in both SL orders

(SL1: 7.4 − 14 µm, SL2: 5.2 − 7.5 µm) and 10 cycles were obtained, for a total

139 exposure time of 1200 sec on source, including the two nod positions (total of 20 images).

To reduce the data, we started from the basic calibrated data (BCD) from the

Spitzer Science Center pipeline (S18.1.0). We constructed a high S/N background image for each order and nod by median combining all the other images. We subtracted the background from the individual 2D images. Rogue pixels in the background-subtracted images were cleaned with IRSCLEAN (v1.9). We used the routines profile, ridge, extract and tune in the Spitzer IRS Custom Extractor

(SPICE) software package in order to extract ﬂux-calibrated 1D spectra. The spectra of each nod were median combined and the two orders were merged together after applying a small multiplicative correction factor of 3.5% to the SL1 spectra.

Figure 5.1 shows the final combined spectrum with ±1σ error bars on the fluxes estimated from the RMS in each pixel. The mean signal-to-noise ratio of the final spectrum is ≃ 35 per pixel.

5.3. Analysis

5.3.1. Spectral Features

The Spitzer spectrum of NGC 300-OT presented in Figure 5.1 shows two prominent broad emission features at ≈ 8.3 µm and ≈ 12.2 µm. There is also

140 a relatively narrow, but resolved, fainter feature at ≈ 9.7 µm. The significance of this faint feature is quite uncertain and depends sensitively on the spectra used to obtain the final combined spectrum. The main properties of the emission features (central wavelength, FWHM, and integrated fluxes) present in the mid-IR spectrum are shown in Table 5.1. They were obtained after fitting Gaussians to the continuum-subtracted spectrum. The continuum was modeled using a high-order

(6th) polynomial ﬁt over the wavelength regions: λ ≤ 7.5 µm, 9.2 − 9.3 µm,

10.25 − 10.6 µm, and λ ≥ 13.4 µm. We obtain consistent results if we use a spline function to model the continuum.

In Figure 5.2 we compare the Spitzer spectrum of NGC 300-OT with mid-IR spectra of type IIP supernovae. The spectra of SN 2004et (Kotak et al. 2009) and

SN 2005af (Kotak et al. 2008) were obtained from the Spitzer archive (PID 237,

20256). Unlike the late-time mid-IR spectra of normal type IIP supernovae (e.g.,

SN 2004dj, Kotak et al. 2007; SN 2005af, Kotak et al. 2008; SN 2004et, Kotak et al. 2009) and SN 1987A (e.g., Roche et al. 1993; Wooden et al. 1993) that are dominated by narrow ﬁne-structure lines of stable Ni, Ar, Ne, Co, and some molecular SiO in emission at ∼ 8 − 9 µm, the mid-IR spectrum of NGC 300-OT presents broad features that are most likely dominated by emission from dust grains in the circumstellar environment. The non-detection of ﬁne-structure lines of Fe-peak elements in emission suggests that the main source of dust heating at this epoch is

141 not the decay of radioactive 56Ni. This is consistent with the low 56Ni production estimated by Botticella et al. (2009) in the case of a supernova explosion.

In Figure 5.3 we compare the Spitzer spectrum of NGC 300-OT with mid-IR spectra of three evolved massive stars that have circumstellar dust. The spectrum of the yellow-hypergiant IRC+10420 (e.g., Humphreys et al. 2002) was obtained from the ISO catalog of SWS spectra (Sloan et al. 2003). The spectra of the yellow-hypergiant M33 Var A (Humphreys et al. 2006) and the LMC B[e] supergiant

R66 (Kastner et al. 2006) are from the Spitzer archive (PID 5, 3426). The spectra of

IRC+10420, M33 Var A and R66 are dominated by the amorphous silicate emission feature at 9.7 µm, characteristic of oxygen-rich dust. The spectrum of R66 also contains emission features of polycyclic aromatic hydrocarbons at 6.2, 7.7, 8.6 and

11.3 µm (PAHs), indicating the presence of carbon-rich dust as well. It is clear from

Figure 5.3 that the mid-IR spectrum of NGC 300-OT does not resemble the spectra of these evolved massive stars with circumstellar dust, even though the optical spectrum of the transient is strikingly similar to IRC+10420 (Bond et al. 2009; see

Smith et al. 2009b for the case of SN 2008S).

In Figure 5.4 we compare the spectrum of NGC 300-OT with mid-IR spectra of

Galactic proto-planetary nebulae (pPNe). Two of the pPNe have carbon-rich dust

(IRAS 20000+3239 and IRAS 13416-6243) and the others have oxygen-rich dust

(IRAS 15452-5459 and IRAS 17150-3224). The spectra are all from the ISO/SWS catalog (Sloan et al. 2003). The spectra of oxygen-rich pPNe contain SiO absorption

142 at 7.9 µm and a strong silicate absorption feature at 9.7 µm, which do not appear to be present in the spectrum of NGC 300-OT. Also the central wavelengths and

FWHM of the two “bumps” at ∼ 8.5 µm and ∼> 12 µm are inconsistent with the features in the spectrum of NGC 300-OT.

The spectrum of NGC 300-OT is most similar to the spectra of carbon-rich pPNe in Figure 5.4. They contain broad emission features at ∼ 8 µm and

∼ 12 µm, which have been associated with C-C and C-H bending and stretching modes identiﬁed as the carriers of PAHs (e.g., Duley & Williams 1981; Peeters et al. 2002). Note, however, that the spectrum of NGC 300-OT does not contain the 6.2 µm PAH feature that is clearly present in the two carbon-rich pPNe. We can put a 3σ upper limit on the integrated ﬂux of the 6.2 µm PAH feature of

−14 −2 −1 I6.2 < 2.1 (FWHM/0.2 µm) × 10 erg cm s . This gives a 3σ limit on the

ﬂux ratio of I6.2/I8.3 < 0.09 for an assumed FWHM = 0.2 µm. The spectrum of

IRAS 20000+3239 also has the 6.9 µm PAH feature, which is not present in the spectrum of NGC 300-OT. We can put a 3σ upper limit on the integrated ﬂux of the

−14 −2 −1 6.9 µm PAH feature of I6.9 < 1.9 (FWHM/0.2 µm) × 10 erg cm s , which gives a limit on the ﬂux ratio I6.9/I8.3 < 0.08 for an assumed FWHM = 0.2 µm.

143 5.3.2. Spectral Energy Distribution

We can construct the full spectral energy distribution (SED) of NGC 300-OT at the epoch of the Spitzer spectrum using the optical and near-IR photometry presented in Bond et al. (2009). Figure 5.5 shows the optical to mid-IR SED of the transient 93 days after the discovery date. We have corrected all the ﬂuxes for a total extinction along the line-of-sight of E(B − V ) = 0.25 mag, which is the mean of the extinction values reported in Bond et al. (2009). We assume RV = 3.1 and use

−1.6 the Schlegel et al. (1998) reddening law, with Aλ ∝ λ in the near-to-mid infrared range. The ﬁlled circles are the optical (BV RI) and near-IR (JHK) ﬂuxes from

Bond et al. (2009). The thick line is the Spitzer mid-IR spectrum. For comparison, we show the SED of the luminous dust-enshrouded progenitor of NGC 300-OT (ﬁlled squares) obtained from pre-explosion Spitzer IRAC (3.6 − 8 µm) and MIPS (24 µm) photometry (see Table 5.2; these data were used in Thompson et al. 2008 and Bond et al. 2009). At the epoch of the Spitzer observation the transient is ∼ 20 times more luminous than the progenitor at 8 µm. The results of black-body ﬁts to the transient and progenitor SEDs are presented in Table 5.3.

The evolution of the light curves of NGC 300-OT in diﬀerent ﬁlters presented by Bond et al. (2009) shows that the transient becomes redder in time, with the color evolving from V − K ≃ 3.1 mag at discovery to V − K ≃ 5.2 mag at the time of the

Spitzer spectrum. This fast evolution in the V − K color, while the B − V color only

144 changes from ≃ 0.8 mag to ≃ 1.1 mag in the same time period, suggests the presence of warm circumstellar dust formed in the explosion or heated pre-existing dust.

Botticella et al. (2009) analyzed the SED of SN 2008S and showed that the evolution in optical+near-IR ﬂuxes could be explained with a single “hot” black-body until ∼ 120 days after explosion, but they needed a second “warm” black-body component at later times. They concluded that the near-IR ﬂux excess of SN 2008S at ∼> 120 days after explosion was possibly due to newly-formed dust in the ejecta or shock-heated dust in the circumstellar environment.

As in the case of SN 2008S, we find that the BV RIJHK fluxes of NGC 300-OT at 93 days after discovery (113 days after the first detection) can be well-fit by

2 the sum of two black-bodies (χν = 1.8), a hot component with T1 = 3893 K,

R1 = 10.7 AU and a warm component with T2 = 1511 K, R2 = 67.3 AU (see the

6 dotted line in Figure 5.5). The total luminosity of these components is 2.1 × 10 L⊙.

However, the mid-IR SED of NGC 300-OT traced by the Spitzer spectrum is a factor of ∼ 2 − 9 brighter between 5 − 14 µm than the extrapolated sum of the two black-bodies. We add a third black-body component with lower temperature, while keeping the fit to the optical and near-IR fluxes fixed, to account for the mid-IR

2 excess. We ﬁnd a good ﬁt (χν = 1.5) to the continuum of the Spitzer spectrum

(deﬁned in §5.3.1) with T3 = 485 K and R3 = 515 AU. The luminosity of this

5 component is 6.0 × 10 L⊙, which is 29% of the optical+near-IR luminosity and 22% of the total integrated luminosity of the transient.

145 The two black-body components that can reproduce the near-IR and mid-IR excesses in the SED of NGC 300-OT are likely due to emission from circumstellar dust. The hotter component (T2 = 1511 K) can be reasonably explained with newly formed dust in the ejecta, as proposed by Botticella et al. (2009) for SN 2008S.

The velocity inferred from the black-body radius R2 = 67 AU and the time after the ﬁrst detection of the transient1 is ∼ 1000 km s−1. This velocity is in the range of velocities measured from emission lines in the optical spectra of NGC 300-OT of

∼ 100 − 1000 km s−1 (Bond et al. 2009; Berger et al. 2009).

Another way to explain this warm dust component would be emission from pre-existing progenitor dust, although this seems less likely because of dust destruction from the initial outburst light (e.g., Dwek 1983). Assuming a luminosity

6 at maximum light of Lmax ≈ 5 × 10 L⊙ (luminosity at discovery) and a dust sublimation temperature of Tsub ≃ 1500 K, we obtain a radius of the dust-free cavity

0 −1 of ∼ 80 − 250 AU depending on the assumed dust emisivity law, Qλ ∝ λ − λ .

This radius is a factor of ∼ 1.2 − 3.7 times larger than the black-body scale R2 of the ∼ 1500 K temperature dust, which suggest the near-IR emitting dust may have been formed in the ejecta.

1We assume this is a good estimate of the time after explosion. This seems like a reasonable assumption given that the light curve of the transient is rising fast at that time (see Figure 1 in

Bond et al. 2009).

146 We can estimate the total dust mass needed to account for the luminosity of the warm black-body component using equation (3) in Smith et al. (2005) and assuming

−3 an average carbon grain density of ρdust ≃ 2.24 g cm ,

−6 −6 Td Ld ≈ × ⊙ Md 2.5 10 6 M (5.1) 1000 K 10 L⊙ !

where Td is the dust temperature in Kelvin, and Ld is the dust luminosity in L⊙.

This equation assumes that the dust grains have a constant radius a ∼< 0.2 µm, and that the dust emissivity is approximated by the function presented in Gilman

−7 (1974). Using T2 and L2 in Table 5.3, we obtain Md ≈ 2 × 10 M⊙ for the mass of newly formed dust at 93 days after discovery. We can compare this dust mass to that in SN 2008S. Under the same assumptions about dust properties considered in equation (5.1), this dust mass is ∼ 3 times larger than was needed to explain the near-IR excess in SN 2008S ∼ 120 days after explosion, but comparable to the mass needed ∼ 230 days after explosion (using the data in Table 8 of Botticella et al. 2009).

The mid-IR excess revealed by the Spitzer spectrum of NGC 300-OT cannot be explained by newly formed dust. The constant expansion velocity needed to reach

−1 a black-body radius of R3 = 515 AU at this epoch is v ∼ 8000 km s , far larger than the velocities observed in the optical spectra of the transient. This component must be emission from pre-existing dust from the progenitor. Dust grains that were not destroyed by the initial outburst will absorb the outburst light, warm up, and

147 re-radiate at mid-IR wavelengths. Using equation (5.1) we estimate that a dust

−4 mass of Md ≈ 10 M⊙ and dust optical depth of τV ≈ 0.4 is needed to account for the mid-IR excess. For a gas-to-dust mass ratio of 200 that is consistent with observations of evolved stars with carbon-rich dust in the LMC (e.g., Matsuura et

−4 al. 2009), we ﬁnd a gas mass of ≈ 0.02 M⊙. A similar mass of dust of ∼ 10 M⊙ is needed to explain the SED of the progenitor of NGC 300-OT, suggesting that a substantial fraction of the dust in the progenitor wind survives the explosion.

Interestingly, a similar mass of dust is also needed to explain the mid-IR excess observed in SN 2008S ∼ 17 days after explosion (Botticella et al. 2009).

We performed radiation transport calculations using DUSTY (Ivezic &

Elitzur 1997) to check if the SED of NGC 300-OT can also be explained by radiation through a spherical shell of dust. We found a reasonable ﬁt to the optical to mid-IR

SED using a central black-body with T= 4000 − 5000 K illuminating a dust shell

−1 with density proﬁle ρ(r) ∝ r , inner and outer radius of the shell Rin ≃ 100 AU and

Rout ≃ 10000 AU, temperature of the dust at the inner radius T = 1500 K, and dust optical depth τV ≃ 1.2. We note that this solution is not unique since we also obtain a reasonable ﬁt to the SED with ρ(r) = constant, Rin ≃ 100 AU, and Rout/Rin ∼ 10, with all the other parameters being equal. Wesson et al. (2009) analyzed the optical to mid-IR SED of the progenitor of SN 2008S and the transient ∼ 17 days and

∼ 180 days after explosion using a radiative transfer code. They ﬁnd that the SED of SN 2008S can be explained with a central source illuminating a spherical dust

148 −2 shell with density proﬁle ρ ∝ r , inner radius Rin ≃ 1250 AU, and τV ≃ 0.8, where

−5 −3 ∼> 98% of the pre-existing progenitor dust (Mdust ≃ 1.2 × 10 − 3.5 × 10 M⊙, depending on Rout) survives the explosion. This is qualitatively similar to our results for NGC 300-OT.

5.4. Discussion & Conclusions

We have presented a low-resolution mid-IR spectrum of NGC 300-OT obtained with Spitzer on August 14.4, 2008, 93 days after the discovery of the transient. We now present our discussion and interpretation of the results.

5.4.1. Mid-IR Spectrum and SED of NGC 300-OT

The mid-IR spectrum of NGC 300-OT shows broad emission features at 8.3 µm and 12.2 µm that are similar to the broad features seen in the spectra of carbon-rich pPNe in the Galaxy (e.g., Kwok 1993; Kwok et al. 2001), called “Class C” PAH sources by Peeters et al. (2002). Joblin et al. (2008) derive proﬁles for the broad

8 µm and 12 µm features from the spectrum of the archetypal “Class C” PAH source, the pPN IRAS 13416-6243 (see Figure 5.4). These broad features are attributed to hydrocarbons with a predominantly aliphatic nature, which undergo photochemical processing in proto-planetary nebula to transform into the more aromatic material observed in carbon-rich planetary nebulae (e.g., Kwok et al. 2001; Sloan et al. 2007).

149 Joblin et al. (2008) show that these broad features are also observed in young planetary nebulae, and are distinct from the spectral features of neutral PAHs, ionized PAHs, and very small grains.

It is interesting to note that the position of the center of the PAH complex at 7 − 8 µm, observed in many astrophysical environments including the ISM and evolved stars (e.g., Tielens 2008 and references therein), has been shown to correlate with the effective temperature of the host star in Herbig A/Be stars, planetary-nebulae, and pPNe (e.g., Sloan et al. 2007; Keller et al. 2008; Boersma et al. 2008). The correlation goes in the sense that stars with lower effective temperatures (i.e., weaker UV-optical radiation field) show the central peak of this complex at longer wavelengths. The central wavelength at 8.3 µm detected in the

Spitzer spectrum of NGC 300-OT would imply an eﬀective temperature of ∼ 4000 K

(see Fig. 8 in Keller et al. 2008), which is consistent with the temperature of the hot black-body (T ≃ 3900 K) derived from the optical SED of the transient. This provides indirect and independent support for our interpretation of the mid-IR spectrum, and also evidence that UV processing has not yet converted predominantly aliphatic hydrocarbons into PAHs in NGC 300-OT.

A noteworthy diﬀerence between the Spitzer spectrum of NGC 300-OT and the spectra of carbon-rich pPNe is the non-detection of the 6.2 µm PAH emission feature to fairly deep limits. In the carbon-rich pPN IRAS 13416-6243, for example, the ratio of the integrated ﬂuxes of the 6.2 µm and 8 µm features is I6.2/I8 ≃ 0.13, which

150 is a factor of 1.3 higher than the 3σ limit for NGC 300-OT. The 6.2 µm emission feature is thought to be produced by the C-C stretching mode in ionized PAHs. The astronomical and laboratory spectra of PAHs and PAH-like molecules show such a wide variety that the absence of the 6.2 µm feature may be explained by diﬀerences in shape, ionization state, impurities, and size of the molecules (e.g., Pathak &

Rastogi 2008; Bauschlicher et al. 2009). On the other hand, the non-detection of the

6.9 µm PAH emission feature in the spectrum of NGC 300-OT seems to be consistent with the integrated ﬂux of this feature measured in some Galactic carbon-rich pPNe like IRAS 20000+3239. This emission feature is thought to be associated with aliphatic material (C-H bending mode) and is detected only in a fraction of pPNe

(e.g., Kwok et al. 1999).

The mid-IR excess traced by the Spitzer spectrum can be well explained by the

−4 presence of warm circumstellar dust (T ∼ 500 K) with mass Md ∼ 10 M⊙. This dust was most likely part of the dusty progenitor wind, pre-existing the luminous explosion that produced the optical transient. The SED of NGC 300-OT at the epoch of the Spitzer spectrum also shows a near-IR excess, which can be explained

−7 with a small mass (∼ 2 × 10 M⊙) of warm circumstellar dust (T ∼ 1500 K) formed in the ejecta.

Alternatively, the SED of NGC 300-OT can be reasonably well explained with a T = 4000 − 5000 K black-body illuminating a spherical shell of pre-existing progenitor dust that extends from ∼ 100 AU to ∼ 10000 AU, where the inner radius

151 of the dust shell marks the destruction of dust by the initial outburst light. The presence of a substantial mass of pre-existing dust from the progenitor wind in the overall SED of NGC 300-OT was also characteristic of SN 2008S (Wesson et al. 2009) and indicates that most of the dust survived the explosion.

5.4.2. NGC 300-OT and SN 2008S: Connection to

Proto-Planetary Nebulae

The similarity of the mid-IR spectrum of NGC 300-OT with carbon-rich pPNe is very striking and may shed new light on the nature of this transient,

SN 2008S, and the other optical transients (e.g., SN 1999bw, M85-OT) that show similar characteristics and appear to be part of the same class (Prieto et al. 2008e;

Thompson et al. 2008). The optical spectra of NGC 300-OT and SN 2008S transients were compared with the spectrum of the massive Galactic yellow-hypergiant star

IRC+10420 (Smith et al. 2009b; Bond et al. 2009; Berger et al. 2009), which shows an F-type supergiant spectrum with Balmer lines in emission, as well as strong Ca II triplet, [Ca II] doublet, and [Fe II] lines in emission. Given the similarity of the mid-IR spectrum of NGC 300-OT with pPNe, we have searched in the literature for their optical spectra. We found several examples of pPNe with optical spectra that are remarkably similar to NGC 300-OT and SN 2008S. In the sample of echelle long-slit spectra of evolved stars of S´anchez Contreras et al. (2008), the Galactic

152 pPNe IRAS 17516-2525 (O-B spectral type), M1-92 (B2-F5), Hen 3-1475 (Be),

IRAS 22036+5306 (F4-7), and IRAS 08005-2356 (F4 Ie) show Balmer lines and also strong Ca II triplet and [Ca II] doublet in emission. Other examples can be found in the atlas of optical spectra of post-AGB stars presented in Su´arez et al. (2006).

The presence of forbidden Ca II in emission in the spectra of NGC 300-OT and

SN 2008S, pPNe, and IRC+10420, which is rarely present in stellar spectra, means that calcium is not depleted onto dust grains, most likely due to the destruction of grains by relatively fast shocks (e.g., Hartigan et al. 1987).

Another similarity between NGC 300-OT and pPNe is revealed by the kinematics and the detection of double-peaked Balmer and Ca II triplet lines in the spectrum of NGC 300-OT. Bond et al. (2009) interpreted these double features as the presence of a bipolar outﬂow with an expansion velocity of ≈ 75 km s−1, and possibly faster components moving at ∼ 200 km s−1. Berger et al. (2009) discussed evidence for even faster velocity components (including inﬂow) going up to ∼ 1000 km s−1.

Aspherical winds or outﬂow moving at velocities of a few × 100 km s−1 up to

∼ 1000 km s−1 are observed in pPNe (e.g., Balick & Frank 2002). Multiple studies using high-resolution imaging of Galactic pPNe with HST have shown a variety of complex morphologies, with bipolar, multipolar and point-symmetric structures (e.g.,

Sahai et al. 1999). In particular the pPN Hen 3-1475 (also classiﬁed as a young PN in some studies), which is in the spectroscopic sample of S´anchez Contreras et al. (2008) and whose spectrum shares many features with the spectra of NGC 300-OT, has

153 a bipolar morphology and velocity components up to ∼ 1200 km s−1 (e.g., Riera

4 et al. 1995, 2003). Given the inferred luminosity of ∼ 10 L⊙, chemistry, and kinematics, Riera et al. (1995, 2003) proposed that Hen 3-1475 was a relatively high-mass star (∼ 3−5M⊙) in the post-AGB phase of evolution. Another interesting example is the Red Rectangle, an extensively studied intermediate-mass pPNe with carbon-rich dust chemistry and a fast (∼ 560 km s−1) bipolar outﬂow traced by Hα in emission (e.g., Witt et al. 2009).

Several studies have proposed that SN 2008S and NGC 300-OT were the result of an energetic eruption in a dust-enshrouded 10 − 20 M⊙ star, where the star survives the eruption. Smith et al. (2009b) discussed a super-Eddington wind as the physical mechanism that produced SN 2008S, similar to the super-outbursts of massive LBVs (e.g., van Marle et al. 2008). Berger et al. (2009) presented possible observational evidence for this model from the complex kinematics that they inferred from their high-resolution spectra of NGC 300-OT. Bond et al. (2009) did not require that the progenitor of the transient was LBV-like, but rather an

OH/IR star (e.g., Wood et al. 1992) that was evolving to warmer temperatures (in a blue-loop) at the time of the eruption. These studies relied heavily on comparing the optical spectra of the transients with the spectrum of the massive yellow-hypergiant

IRC+10420 (e.g., Humphreys et al. 2002; Davies et al. 2007). However, as discussed here, there are examples of pPNe in the Galaxy that share very similar optical spectroscopic characteristics with NGC 300-OT and SN 2008S. In fact, the complex

154 model of inﬂow-outﬂow put forth in Berger et al. (2009) to explain the spectra of

NGC 300-OT has been discussed in the context of fast winds of AGB and post-AGB stars in binaries (Soker 2008).

Finally, in a mid-IR study of massive stars in the LMC, Bonanos et al. (2009) argued that B[e] supergiants (e.g., R66 in Figure 5.3) may share a common origin with NGC 300-OT and SN 2008S. Supergiant B[e] stars in the LMC are

4 very rare (only ∼ 10 discovered), have luminosities Lbol ∼> 10 L⊙, and dusty circumstellar envelopes, properties that are broadly consistent with the properties of the progenitors of NGC 300-OT and SN 2008S. However, the circumstellar dust around B[e] supergiants in the LMC is signiﬁcantly hotter (∼> 800 K) than the dust around the progenitors, probably because they are oxygen-rich.

In summary, we have shown that NGC 300-OT and SN 2008S have several properties (mid-IR spectrum, optical spectra, kinematics, and dusty circumstellar medium) that are characteristic of pPNe in the Galaxy; they are not unique to massive stars like IRC+10420.

5.4.3. The Progenitors of NGC 300-OT and SN 2008S:

Massive Carbon-rich AGB/post-AGB stars ?

4 The progenitors of NGC 300-OT and SN 2008S were luminous (∼ 4−6×10 L⊙) dust-enshrouded stars with warm (T∼ 300 − 450 K) circumstellar dust, found at the

155 red extremum of the AGB sequence in a mid-IR color-magnitude diagram (Thompson et al. 2008). They are part of the extreme-AGB (EAGB) sequence, which has been identiﬁed as a continuation of the AGB to redder mid-IR colors in resolved stellar population studies of nearby galaxies using Spitzer (e.g., LMC, Blum et al. 2006;

M33, Thompson et al. 2008). Their location in the mid-IR color-magnitude diagram indicates extreme mass-loss and relatively cool circumstellar dust (e.g., Srinivasan et al. 2009). Interestingly, Matsuura et al. (2009) ﬁnd that most EAGB stars in the LMC sample for which they have obtained mid-IR spectra have carbon-rich dust, which is consistent with the evidence presented here for carbon-rich dust in

NGC 300-OT. Even though the number of carbon-rich AGBs in the LMC declines as a function of luminosity with respect to oxygen-rich AGBs, interpreted as evidence of Hot-Bottom-Burning which converts carbon into nitrogen and oxygen, there are carbon stars with luminosities approaching those of the progenitors of NGC 300-OT and SN 2008S (e.g., van Loon et al. 1997; Frost et al. 1998). One example is

4 IRAS 05278-6942, a carbon-rich AGB star in the LMC that has Lbol ∼ 4 × 10 L⊙

−5 −1 and M˙ ∼ 3 × 10 M⊙ yr (Groenewegen et al. 2007). Indeed, Kastner et al. (2008) in their study of the most luminous 8 µm sources in the LMC, point out that “more high-Lbol carbon stars may lurk among the very red, unclassiﬁed objects” in their sample.

The high-mass counterparts of AGB stars with MZAMS ≃ 8 − 10 M⊙, so-called super-AGB stars, have been proposed as good candidates for the progenitors of

156 NGC 300-OT and SN 2008S (see Thompson et al. 2008 and references therein;

Botticella et al. 2009). These stars end up with an O-Ne core and, depending on the competing eﬀects of core-growth after carbon ignition and strong mass-loss, they can explode as electron-capture supernovae in a narrow and uncertain mass range around

∼ 9 M⊙ or end-up as O-Ne white dwarfs at lower masses (e.g., Nomoto 1984a;

Poelarends et al. 2008). The luminosities of SAGB stars in theoretical models

5 can reach ∼ 10 L⊙ at the end of their evolution (e.g., Siess 2007), comparable to the luminosities of the progenitors of NGC 300-OT and SN 2008S. These models also predict that the photospheric abundances of SAGB stars should be oxygen-rich (C/O < 1) at the end of their evolution, through a combination of

Hot-Bottom-Burning and the occurrence of the third dredge-up. However, the modeling of these processes in the AGB and SAGB evolution is very uncertain and depends on several important factors like metallicity, the treatment of convection, mass-loss, and the input opacities (e.g., Marigo 2008). In fact, there are theoretical studies that have discussed the possibility of carbon-rich photospheres in massive

AGBs (e.g., Nomoto 1987; Marigo 2007).

An important diﬀerence between the progenitors of the luminous transients and carbon-rich AGB and EAGB stars is that they did not show variability in the mid-IR within 3 − 4 years of explosion (Prieto et al. 2008c; Thompson et al. 2008), whereas most AGB and EAGB stars are highly variable (e.g., Gronewegen et al. 2007; Vijh et al. 2009). Since variability in AGB stars is explained by pulsations that drive the

157 mass-loss (thermal pulses), the lack of variability in the progenitors may indicate that they were at the very tip of the AGB or SAGB phase before the explosion, perhaps past a super-wind phase. If this is the case, the progenitor could be classiﬁed as a pPN (i.e., it was in the post-AGB phase at the time of the explosion).

5.4.4. Progenitors and Transients: Concluding Remarks

The physical mechanism that produced the energetic explosions

(∼ 2 − 6 × 1047 erg in optical to near-IR light) of NGC 300-OT and SN 2008S is still unknown. Although the observations presented here do not directly shed light on the mechanism that produced the transients, we have shown that all the observations of the transients and their progenitors presented thus far are consistent with the explosion of a massive (MZAMS ∼ 6 − 10 M⊙), carbon-rich AGB, super-AGB or post-AGB star, either single or in a binary. An in-depth discussion of some of the mechanisms that could explain the transients can be found in Thompson et al. (2008). Here we brieﬂy comment on the ones that involve a massive AGB or post-AGB star: white dwarf formation and an electron-capture supernova.

In the case of an energetic eruption where the progenitor survives the explosion, the transients could mark the birth of massive white dwarfs (Thompson et al. 2008).

Observations of mass-losing AGB stars show spherically symmetric envelopes, while their descendants (proto-planetary and planetary nebulae) have highly asymmetric

158 and complex morphologies and kinematics. This has been a long standing mystery in stellar evolution for which several mechanisms have been proposed, with magnetic

ﬁelds, rotation and binaries suggested as primary suspects for breaking the symmetry

(e.g., Balick & Frank 2002; de Marco 2009; Sahai et al. 2009 and references therein).

In a recent study, Dennis et al. (2008) argue that pPNe outﬂows may be driven by an explosive MHD launch mechanism similar to the ones discussed in the context of supernovae and gamma-ray bursts. This model seems appealing when applied to

NGC 300-OT and SN 2008S – perhaps we are witnessing the launch of the jet in a massive AGB which is shaping a pPN. In this scenario we expect that the pPN now in formation will become a PN when the central white dwarf left behind ionizes the surrounding gas. The timescale for this is very uncertain, but for a ∼ 8 M⊙ star it can be of the order of ∼ 100 yr (e.g., Stanghellini & Renzini 2000). An interesting prediction of an asymmetric outﬂow that can be tested with new observations is the detection of strongly polaraized light from the transient.

An electron-capture supernova in a massive AGB star has been suggested as a possible mechanism for NGC 300-OT and SN 2008S (e.g., Prieto et al. 2008c;

Thompson et al. 2008; Botticella et al. 2009). Two of the main predictions of this scenario that can be tested with late-time observations are the disappearance of the progenitor star years after the explosion and the detection of radioactive 56Ni decay synthesized in the explosion. Botticella et al. (2009) presented detailed photometric and spectroscopic observations of SN 2008S. Their main argument in favor of a

159 supernova explosion as the origin of the transient was presented in the late-time light curve. They found that the pseudo-bolometric light curve at t ∼> 140 days had a decay slope consistent with radioactive decay of 56Co → 56Fe and inferred the

−3 56 production of ∼ 10 M⊙ of Ni in the explosion. However, Smith et al. (2009b) noted that the late time light curve of SN 2008S was slower than expected from 56Co decay and argued against a supernova interpretation. While a late-time decay slope slower than 56Co is a possibility in SN 2008S, it should be pointed out that slow late-time light curve slopes (compared to 56Co decay) have also been observed in some subluminous type IIP supernovae, including the very well-studied SN 2005cs

(Pastorello et al. 2009). Therefore, we think that a supernova explosion origin cannot be excluded from this result alone. The late-time light curve of NGC 300-OT should give very important clues about the possible supernova origin.

160 Fig. 5.1.— Low-resolution Spitzer/IRS mid-IR spectrum of the NGC300-OT obtained on 2008 August 14.4, 93 days after the transient was discovered. The vertical error bars are the RMS of each pixel obtained after combining the 20 spectra.

161 Fig. 5.2.— Comparison of the Spitzer/IRS spectrum of NGC300-OT (top) with mid- IR spectra of the type IIP supernovae SN 2004et and SN 2005af. In parenthesis are the days with respect to the explosion date for each spectrum. We have subtracted a linear continuum ﬁt in the wavelength region shown to each spectrum and scaled the ﬂux arbitrarily.

162 Fig. 5.3.— Comparison of the Spitzer/IRS spectrum of NGC300-OT (top) with mid- IR spectra of evolved massive stars that show circumstellar dust emission. These include the yellow-hypergiants IRC+10420 and M33 Var A, and the B[e] supergiant R66. We have subtracted a linear continuum ﬁt in the wavelength region shown to each spectrum and scaled the ﬂux arbitrarily.

163 Fig. 5.4.— Comparison of the Spitzer/IRS spectrum of NGC300-OT (top) with mid-IR spectra of Galactic proto-planetary nebulae. The pPNe have different dust chemistry: carbon-rich (IRAS 2000+3239 and IRAS 13416-6243) and oxygen-rich (IRAS 1545-549 and IRAS 17150-3224) dust. We have subtracted a linear continuum fit in the wavelength region shown to each spectrum and scaled the flux arbitrarily.

164 Fig. 5.5.— SED of NGC 300-OT at 93 days after discovery. The filled circles are the optical (BV RI) and near-IR (JHK) fluxes of NGC 300-OT from Bond et al. (2009). The Spitzer/IRS spectrum is shown as the thick black line. The dotted line is a fit to the optical+near-IR SED of the transient using the sum of two black-bodies with temperatures T1 ≈ 3890 K and T2 ≈ 1500 K. The grey line is an extension of the previous black-body fits to mid-IR by adding a cooler component with T3 ≈ 485 K. We show the SED of the progenitor of NGC 300-OT for comparison (filled squares). The dashed line is a single black-body fit with T ≈ 335 K to the progenitor SED.

165 λc Intensity FWHM (µm) (10−13 erg cm−2 s−1) (µm)

8.33 ± 0.01 2.27 ± 0.20 0.94 ± 0.03 9.71 ± 0.06 0.12 ± 0.11 0.49 ± 0.14 12.16 ± 0.07 0.39 ± 0.14 1.47 ± 0.16

Table 5.1. Features in the Spitzer/IRS Spectrum of NGC300-OT

166 λ λFλ Telescope/Instrument (10−14 erg cm−2 s−1)

3.6 µm 0.75 ± 0.15 Spitzer/IRAC 4.5 µm 4.67 ± 0.43 Spitzer/IRAC 5.8 µm 16.83 ± 1.10 Spitzer/IRAC 8.0 µm 31.49 ± 1.67 Spitzer/IRAC 24 µm 15.03 ± 1.37 Spitzer/MIPS

Table 5.2. Spectral Energy Distribution of the Progenitor of NGC 300-OT

167 Parameter Value

NGC 300-OT BV RIJHK ﬂuxes

T1 (K) 3893

R1 (AU) 10.7 6 L1 (L⊙) 1.1 × 10

T2 (K) 1511

R2 (AU) 67.3 5 L2 (L⊙) 9.8 × 10 NGC 300-OT adding ﬁt to Spitzer data

T3 (K) 485

R3 (AU) 515 5 L3 (L⊙) 6.0 × 10 Progenitor SED T (K) 335 R (AU) 332 4 L (L⊙) 5.7 × 10

Table 5.3. Black-body Fits to the Transient and Progenitor SEDs

168 Chapter 6

A Study of the Type Ia/IIn Supernova 2005gj from X-ray to the Infrared

6.1. Introduction

Thermonuclear supernova explosions (Type Ia supernovae, SN Ia hereafter) are believed to be the detonation or deﬂagration of a white dwarf accreting matter from a companion star (Arnett 1982). The mass of the white dwarf slowly increases until it approaches the Chandrasekhar limit where the star becomes thermally unstable.

At this point fusion of Carbon and Oxygen begins near the center and quickly moves through most of the star before degeneracy is lifted. The result is a spectacular and powerful explosion that is visible across much of the Universe. Since SN Ia arise from a narrow range of white dwarf masses, their peak luminosities are very consistent and they make excellent distance indicators (e.g., Phillips 1993). SN Ia are powerful probes of cosmology and have been instrumental in narrowing the uncertainty in the Hubble parameter, discovery of the accelerating universe, and constraining dark energy models (Hamuy et al. 1995, 1996b; Riess et al. 1998; Perlmutter et al. 1999;

Astier et al. 2006; Wood-Vasey et al. 2007).

169 But the use of SN Ia as reliable distance indicators will always be questioned until the progenitor and explosion physics are well-understood. What types of binaries create SN Ia? How is matter transferred to the white dwarf without causing thermonuclear runaways on the surface? Are there several types of progenitors?.

These big questions remain to be answered and detailed observations of hundreds of events have yielded few clues.

In 2002, Hamuy et al. (2003) identiﬁed a new kind of supernova. The early spectrum of SN 2002ic was a cross between a Type Ia event and a Type IIn (Deng et al. 2004, have called this type of supernova a “IIa”), showing P-Cygni features similar to SN Ia and resolved Balmer lines in emission. Type IIn supernovae are core-collapse explosions going oﬀ in dense circumstellar environments (Schlegel 1990;

Chevalier & Fransson 1994). They are relatively common since the massive stars that create core-collapse supernovae often have thick winds. If the interpretations of the pre-explosion observations of SN 2005gl are correct, SN IIn could be associated in some cases with luminous blue variables (Gal-Yam et al. 2007).

In the case of SN 2002ic, the presence of Balmer lines with proﬁles characteristic of SN IIn, and the high luminosities and slow decline after maximum lead to the conclusion that most of the energy came from the interaction of the ejecta with a dense circumstellar medium (CSM). Other Type IIn events (SN 1997cy, Germany et al. 2000; SN 1999E, Rigon et al. 2003) have been re-classiﬁed as SN 2002ic-like that were caught late in their evolution (Hamuy et al. 2003; Wood-Vasey et al. 2004).

170 SN 2002ic provided the ﬁrst direct evidence that thermonuclear explosions can also occur in a dense medium, but in this case the circumstellar medium is probably generated by an Asymptotic Giant companion (Hamuy et al. 2003; Wang et al. 2004;

Han & Podsiadlowski 2006). However, there is still debate in the literature about the origin of SN 2002ic. Livio & Riess (2003) proposed the merger of two white dwarfs as a possible progenitor, with the explosion occurring in the common envelope phase.

Chugai et al. (2004) concluded that the properties of the circumstellar interaction in 2002ic-like events can be broadly explained by the SN 1.5 scenario (Iben &

Renzini 1983): the thermonuclear explosion of the degenerate core of a massive AGB star. Recently, Benetti et al. (2006b) questioned the earlier interpretation of the observations and proposed that SN 2002ic can be equally well explained by the core collapse of a stripped-envelope massive star in a dense medium.

SN 2005gj was discovered on 2005 September 28.6 (UT) by the SDSS-II

Supernova Collaboration (Frieman et al. 2008) in gri images obtained with the

SDSS-2.5m telescope at Apache Point Observatory (APO). The new supernova

(Barentine et al. 2005) was ∼1′′ from the center of its host galaxy at the position α = 03h01m12s.0, δ = +00◦33′13′′.9 (J2000.0). It had SDSS magnitudes

(g,r,i) = (18.6, 18.6, 18.7) mag, obtained from PSF photometry after kernel-matching and subtraction of a template image in each band. The SN was independently discovered by the Nearby Supernova Factory on 2005 September 29 (Aldering et al.

2006).

171 SN 2005gj was classified as a Type Ia candidate from the first three epochs of the gri light curve using a light curve fitting program, and was sent to the queue of the MDM-2.4m telescope for spectroscopic confirmation. The optical spectrum obtained on 2005 October 1 (UT) showed a blue continuum with resolved Hydrogen

Balmer lines in emission, very similar to the spectrum of a young Type IIn supernova, but also with an unusual continuum showing broad and weak absorption features.

Further spectroscopic follow-up showed a dramatic evolution. The continuum became substantially redder and developed broad, P-Cygni features probably associated with blended lines of Fe-peak mass elements, similar to a Type Ia SN a few weeks after maximum. The spectrum obtained on 2005 Nov. 12 (UT) was remarkably similar to that of the unusual Type Ia supernova SN 2002ic obtained on

2002 Dec. 27 (UT) (Prieto et al. 2005).

Aldering et al. (2006) presented optical photometry and spectroscopy of this

SN. Through detailed analysis they conﬁrmed its photometric and spectroscopic resemblance to SN 2002ic, conﬁrming it was a new case of a Type Ia explosion interacting with a dense circumstellar environment. From a spectrum obtained with the slit oriented to overlap with its host galaxy, they calculated a redshift for the host of z = 0.0616 ± 0.0002. SN 2005gj was not detected in the radio with the Very

Large Array (Soderberg & Frail 2005) or in the X-ray with Swift (Immler et al.

2005).

172 Here we present extensive follow-up photometry and spectroscopy of the

Type Ia/Type IIn SN 2005gj during the ﬁrst ∼150 days after discovery. These are the most detailed observations ever obtained of a SN 2002ic-like event, and provide insight into the early evolution, progenitor, and variety of these events. We also present a sensitive, early X-ray observation with Chandra that gives an upper limit on the X-ray luminosity of this peculiar object. We describe the optical and NIR photometry of SN 2005gj in §6.2 and the optical spectroscopy in §6.3. We describe the X-ray observation with Chandra in §6.4. An analysis of the photometric and spectroscopic data are presented in §6.5. Finally, we discuss the signiﬁcance of the

−1 −1 results in §6.6. We adopt a cosmology with H0 = 72 ± 8 km s Mpc , ΩM = 0.3, and ΩΛ = 0.7 throughout the paper (see Spergel et al. 2007), which yields a distance modulus of µ = 37.15 mag to the host of SN 2005gj.

6.2. Photometry

6.2.1. SDSS and MDM

The Sloan Digital Sky Survey (SDSS) uses a wide-ﬁeld, 2.5-meter telescope

(Gunn et al. 2006) and mosaic CCD camera (Gunn et al. 1998) at APO to survey the sky. The SDSS-II Supernova Survey, part of a 3-year extension of the original

SDSS, uses the APO-2.5m telescope to detect and measure light-curves for a large number of supernovae through repeat scans of the SDSS Southern equatorial

173 stripe (about 2.5 deg wide by ∼120 deg long) over the course of three 3-month campaigns (Sept-Nov. 2005-2007). SN 2005gj was discovered in the second month of the ﬁrst campaign (October 2005). Twenty epochs of ugriz photometry were obtained between 2005 Sep 26-Nov 30 (U.T.). Details of the photometric system and calibration are given in Fukugita et al. (1996), Hogg et al. (2001), Smith et al. (2002),

Ivezi´cet al. (2004), Tucker et al. (2006). Additional griz imaging of SN 2005gj was obtained with the MDM Observatory 2.4m telescope using a facility CCD imager

(RETROCAM; see Morgan et al. 2005, for a complete description of the imager).

Photometry of SN 2005gj on the SDSS images was carried out using the scene modeling code developed for SDSS-II as described in Holtzman et al. (2008). A sequence of stars around the supernova were taken from the list of Ivezi´cet al.

(2007), who derived standard SDSS magnitudes from multiple observations taken during the main SDSS survey under photometric conditions. Using these stars, frame scalings and astrometric solutions were derived for each of the supernova frames, as well as for the twenty five pre-supernova frames taken as part of either the main SDSS survey or the SN survey. Finally, the entire stack of frames was simultaneously fit for a single supernova position, a fixed galaxy background in each

ﬁlter (characterized by a grid of galaxy intensities), and the supernova brightness in each frame.

The supernova photometry in the MDM frames was also determined using the scene modeling code. Since the MDM observations had diﬀerent response

174 functions from the standard SDSS bandpasses, the photometric frame solutions included color terms from the SDSS standard magnitudes. To prevent uncertainties in the frame parameters and color terms from possibly corrupting the galaxy model

(here aﬀecting the SDSS photometry), the MDM data were not included in the galaxy determination, but the galaxy model as determined from the SDSS was used

(with color terms) to subtract the galaxy in the MDM frames. The resulting SN photometry from the MDM frames is reported on the native MDM system, since the color terms derived from stars are likely not to apply to the spectrum of the supernova.

Figure 6.1 shows a 3.5′ × 3.5′ ﬁeld around SN 2005gj and 16 comparison stars used for calibration of the SN magnitudes by SDSS and the Carnegie Supernova

Group (CSP; see §6.2.2 below for details). In Table 6.1 we present the SDSS ugriz and CSP u′g′r′i′ photometry of the comparison stars in common. The ﬁnal SDSS and MDM griz photometry is given in Table 6.2.

6.2.2. CSP

Optical photometry from the CSP was obtained with the Swope-1m telescope at

LCO, using a SITe CCD and a set of u′g′r′i′ ﬁlters. A subraster of 1200 × 1200 pixels was read from the center of the CCD, which, at a scale of 0.435 ′′ pixel−1, yielded a

ﬁeld of view of 8.7′ × 8.7′. Typical image quality ranged between 1′′ and 2′′ (FWHM).

175 A photometric sequence of comparison stars in the SN ﬁeld was calibrated with the

Swope telescope from observations of SDSS standard stars from Smith et al. (2002) during four photometric nights. SN magnitudes in the u′g′r′i′ system were obtained diﬀerentially relative to the comparison stars using PSF photometry. In order to minimize the contamination from the host galaxy light in the SN magnitudes,

PSF-matched ugri template images from SDSS were subtracted from the SN images.

On every galaxy-subtracted image, a PSF was ﬁtted to the SN and comparison stars within a radius of 3′′. See Hamuy et al. (2006) for further details about the measurement procedures.

The NIR photometry of SN 2005gj was obtained by the CSP using three diﬀerent instruments/telescopes. A total of 15 epochs in Y , J and H ﬁlters were obtained using RetroCam, mounted on the Swope-1m telescope at LCO. A few additional epochs in YJHKs were obtained with the Wide Infrared Camera (WIRC;

Persson et al. 2002) mounted on the duPont-2.5m telescope, and the PANIC camera

(Martini et al. 2004) mounted on the Magellan-6.5m Baade telescope, both at LCO.

We refer to Hamuy et al. (2006) and Phillips et al. (2007) for details of the imagers and procedures to extract the SN photometry. The host galaxy was not subtracted from the NIR frames; therefore the SN photometry contains an unknown galaxy contamination component.

The ﬁnal CSP u′g′r′i′ photometry of SN 2005gj is given in Table 6.3 and

YJHKs photometry in Table 6.4. A minimum uncertainty of 0.015 mag in the

176 optical bandpasses and 0.02 mag in the NIR is assumed for a single measurement based on the typical scatter in the transformation from instrumental to standard magnitudes of bright stars (Hamuy et al. 2006).

6.3. Spectroscopy

The spectroscopic observations of SN 2005gj are summarized in Table 6.6.

They were obtained using ﬁve diﬀerent telescopes/instruments at four observatories.

A total of twelve spectra were obtained between early-October 2005 and late-January 2006 with the Boller & Chivens CCD Spectrograph (CCDS) mounted at the MDM-2.4m telescope. This instrument uses a Loral 1200×800 pixel CCD with 15 µm pixel−1 and a 150 l/mm grating (blazed at 4700 A).˚ We used a 2′′ slit which gives a dispersion of 3.1 A˚ pixel−1 in the wavelength range ∼3800–7300 A.˚

Eight spectra were obtained at LCO with the Wide Field Reimaging Camera

(WFCCD) and the Modular Spectrograph (ModSpec) at the duPont-2.5m telescope, and LDSS-3 at the Clay-6.5m telescope. In the case of WFCCD a 400 l/mm grism and a 1.6′′ slit were used, reaching a dispersion of 3 A˚ pixel−1 in the wavelength range ∼3800–9200 A.˚ For ModSpec we used a 300 l/mm grating and a 1′′ slit that gave a dispersion of 2.45 A˚ pixel−1 in the range ∼3800–7300 A.˚ Further details of

WFCCD and ModSpec can be found in Hamuy et al. (2006). For LDSS-3, which employs a STA0500A 4064 × 4064 pixel CCD, we used the VPH blue and red grisms,

177 the latter with an OG590 ﬁlter to block second order contamination, with a 0.75′′ slit reaching a dispersion of 0.70 A˚ pixel−1 and 1.12 A˚ pixel−1 in the blue and red sides of the spectrum, respectively.

Additionally we obtained two early spectra with the Double Imaging

Spectrograph (DIS) mounted on the ARC-3.5m telescope at APO, and one spectrum with the Intermediate dispersion Spectrograph and Imaging System (ISIS) at the

WHT-4.2m telescope at the Roque de Los Muchachos Observatory in La Palma,

Spain. The DIS spectrograph has blue and red detectors, each uses a Marconi

2048 × 1024 pixel CCD with 13.5 µm pixel−1. We used a 300 l/mm grating and a

1.5′′ slit which gives a dispersion of 2.4 A˚ pixel−1. The ISIS spectrograph has a blue and red arm, the blue arm using a EEV12 CCD and the red using a MARCONI2 detector. We used a 300 l/mm grating in both the blue and red arm and a 1′′ slit, which gives a dispersion of 0.86 A˚ pixel−1 in the blue and 1.47 A˚ pixel−1 in the red.

Most of the spectra were obtained close to the parallactic angle to minimize relative changes in the calibration of the blue and red parts of the spectrum due to differential refraction through the atmosphere. The spectroscopic reductions were performed using standard IRAF tasks and included: bias and overscan subtraction, flat-fielding, combination of 2-4 individual 2D spectra to reach the best signal-to-noise ratio in the final image, tracing and extraction of a 1D spectrum from the combined 2D image, subtracting the background sky around the selected aperture, wavelength calibration using an arc-lamp, and flux calibration. In order

178 to ﬂux calibrate the spectra we observed 1-2 spectrophotometric standard stars per night. The spectra from LCO, WHT and MDM were corrected by atmospheric telluric lines using the spectrum of a hot spectrophotometric standard star and the spectra from APO were corrected using a model atmosphere. This correction is not optimal for some of the spectra and there are evident residuals left in the corrected spectra.

Figure 6.2 and 6.3 show a montage of the optical spectra of SN 2005gj obtained from October 2005 to March 2006. We have split them in two ﬁgures to avoid crowding. The position of the most conspicuous spectral features have been indicated in this ﬁgure.

6.4. X-ray Observation

SN 2005gj was observed under Director’s Discretionary Time for 49.5 ks on

2005 Dec 11/12 (ObsID 7241) with the Chandra X-ray Observatory’s Advanced

CCD Imaging Spectrometer (ACIS). The data were taken in timed-exposure mode with an integration time of 3.2 s per frame, and the telescope aimpoint was on the back-side illuminated S3 chip. The data were telemetered to the ground in “very faint” mode.

179 Data reduction was performed using the CIAO 3.3 software provided by the

Chandra X-ray Center1. The data were reprocessed using the CALDB 3.2.2 set of calibration files (gain maps, quantum efficiency, quantum efficiency uniformity, effective area) including a new bad pixel list made with the acis run hotpix tool.

The reprocessing was done without including the pixel randomization that is added during standard processing. This omission slightly improves the point spread function. The data were filtered using the standard ASCA grades (0, 2, 3, 4, and 6) and excluding both bad pixels and software-flagged cosmic ray events. Intervals of strong background flaring were searched for, but none were found.

Absolute Chandra astrometry is typically good to ∼0′′.5, and we sought to improve it by registering the Chandra image with an SDSS image. Chandra point sources were found using the wavdetect tool, and their positions were reﬁned using

ACIS Extract version 3.101. Fourteen X-ray sources had SDSS counterparts, which we used to shift the Chandra frame by a small amount (0′′.15 in RA and 0′′.07 in

DEC). After the shift, the residual diﬀerences between the Chandra and SDSS sources had rms values of 0′′.19 in RA and 0′′.12 in DEC.

We extracted counts in the 0.5–8 keV bandpass from the position of the supernova using a standard extraction region (∼1′′ radius), and we constructed response ﬁles with the CAIO tools and ACIS Extract. The background region is a source-free annulus centered on the position of the supernova with inner and outer

1http://asc.harvard.edu

180 radii of 6′′ and 32′′. Based on the 300 photons detected in this region, we expect 0.3 background counts in our source extraction region.

We detect only two counts from the location of the supernova, but neither may be associated with the supernova itself. The counts had energies of 4.0 keV and

6.5 keV, but one would expect some emission in the 0.5–2.5 keV range since this is where Chandra has the most collecting area. For example, the average eﬀective area in each of the 0.5–2.5 keV, 4.0–6.0 keV, and 6.0–8.0 keV bands is 470 cm2, 270 cm2, and 90 cm2, respectively.

We calculate our upper limits using the Bayesian method of Kraft et al. (1991).

For the 0.5–4 keV band, the 68% (95.5%) upper limit to the source counts is 1.14

(3.05). For the 0.5–8 keV band, the 68% (95.5%) upper limit is 3.52 (6.14).

Since no Type Ia supernova has been conclusively detected in X-ray, we have no a priori expectation of the spectral shape. We therefore adopt a simple absorbed

20 power law with a photon index of 2 and an absorbing column of nH = 7.08 × 10 cm−2. For this choice of spectrum, the count rate to ﬂux conversion is 5.2 × 10−12 erg/count in the 0.5–4 keV band and 6.9 × 10−12 erg/count in the 0.5–8 keV band.

We therefore arrive at 68% (95.5%) upper limits on the X-ray luminosity of

1.1 (2.9) × 1039 erg s−1 in the 0.5–4 keV band and 4.3 (7.6) × 1039 erg s−1 in the 0.5–8 keV band. Based on the above statements concerning the Chandra eﬀective area, we feel the 0.5–4 keV limit is a more appropriate limit.

181 6.5. Results

6.5.1. Optical light curves and colors

Figure 6.4 shows the early SDSS and MDM ugriz light curves combined with

′ ′ ′ ′ the late time coverage given by CSP u g r i YJHKs photometry. They give excellent multi-wavelength optical and NIR coverage and sampling of the ﬁrst ∼150 days after discovery (∼140 rest-frame days after the time of explosion).

In Table 6.5 we give important parameters derived from the light curves. We have a good estimate of the time of explosion of the SN at JD 2,453,637.93±2.02

(September 24.4 UT, 2005) calculated as the average time between the epoch when the supernova is detected at > 5σ in all filters at JD 2,453,639.94, and the last pre-discovery observation of the field at JD 2,453,635.91. The times and observed magnitudes at maximum in different filters are presented in Table 6.5. They are calculated from a high order polynomial fit to each light curve. To estimate the errors we assume a gaussian distribution for the magnitude uncertainty at each epoch and filter (assuming they are not correlated from epoch to epoch), then we draw randomly ∼1000 simulated light curves and fit each one with the same high order polynomial. The 1σ uncertainties are taken as the rms deviation of the mean values calculated from each simulated light curve.

182 The risetimes, deﬁned as the time between explosion and maximum light, become longer at redder wavelengths. They are 13.5, 19.7, 33.7, and 46.9 days in ugri ﬁlters, respectively, which correspond to 12.7, 18.5, 31.7, and 44.2 days in the rest-frame of the supernova.

We also give in Table 6.5 the rest-frame magnitudes at maximum in diﬀerent

ﬁlters. They have been corrected by Galactic extinction in the line of sight, using

E(B − V )Gal = 0.121 (SFD: Schlegel et al. 1998) and the Cardelli et al. (1989)

(CCM) reddening law assuming RV = 3.1, and K-corrections (see below), which are not negligible in this object since it is at redshift z ∼ 0.062. We assume that the host galaxy extinction is negligible (Aldering et al. 2006).

K-corrections have been calculated from the multi-epoch spectra presented in

§6.3. In order to estimate accurate K-corrections we need good spectrophotometric calibration in all the wavelength range. Figure 6.5 shows the differences between the observed g − r photometric colors obtained from the light curves and the synthetic colors calculated directly from the spectra, as a function of the observed photometric colors. We use an 8th order polynomial fit of the light curves to obtain the observed g − r color at the epoch of a given spectrum to better than ∼ 0.03 mag. We can see that most of the spectra have good spectrophotometric calibration in the wavelength range of g , r filters (3800–7000 A),˚ with a residual color of ∼0.05 mag rms (difference between observed and synthetic colors), but there are some obvious outliers. To correct the spectra and produce a better spectrophotometric calibration consistent

183 with the observed spectral energy distribution (SED) obtained from broadband photometry we warp the spectra multiplying by a smooth function to match the observed colors, a technique commonly used for calculating K-corrections in Type Ia

SN (Nugent et al. 2002).

First, we extrapolate the continuum in the blue and red sides of each spectrum presented in §6.3 with a low order polynomial to have complete coverage of the SDSS ugriz ﬁlters (2000–11000 A).˚ Using this extended version of the spectra we apply the

CCM reddening law iteratively until the synthetic g − r color matches the observed color in each spectrum. This procedure does not ensure that the calibration is good in the complete wavelength range; therefore we multiply by a smooth spline with knots at the eﬀective wavelength of the SDSS ﬁlters until the synthetic u − g, g − r, r − i and i − z colors match the observed colors obtained, again by using polynomial interpolation of the light curves.

The K-corrections for the same ﬁlter (Hamuy et al. 1993) calculated from the modiﬁed, spectrophotometrically calibrated spectra using SDSS passbands are listed in Table 6.7. The K-corrections are probably accurate to ±0.05 mag for g and r

ﬁlters, and to ±0.1 mag for u, i and z where we had to extrapolate the spectra.

However, our estimate is not precise and we can not exclude the possibility of even larger errors. After ﬁtting with a low order polynomial we use the results to transform the observed SDSS magnitudes in Table 6.2 to the rest-frame.

184 We corrected the CSP u′g′r′i′ magnitudes to rest-frame SDSS ugri magnitudes using cross-ﬁlter K-corrections according to the prescription of Nugent et al. (2002).

These take into account the difference between the CSP and SDSS passbands convolved by the SED of SN 2005gj and allow us to put all the rest-frame magnitudes in the same system. We find that the differences between the same- and cross-filter

K-corrections are small (∼< 0.03 mag) at all epochs.

Figure 6.6 shows the evolution of the rest-frame colors of SN 2005gj as a function of time after explosion. We have corrected the magnitudes by K-corrections and Galactic extinction in the line of sight. As a comparison we also plot the color evolution of the overluminous Type Ia SN 1991T, a typical Type IIn SN 1999el (Di

Carlo et al. 2002) (both obtained from spectral templates of P. Nugent 2 ), and two previous cases that are thought to be Type Ia explosions in a very dense environment:

SN 2002ic and SN 1997cy. For SN 2002ic we use the published BVI photometry

(Hamuy et al. 2003) corrected by Galactic extinction (E(B − V ) = 0.06; SFD).

We calculate cross-ﬁlter K-corrections using the calibrated spectra of SN 2005gj to transform observed magnitudes of the SN at z = 0.0667 (Kotak et al. 2004) to rest-frame SDSS magnitudes: B → u, V → g, I → i. The time of explosion of

SN 2002ic is assumed to at ≈JD 2,452,581.5 (2002, Nov. 3 UT; Deng et al. 2004).

We obtain rest-frame colors in the SDSS system for SN 1997cy (Germany et al. 2000;

Turatto et al. 2000) by transforming the K-corrected V RI magnitudes in Germany

2 http://supernova.lbl.gov/∼nugent/nugent templates.html

185 et al. (2000) to gri magnitudes using S-corrections calculated directly from spectra of SN 1997cy available online in the SUSPECT database 3. We supplement this with synthetic colors from the spectra. We also correct by Galactic extinction in the line of sight (E(B − V ) = 0.02; SFD) and assume the time of explosion of SN 1997cy to be JD 2,450,582.5 (1997, May. 14 UT; Germany et al. 2000), which is very uncertain, since it is taken as the time of detection of the gamma-ray burst GRB 970514, which may not have been associated with the SN. Magnitudes for SN 2002ic and SN 1997cy are not corrected by extinction in their host galaxies, which is unknown.

Initially the evolution in rest-frame u − g and g − r colors of SN 2005gj, up to ∼30-40 days after explosion (∼10-20 days after maximum for a SN Ia), is roughly consistent with the colors of SN 1991T but ∼ 0.2 mag redder in g − r, and evolves to redder colors at later times. SN 1991T reaches its maximum colors of (u − g)=1.5 mag and (g − r)=1.0 mag at ∼50 days (30 days after maximum), and after that it enters the nebular phase and becomes bluer. At late times the u − g color of SN 2005gj has a slow linear increase and becomes systematically redder than SN 1991T at > 70 days, while the g − r color stays approximately

ﬂat at (g − r)=0.5 mag (and bluer than SN 1991T) between 60 − 110 days. The evolution in u − g color of SN 2002ic is very similar and consistent with SN 2005gj.

SN 1997cy has a similar color evolution but it is ∼0.2 mag bluer in g − r color between 60 − 100 days. However, Germany et al. (2000) give K-correction errors of

3 http://suspect.nhn.ou.edu/∼suspect/

186 ∼0.15 mag and we have further applied S-corrections, therefore this does not imply a signiﬁcant diﬀerence between SN 1997cy and the other two Type Ia/IIn supernovae.

The rest-frame r − i and i − z color evolution of SN 2005gj is very diﬀerent from SN 1991T and closely follows the evolution of a SN IIn. The r − i colors of

SN 1997cy are also consistent with SN 2005gj. From these comparisons it is clear that the colors, a proxy for the temperature of the photosphere, of SN 2005gj and two earlier cases of SN Ia strongly interacting with their circumstellar medium are dominated by the radiation coming from the ejecta-CSM interaction.

In Figure 6.7 we present the ugri light curves of SN 2005gj in absolute magnitudes. We also show the light curves of SN 1991T, SN 2002ic and SN 1997cy obtained from the literature and corrected to SDSS rest-frame magnitudes as explained above.

SN 2005gj has peak ugriz absolute magnitudes in the range −20.0 to

−20.3 mag, this is ∼ 0.7 − 1.4 mag brighter than the overluminous Type Ia

SN 1991T. The u light curve is consistent with a linear decline after peak luminosity at a constant rate of 0.027 ± 0.001 mag day−1. The g light curve has a

∼ 20 day plateau with roughly constant luminosity after maximum (20 − 40 days after explosion), then the light curve declines linearly between 40 − 100 days at

0.018 ± 0.001 mag day−1 and continues its linear decay at later times, but with a shallower slope of 0.007 ± 0.002 mag day−1. The r and i light curves have a similar

187 plateau shape between 20 − 60 days and a constant linear decay at later times of

0.013 ± 0.001 mag day−1. The change in slope observed in the g-band at late times is less clear in ri, but still present. The secondary maximum present in ri light curves of SN 1991T and other SN Ia is completely absent in SN 2005gj.

The light curves of SN 2002ic are fainter than SN 2005gj at all times by

0.3−0.6 mag depending on the ﬁlter (0.3−0.8 mag brighter than SN 1991T at peak).

The initial decline rates from maximum of the u and g light curves of SN 2002ic are intermediate between those of SN 1991T and SN 2005gj until around 40 days after explosion; after day 50 when the ejecta-CSM interaction had become dominant in

SN 2002ic (Hamuy et al. 2003), they closely resemble the decline rates of SN 2005gj in the same bands. The i band light curve of SN 2002ic showed deﬁnite evidence for a weak secondary maximum, which is again intermediate in morphology between the strong secondary maximum observed in SN 1991T, and the absence of such a feature in SN 2005gj. The gr light curves of SN 1997cy are consistent with linear decay of ∼ 0.008 mag day−1, being ∼ 0.4 mag brighter than SN 2005gj at 150 days.

Unfortunately, the light curves of SN 1997cy start at ∼ 60 days after explosion and there is no information near peak to compare with the luminosities of SN 2005gj and SN 2002ic. However, if we extrapolate the gr light curves to the time of peak luminosity we ﬁnd that the absolute magnitudes at maximum of SN 1997cy would be within ∼ 0.2 mag of SN 2005gj. This has to be taken with caution because of the extrapolation and uncertain time of explosion of SN 1997cy.

188 6.5.2. NIR light curves

We used the CSP JHKs photometry obtained between 59-166 days after the explosion (rest-frame) to construct the absolute magnitude light curves of SN 2005gj.

The observed magnitudes were corrected by Galactic reddening in the line of sight

(AJ = 0.108, AH = 0.070, and AKs = 0.044 mag) and K-corrections. We calculated

K-corrections for the same ﬁlter using the spectral templates of P. Nugent for the

Type IIn SN 1999el (which are derived from black-body curve ﬁts to the photometry), because as we showed in §6.5.1 the synthetic colors obtained from the spectral templates approximate reasonably well the evolution of the redder optical colors of

SN 2005gj (i.e., r − i and i − z). The values of the K-corrections are consistent with being constant in this time range: KJ ≈ −0.12 mag ,KH ≈ −0.14 mag,

KKs ≈ 0.16 mag.

The JHKs absolute magnitude light curves of SN 2005gj are presented in

Figure 6.8. For comparison we show the NIR light curves of a normal Type Ia obtained from synthetic photometry of spectral templates and the Type IIn SN 1999el

(Di Carlo et al. 2002). The Type Ia light curves have been shifted in magnitudes to match the mean absolute magnitude at maximum of SN Ia (Krisciunas et al. 2004).

SN 2005gj is 1.7 − 3 mag brighter than a normal SN Ia and SN IIn at 60 days after the explosion and declines in a slower fashion at later times. A linear decay is a good

ﬁt in all NIR bandpasses at this late times (> 60 days after explosion), with decline

189 rates of ∼ 0.014 mag day−1 in J, ∼ 0.013 mag day−1 in H, and ∼ 0.011 mag day−1 in Ks. These values are similar to the decline rates in the optical ri bands.

Since there are no template images of the host galaxy obtained in the NIR bands, the light curves are preliminary and the analysis has to be taken with caution.

6.5.3. Bolometric light curve

The SDSS and CSP magnitudes were used to produce a quasi-bolometric light curve of SN 2005gj covering the optical wavelength range from

3000 − 10, 000 A(˚ u → z). We corrected the magnitudes by Galactic extinction in the line of sight and K-corrections to obtain magnitudes in the rest-frame ugriz ﬁlters

(see §6.5.1 for details). We applied small corrections to transform the magnitudes based in the SDSS photometric system into the AB system obtained from the SDSS website 4. The AB magnitudes derived in this way are transformed directly to bandpass averaged fluxes using the definition of the AB system (Oke & Gunn 1983) and they are assigned to the frequencies that correspond to the effective wavelengths of the SDSS ugriz filters, calculated from the filters using the definition in Fukugita et al. (1996): 3567, 4735, 6195, 7510, and 8977 A.˚ We use the trapezoidal rule to obtain the integrated flux from u to z, this is between λ1 = 3340 A˚ and λ2 = 9596 A,˚ where the limits of the wavelength coverage are obtained from λ1 = λeff,u − ∆λu/2

4 http://www.sdss.org/dr5/algorithms/fluxcal.html.

190 and λ2 = λeff,z +∆λz/2. We extrapolate linearly the u and z light curves at late times to ﬁll in the lack of coverage of the SDSS in z, and CSP in u and z bands, including the MDM z-band data at these epochs. The integrated ﬂuxes are converted to luminosity assuming a luminosity distance to SN 2005gj of 268.5 Mpc and a spherically symmetric distribution of the output energy. We present the integrated, quasi-bolometric luminosities from u to i (L(u→i)) and from u to z (L(u→z)) as a function of time in Table 6.8.

In order to estimate bolometric UVOIR luminosities we calculate time dependent bolometric corrections to include the energy output of the SN at wavelengths bluer than u-band (λ < 3340 A)˚ and redder than z-band (λ > 9596 A).˚ We find that black-body distributions with different temperatures are a reasonable approximation of the spectral energy distribution of SN 2005gj. We use χ2-minimization to fit the optical SED with a two parameter black-body function: temperature and a multiplicative scaling factor. The scaling factor is a combination of fundamental

2 constants and the square of the angular radius of the spherical black-body: (Rbb/d) , where Rbb is the radius of the black-body and d is the distance to the SN. We calculate time-dependent bolometric corrections by integrating the black-body distributions in the ultraviolet and NIR/IR regions, and converting the integrated

ﬂuxes to luminosities as explained above. For the CSP data we also include NIR

ﬂux densities derived from the reddening and K-corrected YJHKs magnitudes after

ﬁtting the light curves with low order polynomials.

191 At early phases before peak the bolometric corrections account for ∼ 53 − 65% of the total integrated luminosity, of which 85 − 93% is from the ultraviolet part of the spectrum and only 7 − 15% from the NIR/IR. As the supernova evolves, the ejecta expands and shocks the circumstellar gas. The energy emitted in the ultraviolet/blue part of the spectrum declines quickly after maximum light and most of the energy is emitted in the optical region, coinciding with the appearance in the spectrum of emission/absorption features of the intermediate and iron-peak elements

(see §6.5.4). Between 25 − 150 days after explosion the bolometric corrections account for ∼ 32 − 45% of the total output luminosity, with the NIR/IR correction dominating completely over the blue/ultraviolet at > 60 days.

The bolometric UVOIR luminosities, black-body temperatures and radii derived from the ﬁts are presented in Table 6.8. The uncertainties in black-body temperatures and radii are calculated using the diagonal terms of the covariance matrix obtained from the χ2 minimization. We add a 10% error in the distance to the SN to the error in the black-body radii, which comes from the random and systematic uncertainties in the value of the Hubble constant (Freedman et al. 2001). The uncertainties in the bolometric luminosities were estimated by propagating errors through the trapezoidal integration of the SED, taking into account: uncertainties in the photometry, light curve interpolation and ﬁtting,

Galactic extinction, K-corrections, and distance to the SN. To approximately take

192 into account the errors introduced by the bolometric corrections we multiplied these

2 2 values by χν when the reduced χ is greater than 1. q

In Figure 6.9 we show some examples of black-body fits to the optical SED at different epochs. At early times, shortly after explosion, the SED is very well fit by a hot ∼13000 K black-body. The temperature starts to decrease steadily close to the time of explosion to a constant value of ∼ 6500 K at 60 days after. A black-body is still a reasonable approximation of the SED at later times, but the fits become poorer when emission/absorption features start to dominate the spectrum, which is

2 represented in the χν of the ﬁts (see Table 6.8).

Figure 6.10 shows the bolometric light curve of SN 2005gj in the top panel and the evolution in temperature and radius from the black-body ﬁts in the lower panels. The early data of the bolometric light curve are well ﬁt by an exponential rise in luminosity, L(t) ∝ e0.17 t. The time of maximum bolometric luminosity occurs between 6.6–18.8 days after explosion. After maximum, the bolometric light curve is very well approximated by an exponential decay in luminosity, linear in the logarithmic scale shown in Figure 6.10, L(t) ∝ e−0.013 t (0.014 mag day−1). This is consistent with the exponential density distribution of the ejecta of Type Ia

SN (Dwarkadas & Chevalier 1998), whereas the distribution of ejecta around core-collapse supernovae is better approximated by a power-law (Chevalier &

Fransson 2003). Extrapolating the pre- and post-maximum ﬁts we get a maximum

max 43 −1 bolometric luminosity of Lbol = 5.6 × 10 ergs , which is ∼3 times more luminous

193 than the Type Ia SN 1991T at maximum light (Contardo et al. 2000; Stritzinger et al. 2006). Assuming a bolometric correction of 50% at maximum, we ﬁnd that

SN 2005gj was ∼1.5 times more luminous than SN 2002ic.

6.5.4. Optical spectroscopy

In Figure 6.11 we show a comparison of the spectra of SN 2005gj with spectra of SN 2002ic and SN 1997cy obtained at similar times after explosion. The spectra of SN 2005gj and SN 2002ic are very similar at all times. They are characterized by strong and broad Hydrogen-Balmer lines Hα and Hβ in emission5 and a blue continuum at early times that becomes redder and increasingly dominated by absorption/emission P-cygni proﬁles from Fe-peak ions (e.g., Fe II, Fe III, Ni III,

Si II, S II).

Classification

Benetti et al. (2006b) proposed that SN 2002ic-like events are well explained by the core-collapse of a massive star in a dense medium, casting doubt in the previous classiﬁcation of SN 2002ic as a Type Ia supernova. The authors found relatively good agreement at all times between the spectra of SN 2004aw (Taubenberger et al.

2006), a Type Ic supernova, and SN 2002ic.

5As shown in Figure 6.2, Hγ is also visible in the early spectra of SN 2005gj.

194 We used the SuperNova IDentiﬁcation code, SNID (Matheson et al. 2005;

Miknaitis et al. 2007; Blondin & Tonry 2007b), to ﬁnd the spectra that best match

SN 2005gj at different epochs. SNID cross-correlates an input spectrum with a library of supernovae spectra. In the library we included spectra of 5 normal SN Ia, two 1991T-like objects, two 1991bg-like objects, 4 broad-lined SN Ic (or hypernovae), and 3 normal SN Ic (including SN 2004aw), that were chosen to span a wide range of observed properties of SN Ia and SN Ic. In Table 6.9 we present the supernovae and the epochs of the spectra in the library. We fixed the redshift of SN 2005gj at z = 0.0616 and allowed for a range around the mean redshift of ∆z = 0.02 to find cross-correlation peaks. The Balmer lines in emission were clipped from the input

SN 2005gj spectra to avoid spurious cross-correlation signal with library spectra that contain emission lines from the host galaxy.

Figure 6.12 shows the spectra of SN 2005gj at four epochs and the library spectra with the highest cross-correlation signiﬁcance from SNID. Type Ia supernovae spectra are a better match to SN 2005gj at most epochs, with 20 of the 26 (77%) epochs having best matched a type Ia spectrum (45%-91T, 45%-normal, 10%-91bg) and a similar fraction for the next-best matches. The broad-lined SN Ic 1997ef and

2002ap are the best matching spectra for 6 epochs, all of them between 26–46 days after explosion. We repeated the same procedure using ﬁve spectra of SN 2002ic obtained between 24–84 days after explosion (Hamuy et al. 2003). All the spectra of

195 SN 2002ic are better matched with SN Ia, in contradiction with the results obtained by Benetti et al. (2006b).

The continuum of SN 2005gj is well approximated at all times by the sum of a scaled spectrum of the overluminous Type Ia SN 1991T, at the same epoch after explosion as SN 2005gj, and a fourth order polynomial. A normal SN Ia does not ﬁt as well as SN 1991T. This procedure is very similar to the ﬁts to SN 2002ic (Hamuy et al. 2003) and SN 2005gj (Aldering et al. 2006) presented in previous studies. In

Figure 6.13 we show examples of the spectra decomposition at four epochs. We excluded from the ﬁt a region of ±100 A˚ around the Hα and Hβ lines and obtained a good ﬁt for the remainder of the spectrum.

Balmer lines

We analyzed the Balmer emission features in the spectra using the sum of two Gaussian components to model the line profiles. This decomposition gives much better fits for Hα at all epochs than a single Gaussian and it is physically motivated (Chugai 1997b,a). The spectra of Type IIn SN show Balmer features with both a narrow and broad component that can be explained by radiation coming from different regions of the ejecta/CSM, whether it is direct emission from the shock-heated CSM (broad component) or emission from un-shocked gas photoionized by the SN radiation (narrow component). The Hβ line is unresolved or only marginally resolved for most of the spectra. Therefore a single Gaussian component

196 was used to fit the line profile. We used a third order polynomial to model the local continuum around each line that was included in the Gaussian fits. It is important to stress that at late times there is a broad Fe II feature intrinsic to the supernova spectrum in the region of Hα (see spectra in Figure 6.13) that makes the definition of the continuum less reliable and may affect the line measurements.

The results of the Gaussian ﬁts to the Hα and Hβ emission features, integrated

ﬂuxes and FWHM, are shown in Table 6.10 as a function of epoch of the spectra.

We have excluded the two spectra with better resolution because they show P-Cygni profiles (see below). We used the flux calibrated spectra corrected to match the observed g, r magnitudes (as explained in §6.5.1) and corrected for Galactic reddening in the line of sight. The FWHM of the Gaussian profiles listed in

Table 6.10 were corrected by the resolution of the spectrographs (from Table 6.6).

We do not present the values when the component is unresolved. The errors quoted for the integrated ﬂuxes are obtained by adding in quadrature an estimate of 10% error assigned to the absolute ﬂux calibration and the rms deviation of the Gaussian

ﬁts.

Figure 6.14 shows the evolution in time of the FWHM (top left panel), Hα and Hβ luminosities (top right and bottom left panels) and the Balmer decrement

(bottom right panel). The FWHM of Hα varies between ∼ 130 − 500 km s−1

(narrow component) and ∼ 1200 − 3800 km s−1 (broad component), with the broad component showing a slow increase in time. The FWHM of Hβ varies between

197 ∼ 470 − 1700 km s−1 and does not show evident evolution. The luminosities of

Hα-narrow and Hβ lines evolve in a similar fashion, increasing at early times to peak at ∼ 12 days with luminosities 6 − 6.5 × 1040 ergs−1, then they decay and stay roughly constant after 50 days. The evolution of Hα-broad is similar during the

ﬁrst 50 days, peaking at 1.1 × 1041 ergs−1, but it shows an increase at later times.

Compared with the Hα luminosities observed in SN 2002ic, both components are

∼ 4 times more luminous.

The Balmer decrement, the ratio between Hα (sum of narrow and broad components) and Hβ ﬂuxes, stays approximately constant during the ﬁrst 30−40 days

(mean = 2.5 and rms = 0.5) and is consistent with the theoretical value in Case

B recombination of Hα/Hβ = 2.86 (Osterbrock 1989). At later times it shows an steady increase, reaching Hα/Hβ ∼ 7 − 13 at ∼ 80 days. In Case B recombination a

Balmer decrement Hα/Hβ > 2.86 is usually interpreted as evidence for the presence of internal extinction in the host; however, the large values observed at late times would produce an Na I D interstellar absorption doublet easily detectable in the spectra, which is not observed (see also Aldering et al. (2006)), and the evolution in time is not expected. Aldering et al. (2006) proposed that the H level populations are in Case C recombination, where the optical depth in the Hα line is high implying high densities and greater importance of collisional processes. In this scenario, the observed change in the Balmer decrement could indicate that collisional excitation becomes increasingly important at later times (Branch et al. 1981; Turatto, et al.

198 1993). SN 2002ic (Deng et al. 2004) and other SN IIn, like SN 1988Z (Aretxaga et al. 1999) and SN 1995G (Pastorello et al. 2002), have also shown large values of the Balmer decrement and therefore may have similar physical processes aﬀecting the formation of the Balmer lines.

In Figure 6.15 we show the regions around Hα and Hβ features in the best resolution spectra from ISIS and LDSS-3, obtained at 44 and 115 days after explosion, respectively. We clearly detect P-Cygni proﬁles in all these features, which indicates the presence of an outﬂow moving at ∼150-200 km s−1; however, these measurements are limited by the resolution of the spectra between ∼130-180 km s−1 (FWHM).

After correcting for the resolution we obtain an outﬂow velocity of 60-70 km s−1.

The detection of P-Cygni-like absorption rules out an H II region in the line of sight that could be producing the narrow emission/absorption features; the line proﬁles are intrinsic to the SN. Aldering et al. (2006) detected P-Cygni proﬁles in He I,

Hα and Hβ, in a high resolution spectrum obtained with LRIS+Keck 71 days after

−1 the explosion. They derived a wind velocity of vw ≈ 60 km s consistent with our estimate. Kotak et al. (2004) also detected a P-Cygni proﬁle in the a spectrum of

SN 2002ic obtained 256 days after explosion.

Line identification near maximum

We used the parameterized resonance-scattering code SYNOW (Fisher et al.

1997; Fisher 2000) to identify the lines in the spectra obtained near maximum light

199 of SN 2005gj. SYNOW is a fast supernova spectrum-synthesis code used for direct

(empirical) analysis of supernova spectra, mainly to identify the lines, their formation velocities and optical depths. The code is based on simple assumptions: spherical symmetry, homologous expansion, a sharp photosphere that emits a black-body continuous spectrum, and line formation by resonance-scattering, treated in the

Sobolev approximation. We have used the latest version of the code that includes a

Gaussian distribution of optical depths.

Figure 6.16 shows the spectrum of SN 2005gj at 17 days after explosion (2 days before g maximum) and the best synthetic spectrum obtained with SYNOW. We also show for comparison the spectrum of SN 1991T obtained at -3 days with respect to the time of B maximum. The spectra have been locally normalized as in Jeﬀery et al. (2006). The synthetic spectrum has a black-body continuum

−1 temperature Tbb = 11000 K, photospheric velocity vphot = 10000 km s , and excitation temperature Texc = 10000 K. We ﬁnd a reasonably good match with the spectrum of SN 2005gj using the following lines/multiplets: Fe III λ4404 and

λ5129, Si III λ4561, Ni III, S II λ5468 and λ5633, and Si II λ6355. These lines are characteristic of the overluminous and spectroscopically peculiar, Type Ia SN 1991T around maximum light with strong Fe III features and weak S II doublet and Si II

(Jeﬀery et al. 1992; Mazzali et al. 1995; Fisher et al. 1999).

The main discrepancy between the SYNOW modeling of SN 2005gj and

SN 1991T is in the optical depths of the lines. The ﬁt to SN 2005gj needs

200 unphysically small optical depths, approximately 1/10th of the values used for

SN 1991T around maximum light. We interpret this as an eﬀect of the extra continuum radiation that is added by the ejecta-CSM interaction, which is veiling

(Branch et al. 2000) the supernova lines (e.g., Hamuy et al. 2003; Aldering et al.

2006). This interpretation is supported by the good agreement obtained from ﬁtting the spectra of SN 2005gj using a simple polynomial continuum added to the spectra of SN 1991T at the same epochs after explosion (see Figure 6.13).

6.6. Discussion

We have presented extensive spectroscopy and optical/NIR photometry of

SN 2005gj obtained by the SDSS-II and CSP supernova groups during the ﬁrst

∼150 days after explosion, and also an X-ray observation at 74 days that gives an upper limit on the X-ray luminosity. We have shown the remarkable similarity in spectroscopic and photometric properties between SN 2005gj and SN 2002ic, which is thought to be the ﬁrst clear case of a thermonuclear supernova explosion embedded in a dense CSM. The observational properties of SN 2005gj support this interpretation, they are summarized as follows:

• Spectroscopic evidence for a shock propagating into an Hydrogen-rich medium

close to the site of the explosion inferred from the presence of Balmer lines with

narrow (FWHM∼ 200 − 500 km s−1) and broad (FWHM∼ 1500 − 3000 km s−1)

201 components at all times. The Balmer lines show P-Cygni proﬁles in the highest

resolution spectra obtained at 44 and 115 days after explosion, these detections

show the presence of a slow (∼ 100 km s−1) moving outﬂow. Both observations

support the interpretation of the supernova ejecta interacting with a dense

circumstellar material.

• Spectrum evolves from a very blue continuum (13000 K black-body) similar

to SN IIn at ∼7 days after explosion to a redder continuum at later times

with P-Cygni absorption/emission proﬁles. The strongest lines present around

maximum are identiﬁed with singly and doubly ionized iron-peak elements

(especially strong Fe III, weak S II and Si II) and the spectra are well matched

by the overluminous Type Ia SN 1991T diluted with a polynomial continuum

at similar times after explosion.

• Very luminous and slowly declining bolometric light curve. The linear decay

in luminosity after peak (∼ 0.014 mag day−1) suggests an exponential density

distribution of the ejecta, which is consistent with the ejecta-density proﬁles

obtained from simulations of SN Ia.

The data presented here on SN 2005gj makes the interpretation of 2002ic-like events as thermonuclear supernovae in a dense CSM, initially proposed by Hamuy et al. (2003), stronger. In contrast with Benetti et al. (2006b), we ﬁnd that the overall shape of the spectra of SN 2005gj are more consistent with spectra of SN Ia

202 at diﬀerent epochs. Speciﬁcally, Type Ic SNe usually do not show the S II doublet at ∼ 5500 A˚ around maximum light; in fact SN 2004aw shows only two very weak notches at the wavelengths of S II near maximum (Taubenberger et al. 2006). This is one of the identifying features in SN Ia spectra, also present in the overluminous

SN 1991T (Phillips et al. 1992). In the spectrum of SN 2005gj obtained at 17 days

(see Figure 6.16) we detect a weak double absorption that we identify with S II, that is much stronger in the spectrum of SN 2002ic around maximum light. We can see on the top of Figure 6.11 that the spectrum of SN 2002ic obtained 24 days after explosion clearly shows this feature. Other conspicuous features observed in

SN 2005gj and SN 2002ic around maximum are Fe III and Si II. These features are present in SN 1991T, but Fe III is not observed and Si II is generally weaker in SN Ic.

SN 2005gj has stronger ejecta-CSM interaction than SN 2002ic. The peak bolometric luminosity is ∼ 1.5 times brighter and the broad and narrow components of Hα are ∼ 4 times more luminous in SN 2005gj. The fact that the SN 1991T features are weaker in SN 2005gj compared with SN 2002ic at similar epochs is consistent with this interpretation, because the supernova features are more diluted by the stronger continuum. The absence of evidence for a secondary maximum in

SN 2005gj, whereas the i band light curve of SN 2002ic does show a hint of such a feature, is likewise consistent with the ejecta-CSM interaction in SN 2005gj having been stronger than in SN 2002ic.

203 6.6.1. Structure of the CSM

The circumstellar interaction of core-collapse supernovae with a circumstellar medium has been studied in detail in the literature (see Chevalier & Fransson

(2003) for a review). When the fast moving ejecta encounters the approximately stationary CSM, a forward shock moving into the CSM (also called circumstellar shock) and a reverse shock develops. The fast-moving shockwave implies large post-shock temperatures, therefore radiating energy in the X-ray regime. The density distribution of the ejecta and the CSM can be well described by power-laws in radius, which leads to a set of self-similar analytical solutions for the evolution of the shock radius in time (Chevalier 1982). The physics of the ejecta-CSM interaction in the case of thermonuclear supernovae is basically the same, the main difference is in the distribution of the ejected material which follows an exponential function in velocity (Dwarkadas & Chevalier 1998). In this case the solutions are no longer analytic. The density profile of the shocked region is different in the case of exponential ejecta expanding into a constant density medium, but the similarity increases for expansion into a wind profile whose density decreases as ∝ r−2.

A simple self-similar model of a SN shock expanding into a medium with a power-law density decline, as suggested for core-collapse SNe by Chevalier (1982), is ruled out for this object by several observations: the exponential decrease in luminosity, suggesting an exponential ejecta density proﬁle; the strange behavior of

204 the broad and narrow Hα components; and the decrease in the blackbody radius at later times. Detailed calculations of the SN-CSM interaction would require highly detailed hydrodynamic modeling, which are beyond the scope of this paper. Instead herein we focus on trying to explain the basic features of SN-CSM interaction as deduced from the observational data.

The initial velocity of a SN shock wave as it breaks out from the surface is at least of the order of 2 × 104 km s−1. The broad Hα velocities that are seen in the

ﬁrst week or so are of the order of 1500 km s−1, and increase to more than twice this value after ∼ 50 days. These velocities are almost an order of magnitude smaller than expected SN blast wave velocities in the early stages, and a factor of few smaller even after ∼ 50 days. Furthermore, the SN shock velocity would be expected to gradually decrease as the shock moves outwards, whereas the Hα proﬁle actually indicates an increasing velocity after ∼ 50 days.

For these reasons, we suggest that the broad Hα lines do not indicate the

SN velocity. Instead, we put forward a scenario of a shock expanding into a two-component ambient medium: a low density wind in which are embedded high-density clumps. In this picture, there should theoretically exist three diﬀerent velocity components: a broad velocity component, which is not easily seen in this case, and is related to the velocity of the blast wave itself; an intermediate velocity component, which is what we have referred to as the broad Hα and is related to the velocity of the shock driven into the clumped material; and a narrow velocity

205 component, which may be related to the narrow Hα and is representative of the velocity of the ambient medium. This scenario is like the scenario put forward by

(Chugai & Danziger 1994) to explain the origin of the broad, intermediate and narrow line components in SN 1988Z. The large Hα luminosity of SN 2005gj at late times is very similar to that seen in other Type IIn SNe, and is especially large considering that this was a Type Ia. However, there are significant differences. We do not see a really broad line component representative of the SN velocity, although there are some suggestions that this may be appearing at late times. In particular, the Hα profile of the spectrum obtained at ∼ 150 days is better fitted by three components, including a very broad component with FWHM ≈ 7000 kms−1.

Our scenario envisions the Type Ia SN shock wave expanding in a clumped medium presumably formed by mass-loss from a companion star. The broad component is not easily visible in Hα initially because the forward shock is not radiative. The density of the clumps is much higher than that of the interclump

(ambient) medium. When the SN shock wave interacts with a dense cloud or clump, it drives a strong shock into the clump. A reﬂected shock is driven back into the expanding ejecta (Klein et al. 1994). Assuming pressure equilibrium, the ratio of the velocity of the clump shock to that of the blast wave is inversely proportional to the square-root of the ratio of the clump density to that of the interclump medium. The optical emission arises from behind the clump shock, probably by reprocessing of the

X-ray emission.

206 In this model, the intermediate component represents the velocity of the clump shock, which is probably radiative. If we assume that the initial velocity of the SN shock wave is ∼ 20000 km s−1 and the broad Hα emission velocity is ∼ 1500 km s−1, then the ratio of velocities is 13 − 14. This indicates that the clump density is about

142, or ∼ 200 times the interclump density. Note that the optical emission, which goes as density squared, will then be 2002 times, or about 40,000 times greater compared to that from the interclump medium. This is consistent with the fact that no broad line emission is seen from the interclump medium. If the initial velocity is much higher, as is conceivable, the clump density could be up to ∼ 50% higher, and the ratio between the emission from the dense clumps and interclump medium even larger.

What value of the clump density is suggested? A shock wave traveling at

1500 km s−1 would be radiative if it were expanding in a medium whose density is greater than ∼ 106 cm−3, whereas a 2500 km s−1 shock would require minimum densities of the order of 107 cm−3 (Draine & McKee 1993) in order to be radiative.

The CSM density, being two orders of magnitude smaller, would then to be

4 −3 ∼> 10 cm . These are just minimum values, and it is conceivable that the actual clump density is much higher. This result is consistent with the conclusion of

Aldering et al. (2006).

The observations show that the broad Hα width increases after 50 days, suggesting an increase in the clump shock velocity at later times, which could

207 perhaps be due to a decrease in the clump density. Conversely, however, the luminosity of the Hα also increases, suggesting an increase in the electron density.

At the same time, we would expect the SN shock to be decreasing in velocity as it continues its outwards expansion.

We suggest that the way to reconcile these observations is a scenario in which the density within the clump medium starts out higher than 108 cm−3, probably as high as 1010 cm−3 in the first few days, and decreases gradually outwards. The almost constant behavior of the FWHM of the broad Hα suggests that the density profile of the ambient medium is flatter than r−2. Since we want the clump shock to be radiative even when the shock velocity is almost 3000 km s−1, this suggest that the density at ∼ 150 days is greater than about 107 cm−3. And since the density is decreasing outwards, we infer that the density close in is even larger. Over the entire period of observations the clump density is large enough that the shock driven into these clumps is always radiative. The density of the ambient medium is two orders of magnitude smaller, as discussed above. The high bolometric luminosity is consistent with these values.

For the ﬁrst ∼ 50 days the Hα emission arises only from the radiative shock driven into the dense clumps. However, by ∼ 50 days the SN forward shock, which is decreasing in velocity, enters the radiative regime, and the cooling shell of material begins to contribute to the Hα luminosity. The velocity of the SN shock is quite large, and its contribution initially is not a large fraction of the total Hα luminosity.

208 But as it expands outwards, its velocity decreases and the shock becomes more radiative, and the contribution to the total Hα luminosity increases, more than compensating for the decreasing density. If this conclusion is correct, then we would expect that a broad velocity component would be visible in the Hα spectra, whose intensity would gradually increase with time even as the FWHM decreases.

Although the underlying supernova contamination makes it hard to isolate a broad component, it is suggestive that by day ∼ 150 the spectrum is best ﬁt by a third, much broader component of the velocity, thus providing support for this line of reasoning.

Finally, in this model the narrow line emission arises from the unshocked slowly expanding ambient material, presumably the outflow that we find expanding at ∼60 km s−1. We note that although the width of the narrow line Hα emission as listed is higher, it is still unresolved, and it is possible that within the limits of resolution the narrow line component and outflow velocity are indeed the same.

To summarize, in this model the Type Ia SN expands in a clumped ambient medium, with the clump density about ∼ 200 times that of the surrounding medium close in to the star, and decreasing as we go outwards. The Hα emission initially arises mainly from the shock driven into the dense clumps. The SN shock propagating into the interclump medium begins to enter the radiative regime around day 50, and its contribution to the Hα emission gradually increases beyond that coming from the clumped medium, leading to the gradual rise in the Hα emission.

209 We note that several features of this model are similar to the model presented by

Chugai et al. (2004) for SN 2002ic, thus further supporting the similarity between the two supernovae.

The upper limit on the X-ray luminosity obtained at 74 days after the explosion can put a constrain on the mass loss rate from the precursor or companion (e.g.,

Immler et al. 2006). Assuming that the X-ray luminosity is dominated by emission

˙ −4 −1 from the reverse shock we obtain M ∼< 2 × 10 M⊙ yr (2σ) using Equation 3.10 in Fransson et al. (1996). This value has to be taken as an approximate estimate because we are making several assumptions about the physical properties of the ejecta-CSM interaction that should be calculated using detailed hydrodynamical

−1 simulations: a constant velocity of the shock, Vsh ≈ 8000 km s ; a solar composition

7 of the CSM material; an electron temperature at the reverse shock of Te = 10 K, which comes from the modeling of SN 2002ic (Nomoto et al. 2005); a flat density profile of the CSM, ρ ∝ r−2; and a power-law ejecta density profile with index n = 7

(Nomoto et al. 1984b).

We can also estimate a mass loss rate from the companion using the density of the ambient medium (n ∼ 108 cm−3), the initial optical radius of the CSM

(R ≈ 1015 cm), and the velocity of the wind: M˙ = 4π R2 vρ, this is assuming a

−4 −1 ﬂat density proﬁle for the CSM. We obtain: M˙ ≈ 2 × 10 M⊙ yr , which is in agreement with the 2σ upper limit calculated from the X-ray luminosity.

210 The presence of Balmer lines in emission in the first spectrum obtained 6.6 days after explosion shows that the ejecta started to interact with the CSM at an earlier epoch (Aldering et al. 2006). Extrapolating linearly to zero flux the early increase of Hα and Hβ fluxes we find that the ejecta-CSM interaction started 3 ± 1 days

15 after explosion, which gives an internal radius of the CSM Ri ≈ 1.1 × 10 cm. The outer radius of the CSM can be estimated assuming a constant velocity of the shock

−1 of Vsh ≈ 8000 km s over the ﬁrst year. We detect Hα in emission in a spectrum obtained at 368 days after explosion, which will be presented elsewhere, putting a

16 lower limit on the outer radius of the CSM, Ro ∼> 3 × 10 cm. This is also consistent with a Type Ia SN with an exponential ejecta density proﬁle expanding outwards in

7 −3 a medium of average density ∼> 10 cm .

In the interpretation above we assume that the broad component of the

Hydrogen Balmer lines originate in the dense clumps, while the narrow component arises from the photoionized un-shocked gas. However, Thompson scattering of the lines has been considered as an alternative mechanism that can explain relatively well the symmetric line proﬁles of SN 2002ic (Wang et al. 2004) and SN 2005gj

(Aldering et al. 2006). In this scenario, both components would arise from a single high density region. The total mass of hydrogen in the emitting region would

−2 10 be MH ≈ 2 × 10 (10 /ne) M⊙, where ne is the average electron density in the emitting zone, as calculated from the luminosity of the Hα line at maximum using the Case B recombination coeﬃcient. The electron density must be suﬃciently

211 10 −3 high, ne ≈ 10 cm , to be consistent with the line ratios of He lines observed in the spectra (Aldering et al. 2006), and a high electron density would explain the non-detection in X-ray and radio (Soderberg & Frail 2005). However, it is unlikely that this model would be able to explain the initial constancy and then rise of the broad Hα luminosity.

6.6.2. Rates, hosts galaxies and possible progenitors of

SN 2002ic-like supernovae

The SDSS-II Supernova Survey has a well understood discovery eﬃciency of

SN Ia at low redshift (z ∼< 0.1), which allows us to obtain an accurate supernova rate measurement controlling systematic errors (Dilday et al. 2008). In the fall

2005 season there were a total of 16 spectroscopically conﬁrmed SN Ia (including

1991T-like and 1991bg-like objects) at z < 0.12, one photometric identification, and the spectroscopically confirmed peculiar events: SN 2005hk (Phillips et al. 2007) and SN 2005gj. Since the detection efficiency of 2002ic-like objects has not been carefully modeled, we can only put a lower limit on the fraction of these events.

+7 The spectroscopic conﬁrmation of one object at z < 0.12 puts a lower limit of 5−4%

(68% conﬁdence) in the fraction of 2002ic-like events among SN Ia at low redshift.

From the previously known (2002ic) and probable events (1997cy and 1999E) the

212 estimated fraction is ∼ 1% of SN Ia discovered between 1997 and 2002, which is consistent with our limit.

In the fall of 2006 we obtained the spectrum of a slowly declining supernova that was discovered in 2005, but did not have a spectroscopic classiﬁcation, SN 70176.

To our surprise, the late spectrum of SN 7017 resembles that of SN 2005gj one year after explosion and the early photometry also shows similarities which lead us to classify it as the highest redshift SN 2002ic-like object observed to date, at z = 0.27.

Considering SN 7017 in the spectroscopically conﬁrmed sample of SN Ia during the

2005 season, a total of 129 SNe at z ∼< 0.42, we have that 2/130 (1.5%) are SN 2002ic like objects, which is consistent with the low limit on the fraction at low redshift estimated before. However, this fraction has to be taken with extreme care and probably does not reflect the true fraction. This is because the discovery efficiency of SN Ia declines as a function of redshift and the total number of spectroscopically confirmed SN Ia does not include SNe with good Ia-like light curves that were not spectroscopically classified. A more careful study of the rates of SN 2005gj-like supernovae in the SDSS-II is planned for a future publication.

The host galaxies of supernovae can provide important clues about their progenitors. The host of SN 2005gj is a very blue, low-luminosity dwarf

(MB ≈ −17), and has an irregular morphology with no well deﬁned core. Aldering

6This is the internal name given by the SDSS-II Collaboration. It was not announced in an IAU circular because of the late spectroscopic classiﬁcation.

213 et al. (2006) combined the SDSS photometry with UV imaging from GALEX to construct an SED of the galaxy. They constrained the metallicity to Z < 0.3 Z⊙, with a burst of star formation ∼ 200 Myr ago. SN 2002ic has a late type (Sbc) spiral host with a well deﬁned core. The host of SN 1997cy is also a blue, low-luminosity

(MV ≈ −18.2), and low surface brightness dwarf irregular galaxy (Germany et al.

2000). GALEX has imaged the positions of SN 1997cy, 2002ic and 2005gj, and their hosts galaxies are all detected in the Near-UV (NUV) band. Their absolute magnitudes in the NUV (AB magnitudes) are between -16.6 (SN 2005gj) and -17.3

(SN 1997cy and 2002ic). They are low-luminosity late type galaxies, ∼ 1 − 1.7 magnitudes fainter than L∗ galaxies observed by GALEX at redshift z < 0.1 (Wyder et al. 2005). The host galaxy of SN 1999E is a late spiral with a nuclear starburst

ﬁrst observed by the IRAS satellite (Allen et al. 1991). From 2MASS photometry, its absolute magnitude is 1 mag brighter than an L∗ galaxy in the K-band (Kochanek et al. 2001). SN 7017 at redshift z = 0.27, has a blue, dwarf-like host galaxy with absolute magnitude in B of −17.9.

The host galaxies of the ﬁve SN 2002ic-like objects known share some common properties: they are late type galaxies, irregulars and late spirals, most likely with recent star formation. Four of the host galaxies have low luminosities, similar to the

Magellanic clouds, which indicates they are low metallicity systems. For example, a dwarf galaxy with intrinsic luminosity MB = −18 has an Oxygen abundance of 12 + log(O/H) ≈ 8.4 (van Zee et al. 2006), which corresponds to 1/3 the solar

214 Oxygen abundance (Delahaye & Pinsonneault 2006). On the other hand, the host galaxy of SN 1999E has a K-band luminosity, that when converted to metallicity using the luminosity-metallicity relationship derived by Salzer et al. (2005), makes it consistent with the solar value. The host luminosities are only an approximate indicator of their metallicities, therefore spectra of the hosts are needed to infer the metallicities and star formation rates (SFRs) of these galaxies. However, it is interesting to note that the range of host galaxy properties of SN 2002ic-like events seem to be inconsistent with the host galaxies of GRBs associated with supernovae

(Stanek et al. 2006) and broad-lined type Ib/c SNe (Modjaz et al. 2008).

Type Ia supernovae are observed in all types of galaxies. There is a well established correlation between the morphology of their host galaxies and the peak luminosity of the SNe: brighter supernovae (1991T-like) tend to explode in late type spirals and irregulars with recent star formation, while intrinsically fainter events (1991bg-like) are observed mainly in early type galaxies with an old stellar population (Hamuy et al. 1995, 1996a; Branch et al. 1996; Hamuy et al. 2000;

Gallagher et al. 2005). This environmental eﬀect and observations of the local supernovae rate as a function of host galaxy properties (Cappellaro et al. 1999;

Mannucci et al. 2005), motivated Scannapieco & Bildsten (2005) to parametrize the delay time distribution, time between star formation and the appearance of SNe, and the rates with a two-component model having a piece proportional to the SFR of the host galaxy (or prompt, they explode ∼ few × 108 yr after an episode of

215 star formation), and a second piece proportional to the total stellar mass (delayed component, they explode on scales of a few Gyr after the onset of star formation).

The difference in age of the stellar populations of these subclasses suggests that the progenitors may also be different: prompt SN Ia would come from more massive progenitors. The host galaxies of all five SN 2002ic-like events known are broadly consistent with the properties of the hosts of prompt SN Ia, which suggest a real association given that the best studied SN IIa to date, SN 2002ic and SN 2005gj, have spectral characteristics similar to 1991T-like events.

Several progenitors have been discussed in the literature for SN 2002ic and

SN 2005gj. Livio & Riess (2003) proposed that SN 2002ic is a rare case of a double-degenerate binary system, a white dwarf (WD) and the core of an AGB star spiraling-in through gravitational wave losses, in which the explosion occurs during or immediately after the common-envelope phase (a few hundred to a few thousand years of duration). The diﬀerence in line strengths of the Balmer emission lines observed for SN 2002ic and SN 2005gj makes this scenario unlikely. Also, as Aldering et al. (2006) points out, in both SN 2002ic and SN 2005gj the mass loss stopped only a few years before explosion, which is too short compared with the timescale for gravitational wave radiation to produce the merger of the core and the WD.

Another possible progenitor initially proposed by Hamuy et al. (2003) and favored by the models of Chugai et al. (2004), is the explosion of the Chandrasekhar- mass Carbon-Oxygen core of a massive AGB star in a degenerate medium, a

216 supernova Type 1.5 (Iben & Renzini 1983), where the dense Hydrogen-rich CSM would come from the outer layers of the AGB. In order for the core to grow to the

Chandrasekhar mass, the radiatively driven winds from the AGB have to be weak enough, a condition that is only met in a very low-metallicity environment like the

Galactic halo (Zijlstra 2004). At face value, the range of host galaxy metallicities for SN 2002ic-like events inferred from the luminosity-metallicity relation does not support the SN 1.5 scenario, although admittedly these are average metallicities and do not tell us the actual range of metallicities of the progenitors.

Han & Podsiadlowski (2006) proposed that SN 2002ic could be produced through the “super-soft channel”, the most common single-degenerate model for the progenitors of SN Ia. In this scenario the white dwarf is accreting material from a main sequence, or slightly evolved, relatively massive companion (∼ 3 M⊙) and experiences a delayed dynamical instability that leads to a large amount of mass-loss from the system in the last few × 104 yr before the explosion. Aldering et al. (2006) notes that the estimated main-sequence mass of the progenitor of SN 2005gj of

∼ 2 M⊙, calculated using the age of the starburst of its host galaxy, is consistent with the Han & Podsiadlowski (2006) model. Also, the predicted fraction of SN Ia that would be produced through the “delayed dynamical” channel is 0.1-1%, consistent with the limits we have obtain from the detection of SN 2005gj in the SDSS-II

Survey.

217 In general terms, the progenitor model proposed by Han & Podsiadlowski

(2006) successfully reproduces the observational properties of SN 2002ic and

SN 2005gj. However, it is still very early in the study of this new sub-class of SN Ia.

It would be interesting to see in the near future the results of theoretical modeling exploring other single degenerate conﬁgurations (e.g., AGB donor) and detailed hydrodynamical modeling of the ejecta-CSM interaction of SN 2005gj using the observations of the early photometric and spectroscopic evolution presented in this work.

218 N 16

15 3

6 2 SN E 1

12 8 4

13 9 14 10

Fig. 6.1.— r′-band image (3.5′ ×3.5′) of the field around SN 2005gj obtained with the Swope-1m telescope at LCO. North is up and east is to the left. Sixteen comparison stars in common between SDSS and CSP used to derive differential photometry of the SN are labeled as in Table 6.1. The SN is close to the center of the field.

219 Fig. 6.2.— Spectra of SN 2005gj obtained from Oct. 1 (∼7 days after explosion) to Nov. 28 (∼61 days after explosion) of 2005. The sequence show the dramatic spectral evolution of the SN from a very blue continuum with strong Hydrogen- Balmer lines in emission in the early phases, resembling the spectrum of a Type IIn SN, to a Type Ia supernova-dominated continuum with broad absorption and emission features (P-cygni profiles) of blended Fe II and Fe III profiles. The spectra are shown in logarithmic flux scale and a constant shift has been applied for clarity. The wavelength is in the rest-frame corrected using z = 0.0616 for the host galaxy. We show the UT date when the spectra were obtained and the epoch (rest-frame days after explosion) in parenthesis.

220 Fig. 6.3.— Late time spectra of SN 2005gj obtained between Dec. 8, 2005 (∼71 days after explosion) and Mar. 6, 2006 (∼152 days after explosion). The labels, axis and symbols are the same as in Figure 6.2. The earth symbol shows the position of a telluric feature present in some of the spectra.

221 Fig. 6.4.— Observed light curves of SN 2005gj from SDSS (open circles), MDM (open squares) and CSP/Swope (ﬁlled triangles). The error bars are smaller than the symbols. For clarity, the light curves have been shifted by an arbitrary constant.

222 Fig. 6.5.— Diﬀerence between the synthetic g − r color calculated from the spectra and the observed color from the photometry. We do not include the latest spectra obtained on Jan 24. and Mar. 6 because there is no contemporaneous photometric data.

223 Fig. 6.6.— Time evolution of the colors of SN 2005gj (ﬁlled circles). For comparison we also show the color evolution of the overluminous Type Ia SN 1991T (solid line), the Type IIn SN 1999el (dashed line), and two previous cases of Type Ia strongly interacting with its circumstellar medium, SN 2002ic (stars) and SN 1997cy (triangles).

224 Fig. 6.7.— Absolute ugri light curves of SN 2005gj (ﬁlled circles). For comparison we also show the absolute light curves of the overluminous Type Ia SN 1991T (solid line), SN 2002ic (stars) and SN 1997cy (triangles). The error bar in the lower right pannel represents the typical error in the absolute magnitudes dominated by a 10% uncertainty in the Hubble constant.

225 Fig. 6.8.— Absolute light curves of SN 2005gj in the NIR: J (top panel), H (middle panel) and Ks (bottom panel). For comparison we also show the absolute light curves of a normal Type Ia (solid line) and the Type IIn SN 1999el (open triangles). The error bar in the lower right of each panel represents the typical error in the absolute magnitudes dominated by a 10% uncertainty in the Hubble constant.

226 Fig. 6.9.— Examples of black-body fits (solid line) to the SED of SN 2005gj obtained by transforming the rest-frame ugriz magnitudes to monochromatic fluxes at the effective wavelength of the filters (filled circles). These examples show the quality 2 (i.e., goodness-of-fit) range of the black body fits at different epochs: χν =0.3 (left panel), 2.4 (middle), 4.4 (right). The units of flux density in the y-axis are mJy = 10−26 erg s−1 cm−2 Hz−1.

227 Fig. 6.10.— Top panel: Quasi-bolometric (open circles) and bolometric light curves of SN 2005gj (filled circles). The bolometric light luminosities were obtained after applying bolometric corrections calculated from black-body fits to the optical SED obtained from the ugriz photometry. The dashed line shows the best-fit linear decay of 0.014 mag day−1. The middle and bottom panels show the evolution of the black- body temperature and radius, respectively.

228 Fig. 6.11.— Comparison of spectra of SN 2005gj at 26, 46, 62 and 83 days after explosion with comparable epoch spectra of SN 2002ic (Hamuy et al. 2003) and SN 1997cy (from SUSPECT database). The spectra are plotted on a logarithmic ﬂux scale and shifted by an arbitrary constant. The wavelength was shifted to the restframe using z = 0.0616 of the host galaxy (Aldering et al. 2006).

229 Fig. 6.12.— Results of SNID. We show the spectra of SN 2005gj at four epochs (26, 37, 61, and 85 days after explosion) and their best three cross-correlation library spectra. The spectra are plotted on a logarithmic ﬂux scale and shifted by an arbitrary constant.

230 Fig. 6.13.— Fits to the spectra of SN 2005gj. We model the spectra as the sum of two components: (1) SN 1991T spectrum at the same epoch after explosion as SN 2005gj scaled by an arbitrary constant (blue line); (2) fourth order polynomial (green line). The results of the ﬁts are in red and the spectra of SN 2005gj, corrected by Galactic extinction in the line of sight are in black. The epochs of the spectra are shown in the upper right of each panel.

231 Fig. 6.14.— Results from the Gaussian ﬁts to Hα and Hβ emission features as a function of time. Top left panel: FWHM of the Hα-broad and Hβ. Top right panel: Luminosity of the narrow and broad Gaussian components of Hα. Bottom left panel: Luminosity of Hβ. Bottom right panel: Balmer decrement, ratio of total ﬂuxes in Hα and Hβ lines.

232 Fig. 6.15.— Line profiles of Hβ (top) and Hα (bottom) in the two highest resolution spectra of SN 2005gj obtained with WHT+ISIS on day 44 (left) and Magellan+LDSS- 3 on day 115 (right). The features show clear P-cygni profiles with weak absorption minima at ∼ −200 kms−1, demonstrating the presence of a slowly moving outflow.

233 Fig. 6.16.— Identiﬁcation of lines in the spectrum of SN 2005gj obtained at 17 days after explosion (2 days before g maximum). The red line shows the best ﬁt synthetic spectrum generated with the SYNOW code. The lines of SN 2005gj are typical of SN 1991T around maximum light (blue line), and very similar to the spectrum of SN 2002ic around maximum (green line). All the spectra have been locally normalized. We have subtracted a constant value to the spectra of SN 2002ic and SN 1991T for clarity.

234 Star u g r iz ID SDSS CSP SDSS CSP SDSS CSP SDSS CSP SDSS

1 18.639(027) 18.690(054) 17.367(018) 17.362(009) 16.837(022) 16.823(013) 16.643(018) 16.586(010) 16.549(020) 2 ······ 19.122(027) 19.043(018) 17.609(014) 17.575(015) 16.191(017) 16.051(017) 15.403(018) 3 19.843(039) 19.899(097) 17.676(016) 17.653(023) 16.745(012) 16.726(015) 16.404(015) 16.362(010) 16.215(017)

235 4 ······ 20.116(026) 20.038(045) 19.579(018) 19.556(062) 19.356(025) 19.113(067) 19.207(040) 5 ··· 20.262(436) 18.856(021) 18.770(015) 17.345(009) 17.323(009) 16.595(014) 16.505(010) 16.145(064) 6 16.074(030) 16.102(011) 14.556(010) 14.544(011) 13.965(016) 13.961(015) 13.774(007) 13.728(015) 13.711(023) 7 19.915(041) 20.347(285) 18.385(019) 18.384(017) 17.741(020) 17.735(010) 17.502(016) 17.451(013) 17.372(021) 8 18.486(027) 18.533(030) 17.214(011) 17.204(009) 16.694(017) 16.685(009) 16.508(018) 16.473(009) 16.418(016)

(cont’d) Table 6.1. SDSS ugriz and CSP u′g′r′i′ photometry of comparison stars in common in the ﬁeld of SN 2005gj. Table 6.1—Continued

Star u g r iz ID SDSS CSP SDSS CSP SDSS CSP SDSS CSP SDSS

9 18.847(021) 18.923(041) 17.458(013) 17.453(009) 16.940(011) 16.936(011) 16.765(018) 16.732(011) 16.679(024) 10 18.285(025) 18.306(026) 16.956(018) 16.935(011) 16.412(011) 16.404(009) 16.208(017) 16.181(009) 16.125(021) 11 ······ 20.227(033) 20.118(058) 18.717(014) 18.715(017) 17.861(021) 17.787(014) 17.347(020) 236 12 ······ 18.927(020) 18.860(015) 17.433(014) 17.404(020) 16.669(018) 16.611(010) 16.243(024) 13 18.510(035) 18.529(031) 16.922(017) 16.897(009) 16.249(016) 16.226(011) 15.992(017) 15.958(009) 15.868(016) 14 20.030(065) 20.081(196) 17.833(019) 17.819(010) 16.882(007) 16.864(013) 16.515(017) 16.470(009) 16.306(015) 15 18.748(029) 18.314(545) 16.915(010) 16.888(009) 16.142(013) 16.117(009) 15.845(022) 15.798(009) 15.708(015) 16 18.767(032) 18.834(109) 17.458(013) 17.454(011) 16.896(015) 16.882(009) 16.692(013) 16.652(009) 16.599(018)

Note. — Uncertainties given in parentheses are in thousandths of a magnitude. For the CSP photometry with the Swope they correspond to the rms of the magnitudes obtained on four photometric nights, with a minimum uncertainty of 0.015 mag for an individual measurement. JD Epocha −2, 453, 000 (days) u g r i z Source

616.94 . . . 24.00±1.24 25.19±0.80 23.81±0.61 23.36±0.54 22.15±0.53 SDSS 626.91 . . . 21.12±0.70 27.83±0.86 22.85±0.45 24.92±0.93 22.28±1.77 SDSS 628.90 . . . 23.88±1.45 25.72±1.23 22.92±0.41 25.98±0.45 21.72±0.54 SDSS 635.91 . . . 22.00±0.49 26.38±1.11 23.32±0.62 23.27±0.59 22.15±0.53 SDSS 639.95 1.9 18.887(045) 18.592(025) 18.621(010) 18.718(016) 18.924(043) SDSS 641.95 3.8 18.468(034) 18.154(011) 18.256(020) 18.340(016) 18.511(041) SDSS 644.89 6.6 18.121(023) 17.795(011) 17.879(022) 17.935(010) 18.141(022) SDSS 656.94 17.9 . . . 17.355(026) 17.255(029) 17.360(064) . . . MDM

237 657.90 18.8 17.949(039) 17.343(012) 17.214(012) 17.275(016) 17.438(024) SDSS 663.89 24.5 18.149(039) 17.410(050) 17.158(012) 17.163(012) 17.280(027) SDSS 663.92 24.5 . . . 17.368(025) 17.160(020) 17.248(055) . . . MDM 665.91 26.4 ...... 17.146(028) 17.215(054) . . . MDM 666.96 27.3 18.282(043) 17.408(012) 17.121(012) 17.133(013) 17.268(015) SDSS 668.87 29.2 . . . 17.393(014) 17.125(026) 17.186(045) . . . MDM 668.88 29.2 18.355(028) 17.401(009) 17.127(012) 17.096(016) 17.240(016) SDSS 669.97 30.2 18.347(045) 17.430(011) 17.101(014) 17.105(031) 17.231(017) SDSS 670.88 31.0 18.386(024) 17.419(007) 17.125(013) 17.092(014) 17.249(021) SDSS 673.85 33.8 18.479(028) 17.450(009) 17.123(012) 17.081(014) 17.229(015) SDSS

(cont’d) Table 6.2. SDSS and MDM ugriz photometry of SN 2005gj Table 6.2—Continued

JD Epocha −2, 453, 000 (days) u g r i z Source

675.85 35.7 18.551(025) 17.477(010) 17.134(029) 17.088(019) 17.182(019) SDSS 676.90 36.7 . . . 17.464(043) 17.116(026) 17.138(055) . . . MDM 680.86 40.4 18.663(027) 17.578(009) 17.133(011) 17.054(009) 17.156(016) SDSS 683.92 43.3 18.759(062) 17.656(034) 17.220(055) 17.080(015) 17.061(066) SDSS 686.86 46.1 18.887(047) 17.695(014) 17.160(011) 17.060(010) 17.138(021) SDSS

238 687.91 47.1 18.937(105) 17.687(013) 17.206(024) 17.042(018) 17.119(026) SDSS 693.86 52.7 19.018(081) 17.780(009) 17.207(015) 17.084(014) 17.130(018) SDSS 699.85 58.3 . . . 17.864(060) 17.271(034) 17.202(057) 17.166(024) MDM 699.88 58.4 19.083(042) 17.909(013) 17.273(009) 17.124(010) 17.135(020) SDSS 727.70 84.6 . . . 18.368(025) 17.639(043) 17.551(053) 17.407(015) MDM 737.75 94.0 . . . 18.525(038) 17.791(033) 17.709(063) 17.470(057) MDM 739.64 95.8 . . . 18.540(037) 17.822(035) 17.744(065) 17.548(027) MDM

Note. — Uncertainties given in parentheses in thousandths of a magnitude.

aRest-frame days since the time of explosion (JD 2,453,637.93). JD Epoch −2, 453, 000 (days) u′ g′ r′ i′

698.69 57.2 19.159(042) 17.884(017) 17.243(017) 17.106(017) 699.67 58.2 19.196(041) 17.910(017) 17.261(017) 17.100(017) 702.68 61.0 19.286(042) 17.959(017) 17.281(017) 17.130(017) 706.69 64.8 19.312(040) 18.031(017) 17.345(017) 17.173(017) 712.64 70.4 19.619(060) 18.157(017) 17.418(017) 17.249(017) 720.66 77.9 19.721(131) 18.283(023) 17.542(017) 17.363(017) 725.65 82.6 19.964(079) 18.365(017) 17.602(017) 17.412(017) 728.71 85.5 19.938(104) 18.430(017) 17.658(017) 17.453(017) 736.62 93.0 20.104(084) 18.554(017) 17.794(017) 17.586(017) 740.64 96.8 20.229(090) 18.606(017) 17.856(017) 17.641(017) 741.59 97.6 20.238(103) 18.635(017) 17.872(017) 17.651(017) 746.61 102.4 20.532(220) 18.641(025) 17.926(017) 17.725(017) 754.58 109.9 20.509(143) 18.770(017) 18.021(017) 17.803(017) 761.64 116.5 ··· 18.847(017) 18.100(017) 17.910(018) 763.56 118.3 ··· 18.841(017) 18.107(017) 17.960(017) 764.58 119.3 ··· 18.859(017) 18.143(017) 17.947(017) 768.60 123.1 ··· 18.882(019) 18.200(017) 17.978(021) 773.55 127.7 ··· 18.884(026) 18.239(019) 18.050(021) 774.56 128.7 ··· 18.946(031) 18.221(017) 18.079(022) 783.55 137.2 ··· 19.017(018) 18.322(017) 18.117(021) 786.53 140.0 ··· 19.015(017) 18.363(017) 18.178(020) 795.55 148.5 ··· 19.099(019) 18.416(021) 18.227(031)

Note. — Uncertainties given in parentheses in thousandths of a magnitude.

Table 6.3. CSP u′g′r′i′ photometry of SN 2005gj

239 JD Epoch

−2, 453, 000 (days) Y J H Ks Instrument

700.71 59.1 16.565(015) 16.484(034) 16.253(030) ··· Retrocam 704.68 62.9 16.591(015) 16.537(020) 16.315(033) ··· Retrocam 709.66 67.6 16.628(015) 16.550(023) 16.271(029) ··· Retrocam 714.58 72.2 16.673(015) 16.594(020) 16.389(028) ··· Retrocam 718.65 76.0 16.725(016) 16.658(022) 16.364(037) ··· Retrocam 722.62 79.8 16.832(016) 16.725(032) 16.490(017) 16.384(096) WIRC 724.68 81.7 16.781(016) 16.716(028) ······ Retrocam 727.69 84.6 16.872(024) 16.757(036) ······ Retrocam 732.71 89.3 16.920(019) 16.839(025) ······ Retrocam 750.61 106.1 17.307(016) 17.152(016) 16.812(016) ··· WIRC 755.57 110.8 17.337(024) 17.263(024) 16.891(025) 16.745(039) PANIC 756.59 111.8 17.428(016) 17.269(016) 16.928(017) ··· WIRC 757.60 112.7 17.423(024) 17.270(024) 16.938(025) 16.807(042) PANIC 773.57 127.8 17.677(016) 17.476(016) ······ WIRC 776.54 130.6 17.538(026) 17.399(070) 17.081(068) ··· Retrocam 777.55 131.5 17.648(038) 17.417(056) 17.129(119) ··· Retrocam 782.55 136.2 17.513(039) 17.436(073) 17.072(114) ··· Retrocam 783.55 137.2 17.766(024) 17.586(024) 17.271(025) 17.045(042) PANIC 785.53 139.0 17.765(025) 17.570(022) 17.192(049) 17.152(177) WIRC 788.54 141.9 17.869(016) 17.594(016) 17.335(019) ··· WIRC 797.52 150.3 17.855(061) 17.563(118) 17.313(158) ··· Retrocam 800.50 153.1 17.840(041) ········· Retrocam 808.52 160.7 17.871(058) 17.551(079) ······ Retrocam 814.49 166.3 17.864(084) ········· Retrocam

Note. — Uncertainties given in parentheses in thousandths of a magnitude.

Table 6.4. CSP YJHKs photometry of SN 2005gj

240 Parameter Value

Time of explosiona 637.93 ± 2.02

Time of umax 651.77 ± 0.48

Time of gmax 657.80 ± 1.28

Time of rmax 672.49 ± 1.39

Time of imax 684.51 ± 1.00

umax 17.85 ± 0.05

gmax 17.35 ± 0.01

rmax 17.12 ± 0.01

imax 17.05 ± 0.01 0 b umax 17.11 ± 0.12 0 gmax 16.94 ± 0.09 0 rmax 16.83 ± 0.07 0 imax 16.89 ± 0.11 0 Mg,max –20.21 c E(B − V )Gal 0.121 ± 0.019

Ag(Gal) 0.45 ± 0.07

aJD-2, 453, 000 bMagnitudes at maximum in the rest-frame, they have been corrected by Galactic extinction and K-corrections. We assume a negligible extinction in the host galaxy. cFrom Schlegel et al. (1998)

Table 6.5. Light-curve parameters for SN 2005gj

241 JD Epoch Wavelength Resolutiona Exposure −2, 453, 000 (days) Instrument Range (A)˚ (A)˚ (s)

644.92 6.6 MDM-CCDS 3850 – 7308 15 1200 646.95 8.5 ARC-DIS 3824 – 10192 7 1800 650.84 12.2 ARC-DIS 3600 – 9597 7 1000 655.87 16.9 MDM-CCDS 3823 – 7283 15 1800 665.92 26.4 MDM-CCDS 3883 – 7341 15 2700 668.83 29.1 MDM-CCDS 3886 – 7346 15 2700 676.79 36.6 MDM-CCDS 3882 – 7338 15 3600 684.73 44.1 WHT-ISIS 3924 – 8901 4 1800 686.79 46.0 MDM-CCDS 3858 – 7315 15 2700 698.67 57.2 duPont-ModSpec 3780 – 7290 7 2700 699.67 58.2 duPont-ModSpec 3780 – 7290 7 2700 700.76 59.2 MDM-CCDS 3933 – 7391 15 2700 702.73 61.1 MDM-CCDS 3856 – 7310 15 2700 712.73 70.5 MDM-CCDS 3831 – 7286 15 2700 722.71 79.9 NTT-EMMI 4000 – 10200 9 2700

(cont’d) Table 6.6. Spectroscopic observations of SN 2005gj

242 Table 6.6—Continued

JD Epoch Wavelength Resolutiona Exposure −2, 453, 000 (days) Instrument Range (A)˚ (A)˚ (s)

724.66 81.7 duPont-WFCCD 3800 – 9235 6 2700 725.65 82.6 duPont-WFCCD 3800 – 9235 6 2700 726.66 83.6 duPont-WFCCD 3800 – 9235 6 3600 727.67 84.5 duPont-WFCCD 3800 – 9235 6 3600 728.67 85.5 duPont-WFCCD 3800 – 9235 6 3600 729.67 86.4 MDM-CCDS 3915 – 7373 15 2700 737.70 94.0 MDM-CCDS 3909 – 7368 15 2700 751.60 107.1 NTT-EMMI 3200 – 10200 9 2700 755.62 110.9 MDM-CCDS 3844 – 7299 15 3600 759.61 114.6 Magellan-LDSS-3 3788 – 9980 3 3600 799.52 152.2 duPont-WFCCD 3800 – 9235 6 1200

Note. — Most of the spectra are the combination of multiple observation, the total exposure is given. aAverage resolution obtained from the FWHM of arc-lamp lines.

243 Epoch

(days) Ku Kg Kr Ki Kz

6.6 0.039 –0.077 –0.053 –0.141 –0.133 12.2 0.162 –0.097 –0.005 –0.196 –0.047 16.9 0.226 –0.052 –0.055 –0.135 –0.223 26.4 0.276 –0.015 –0.058 –0.116 –0.228 29.1 0.255 0.008 –0.056 –0.104 –0.239 36.6 0.296 0.034 –0.030 –0.118 –0.096 46.0 0.251 0.063 –0.025 –0.079 –0.201 57.2 0.195 0.106 0.049 –0.123 –0.152 58.2 0.186 0.101 0.013 –0.068 –0.216 59.2 0.313 0.090 0.006 –0.072 –0.215 61.1 0.241 0.091 0.011 –0.096 –0.136 70.5 0.505 0.086 0.055 –0.113 –0.054 79.9 0.439 0.112 0.046 –0.090 –0.115 81.7 0.328 0.118 0.063 –0.096 –0.057 82.6 0.340 0.115 0.063 –0.104 –0.102 83.6 0.378 0.105 0.079 –0.099 –0.101 84.5 0.326 0.104 0.071 –0.104 –0.026 85.5 0.391 0.112 0.078 –0.110 –0.036 86.4 0.330 0.124 0.066 –0.094 –0.071 94.0 0.249 0.126 0.094 –0.081 –0.092 107.1 0.266 0.141 0.082 –0.084 –0.089 110.9 0.337 0.138 0.097 –0.091 –0.007

Table 6.7. K-corrections of SN 2005gj

244 a b Epoch log L(u→i) log L(u→z) log Lbol Tbb Rbb −1 −1 −1 15 2 c (days) (erg s ) (erg s ) (erg s ) (K) (10 cm) χν

1.9 42.822 42.862 43.285(0.109) 12898(1193) 1.02(0.16) 0.7 3.8 42.990 43.029 43.485(0.105) 13552(1316) 1.16(0.18) 0.3 6.6 43.141 43.180 43.637(0.098) 13569(1288) 1.38(0.21) 0.2 18.8 43.344 43.389 43.713(0.096) 10955(831) 2.32(0.34) 1.2 24.5 43.335 43.387 43.644(0.125) 9468(612) 2.87(0.42) 1.2 27.3 43.335 43.388 43.628(0.135) 9069(720) 3.06(0.50) 2.2 29.2 43.335 43.389 43.617(0.140) 8772(683) 3.22(0.53) 2.3 30.2 43.334 43.388 43.617(0.137) 8761(692) 3.23(0.54) 2.3 31.0 43.332 43.386 43.611(0.122) 8697(705) 3.25(0.55) 2.6 33.8 43.325 43.381 43.595(0.140) 8392(685) 3.43(0.59) 2.9 35.7 43.316 43.374 43.581(0.160) 8100(582) 3.61(0.59) 2.4 40.4 43.299 43.361 43.559(0.152) 7684(526) 3.92(0.64) 2.5 43.3 43.270 43.338 43.532(0.143) 7284(335) 4.24(0.59) 0.6 46.1 43.269 43.336 43.526(0.101) 7142(477) 4.36(0.73) 2.8 47.1 43.263 43.332 43.523(0.139) 7115(450) 4.38(0.70) 2.1 52.7 43.247 43.317 43.506(0.123) 6889(425) 4.58(0.74) 2.4 57.2 43.219 43.294 43.480(0.095) 6958(247) 4.32(0.54) 2.0 58.2 43.212 43.288 43.473(0.096) 6913(248) 4.34(0.55) 2.0 58.4 43.215 43.290 43.480(0.133) 6656(322) 4.78(0.70) 1.6 61.0 43.198 43.274 43.458(0.107) 6840(261) 4.34(0.56) 2.4 64.8 43.176 43.254 43.438(0.087) 6830(249) 4.27(0.54) 2.2 70.4 43.133 43.214 43.394(0.109) 6570(278) 4.33(0.58) 3.4 77.9 43.088 43.171 43.355(0.098) 6633(280) 4.10(0.54) 2.8 82.6 43.057 43.143 43.321(0.127) 6380(293) 4.18(0.58) 4.2

(cont’d) Table 6.8. Derived integrated luminosity and black-body ﬁts.

245 Table 6.8—Continued

a b Epoch log L(u→i) log L(u→z) log Lbol Tbb Rbb −1 −1 −1 15 2 c (days) (erg s ) (erg s ) (erg s ) (K) (10 cm) χν

85.5 43.038 43.125 43.305(0.096) 6459(282) 4.04(0.54) 3.5 93.0 42.987 43.076 43.256(0.103) 6360(272) 3.92(0.53) 3.6 96.8 42.963 43.054 43.232(0.101) 6305(279) 3.87(0.53) 3.9 97.6 42.955 43.047 43.226(0.105) 6318(279) 3.83(0.52) 3.8 102.4 42.935 43.026 43.205(0.133) 6372(332) 3.68(0.52) 4.4 109.9 42.896 42.989 43.165(0.117) 6339(313) 3.53(0.50) 4.4 116.5 42.859 42.953 43.135(0.123) 6549(350) 3.25(0.46) 3.8 118.3 42.855 42.947 43.128(0.138) 6566(362) 3.21(0.46) 4.0 119.3 42.847 42.940 43.122(0.117) 6548(357) 3.20(0.46) 4.0 120.2 42.839 42.932 43.114(0.120) 6536(352) 3.18(0.45) 3.9 123.1 42.831 42.924 43.104(0.127) 6533(370) 3.14(0.45) 4.3 127.7 42.818 42.909 43.087(0.155) 6523(399) 3.06(0.46) 5.0 128.7 42.809 42.901 43.078(0.153) 6486(399) 3.07(0.46) 5.2 137.2 42.776 42.869 43.043(0.169) 6404(418) 2.99(0.46) 6.3 140.0 42.766 42.857 43.030(0.197) 6377(421) 2.96(0.46) 6.5 148.5 42.738 42.830 42.999(0.205) 6246(438) 2.94(0.48) 8.0

aBlack-body temperature from the fits to the broadband photometry; 1σ uncertainty are given in parenthesis. bBlack-body radius from the fits to the broadband photometry; 1σ uncertainty are given in parenthesis and include a 10% uncertainty in the distance to SN 2005gj. cχ2 per degree of freedom of the black-body fits.

246 SN Name Class Epochs Reference

1990N Ia normal –14, –13, –8, –7, –6, 0, 4, 8, 15, 18, 39 1 1991T Ia 91T –9, –8, –7, –6, –5, –4, –2, –1, 9, 10, 11, 12, 15, 16, 17, 18, 19, 20, 21, 22, 23, 27, 43, 44, 47, 48, 51, 69, 77 2, 3 1991bg Ia 91bg 1, 3, 16, 18, 25, 32, 33, 46, 54, 85 4 1992A Ia normal –5, –1, 3, 5, 6, 7, 9, 11, 16, 17, 24, 28 5 1994I Ic normal –6, –4, –3, 0, 1, 2, 21, 22, 23, 24, 26, 30, 36, 38 6 1997ef Ic broad –14, –12, –11, –10, –9, –6, –5, –4, 7, 13, 14, 16, 17,

247 19, 22, 24, 27, 41, 45, 47, 49, 75, 80, 81 7 1998aq Ia normal –9, –8, –3, 0, 1, 2, 3, 4, 5, 6, 7, 19, 21, 24, 31, 32, 36, 51, 55, 58, 60, 63, 66, 79, 82, 91, 211, 231, 241 8 1998bu Ia normal –3, –2, –1, 9, 10, 11, 12, 13, 14, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 57 9 1998bw Ic broad –9, –7, –6, –3, –2, –1, 1, 3, 4, 6, 9, 11, 12, 13, 19, 22, 29, 45, 64, 125, 200, 337, 376 10 1999aa Ia 91T –11, –7, –3, –1, 5, 6, 14, 19, 25, 28, 33, 49, 47, 51 11

(cont’d) Table 6.9. Library of spectra used in SNID Table 6.9—Continued

SN Name Class Epochs Reference

1999ee Ia normal –9, –7, –2, 0, 3, 8, 10, 12, 17, 20, 23, 28, 33, 42 12 1999ex Ic normal –1, 4, 13 12 1999by Ia 91bg –4, –3, –2, –1, 2, 3, 5, 6, 7, 8, 10, 11, 25, 29, 31, 33, 42 13 2002ap Ic broad –5, –4, 3, 8, 10, 17, 19 14

248 2004aw Ic normal 1, 5, 6, 8, 15, 21, 22, 26, 28, 39, 35, 44, 49, 63, 64, 236, 260, 413 15 2006aj Ic broad –6, –5, –4, –3, –2, –1, 0, 2, 3 16

References. — (1) Leibundgut et al. (1991); (2) Jeﬀery et al. (1992); (3) Schmidt et al. (1994); (4) Leibundgut et al. (1993); (5) Kirshner et al. (1993); (6) Millard et al. (1999); (7) Iwamoto et al. (2000); (8) Branch et al. (2003); (9) Jha et al. (1999); (10) Patat et al. (2001); (11) Garavini et al. (2004); (12) Hamuy et al. (2002); (13) Garnavich et al. (2004); (14) Gal-Yam et al. (2002); (15) Taubenberger et al. (2006); (16) Modjaz et al. (2006). Hα (narrow) Hα (broad) Hβ

JD Epoch FWHMa fluxb FWHMa fluxb FWHMa fluxb −2, 453, 000 (days)

644.92 6.6 . . . 0.24(0.02) 1575 0.58(0.06) 776 0.50(0.05) 646.95 8.5 137 0.37(0.04) 1481 0.83(0.08) 1307 0.52(0.06) 650.84 12.2 314 0.69(0.07) 1731 1.25(0.13) 1462 0.75(0.10) 655.87 16.9 . . . 0.57(0.06) 1555 1.11(0.11) 1339 0.75(0.09) 249 665.92 26.4 . . . 0.53(0.05) 1569 1.03(0.10) 523 0.48(0.06) 668.83 29.1 . . . 0.37(0.04) 1234 0.99(0.10) 1275 0.55(0.07) 676.79 36.6 . . . 0.45(0.05) 1513 0.77(0.08) 884 0.28(0.04) 686.79 46.0 . . . 0.41(0.04) 1836 0.71(0.07) . . . 0.15(0.02) 698.67 57.2 . . . 0.45(0.04) 2115 0.91(0.10) . . . 0.15(0.03) 699.67 58.2 . . . 0.34(0.04) 2053 0.76(0.08) . . . 0.14(0.03) 700.76 59.2 . . . 0.36(0.04) 1830 0.77(0.08) 620 0.16(0.02)

(cont’d) Table 6.10. Results of the Gaussian ﬁts to Hα and Hβ features Table 6.10—Continued

Hα (narrow) Hα (broad) Hβ

JD Epoch FWHMa fluxb FWHMa fluxb FWHMa fluxb −2, 453, 000 (days)

702.73 61.0 . . . 0.41(0.04) 1978 0.67(0.07) . . . 0.11(0.02) 712.73 70.5 . . . 0.41(0.04) 2357 1.02(0.11) . . . 0.11(0.02) 722.71 79.9 . . . 0.34(0.04) 2413 1.03(0.11) 490 0.12(0.03) 724.66 81.7 . . . 0.34(0.04) 2137 0.99(0.10) 1067 0.19(0.03)

250 725.65 82.6 . . . 0.34(0.03) 2260 1.02(0.11) 568 0.11(0.02) 726.66 83.6 . . . 0.35(0.04) 2322 1.10(0.11) . . . 0.08(0.02) 727.67 84.5 160 0.32(0.03) 2364 1.02(0.10) . . . 0.10(0.02) 728.67 85.5 . . . 0.20(0.02) 1802 1.24(0.14) 1127 0.19(0.03) 729.67 86.4 . . . 0.37(0.04) 2687 1.15(0.12) 680 0.14(0.02) 737.70 94.0 . . . 0.29(0.03) 1941 0.85(0.09) 459 0.09(0.01) 755.62 110.9 . . . 0.32(0.03) 2236 1.14(0.12) 1031 0.16(0.02) 799.52 152.2 525 0.51(0.05) 3809 2.18(0.22) 1669 0.23(0.03)

aUnits of FWHM are in km s−1; FWHM is not presented when the spectral resolution is bigger than the measured value. bUnits of ﬂux are in 10−14 erg s−1 cm−2; 1σ uncertainties are given in parentheses. Chapter 7

LBT Discovery of a Yellow Supergiant Eclipsing Binary in the Dwarf Galaxy Holmberg IX

7.1. Introduction

Although small in number, massive stars are critical to the formation and evolution of galaxies. They shape the ISM of galaxies through their strong winds and high UV ﬂuxes, and are a major source of the heavy elements enriching the ISM

(e.g, Massey 2003; Zinnecker & Yorke 2007, and references therein). A large fraction of massive stars are found in binaries (e.g., Kiminki et al. 2007). Eclipsing binaries are of particular use because they allow us to determine the masses and radii of the components and the distance to the system. Many young, massive eclipsing binaries have been found and studied in our Galaxy, the LMC, and the SMC, primarily in OB associations and young star clusters (e.g., Bonanos et al. 2004; Peeples et al. 2007;

Gonz´alez et al. 2005; Hilditch et al. 2005). The study of massive eclipsing binaries beyond the Magellanic clouds has been limited until very recently, when variability searches using medium-sized telescopes with wide-ﬁeld CCD cameras, coupled with

251 spectroscopy using 8-meter class telescopes, have yielded the ﬁrst systems with accurately measured masses in M31 (Ribas et al. 2005) and M33 (Bonanos et al.

2006).

We conducted a deep variability survey of M81 and its dwarf irregular companion, Holmberg IX, using the Large Binocular Camera (LBC) mounted on the

Large Binocular Telescope (LBT), between January and October 2007. Holmberg IX is a young dwarf galaxy (age ∼< 200 Myr), with a stellar population dominated by blue and red supergiants with no signs of old stars in the red giant branch (Makarova et al. 2002). The dwarf may have formed during a recent tidal interaction between

M81 and NGC 2976 (e.g., Boyce et al. 2001). The gas-phase metal abundance of

Holmberg IX of between 1/8 and 1/3 solar (e.g., Miller 1995; Makarova et al. 2002) is consistent with this hypothesis (e.g., Weilbacher et al. 2003). A normal, isolated dwarf on the luminosity-metallicity relationship would have a metallicity of ∼ 1/20 solar (Lee et al. 2006).

In this paper, we report on the discovery of a 271 day period, evolved, massive eclipsing binary in Holmberg IX using data from the LBT. The overcontact system is the brightest periodic variable discovered in our LBT variability survey. It has

h m s an out-of-eclipse magnitude of Vmax = 20.7 mag and is located at α = 09 57 37.14,

δ = +69◦02′11′′ (J2000.0). In §7.2 we discuss the observations and data reduction, in §7.3 we present the light curve and the best-ﬁt eclipsing binary model, and note a similar eclipsing binary in the SMC. In §7.4 we discuss the results and their possible

252 implications for Type II supernovae. Throughout this paper, we assume the HST

Key Project distance to M81 of µ = 27.80 mag (3.6 Mpc; Freedman et al. 2001) as the distance to Holmberg IX, and correct only for a foreground Galactic extinction of E(B − V ) = 0.08 mag (Schlegel et al. 1998).

7.2. Observations

Holmberg IX was observed as part of a variability survey of the entire M81 galaxy conducted between January and October 2007 with the LBT 8.4-meter telescope (Hill et al. 2006), using the LBC-Blue CCD camera (Ragazzoni et al. 2006; Giallongo et al. 2008) during Science Demonstration Time. The survey cadence and depth (1 min single exposures, with ≥ 3 consecutive exposures) are optimized to detect and follow-up Cepheid variables with periods between 10–100 days (V ∼< 24 mag), getting better than 10% photometry in the B and V ﬁlters. We obtained 168 V -band images on 24 diﬀerent nights, and 87 B-band images on 13 nights. We coadded the B-band images from each night (usually 3–4) to improve the signal-to-noise in the combined images. The seeing (FWHM) varied between

0′′.7 − 2′′.0 in V (median 1′′.4), and between 1′′.0 − 3′′.3 in B-band (median 1′′.9). Our program did not request especially good image quality for these queue scheduled

SDT observations.

253 We also observed Holmberg IX as part of a variability survey of M81 conducted with the 8K Mosaic imager mounted on the MDM 2.4-meter telescope.

The observations were obtained in 5 one-week runs between February 2006 and

February 2007. All the images were obtained in V -band using 15 min exposures.

Due to weather loses and bad seeing, we ended up using only 36 images from 12 diﬀerent nights. The typical seeing was ∼ 1′′.1.

7.3. Light Curve

We used the ISIS diﬀerence image analysis package (Alard 2000; Hartman et al. 2004) to obtain the V -band light curves of all the point sources detected in the

LBT reference image. The detection of all point sources and the transformation of diﬀerence-ﬂux light curves to instrumental magnitudes were done using the

DAOPHOT/ALLSTAR package (Stetson 1987, 1992). After visual inspection of all the light curves of variable point-sources selected by standard criteria (rms and

AoV signiﬁcance; Hartman et al. 2007), we detected ∼ 20 periodic variables in the

ﬁeld of Holmberg IX. These include Cepheids with periods of 10 − 60 days and one long-period variable. The analysis of the Cepheid PL-relation and the distance to

Holmberg IX will be presented in a future paper. The brightest periodic variable is the peculiar, long-period (P = 270.7 days) eclipsing binary we discuss here (hereafter

V1).

254 After discovering the binary in the LBT data, we also ran ISIS and

DAOPHOT/ALLSTAR on the MDM data to extract the light curve of the long-period binary. The variability data from LBT and MDM were complemented with single-epoch archival imaging of the ﬁeld obtained from the SDSS Data Release

6 (Adelman-McCarthy et al. 2008) in the gri bands (UT Nov. 30, 2003), and the

HST/ACS Wide Field Camera (GO proposal 10605, PI E. Skillman) in the F 555W and F 814W ﬁlters (UT Mar. 23, 2006). The high resolution HST/ACS images

(FWHM ∼ 0′′.1, corresponding to ∼ 2 pc at the distance of Holmberg IX) show that the binary is spatially coincident with a stellar association in the dwarf galaxy.

Figure 7.1 shows the phased V -band light curve and B − V color curve of the eclipsing binary system. We include all the LBT, MDM, SDSS and HST/ACS

V -band photometry. The LBT and MDM photometry have been calibrated using

SDSS photometry of several relatively bright (r ∼< 21.0 mag) and unsaturated stars in the ﬁeld, transforming the gr magnitudes to standard BV magnitudes with the transformations presented in Ivezi´cet al. (2007). The rms deviations of the absolute calibration are 0.02 − 0.03 mag for LBT-BV , and ∼ 0.05 mag for MDM-V . The

SDSS gri photometry of the binary was extracted in the same way as for LBT and

MDM, using the DAOPHOT/ALLSTAR package to obtain instrumental magnitudes calibrated using absolute photometry of the bright stars in the ﬁeld. Our g and r magnitudes of the binary from the SDSS data are 0.2 and 0.5 mag brighter, respectively, than the magnitudes reported in the SDSS-DR6 catalog, while the

255 i-band magnitude agrees at the 1% level. We think this is due to problems in the

SDSS photometry for faint sources in a crowded ﬁeld (e.g., Smolˇci´cet al. 2007). The details of the HST photometry can be found in Weisz et al. (2008, in preparation).

In Figure 7.2 we show the position of the binary in the color magnitude diagrams

(CMD), obtained from calibrated LBT and HST/ACS photometry. The CMDs show the well-populated blue and red supergiant sequences in Holmberg IX. The binary is among the most luminous stars in this dwarf galaxy, with MV ∼ −7.1 mag, and it has clearly evolved from the main-sequence. With such a high intrinsic luminosity, the binary is bound to be massive. After correcting for Galactic foreground extinction, the B − V and V − I colors are consistent with an eﬀective temperature of Teﬀ = 4800 ± 150 K (Houdashelt et al. 2000). Both components seem to be

G-type yellow supergiants given the equal depths of the eclipses and the lack of color variations (see Figure 7.1).

We used the eclipsing binary model-fitting program NIGHTFALL1 to model the V -band light curve. As shown in Figure 7.1, we obtain a good fit to the light curve with an overcontact configuration where both stars are overflowing their Roche lobes. We fixed the effective temperature of the primary at T1 = 4800 K obtained from the colors. We assumed equal masses for the stars, a linear limb-darkening law, circular orbits, and synchronous rotation. We fit for four parameters: the Roche lobe filling factors, the inclination, and the temperature of the secondary. The time

1 http://www.hs.uni-hamburg.de/DE/Ins/Per/Wichmann/Nightfall.html

256 of the primary eclipse and the period were determined externally and were fixed for these fits. The main parameters of the binary are listed in Table 7.1. The light curve shows a hint of the O’Connell (1951) effect, in which the maxima (out-of-eclipse regions) show a difference in brightness (e.g., Pilecki et al. 2007).

We searched the literature for other examples of evolved, massive eclipsing binaries in the yellow supergiant phase and found none2. We also searched the available catalogs of eclipsing binaries in the LMC and SMC. The MACHO catalog of eclipsing binaries in the LMC (Derekas et al. 2007) contains 25 contact systems with red colors (i.e., evolved), (V − R) > 0.5 mag, and periods > 200 days. However, these systems have absolute magnitudes MV ∼> −4 (V ∼> 14.5 mag), that are

∼ 3 mag fainter than the yellow supergiant eclipsing binary in Holmberg IX3. The

All Sky Automated Survey (ASAS; Pojmanski 2002) contains complete Southern sky coverage for V < 15 mag. To our surprise, we found in the ASAS catalog a luminous (MV ∼ −7.5, Vmax ∼ 11.5 mag), 181 day period contact eclipsing binary in the SMC. The star, SMC R47 (α = 01h29m17s.26, δ = −72◦43′20.2′′), had been spectroscopically classiﬁed as an FO supergiant (Teﬀ ≃ 7500 K) with emission lines

2Note, however, that a possible Galactic counterpart is the contact binary BM Cas (Fernie &

Evans 1997), composed by an A7 Iab supergiant (MV ≃ −6.3) and a late-type giant. 3Mennickent et al. (2006) obtained spectroscopy of 17 “peculiar” periodic variables in the

SMC from the OGLE database, and found a 184 day period eclipsing binary composed by two yellow supergiants (F5Ie + G5-K0I). However, this system is ∼ 2 mag fainter than the binary in

Holmberg IX.

257 by Humphreys (1983). The ASAS V -band light curve of SMC R47, obtained between

December 2002 and June 2006, and the ﬁt obtained with NIGHTFALL are shown in

Figure 7.3. The best-fit eclipsing binary model requires a contact configuration, with non-zero eccentricity to account for the difference in timing between the eclipses.

Even though we selected a clean part of the full ASAS light curve, there seems to be intrinsic variability from the binary components. The main parameters of this eclipsing binary are in Table 7.1. While substantially hotter than the Holmberg IX binary, it does not lie on the SMC blue supergiant sequence (Grieve & Madore 1986).

7.4. Discussion and Conclusions

An eclipsing binary is the best explanation for the light curve of the brightest variable we have discovered in our LBT variability survey of the dwarf irregular companion of M81 Holmberg IX. The other possible explanation for the periodic variability of V1 is a long-period (P = 135 days) Cepheid. Such long-period

Cepheids (P > 100 days) have been observed in dwarf galaxies like the LMC and

SMC (e.g., Freedman et al. 1985), NGC 55 ( e.g., Pietrzy´nski et al. 2006), NGC 300

(e.g., Pietrzy´nski et al. 2002), NGC 6822 (e.g., Pietrzy´nski et al. 2004), IC 1613

(e.g., Antonello et al. 1999), and I Zw 18 (Aloisi et al. 2007). The magnitude of

V1 is consitent with the magnitude of a Cepheid with P = 135 days (MV ≃ −7.0), extrapolating the period-luminosity relationship of Fouque et al. (2007). However,

258 while a few of these long period Cepheids have quasi-sinusoidal light curves that are nearly symmetric under a time reversal, they all have larger amplitudes in bluer bands (B amplitudes 1.3 − 1.6 times V amplitudes) due to the changes in the eﬀective temperature as the star pulsates (e.g., Freedman et al. 1985; Madore

& Freedman 1991). Spectroscopy of V1, while challenging, would eliminate any remaining ambiguities in classifying this system.

We can safely rule out the possibility that the eclipsing binary is in our Galaxy.

Using the period, estimated eﬀective temperature, magnitude at maximum, and assuming that the stars are of similar size in a contact conﬁguration, we can use

Kepler’s law and their total surface brightness to estimate a distance to the system

1/3 of D ≃ (Mtotal/M⊙) Mpc, where Mtotal = M1 + M2 (e.g., Gaposchkin 1962).

Conversely, we can estimate the total mass of the binary system by assuming the

3 distance, Mtotal ≃ 45 (D/3.6Mpc) M⊙. This should be taken as a rough estimate because of the overly simple model − to accurately constrain the total mass of the system, and its components, we need radial velocity measurements. Another piece of evidence that puts the binary system in Holmberg IX is its spatial coincidence with a stellar overdensity in the dwarf, observed in the HST/ACS images.

259 We expected that such systems were rare4, but were surprised to ﬁnd none in the literature. However, we found a similar eclipsing binary system in the SMC

(SMC R47) searching through the ASAS catalog. From the absolute magnitudes of both binaries and their colors, we estimate that at least one of the stars in each binary is ∼ 15 − 20 M⊙ (main-sequence age ∼ 10 − 15 Myr) using the evolutionary tracks for single stars of Lejeune & Schaerer (2001) (see Figure 7.4).

The stellar evolutionary path of stars of a given mass in binary systems can differ significantly from their evolution in isolation (e.g., Paczyński 1971). In particular, binary interactions through mass loss, mass accretion, or common-envelope evolution, play a very important role in the pre-supernova evolution (e.g., Podsiadlowski et al. 1992). Most of the massive stars with masses 30 M⊙ ∼> M ∼> 8 M⊙ are expected to explode as supernova when they are in the red supergiant stage, with a small contribution from blue supergiants (e.g., SN 1987A; West et al. 1987). Surprisingly,

Li et al. (2005) identiﬁed the progenitor of the Type IIP supernova 2004et in pre-explosion archival images and determined that it was a yellow supergiant with a main-sequence mass of ∼ 15 M⊙. Also, the position in the CMD of the likely progenitor of the Type IIP supernova 2006ov (see Fig. 10 in Li et al. 2007) is

4While the relative numbers of eclipsing binaries is a much more complicated problem, we note that the relative abundances of red, blue and yellow supergiants is 4:13:1 for the Geneva evolutionary track (Lejeune & Schaerer 2001) of a single, non-rotating star with M = 15 M⊙ and Z = 0.004.

260 remarkably similar to the position of the eclipsing binary in Holmberg IX (see

Figure 7.4).

We propose that the binary we discovered in Holmberg IX and the binary found in the SMC5 are the kind of progenitor objects of supernovae like SN 2004et and

SN 2006ov that appeared to be the explosions of yellow supergiants. A close binary provides a natural means of slowing the transition from blue to red, allowing the star to evolve and then explode as a yellow supergiant. As the more massive star evolves and expands, the Roche lobe limits the size of the star forcing it to have a surface temperature set by the uncoupled core luminosity and the size of the Roche lobe. It can expand further and have a cooler envelope only by becoming a common envelope system, which should only occur as the secondary evolves to ﬁll its Roche lobe. This delayed temperature evolution allows the core to reach SN II conditions without a red envelope.

5A possible earlier stage of these yellow supergiant binaries might be represented by the B- supergiant pair HD 1383 (Boyajian et al. 2006).

261 Fig. 7.1.— Phased V -band light curve of the 270.7 day period eclipsing binary in Holmberg IX (top panel) and its B − V color evolution (lower panel). The different symbols correspond to photometry from different telescopes: LBT (filled circles), MDM (open triangles), HST/ACS (filled square) and SDSS (open square). The solid line shows an overcontact eclipsing binary model that best fits the light V light curve.

262 Fig. 7.2.— Color Magnitude Diagrams of stars in the ﬁeld of Holmberg IX obtained from the LBT-BV reference images (left panel) and HST/ACS-VI single-epoch observations (right panel). The ACS CMD shows well-deﬁned stellar sequences for the main sequence (MS) and the evolved blue (BSG) and red supergiants (RSG). The eclipsing binary (pentagon) lies between the blue and red supergiant sequences.

263 Fig. 7.3.— Phased V -band light curve of the long-period, evolved eclipsing binary SMC R47 obtained from the ASAS catalog (Pojmanski 2002). The star was classified spectroscopically as an F0 supergiant by Humphreys (1983). The solid line shows the contact eclipsing binary model that best fits the light curve. The lower panel shows the residuals of the fit (observed-calculated).

264 Fig. 7.4.— CMD of Holmberg IX from HST/ACS V and I photometry. The connected filled symbols show the position of the evolved eclipsing binaries in Holmberg IX (filled pentagons) and the SMC (filled triangles) at maximum and minimum. The lines show evolutionary tracks with extended mass-loss from the Geneva group (Lejeune & Schaerer 2001) for single stars with masses between 12– 25 M⊙, assuming two different metallicities: 1/3 solar (dashed) and 1/5 solar (solid). We use a distance modulus of µ = 27.80 mag to Holmberg IX, and Galactic color- excess E(B − V ) = 0.08 mag, to put the evolutionary tracks in the diagram.

265 Parameter Holmberg IX V1 SMC R47

Period, P 270.7 ± 2.3 days 181.58 ± 0.16 days

Time of primary eclipse, Tprim 2454186.0 ± 0.6 2452073.1 ± 0.2 Inclination, i 55.7 ◦ ± 0.6 ◦ 82.2 ◦ ± 0.2 ◦

Primary temperature, T1 4800 ± 150 K 7500 ± 100 K

Temperature ratio, T2/T1 1.05 ± 0.05 1.17 ± 0.02 Eccentricity, e 0.00 0.039 ± 0.002a Roche Lobe Filling factorsb 1.23 ± 0.02 1.02 ± 0.02 c Semi-major axis, a 547 R⊙ 418 R⊙

Note. — The mass ratio was fixed at q = 1 for fitting both light curves. aA non-zero eccentricity is required to fit the difference in timing between the primary and secondary eclipses. The best-fit longitude of the periastron is w = 168.6 ◦ ± 1.1 ◦. bRatio of stellar to Roche lobe polar radius for each star. c Separation between the stars assuming a total mass of 30 M⊙ for each system.

Table 7.1. Best-ﬁt Binary Model Parameters.

266 Appendix A

A. Extreme-AGB Star Variability

Section 4.4.4 presents a discussion of the variability of the 18 reddest sources selected as EAGB stars (see Table 4.2 for their photometry). In this Appendix (in

Figs. A.1 & A.2) we present the lightcurves for all 18 sources. See the large open circles in Figures 4.2, 4.4, and 4.7, as well as the left panel of 4.5 for a summary of their colors, SEDs, and RMS variability.

267 Fig. A.1.— Lightcurves for sources S1−S9 of the 18 reddest sources (sorted by color) with M4.5 < −10 and [3.6] − [4.5] > 1.5 in the M33 MIR color-magnitude diagram (open circles in Fig. 4.2; remaining source lightcurves are shown in Fig. A.2). For each source, the top and bottom panels show the absolute 4.5µm magnitude and the [3.6]−[4.5] color variation, respectively, as a function of time. More than a magnitude variation on a timescale of 100 − 1000 days is common.

268 Fig. A.2.— Same as Figure A.1, but for sources S10-S18 of the 18 reddest sources with M4.5 < −10 in the M33 MIR color-magnitude diagram (see Figs. 4.2 & 4.3).

269 Appendix B

LBV Candidate Variability

Like Appendix A, here we present the lightcurves for the 16 LBV candidates from Massey et al. (2007) that have been matched to the MIR point source catalog, as described in §4.4. Table 4.3 lists photometry for these sources. See Figures 4.2,4.4,

4.5, and 4.7 for a summary of their colors, SEDs, and RMS variability properties.

270 Fig. B.1.— Lightcurves at 4.5 µm for 9 of the 16 LBV candidates of Massey et al. (2007). Although there are several exceptions (notably J013335.14+303600.4 above and J013429.64+303732.1 in Fig. B.2), the LBV candidates do not vary signiﬁcantly in absolute magnitude or color (see also Fig. 4.7). As shown in Fig. 4.2 (open triangles), the LBV candidates with high luminosities at 4.5µm are characteristically more red than those with larger M4.5.

271 Fig. B.2.— Same as Figure B.1, but for the remaining 7 LBV candidates of Massey et al. (2007), matched to our 4.5µm catalog.

272 BIBLIOGRAPHY

Abazajian, K., et al. 2004, AJ, 128, 502

Adelman-McCarthy, J. K., et al. 2006, ApJS, 162, 38 delman-McCarthy, J. K., et al. 2008, ApJS, 175, 297

Alard, C. 2000, A&AS, 144, 363

Aldering, G., et al. 2006, ApJ, 650, 510

Allen, D. A., Norris, R. P., Meadows, V. S., & Roche, P. F. 1991, MNRAS, 248, 528

Aloisi, A., et al. 2007, ApJ, 667, L151

Anderson, J. P., & James, P. A. 2008, MNRAS, 390, 1527

Ando, S., Beacom, J. F., Y¨uksel, H. 2005, Phys. Rev. Lett., 95, 171101

Antonello, E., Mantegazza, L., Fugazza, D., Bossi, M., & Covino, S. 1999, A&A, 349, 55

Arbour, R., & Boles, T. 2008, Central Bureau Electronic Telegrams, 1234, 1

Aretxaga, I., Benetti, S., Terlevich, R. J., Fabian, A. C., Cappellaro, E., Turatto, M., & della Valle, M. 1999, MNRAS, 309, 343

Arnett, W. D. 1982, ApJ, 253, 785

Arnett, W. D. 1987, ApJ, 319, 136

Astier, P., et al. 2006, A&A, 447, 31

Badenes, C., Hughes, J. P., Bravo, E., & Langer, N. 2007, ApJ, 662, 472

Balick, B., & Frank, A. 2002, ARA&A, 40, 439

Barentine, J., et al. 2005, Central Bureau Electronic Telegrams, 247, 1

273 Bauschlicher, C. W., Peeters, E., & Allamandola, L. J. 2009, ApJ, 697, 311

Beacom, J. F., & Y¨uksel, H. 2006, Phys. Rev. Lett., 97, 071102

Benetti, S., et al. 2006a, Central Bureau Electronic Telegrams, 674, 2

Benetti, S., Cappellaro, E., Turatto, M., Taubenberger, S., Harutyunyan, A., & Valenti, S. 2006b, ApJ, 653, L129

Berger, E., & Soderberg, A. 2008, The Astronomer’s Telegram, 1544, 1

Berger, E., et al. 2009, ApJ, 699, 1850

Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393

Blanton, M. R., et al. 2003, AJ, 125, 2348

Blondin, S., Modjaz, M., Kirshner, R., Challis, P.,& Calkins, M. 2007a, Central Bureau Electronic Telegrams, 808, 1

Blondin, S. & Tonry, J. L. 2007b, ApJ, 666, 1024

Blum, R. D., et al. 2006, AJ, 132, 2034

Boersma, C., Bouwman, J., Lahuis, F., van Kerckhoven, C., Tielens, A. G. G. M., Waters, L. B. F. M., & Henning, T. 2008, A&A, 484, 241

Bolatto, A. D., et al. 2007, ApJ, 655, 212

Bonanos, A. Z., et al. 2004, ApJ, 611, L33

Bonanos, A. Z., et al. 2006, ApJ, 652, 313

Bonanos, A. Z., et al. 2009, arXiv:0905.1328

Bond, H. E., et al. 2003, Nature, 422, 405

Bond, H. E., Walter, F. M., & Velasquez, J. 2008, IAU Circ., 8946, 2

Bond, H. E., Bedin, L. R., Bonanos, A. Z., Humphreys, R. M., Monard, L. A. G. B., Prieto, J. L., & Walter, F. M. 2009, ApJ, 695, L154

Botticella, M. T., et al. 2009, arXiv:0903.1286

Boyajian, T. S., Gies, D. R., Helsel, M. E., Kaye, A. B., McSwain, M. V., Riddle, R. L., & Wingert, D. W. 2006, ApJ, 646, 1209

Boyce, P. J., et al. 2001, ApJ, 560, L127

274 Branch, D., Falk, S. W., Uomoto, A. K., Wills, B. J., McCall, M. L., & Rybski, P. 1981, ApJ, 244, 780

Branch, D., Romanishin, W., & Baron, E. 1996, ApJ, 465, 73

Branch, D., Jeﬀery, D. J., Blaylock, M., & Hatano, K. 2000, PASP, 112, 217

Branch, D., et al. 2003, AJ, 126, 1489

Brinchmann, J., Charlot, S., White, S. D. M., Tremonti, C., Kauﬀmann, G., Heckman, T., & Brinkmann, J. 2004, MNRAS, 351, 1151

Buras, R., Rampp, M., Janka, H.-T., & Kifonidis, K. 2003, Physical Review Letters, 90, 241101

Burrows, A., Livne, E., Dessart, L., Ott, C. D., & Murphy, J. 2006, ApJ, 640, 878

Cappellaro, E., Turatto, M., Tsvetkov, D. Y., Bartunov, O. S., Pollas, C., Evans, R., & Hamuy, M. 1997, A&A, 322, 431

Cappellaro, E., Evans, R., & Turatto, M. 1999, A&A, 351, 459

Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245

Cass´e, M., Cordier, B., Paul, J., & Schanne, S. 2004, ApJ, 602, L17

Chandra, P., & Soderberg, A. 2008, The Astronomer’s Telegram, 1382, 1

Charlot, S., & Longhetti, M. 2001, MNRAS, 323, 887

Chevalier, R. A. 1982, ApJ, 258, 790

Chevalier, R. A. & Fransson, C. 1994, ApJ, 420, 268

Chevalier, R. A., & Fransson, C. 2003, Supernovae and Gamma-Ray Bursters, 598, 171

Chugai, N. N. & Danziger, I. J. 1994, MNRAS, 268, 173

Chugai, N. N. 1997a, Ap&SS, 252, 225

Chugai, N. N. 1997b, Astronomy Reports, 41, 672

Chugai, N. N., & Utrobin, V. P. 2000, A&A, 354, 557

Chugai, N. N., Chevalier, R. A., & Lundqvist, P. 2004, MNRAS, 355, 627

Cohen, M., Megeath, S. T., Hammersley, P. L., Mart´ın-Luis, F., & Stauﬀer, J. 2003, AJ, 125, 2645

275 Contardo, G., Leibundgut, B., & Vacca, W. D. 2000, A&A, 359, 876

Crockett, R. M., et al. 2008, ApJ, 672, L99

Crotts, A., Eastman, J., Depoy, D., Prieto, J. L., & Garnavich, P. 2006, Central Bureau Electronic Telegrams, 672, 1

Crowther, P. A. 2007, ARA&A, 45, 177

Delahaye, F., & Pinsonneault, M. H. 2006, ApJ, 649, 529

Dennefeld, M., Benetti, S., Harutyunyan, A., Agnoletto, I., Turatto, M., Navasardyan, H., Troncin, J. P., & Boselli, A. 2007, Central Bureau Electronic Telegrams, 937, 1

Deng, J., et al. 2004, ApJ, 605, L37

Derekas, A., Kiss, L. L., & Bedding, T. R. 2007, ApJ, 663, 249

Davies, B., Oudmaijer, R. D., & Sahu, K. C. 2007, ApJ, 671, 2059 de Marco, O. 2009, PASP, 121, 316

Dennis, T. J., Cunningham, A. J., Frank, A., Balick, B., Blackman, E. G., & Mitran, S. 2008, ApJ, 679, 1327 de Vaucouleurs, G., de Vaucouleurs, A., Corwin, H. G., Jr., Buta, R. J., Paturel, G., & Fouque, P. 1991, Volume 1-3, XII, 2069 pp. 7, Springer-Verlag Berlin Heidelberg New York

Di Carlo, E., et al. 2002, ApJ, 573, 144

Diehl, R., et al. 2006, Nature, 439, 45

Dilday, B., et al. 2008, ApJ, 682, 262

Draine, B. T. & McKee, C. F. 1993, ARA&A, 31, 373

Duley, W. W., & Williams, D. A. 1981, MNRAS, 196, 269

Duszanowicz, G., Boles, T., & Corelli, P. 2007, Central Bureau Electronic Telegrams, 1182, 1

Dwarkadas, V. V. & Chevalier, R. A. 1998, ApJ, 497, 807

Dwek, E. 1983, ApJ, 274, 175

Eldridge, J. J., & Tout, C. A. 2004, MNRAS, 353, 87

276 Eldridge, J. J., Mattila, S., & Smartt, S. J. 2007a, MNRAS, 376, L52

Eldridge, J. J. 2007b, Revista Mexicana de Astronomia y Astroﬁsica Conference Series, 30, 35

Eldridge, J. J., Izzard, R. G., & Tout, C. A. 2008, MNRAS, 384, 1109

Elias-Rosa, N., Benetti, S., Cappellaro, E., Dolci, M., Pastorello, A., Boschin, W., & Pinilla Alonso, N. 2005, IAU Circ., 8498, 1

Ellison, S. 2006, The Fabulous Destiny of Galaxies: Bridging Past and Present, 53

Fazio, G. G., et al. 2004, ApJS, 154, 10

Fernie, J. D., & Evans, N. R. 1997, PASP, 109, 541

Filippenko, A. V. 1997, ARA&A, 35, 309

Fisher, A., Branch, D., Nugent, P., & Baron, E. 1997, ApJ, 481, L89

Fisher, A., Branch, D., Hatano, K., & Baron, E. 1999, MNRAS, 304, 67

Fisher, A. K. 2000, Ph.D. Thesis

Foley, R. J., Smith, N., Ganeshalingam, M., Li, W., Chornock, R., & Filippenko, A. V. 2007, ApJ, 657, L105

Fouqu´e, P., et al. 2007, A&A, 476, 73

Fransson, C., Lundqvist, P., & Chevalier, R. A. 1996, ApJ, 461, 993

Freedman, W. L., Grieve, G. R., & Madore, B. F. 1985, ApJS, 59, 311

Freedman, W. L., et al. 2001, ApJ, 553, 47

Frieman, J. 2005, IAU Circ., 8616, 1

Frieman, J., et al. 2006, Central Bureau Electronic Telegrams, 762, 1

Frieman, J. A., et al. 2008, AJ, 135, 338

Frost, C. A., Lattanzio, J. C., & Wood, P. R. 1998, ApJ, 500, 355

Fruchter, A. S., et al. 2006, Nature, 441, 463

Fukugita, M., Ichikawa, T., Gunn, J. E., Doi, M., Shimasaku, K., & Schneider, D. P. 1996, AJ, 111, 1748

Fryer, C. L., et al. 2007, PASP, 119, 1211

277 Gallagher, J. S., Garnavich, P. M., Berlind, P., Challis, P., Jha, S., & Kirshner, R. P. 2005, ApJ, 634, 210

Gal-Yam, A., Ofek, E. O., & Shemmer, O. 2002, MNRAS, 332, L73

Gal-Yam, A., et al. 2007, ApJ, 656, 372

Gaposchkin, S. 1962, AJ, 67, 358

Garavini, G., et al. 2004, AJ, 128, 387

Garc´ıa-Hern´andez, D. A., Perea-Calder´on, J. V., Bobrowsky, M., & Garc´ıa-Lario, P. 2007, ApJ, 666, L33

Gardan, E., Braine, J., Schuster, K. F., Brouillet, N., & Sievers, A. 2007, A&A, 473, 91

Garnavich, P., Jha, S., Kirshner, R., Calkins, M., & Brown, W. 1999, IAU Circ., 7150, 1

Garnavich, P. M., et al. 2004, ApJ, 613, 1120

Germany, L. M., et al. 2000, ApJ, 533, 320

Giallongo, E., et al. 2008, A&A, 482, 349

Gieren, W., Pietrzy´nski, G., Soszy´nski, I., Bresolin, F., Kudritzki, R.-P., Minniti, D., & Storm, J. 2005, ApJ, 628, 695

Gilman, R. C. 1974, ApJS, 28, 397

Gonz´alez, J. F., et al. 2005, ApJ, 624, 946

Gordon, S. M., Kirshner, R. P., Long, K. S., Blair, W. P., Duric, N., & Smith, R. C. 1998, ApJS, 117, 89

Grieve, G. R., & Madore, B. F. 1986, ApJS, 62, 427

Groenewegen, M. A. T., et al. 2007, MNRAS, 376, 313

Gunn, J. E., et al. 1998, AJ, 116, 3040

Gunn, J. E., et al. 2006, AJ, 131, 2332

Hachisu, I., Kato, M., & Nomoto, K. 1996, ApJ, 470, L97

Hamuy, M., Phillips, M. M., Wells, L. A., & Maza, J. 1993, PASP, 105, 787

278 Hamuy, M., Phillips, M. M., Maza, J., Suntzeﬀ, N. B., Schommer, R. A., & Aviles, R. 1995, AJ, 109, 1

Hamuy, M., Phillips, M. M., Suntzeﬀ, N. B., Schommer, R. A., Maza, J., & Aviles, R. 1996, AJ, 112, 2391

Hamuy, M., Phillips, M. M., Suntzeﬀ, N. B., Schommer, R. A., Maza, J., & Aviles, R. 1996, AJ, 112, 2398

Hamuy, M., Trager, S. C., Pinto, P. A., Phillips, M. M., Schommer, R. A., Ivanov, V., & Suntzeﬀ, N. B. 2000, AJ, 120, 1479

Hamuy, M., et al. 2002, AJ, 124, 417

Hamuy, M., et al. 2003, Nature, 424

Hamuy, M., et al. 2006, PASP, 118, 2

Han, Z. & Podsiadlowski, P. 2006, MNRAS, 368, 1095

Hartman, J. D., Bakos, G., Stanek, K. Z., & Noyes, R. W. 2004, AJ, 128, 1761

Hartman, J. D., Bersier, D., Stanek, K. Z., Beaulieu, J.-P., Kaluzny, J., Marquette, J.-B., Stetson, P. B., & Schwarzenberg-Czerny, A. 2006, MNRAS, 371, 1405

Hartman, J. D., Gaudi, B. S., Holman, M. J., McLeod, B. A., Stanek, K. Z., Barranco, J. A., Pinsonneault, M. H., & Kalirai, J. S. 2008, ApJ, 675, 1254

Hartigan, P., Raymond, J., & Hartmann, L. 1987, ApJ, 316, 323

Harutyunyan, A., Benetti, S., Turatto, M., Pastorello, A., Young, D., Smartt, S. J., Valenti, S., & Andreuzzi, G. 2007, Central Bureau Electronic Telegrams, 948, 1

Hatano, K., Fisher, A., & Branch, D. 1997, MNRAS, 290, 360

Heger, A., Jeannin, L., Langer, N., & Baraﬀe, I. 1997, A&A, 327, 224

Heger, A., Fryer, C. L., Woosley, S. E., Langer, N., & Hartmann, D. H. 2003, ApJ, 591, 288

Hendry, M. A., et al. 2006, MNRAS, 369, 1303

Higdon, J. C., Lingenfelter, R. E., & Rothschild, R. E. 2004, ApJ, 611, L29

Hilditch, R. W., Howarth, I. D., & Harries, T. J. 2005, MNRAS, 357, 304

Hill, J. M., et al. 2006, Proceedings of the SPIE, 6267E, 31H

Hirschi, R., Meynet, G., & Maeder, A. 2005, A&A, 443, 581

279 H¨oﬂich, P., Nomoto, K., Umeda, H., & Wheeler, J. C. 2000, ApJ, 528, 590

Hogg, D. W., Finkbeiner, D. P., Schlegel, D. J., & Gunn, J. E. 2001, AJ, 122, 2129

Holtzman, J. A., et al. 2008, AJ, 136, 2306

Hoopes, C. G., & Walterbos, R. A. M. 2000, ApJ, 541, 597

Hoopes, C. G., Walterbos, R. A. M., & Bothun, G. D. 2001, ApJ, 559, 878

Hopkins, A. M., & Beacom, J. F. 2006, ApJ, 651, 142

Hornoch, K. 2002, IAU Circ., 7923, 5

Houck, J. R., et al. 2004, ApJS, 154, 18

Houdashelt, M. L., Bell, R. A., & Sweigart, A. V. 2000, AJ, 119, 1448

Hughes, J. P., et al. 1995, ApJ, 444, L81

Humphreys, R. M. 1983, ApJ, 265, 176

Humphreys, R. M., Davidson, K., & Smith, N. 2002, AJ, 124, 1026

Humphreys, R. M. 2005, The Fate of the Most Massive Stars, 332, 90

Humphreys, R. M., et al. 2006, AJ, 131, 2105

Iben, Jr., I. & Renzini, A. 1983, ARA&A, 21, 271

Immler, S., Petre, R., & Brown, P. 2005, IAU Circ., 8633, 2

Immler, S., et al. 2006, ApJ, 648, L119

Ivezi´c, Z., & Elitzur, M. 1996, MNRAS, 279, 1019

Ivezic, Z., & Elitzur, M. 1997, MNRAS, 287, 799

Ivezi´c, Z.,ˇ et al. 2004, Astronomische Nachrichten, 325, 583

Ivezi´c, Z.,ˇ et al. 2007, AJ, 134, 973

Iwamoto, K., et al. 2000, ApJ, 534, 660

James, P. A., & Anderson, J. P. 2006, A&A, 453, 57

Jeﬀery, D. J., Leibundgut, B., Kirshner, R. P., Benetti, S., Branch, D., & Sonneborn, G. 1992, ApJ, 397, 304

Jeﬀery, D. J., Branch, D., & Baron, E. 2006, arXiv:astro-ph/0609804

280 Jha, S., et al. 1999, ApJS, 125, 73

Joblin, C., Szczerba, R., Bern´e, O., & Szyszka, C. 2008, A&A, 490, 189

Karachentsev, I. D., Karachentseva, V. E., Huchtmeier, W. K., & Makarov, D. I. 2004, AJ, 127, 2031

Kastner, J. H., Buchanan, C. L., Sargent, B., & Forrest, W. J. 2006, ApJ, 638, L29

Kastner, J. H., Thorndike, S. L., Romanczyk, P. A., Buchanan, C. L., Hrivnak, B. J., Sahai, R., & Egan, M. 2008, AJ, 136, 1221

Keller, L. D., et al. 2008, ApJ, 684, 411

Kelly, P. L., Kirshner, R. P., & Pahre, M. 2008, ApJ, 687, 1201

Kennicutt, R. C., et al. 2003, PASP, 115, 928

Kewley, L. J., & Dopita, M. A. 2002, ApJS, 142, 35

Kewley, L. J., Jansen, R. A., & Geller, M. J. 2005, PASP, 117, 227

Kiminki, D. C., et al. 2007, ApJ, 664, 1102

Kirshner, R. P., et al. 1993, ApJ, 415, 589

Kistler, M. D., Yuksel, H., Ando, S., Beacom, J. F., & Suzuki, Y. 2008, arXiv:0810.1959

Kitaura, F. S., Janka, H.-T., & Hillebrandt, W. 2006, A&A, 450, 345

Klein, R. I., McKee, C. F., & Colella, P. 1994, ApJ, 420, 213

Kn¨odlseder, J., et al. 2005, A&A, 441, 513

Kobayashi, C., Tsujimoto, T., Nomoto, K., Hachisu, I., & Kato, M. 1998, ApJ, 503, L155

Kobayashi, C., Springel, V., & White, S. D. M. 2007, MNRAS, 376, 1465

Kochanek, C. S., et al. 2001, ApJ, 560, 566

Kochanek, C. S., Beacom, J. F., Kistler, M. D., Prieto, J. L., Stanek, K. Z., Thompson, T. A., Y¨uksel, H. 2008, ApJ, 684, 1336

Kotak, R., Meikle, W. P. S., Adamson, A., & Leggett, S. K. 2004, MNRAS, 354, L13

Kotak, R., Meikle, P., van Dyk, S. D., H¨oﬂich, P. A., & Mattila, S. 2005, ApJ, 628, L123

281 Kotak, R., et al. 2006, ApJ, 651, L117

Kotak, R., et al. 2009, arXiv:0904.3737

Kraft, R. P., Burrows, D. N., & Nousek, J. A. 1991, ApJ, 374, 344

Krisciunas, K., Phillips, M. M., & Suntzeﬀ, N. B. 2004, ApJ, 602, L81

Kudritzki, R. P., & Puls, J. 2000, ARA&A, 38, 613

Kulkarni, S. R., et al. 2007, Nature, 447, 458

Kwok, S. 1993, ARA&A, 31, 63

Kwok, S., Volk, K., & Hrivnak, B. J. 1999, A&A, 350, L35

Kwok, S., Volk, K., & Bernath, P. 2001, ApJ, 554, L87

Lee, H., Skillman, E. D., Cannon, J. M., Jackson, D. C., Gehrz, R. D., Polomski, E. F., & Woodward, C. E. 2006, ApJ, 647, 970

Leibundgut, B., Kirshner, R. P., Filippenko, A. V., Shields, J. C., Foltz, C. B., Phillips, M. M., & Sonneborn, G. 1991, ApJ, 371, L23

Leibundgut, B., et al. 1993, AJ, 105, 301

Lejeune, T., & Schaerer, D. 2001, A&A, 366, 538

Li, W. 1999, IAU Circ., 7149, 1

Li, W., Filippenko, A. V., van Dyk, S. D., Hu, J., Qiu, Y., Modjaz, M., & Leonard, D. C. 2002, PASP, 114, 403

Li, W., Filippenko, A. V., Chornock, R., & Jha, S. 2003, PASP, 115, 844

Li, W., van Dyk, S. D., Filippenko, A. V., & Cuillandre, J.-C. 2005, PASP, 117, 121

Li, W., Wang, X., van Dyk, S. D., Cuillandre, J.-C., Foley, R. J., & Filippenko, A. V. 2007, ApJ, 661, 1013

Liebend¨orfer, M., Mezzacappa, A., Thielemann, F.-K., Messer, O. E., Hix, W. R., & Bruenn, S. W. 2001, Phys. Rev. D, 63, 103004

Livio, M. & Riess, A. G. 2003, ApJ, 594, L93

Lupton, R. 2005 (http://www.sdss.org/dr4/algorithms/sdssUBVRITransform.htm)

Lupton, R. H., Gunn, J. E., & Szalay, A. S. 1999, AJ, 118, 1406

282 Maeder, A., & Meynet, G. 1994, A&A, 287, 803

Maeder, A., & Meynet, G. 2004, A&A, 422, 225

MacFadyen, A. I., & Woosley, S. E. 1999, ApJ, 524, 262

Madore, B. F., & Freedman, W. L. 1991, PASP, 103, 933

Makarova, L. N., et al. 2002, A&A, 396, 473

Mannucci, F., Della Valle, M., Panagia, N., Cappellaro, E., Cresci, G., Maiolino, R., Petrosian, A., & Turatto, M. 2005, A&A, 433, 807

Marigo, P. 2007, A&A, 467, 1139

Marigo, P. 2008, Memorie della Societa Astronomica Italiana, 79, 403

Martini, P., Persson, S. E., Murphy, D. C., Birk, C., Shectman, S. A., Gunnels, S. M., & Koch, E. 2004, Proceedings of the SPIE, 5492, 1653

Massey, P. 2003, ARA&A, 41, 15

Massey, P., Olsen, K. A. G., Hodge, P. W., Strong, S. B., Jacoby, G. H., Schlingman, W., & Smith, R. C. 2006, AJ, 131, 2478

Massey, P., McNeill, R. T., Olsen, K. A. G., Hodge, P. W., Blaha, C., Jacoby, G. H., Smith, R. C., & Strong, S. B. 2007, AJ, 134, 2474

Matheson, T., et al. 2005, AJ, 129, 2352

Matsuura, M., et al. 2009, MNRAS, 396, 918

Maund, J. R., Smartt, S. J., Kudritzki, R. P., Podsiadlowski, P., & Gilmore, G. F. 2004, Nature, 427, 129

Maund, J. R., et al. 2006, MNRAS, 369, 390

Mazzali, P. A., Danziger, I. J., & Turatto, M. 1995, A&A, 297, 509

McQuinn, K. B. W., et al. 2007, ApJ, 664, 850

Mennickent, R. E., et al. 2006, RMxAC, 26, 74

Menzies, J. W., et al. 1987, MNRAS, 227, 39P

Meynet G., & Maeder, A. 2003, A&A, 404, 975

Meynet, G., Mowlavi, N., & Maeder, A. 2006, arXiv:astro-ph/0611261

283 Miknaitis, G., et al. 2007, ApJ, 666, 674

Mikuz, H. 2007, Central Bureau Electronic Telegrams, 933, 1

Millard, J., et al. 1999, ApJ, 527, 746

Miller, B. W. 1995, ApJ, 446, L75

Mink, D. J. 1999, Astronomical Data Analysis Software and Systems VIII, 172, 498

Miyaji, S., Nomoto, K., Yokoi, K., & Sugimoto, D. 1980, PASJ, 32, 303

Modjaz, M., et al. 2006, ApJ, 645, L21

Modjaz, M., et al. 2008, AJ, 135, 1136

Monard, L. A. G. 2006, IAU Circ., 8690, 2

Monard, L. A. G. 2008, IAU Circ., 8946, 1

Monet, D. G., et al. 2003, AJ, 125, 984

Morgan, C. W., Byard, P. L., DePoy, D. L., Derwent, M., Kochanek, C. S., Marshall, J. L., O’Brien, T. P., & Pogge, R. W. 2005, AJ, 129, 2504

Mould, J., Barmby, P., Gordon, K., Willner, S. P., Ashby, M. L. N., Gehrz, R. D., Humphreys, R., & Woodward, C. E. 2008, ApJ, 687, 230

Neill, J. D., et al. 2006, AJ, 132, 1126

Nomoto, K. 1984a, ApJ, 277, 791

Nomoto, K., Thielemann, F.-K., & Yokoi, K. 1984b, ApJ, 286, 644

Nomoto, K. 1987, ApJ, 322, 206

Nomoto, K., Suzuki, T., Deng, J., Uenishi, T., & Hachisu, I. 2005, 1604-2004: Supernovae as Cosmological Lighthouses, 342, 105

Nugent, P., Kim, A., & Perlmutter, S. 2002, PASP, 114, 803

O’Connell, D. J. K. 1951, Publications of the Riverview College Observatory, 2, 85

Ofek, E. O., et al. 2008, ApJ, 674, 447

Oke, J. B., & Gunn, J. E. 1983, ApJ, 266, 713 sterbrock, D. E. 1989, Research supported by the University of California, John Simon Guggenheim Memorial Foundation, University of Minnesota, et al. Mill Valley, CA, University Science Books, 1989, 422 p.

284 Paczy´nski, B. 1971, ARA&A, 9, 183

Palacios, A., Meynet, G., Vuissoz, C., Kn¨odlseder, J., Schaerer, D., Cervi˜no, M., & Mowlavi, N. 2005, A&A, 429, 613

Pastorello, A., et al. 2002, MNRAS, 333, 27

Pastorello, A., et al. 2004, MNRAS, 347, 74

Pastorello, A., et al. 2006, MNRAS, 370, 1752

Pastorello, A., et al. 2007, Nature, 447, 829

Pastorello, A., et al. 2009, MNRAS, 394, 2266

Patat, F., et al. 2001, ApJ, 555, 900

Pathak, A., & Rastogi, S. 2008, A&A, 485, 735

Peeples, M. S., et al. 2007, ApJ, 654, L61

Peeters, E., Hony, S., van Kerckhoven, C., Tielens, A. G. G. M., Allamandola, L. J., Hudgins, D. M., & Bauschlicher, C. W. 2002, A&A, 390, 1089

Perlmutter, S., et al. 1999, ApJ, 517, 565

Persson, S. E., et al. 2002, AJ, 124, 619

Phillips, M. M., et al. 1992, AJ, 103, 1632

Phillips, M. M. 1993, ApJ, 413, L105

Phillips , M. M., et al. 2007, PASP, 119, 360

Pier, J. R., et al. 2003, AJ, 125, 1559

Pietrzy´nski, G., et al. 2006, ApJ, 132, 2556

Pietrzy´nski, G., et al. 2004, ApJ, 128, 2815

Pietrzy´nski, G., et al. 2002, ApJ, 123, 789

Pilecki, B., Fabrycky, D., & Poleski, R. 2007, MNRAS, 378, 757

Prantzos, N., & Diehl, R. 1996, Phys. Rep., 267, 1

Prantzos, N., & Boissier, S. 2003, A&A, 406, 259

Prantzos, N. 2004, A&A, 420, 1033

285 Prieto, J. L., et al. 2005, IAU Circ., 8633

Prieto, J. L., et al. 2007, arXiv:0706.4088

Prieto, J. L., et al. 2008a, ApJ, 673, L59

Prieto, J. L., Stanek, K. Z., & Beacom, J. F. 2008b, ApJ, 673, 999

Prieto, J. L., et al. 2008c, ApJ, 681, L9

Prieto, J. L. 2008d, The Astronomer’s Telegram, 1550, 1

Prieto, J. L., et al. 2008e, The Astronomer’s Telegram, 1596, 1

Prieto, J. L., Sellgren, K., Thompson, T. A., & Kochanek, C. S. 2009, arXiv:0907.0230

Petrosian, A., et al. 2005, AJ, 129, 1369

Podsiadlowski, P., Joss, P. C., & Hsu, J. J. L 1992, ApJ, 391, 246

Podsiaklowski, P., Hsu, J. J. L., Joss, P. C., & Ross, R. R. 1993, Nature, 364, 509

Podsiadlowski, P., Mazzali, P. A., Lesaﬀre, P., Wolf, C., & Forster, F. 2006, arXiv:astro-ph/0608324

Poelarends, A. J. T., Herwig, F., Langer, N., & Heger, A. 2008, ApJ, 675, 614

Pojmanski, G. 2002, Acta Astronomica, 52, 397

Puckett, T., & Gauthier, S. 2002, IAU Circ., 7863, 1

Quimby, R., Höflich, P., Kannappan, S. J., Rykoff, E., Rujopakarn, W., Akerlof, C. W., Gerardy, C. L., & Wheeler, J. C. 2006, ApJ, 636, 400

Quimby, R., Rykoﬀ, E., & Yuan, F. 2007, Central Bureau Electronic Telegrams, 927, 1

Ragazzoni, R., et al. 2006, Proceedings of the SPIE, 6267E, 33R

Rampp, M., & Janka, H.-T. 2000, ApJ, 539, L33

Reach, W. T., et al. 2005, PASP, 117, 978

Ribas, I., et al. 2005, ApJ, 635, L37

Rich, D., Yamaoka, H., & Itagaki, K. 2005, IAU Circ., 8497, 3

Richardson, D., Branch, D., Casebeer, D., Millard, J., Thomas, R. C., & Baron, E. 2002, AJ, 123, 745

286 Rieke, G. H., et al. 2004, ApJS, 154, 25

Riera, A., Garcia-Lario, P., Manchado, A., Pottasch, S. R., & Raga, A. C. 1995, A&A, 302, 137

Riera, A., Garc´ıa-Lario, P., Manchado, A., Bobrowsky, M., & Estalella, R. 2003, A&A, 401, 1039

Riess, A. G., et al. 1998, AJ, 116, 1009

Rigon, L., et al. 2003, MNRAS, 340, 191

Ritossa, C., Garcia-Berro, E., & Iben, I. J. 1996, ApJ, 460, 489

Roche, P. F., Aitken, D. K., & Smith, C. H. 1993, MNRAS, 261, 522

R¨opke, F. K., Gieseler, M., Reinecke, M., Travaglio, C., & Hillebrandt, W. 2006, A&A, 453, 203

Sahai, R., Te Lintel Hekkert, P., Morris, M., Zijlstra, A., & Likkel, L. 1999, ApJ, 514, L115

Sahai, R. 2009, Astronomy, 2010, 256

Sahu, D. K., Anupama, G. C., Srividya, S., & Muneer, S. 2006, MNRAS, 372, 1315

Sako, M., et al. 2005, arXiv:astro-ph/0504455

Salzer, J. J., et al. 2005, ApJ, 624, 661

S´anchez Contreras, C., Sahai, R., Gil de Paz, A., & Goodrich, R. 2008, ApJS, 179, 166

Scannapieco, E., & Bildsten, L. 2005, ApJ, 629, 85L

Schaefer, B. E. 1996, ApJ, 464, 404

Schaller, G., Schaerer, D., Meynet, G., & Maeder, A. 1992, A&AS, 96, 269

Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525

Schlegel, E. M. 1990, MNRAS, 244, 269

Schmidt, B. P., Kirshner, R. P., Leibundgut, B., Wells, L. A., Porter, A. C., Ruiz-Lapuente, P., Challis, P., & Filippenko, A. V. 1994, ApJ, 434, L19

Schwartz, M., Li, W., Filippenko, A. V., & Chornock, R. 2003, IAU Circ., 8051, 1

287 Semenov, D., Henning, T., Helling, C., Ilgner, M., & Sedlmayr, E. 2003, A&A, 410, 611

Shields, J. C., Pogge, R. W., & De Robertis, M. M. 1999, Structure and Kinematics of Quasar Broad Line Regions, 175, 353

Siess, L. 2006, A&A, 448, 717

Siess, L. 2007, A&A, 476, 893

Sloan, G. C., Kraemer, K. E., Price, S. D., & Shipman, R. F. 2003, ApJS, 147, 379

Sloan, G. C., et al. 2007, ApJ, 664, 1144

Smartt, S. J., Maund, J. R., Hendry, M. A., Tout, C. A., Gilmore, G. F., Mattila, S., & Benn, C. R. 2004, Science, 303, 499

Smartt, S. J., Eldridge, J. J., Crockett, R. M., & Maund, J. R. 2009, MNRAS, 395, 1409

Smith, J. A., et al. 2002, AJ, 123, 2121

Smith, N., Humphreys, R. M., & Gehrz, R. D. 2001, PASP, 113, 692

Smith, N., et al. 2004, ApJ, 615, 475

Smith, N. 2007, AJ, 133, 1034

Smith, N. 2008, IAU Symposium, 250, 193

Smith, N., Hinkle, K. H., & Ryde, N. 2009a, AJ, 137, 3558

Smith, N., et al. 2009b, ApJ, 697, L49

Smolˇci´c, V., et al. 2007, AJ, 134, 1901

Soderberg, A. M. & Frail, D. A. 2005, The Astronomer’s Telegram, 663, 1

Soderberg, A. M., Nakar, E., Berger, E., & Kulkarni, S. R. 2006, ApJ, 638, 930

Soker, N. 2008, New Astronomy, 13, 491

Spergel, D. N., et al. 2007, ApJS, 170, 377

Srinivasan, S., et al. 2009, AJ, 137, 4810

Stanek, K. Z., et al. 2003, ApJ, 591, 17L

Stanek, K. Z., et al. 2006, Acta Astronomica, 56, 333

288 Stanghellini, L., & Renzini, A. 2000, ApJ, 542, 308

Stanishev, V., Pastorello, A., & Pursimo, T. 2008, Central Bureau Electronic Telegrams, 1235, A260000

Steele, T. N., Silverman, J. M., Ganeshalingam, M., Lee, N., Li, W., & Filippenko, A. V. 2008, Central Bureau Electronic Telegrams, 1275, 1

Stern, D., et al. 2005, ApJ, 631, 163

Stetson, P. B. 1987, PASP, 99, 191

Stetson, P. B. 1992, Astronomical Data Analysis Software & Systems I, 25, 297

Stritzinger, M., Mazzali, P. A., Sollerman, J., & Benetti, S. 2006, A&A, 460, 793

Strolger, L.-G., et al. 2002, AJ, 124, 2905

Su´arez, O., Garc´ıa-Lario, P., Manchado, A., Manteiga, M., Ulla, A., & Pottasch, S. R. 2006, A&A, 458, 173

Sugerman, B., et al. 2004, IAU Circ., 8442, 2

Taubenberger, S., et al. 2006, MNRAS, 371, 1459

Thilker, D. A., et al. 2005, ApJ, 619, L67

Thompson, T. A., Burrows, A., & Pinto, P. A. 2003, ApJ, 592, 434

Thompson, T. A., Prieto, J. L., Stanek, K. Z., Kistler, M. D., Beacom, J. F., & Kochanek, C. S. 2008, arXiv:0809.0510

Tielens, A. G. G. M. 2008, ARA&A, 46, 289

Timmes, F. X., Brown, E. F., & Truran, J. W. 2003, ApJ, 590, L83

Tremonti, C. A., et al. 2004, ApJ, 613, 898

Tsvetkov, D. Y., Pavlyuk, N. N., & Bartunov, O. S. 2004, AstL, 30, 729

Tucker, D. L., et al. 2006, Astronomische Nachrichten, 327, 821

Tully, R. B., et al. 1998, AJ, 115, 2264

Turatto, M., Cappellaro, E., Danziger, I. J., Benetti, S., Gouiﬀes, C., & della Valle, M. 1993, MNRAS, 262, 128

Turatto, M., et al. 2000, ApJ, 534, L57

289 Turatto, M., Benetti, S., & Cappellaro, E. 2003, From Twilight to Highlight: The Physics of Supernovae, 200

Umeda, H., Nomoto, K., Kobayashi, C., Hachisu, I., & Kato, M. 1999, ApJ, 522, L43 van den Bergh, S. 1997, AJ, 113, 197 van Dyk, S. D. 1992, AJ, 103, 1788 van Dyk, S. D., Hamuy, M., & Filippenko, A. V. 1996, AJ, 111, 2017 van Dyk, S. D., et al. 2000, PASP, 112, 1532 van Dyk, S. D., Li, W., & Filippenko, A. V. 2003, PASP, 115, 1 van Dyk, S. D. 2005, The Fate of the Most Massive Stars, 332, 47 van Dyk, S. D., Li, W., Filippenko, A. V., Humphreys, R. M., Chornock, R., Foley, R., & Challis, P. M. 2006, arXiv:astro-ph/0603025 van Loon, J. T., et al. 1998, A&A, 329, 169 van Loon, J. T., et al. 2005, A&A, 438, 273 van Loon, J. T., Marshall, J. R., Cohen, M., Matsuura, M., Wood, P. R., Yamamura, I., & Zijlstra, A. A. 2006, A&A, 447, 971 van der Heyden, K. J., Bleeker, J. A. M. & Kaastra, J. S. 2004, A&A, 421, 1031 van Marle, A. J., Owocki, S. P., & Shaviv, N. J. 2008, MNRAS, 389, 1353 van Zee, L., Skillman, E. D., & Haynes, M. P. 2006, ApJ, 637, 269

Vijh, U. P., et al. 2009, AJ, 137, 3139

Vink, J. S., & de Koter, A. 2005, A&A, 442, 587

Wagner, R. M., et al. 2004, PASP, 116, 326

Wanajo, S., Nomoto, K., Janka, H.-T., Kitaura, F. S., M¨uller, B. 2009, ApJ, 695, 208

Wang, L., Baade, D., H¨oﬂich, P., Wheeler, J. C., Kawabata, K., & Nomoto, K. 2004, ApJ, 604, L53

Weilbacher, P. M., et al. 2003, A&A, 397, 545

Weis, K., & Bomans, D. J. 2005, A&A, 429, L13

290 Welch, D. L., et al. 2008, Central Bureau Electronic Telegrams, 1330, 1

Wesson, R., Fabbri, J., Barlow, M., & Meixner, M. 2008, Central Bureau Electronic Telegrams, 1381, 1

Wesson, R., et al. 2009, arXiv:0907.0246

West, R. M., et al. 1987, A&A, 177, L1

Wheeler, J. C., Cowan, J. J., & Hillebrandt, W. 1998, ApJ, 493, L101

Whelan, J., & Iben, I. J. 1973, 186, 1007

Witt, A. N., Vijh, U. P., Hobbs, L. M., Aufdenberg, J. P., Thorburn, J. A., & York, D. G. 2009, ApJ, 693, 1946

Wood, P. R., Whiteoak, J. B., Hughes, S. M. G., Bessell, M. S., Gardner, F. F., & Hyland, A. R. 1992, ApJ, 397, 552

Wood-Vasey, W. M., Wang, L., & Aldering, G. 2004, ApJ, 616, 339

Wood-Vasey, W. M., et al. 2007, ApJ, 666, 694

Wooden, D. H., Rank, D. M., Bregman, J. D., Witteborn, F. C., Tielens, A. G. G. M., Cohen, M., Pinto, P. A., & Axelrod, T. S. 1993, ApJS, 88, 477

Woosley, S. E., Heger, A., & Weaver, T. A. 2002, Reviews of Modern Physics, 74, 1015

Wyder, T. K., et al. 2005, ApJ, 619, L15

Yee, J. C., Eastman, J. D., & Prieto, J. L. 2008, Central Bureau Electronic Telegrams, 1340, 1

Yoon, S. -C., & Langer, N. 2005, A&A, 443, 643

York, D. G., et al. 2000, AJ, 120, 1579

Zaritsky, D., Kennicutt, R. C., Jr., & Huchra, J. P. 1994, ApJ, 420, 87

Zijlstra, A. A. 2004, MNRAS, 348, L23

Zinnecker, H., & Yorke, H. W. 2007, ARA&A, 45, 481

291