Light Echoes Reveal an Unexpectedly Cool Η Carinae During Its Nineteenth

Home , Eta Carinae, SN 2009ip, Var 83

LETTER doi:10.1038/nature10775

Light echoes reveal an unexpectedly cool g Carinae during its nineteenth-century Great Eruption

A. Rest1, J. L. Prieto2,3, N. R. Walborn1, N. Smith4, F. B. Bianco5,6, R. Chornock7, D. L. Welch8, D. A. Howell5,6, M. E. Huber9, R. J. Foley7, W. Fong7, B. Sinnott8, H. E. Bond1, R. C. Smith10, I. Toledo11, D. Minniti12 & K. Mandel7,13 g Carinae is one of the most massive binary stars in the Milky itself, but in any case it must be weak, if present. By cross-correlating Way1,2. It became the second-brightest star in our sky during its each of our g Car echo spectra with the Ultraviolet and Visual Echelle mid-nineteenth-century ‘Great Eruption’, but then faded from view Spectrograph (UVES) spectral library14 (see Supplementary Figs 2 (with only naked-eye estimates of brightness3,4). Its eruption is and 3), we find the best agreement with supergiant spectral types in unique in that it exceeded the Eddington luminosity limit for ten the range of G2 to G5, with an effective temperature of around 5,000 K. years. Because it is only 2.3 kiloparsecs away, spatially resolved Spectral types of F7 or earlier are ruled out by our analysis (see studies of the nebula have constrained the ejected mass and velocity, Supplementary Information for more details). indicating that during its nineteenth-century eruption, g Car Doppler shifts of absorption features in the echo spectra provide ejected more than ten solar masses in an event that released ten direct information about the outflow speeds during the eruption. The 5,6 per cent of the energy of a typical core-collapse supernova , Ca II infrared triplet absorption features in the spectrum are noticeably without destroying the star. Here we report observations of light blueshifted (see Supplementary Fig. 2). By cross-correlation with echoes of g Carinae from the 1838–1858 Great Eruption. Spectra of G-type15 templates, we determine velocities of –202 6 9, –210 6 14 these light echoes show only absorption lines, which are blueshifted and –237 6 17 km s21 (errors are standard deviation) for our three by 2210 km s21, in good agreement with predicted expansion individual spectra and an average velocity of –210 6 30 km s21, which speeds6. The light-echo spectra correlate best with those of G2- includes an uncertainty for the dust sheet velocity. to-G5 supergiants, which have effective temperatures of around The bipolar nature of the Homunculus nebula shows that the g Car 5,000 kelvin. In contrast to the class of extragalactic outbursts Great Eruption was strongly aspherical. It has been predicted that the assumed to be analogues of the Great Eruption of g Carinae7–12, outflow speeds that one would derive from the spectra of g Car in the effective temperature of its outburst is significantly lower than outburst, looking at the poles and equator of the double lobes, would that allowed by standard opaque wind models13. This indicates that be about 2650 km s21 and 240 to 2100 km s21, (note that the speeds other physical mechanisms such as an energetic blast wave may have are negative because they are blueshifted, that is, the outflow is moving triggered and influenced the eruption. towards us) respectively16 (outflow speeds near the equator have a g-Car-like giant eruptions of luminous blue variable stars are steep latitude dependence). The light echo we investigate here arises characterized by significant mass-loss and an increase in luminosity from latitudes near the equator of g Car (see Supplementary Fig. 1), by several magnitudes8,9. It has been thought that this increase in and the measured blueshifted velocity of 2210 6 30 km s21 is in good luminosity drives a dense wind, producing an optically thick, cooler agreement with expansion speeds within 620u of the equatorial plane. pseudo-photosphere with a minimum effective temperature of 7,000 K We also find a strong asymmetry in the Ca II infrared triplet, extending and an F-type spectrum13. Within this model, g Car has been consid- to a velocity of 2850 km s21—well below the speed of the fastest polar ered the prototype of these ‘‘supernova imposters’’7–12. ejecta found previously6, but in good agreement with speeds observed We obtained images in proximity to g Car (Fig. 1) that, when in the blast wave at lower latitudes6. Future observations of light echoes differenced, show a rich set of light echoes. The largest interval between viewing the g Car eruption from different directions, in particular our images is eight years. We have also found similar echo candidates from the poles, has the potential to observe these very-high-velocity at other positions, which we are currently monitoring. The large ejecta and other asymmetries. brightening and long duration point to the Great Eruption as the A characteristic of luminous-blue-variable outbursts is their trans- source of the light echoes. We have also obtained a composite light ition from a hot quiescent state to a cooler outburst state, although this curve in the Sloan Digital Sky Survey (SDSS) i filter of the light echoes feature is less well observed for the giant eruptions (see Fig. 4). Two (see Fig. 2), showing a slow decline of several tenths of a magnitude potential models for luminous-blue-variable outbursts involve either over half a year. This light curve is most consistent with the historical an opaque stellar wind driven by an increase in luminosity, or a observations4 of a peak in 1843, part of the 1838–1858 Great Eruption, hydrodynamic explosion. The traditional mechanism for g-Car-like although further observations are necessary to be certain (see the giant eruptions has been that an unexplained increase in luminosity Supplementary Information). drives a denser wind, so that an optically thick pseudo-photosphere Spectra of the light echoes (see Fig. 3) show only absorption lines forms at a layer much larger and cooler than the hydrostatic stellar characteristic of cool stellar photospheres, but no evidence of emission surface13. This model predicts a minimum effective temperature of lines. In particular, the Ca II infrared triplet is only observed as absorp- 7,000 K, resembling A- or F-type supergiants8,17, because the wind tion lines in the spectrum. Because of bright ambient nebular emission, opacity depends on the temperature (see Fig. 4). A giant eruption it is difficult to determine whether there is any Ha emission from g Car evidently occurs as a massive star attempts to evolve redward and

1Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, Maryland 21218, USA. 2Carnegie Observatories, 813 Santa Barbara Street, Pasadena, California 91101, USA. 3Department of Astrophysical Sciences, Princeton University, Peyton Hall, Princeton, New Jersey 08544, USA. 4Steward Observatory, University of Arizona, 933 North Cherry Avenue, Tucson, Arizona 85721, USA. 5Las Cumbres Observatory Global Telescope Network, Goleta, California 93117, USA. 6Department of Physics, University of California, Santa Barbara, California 93106, USA. 7Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, Massachusetts 02138, USA. 8Department of Physics and Astronomy, McMaster University, Hamilton, Ontario, L8S 4M1, Canada. 9Department of Physics and Astronomy, Johns Hopkins University, Baltimore, 3400 North Charles Street, Maryland 21218, USA. 10Cerro Tololo Inter-American Observatory, National Optical Astronomy Observatory, Colina el Pino S/N, La Serena, Chile. 11ALMA, KM 121 CH 23, San Pedro de Atacama, II Region, Chile. 12Department of Astronomy and Astrophysics, Pontificia Universidad Catolica de Chile, Santiago 22, Chile. 13Imperial College London, Blackett Laboratory, Prince Consort Road, London SW7 2AZ, UK.

10 March 2003 (A) Difference image C – A

η Carinae 120 ′′ 120 ′′

10 May 2010 (B) Difference image C – B 120 ′′ 120 ′′

6 February 2011 (C) C – A Difference image zoom

8.5′ × 8.5′ N

15′ E 30 ′′ 120 ′′

Figure 1 | g Car light echoes. The left panel shows the positions of the star blue line. For all panels north is up and east is to the left. Applying the vector g Carinae and our images (white box), plotted on an image in the light of three method that previously allowed us to identify the source of the light echoes different emission lines: oxygen (blue), hydrogen (green) and sulphur (red). from the supernovae that produced the supernova remnants SNR 0509–67.5, (Photo taken by N.S.) The middle panels show the images obtained with the Cas A and Tycho24,25, we find that a dramatic brightening of g Car must be the CTIO 4-m Blanco telescope of a region about 0.5u to the south of g Car at 10 origin. In these echoes, unlike those of Galactic supernovae26, there is still March 2003 (epoch A), 10 May 2010 (epoch B), and 6 February 2011 (epoch C), significant spatial overlap even at separations of one light-year, suggesting that from top to bottom. The right panels show the difference images ‘C minus A’ the duration of the event causing them must be significantly longer than one and ‘C minus B’ at the top and middle, respectively. Example light-echo year. We also see brightening of two magnitudes or more within eight years. positions are indicated with blue (epoch A) and red (epochs B and C) arrows. Thus, the Lesser Eruption from 1887 to 1896, which brightened by only a The bottom right panel shows a zoom of the spectrograph slit, indicated with a magnitude, is excluded as the source.

Historic visual light curve –1.0 Light echo light curve, year minus 174.20 13 Light echo light curve, encounters the Humphreys–Davidson limit , beyond which no stable –0.5 year minus 167.95 Light echo light curve, stars are observed. 0.0 year minus 166.28 Surprisingly, our G-type light-echo spectrum of the g Car Great 0.5 Eruption is inconsistent with expectations of an opaque-wind model13 1.0 50 Magnitude (see Fig. 4). With this model, it is difficult to explain the high (10 erg) 1.5 5 6 2.0 kinetic energy and the presence of a fast blast wave at large radii . 1825 1830 1835 1840 1845 1850 1855 1860 Instead, these observations point towards a hydrodynamic explosion that must have influenced the Great Eruption2,5,6. –1.0 The first visual spectroscopic observations of g Car around 1870 –0.5 showed emission lines18,19. A photographic spectrogram obtained dur- 0.0 ing its Lesser Eruption20,21 around 1890 resembles an F-type supergiant 0.5 21 1.0 blueshifted by 2200 km s , with moderate P Cygni hydrogen pro-

Magnitude 13 1.5 files, which is as expected in the opaque-wind model . The difference 2.0 between the 1890 spectrum and our light-echo spectrum of the Great 1837.5 1838.0 1838.5 1842.5 1843.0 1843.5 1844.5 1845.0 Eruption is therefore quite striking, indicating that two distinct phys- Year ical processes may have been involved for two outbursts of the same Figure 2 | Historical and light-echo lightcurve of g Car. The historical light object. However, the 1890 event also produced a mass ejection, the curve4 in visual apparent magnitudes is shown with black circles and lines, with Little Homunculus, with the same axial symmetry (although much error bars indicating approximate uncertainties in these eye estimates. Light smaller mass) as the Great Eruption22. echo brightnesses (SDSS i; error bars are the standard deviation) from our eight Luminous-blue-variable giant eruptions are rare, and have only modern images spanning about eight years are displayed shifted by 174.2 Earth been recorded twice in our Galaxy in the last 400 years: the Great years (green circles), 167.95 years (red circles) and 166.28 years (blue circles), to Eruption of g Car and the giant eruption in the seventeenth century illustrate the best-matching time delays for the 1838, 1843 and 1845 outbursts, respectively. The first epoch is an upper limit indicated with an arrow. The of P Cygni. Because of their considerable intrinsic brightness just upper panel shows the full time range of the Great Eruption and therefore below the luminosity of faint core-collapse supernovae, about two shows all three potential matches, whereas the lower panels show the dozen giant eruption candidates, called supernova imposters because brightnesses from seven of our eight modern epochs in a magnified time period they have often been mistaken for supernovae, have been found in around each peak. various extragalactic supernova searches7–12. Typically, the hotter

7 [N II] Hα SN2009ip

7.5 Eruptions 6 40 SN 2009ip Luminous blue variable stars η SN 2009ip Car (Great Eruption) Luminous blue variable 5 SN 1997bs candidates SN 1997bs UGC2773-OT Cool Scaled F UGC 2773-OT1 UGC 2773-OT1 30 7.0 hypergiants 4

λ η η Car (1890) + constant Car (today) + constant λ

EC1C ] ⊕ 3 20 6.5 [ L / HD 5980 Scaled F 10 Var 83 S Dor instability strip log

2 EC1C AF And Wray 17-96 EC1B AG Car IRAS 10 R 127 18576+0341 EC1A EC1B S 61 G79 29+0.46 GR290 1 6.0 G26.47+0.02 Var B R 143 S Dor P Cyg AE And IRC+10420 SK -69º279 Var C Var A EC1A Wra 751 S119 W243 Hen 3_519 4,000 5,000 6,000 7,000 8,000 6,550 6,600 G24.73+0.69 ρ Cas HR8752 Rest wavelength (Å) HR Car R 40 5.5 160529 v382 Car Figure 3 | Light echo spectra of the Great Eruption of g Car. The three R 71 R110 168625 optical low-resolution spectra of the light echo (black lines) were taken at J2000 2 9 99 position right ascension 10 h 44 min 12.127 s and declination 60u 16 01.69 40,000 20,000 10,000 5,000 in March and April 2011 obtained at the Magellan I 6.5-m and du Pont 2.5-m Temperature (K) telescopes of the Las Campanas Observatory, Chile. A log of the spectroscopic observations and details of the spectra is presented in Supplementary Table 1. Figure 4 | Hertzsprung–Russell diagram with luminous blue variables and The slit positions differ only slightly in slit angle. The spectra were reduced g Car. Adaptation17 of a Hertzsprung–Russell diagram showing luminous blue using standard techniques and then wavelength-calibrated using observations variables, related hypergiant stars, and the peak luminosities of luminous-blue- of an HeNeAr lamp. The wavelength calibration was checked and corrected variable-like eruptions. The grey bands denote the typical locations of luminous using night-sky emission lines, especially [O I] l5,577 A˚ , and OH lines in the blue variables in quiescence (left, diagonal band) and during the S Doradus-like red part of the spectrum. We flux-calibrated the spectra using a flux standard outburst. Temperatures for the Great Eruption and the 1890 eruption of g Car observed the same night as the science observations. The left panel shows the are based on the echo spectra presented here and the F-type spectrum of the spectra from 5,000 to 9,000 A˚ . The spectra are not corrected for reddening nor 1890 event20, respectively. The temperature of 10,000 K for SN 2009ip is based for the blueward scattering by the dust. For comparison, the blue lines show on the observed continuum shape, but this is only a lower limit because of the spectra of three examples of supernova imposters: SN 1997bs, SN 2009ip and possible effects of circumstellar or host galaxy reddening27. Because of the UGC 2773-OT1. The right panel shows the Ha and [N II] emission lines. We presence of He I lines in the spectrum, the true temperature is probably much note that the background emission-line subtraction is incomplete because the hotter. The 8,500 K temperature of UGC2773-OT is indicated by the F-type emission lines vary spatially. Also, EC1A Ha is at the edge of the chip and is absorption features in its spectrum, and this temperature is relatively therefore uncertain. Crossed circles indicate the locations of atmospheric independent of reddening27,28. absorption lines. supernova imposters have steep blue continua, stronger and broader these spectral observations. Alternative models, such as the ones that Balmer lines, and relatively weak absorption, whereas the cooler ones use mass accretion from the companion star during periapsis passage 12 tend to have redder continua, weaker and narrower Balmer lines, as a trigger for the eruption , can be verified or dismissed. strong [Ca II] and Ca II emission, deeper P Cygni absorption features, Received 26 August; accepted 8 December 2011. and in some cases stronger absorption spectra similar to those of 23 F-type supergiants . However, the g Car Great Eruption light-echo 1. Davidson, K. & Humphreys, R. M. Eta Carinae and its environment. Annu. Rev. spectrum is quite different. Its spectral type is G2 to G5, significantly Astron. Astrophys. 35, 1–32 (1997). 2. Damineli, A. The 5.52 year cycle of eta Carinae. Astrophys. J. 460, L49 (1996). later than all other supernova imposters at peak. Furthermore, the Ca II 3. Frew, D. J. The historical record of g Carinae I. The visual light curve, 1595–2000. infrared triplet lines are only in absorption. For the extreme mass-loss J. Astron. Data 10, 6 (2004). rates required for the Great Eruption of g Car, another process must 4. Smith, N. & Frew, D. J. A revised historical light curve of eta Carinae and the timing give rise to the apparent temperature. of close periastron encounters. Mon. Not. R. Astron. Soc. 415, 2009–2019 (2011). 5. Smith, N. et al. Mass and kinetic energy of the Homunculus nebula around The Great Eruption of g Car has been considered the prototype of g Carinae.Astron.J.125, 1458–1466 (2003). the extragalactic supernova imposters or g Car analogues, even though 6. Smith, N. A blast wave from the 1843 eruption of g Carinae. Nature 455, 201–203 it is actually an extreme case in terms of radiated energy (1049.3 erg), (2008). kinetic energy (.1050 erg), and its decade-long duration23. The spectra 7. Goodrich, R. W., Stringfellow, G. S., Penrod, G. D. & Filippenko, A. V. S. N. 1961V—an extragalactic ETA Carinae analog. Astrophys. J. 342, 908–916 (1989). of the light echo indicates now that it is not only extreme, but a 8. Humphreys, R. M. & Davidson, K. The luminous blue variables: astrophysical different, unique object. It is difficult to see how strong emission lines geysers. Publ. Astron. Soc. Pacif. 106, 1025–1051 (1994). could be avoided in an opaque wind where the continuum pho- 9. Humphreys, R. M., Davidson, K. & Smith, N. Eta Carinae’s second eruption and the light curves of the eta Carinae variables. Publ. Astron. Soc. Pacif. 111, 1124–1131 tosphere is determined by a change in opacity, and its temperature (1999). and broad absorption lines are more consistent with the opaque cool- 10. Van Dyk, S. D. et al. SN 1997bs in M66: another extragalactic g Carinae analog? ing photosphere of an explosion. What triggered such an explosion Publ. Astron. Soc. Pacif. 112, 1532–1541 (2000). and the reason that the huge mass-loss did not destroy the star are still 11. Vink, J. S. Eta Carinae and the luminous blue variables. Preprint at http://arxiv.org/ abs/0905.3338 (2009). unknown, but predictions from future radiative transfer simulations 12. Kashi, A., Frankowski, A. & Soker, N. NGC 300 OT2008–1 as a scaled down version trying to explain g Car and its Great Eruption can now be matched to of the eta Carinae great eruption. Astrophys. J. 709, L11–L15 (2010).

13. Davidson, K. The relation between apparent temperature and mass-loss rate in 27. Smith, N. et al. Discovery of precursor luminous blue variable outbursts in two hypergiant eruptions. Astrophys. J. 317, 760–764 (1987). recent optical transients: the fitfully variable missing links UGC 2773-OT and SN 14. Bagnulo, S. et al. The UVES Paranal Observatory Project: a library of high- 2009ip. Astron. J. 139, 1451–1467 (2010). resolution spectra of stars across the Hertzsprung-Russell diagram. The 28. Foley, R. J. et al. The diversity of massive star outbursts. I. Observations of Messenger 114, 10–14 (2003). SN2009ip, UGC 2773 OT2009–1, and their progenitors. Astrophys. J. 732, 32 15. Cenarro, A. J. et al. Empirical calibration of the near-infrared Ca II triplet—I. The (2011). stellar library and index definition. Mon. Not. R. Astron. Soc. 326, 959–980 (2001). Supplementary Information is linked to the online version of the paper at 16. Smith, N. The structure of the Homunculus. I. Shape and latitude dependence www.nature.com/nature. from H2 and [Fe II] velocity maps of g Carinae. Astrophys. J. 644, 1151–1163 (2006). Acknowledgements We thank R. Humphreys, K. Davidson and J. Vink for comments 17. Smith, N., Vink, J. S. & de Koter, A. The missing luminous blue variables and the and discussions. We thank S. Blondin for help with the continuum subtraction. The bistability jump. Astrophys. J. 615, 475–484 (2004). Blanco 4-m telescope is a facility of the Cerro Tololo Inter-American Observatory, 18. Le Sueur, A. On the nebulae of Argo and Orion, and on the spectrum of Jupiter. National Optical Astronomy Observatory, which is operated by the Association of Proc. R. Soc. Lond. I 18, 245–250 (1869). Universities for Research in Astronomy, under contract with the National Science 19. Le Sueur, A. Observations with the Great Melbourne telescope, in a letter to Prof. Foundation. We use data from the UVES Paranal Observatory Project. The Stokes. Proc. R. Soc. Lond. I 19, 18–19 (1870). computations in this paper were run on the Odyssey cluster supported by the FAS 20. Walborn, N. R. & Liller, M. H. The earliest spectroscopic observations of eta Carinae Science Division Research Computing Group at Harvard University. Observations were and its interaction with the Carina nebula. Astrophys. J. 211, 181–183 (1977). obtained at LCOGT, and F.B.B. and D.A.H. acknowledge support from LCOGT. J.L.P. is a 21. Humphreys, R. M., Davidson, K. & Koppelman, M. The early spectra of eta Carinae Hubble Carnegie-Princeton Fellow. R.J.F. is a Clay Fellow. 1892 to 1941 and the onset of its high excitation emission spectrum. Astron. J. 135, 1249–1263 (2008). Author Contributions All authors contributed to the drafting of the paper. A.R., N.S. and R.C.S. imaged the area around g Car. A.R. and M.E.H. reduced the imaging data. H.E.B. 22. Smith, N. Doppler tomography of the Little Homunculus: high-resolution spectra provided images of the echoes that guided our spectroscopic pointings. J.L.P., R.C., of [FeII]l16435 around eta Carinae*. Mon. Not. R. Astron. Soc. 357, 1330–1336 R.J.F. and W.F. obtained the spectra and reduced them. A.R. and J.L.P. performed (2005). spectral analysis and interpretation. A.R., N.R.W. and F.B.B. performed spectral 23. Smith, N., Li, W., Silverman, J. M., Ganeshalingam, M. & Filippenko, A. V. Luminous classification. F.B.B. and K.M. correlated the spectra. A.R., D.L.W. and B.S. modelled the blue variable eruptions and related transients: diversity of progenitors and light echo. I.T. and D.M. provided imaging of g Car. F.B.B. and D.A.H. provided the FTS outburst properties. Mon. Not. R. Astron. Soc. 415, 773–810 (2011). images, and F.B.B. and A.R. reduced them. 24. Rest, A. et al. Light echoes from ancient supernovae in the Large Magellanic Cloud. Nature 438, 1132–1134 (2005). Author Information Reprints and permissions information is available at 25. Rest, A. et al. Scattered-light echoes from the historical galactic supernovae www.nature.com/reprints. The authors declare no competing financial interests. Cassiopeia A and Tycho (SN 1572). Astrophys. J. 681, L81–L84 (2008). Readers are welcome to comment on the online version of this article at 26. Rest, A. et al. On the interpretation of supernova light echo profiles and spectra. www.nature.com/nature. Correspondence and requests for materials should be Astrophys. J. 732, 2 (2011). addressed to A.R. ([email protected]).