We present a 0.8–2.5 μm spectrum of the peculiar variable V838 Mon, obtained on 2002 October 29. We see one of the coolest stars — certainly the coolest supergiant — ever observed, with deep water bands and other spectral signatures normally associated with very late L and T brown dwarfs. We suggest that V838 Mon may be the first known ‘L supergiant’.
The peculiar variable V838 Mon was discovered in eruption on 2002 January 6 (Brown 2002). Its visual light curve (Munari et al. 2002; Kimeswenger et al. 2002; Crause et al. 2003) displayed three distinct maxima; as of 2003 January, it had faded by about 8 mag. Optical imaging of the object (Munari et al. 2002; Bond et al. 2002, 2003; Henden, Munari & Schwartz 2002) has shown the presence of a ‘light echo’, leading to distance estimates of ∼700–800 pc (Munari et al. 2002; Kimeswenger et al. 2002) and ≳6 kpc (Bond et al. 2003). The suggested progenitor (Munari et al. 2002) had blue and red apparent magnitudes of B≃ 16.0 and R≃ 15.3, respectively; further, there is good evidence (Munari et al. 2002; Kimeswenger et al. 2002) that the progenitor was detected by both the IRAS (Neugebauer et al. 1984) and 2MASS (Beichman et al. 1998) surveys. These data, combined with an interstellar reddening of E(B−V) = 0.5, suggest a progenitor with a temperature of 7300 K and a luminosity of 1.6 L⊙ (92 L⊙ at 6 kpc) (Munari et al. 2002).
The post-eruption evolution of the object on the Hertzsprung–Russell diagram has been extremely rapid (Munari et al. 2002; Kimeswenger et al. 2002). Optical and infrared (IR) photometry (Munari et al. 2002; Kimeswenger et al. 2002; Crause et al. 2003) shortly after discovery suggested that both the effective temperature and the luminosity of the star increased initially, but that subsequent evolution was to lower the effective temperature at nearly constant luminosity (∼4 × 103 L⊙ for a distance of 790 pc; ∼2.3 × 105 L⊙ at 6 kpc) (Munari et al. 2002). IR spectroscopy (Geballe et al. 2002a,b,c; Banerjee & Ashok 2002) showed dramatic changes in individual spectral features, and indicated a rapidly cooling star. The nature of V838 Mon remains an enigma, with similarities to final thermal-pulse objects such as Sakurai's object (papers in Evans & Smalley 2002) and to the luminous red variables V4332 Sgr and M31 RV being noted (Munari et al. 2002; Kimeswenger et al. 2002; Crause et al. 2003); Soker & Tylenda (2003) have suggested the merger of two main-sequence stars.
Here, we present a 0.8–2.5 μm spectrum of V838 Mon, obtained on October 29.125 ut, which shows one of the coolest stars ever observed.
The observations were carried out with the cooled-grating spectrometer CGS4 on the United Kingdom Infrared Telescope (UKIRT). The wavelength coverage was 0.80–1.34 μm (with resolution R in the range of 500–900, Δλ= 0.0015 μm) and 1.45–2.52 μm (R in the range of 500–850, Δλ= 0.0030 μm). Telluric features were removed — and flux calibration was achieved — by observing the F2 V star HD 2530 at the same airmass; the hydrogen lines in the calibration star spectrum were removed prior to ratioing. The 0.8–2.5 μm spectrum is shown in Fig. 1, with the spectrum in the region 0.8–1.35 μm region shown in greater detail in Fig. 2. The signal-to-noise ratio in the spectra was typically several hundred in the centres of the atmospheric windows.
The 0.8–2.5 μm spectrum of V838 Mon bears some resemblance to those of the coolest L dwarfs (see e.g. Geballe et al. 2002d). As in these objects, bands of H2O and CO dominate the spectrum. However, in V838 Mon, these bands were — as of 2002 October — far deeper than those seen in any L dwarf spectrum. In particular, the H2O bands (1.3–1.6 μm, 1.8–2.1 μm), which were weak in 2002 January and absent in 2002 March (Geballe et al. 2002a,b), were nearly saturated in 2002 October and comparable in strength to those seen in the latest and coolest T-type brown dwarfs (Geballe et al. 2002d). Whereas the CO bands weaken in T dwarfs, and by mid-T the wavelength of those bands are dominated by overtone and combination bands of methane, there is no evidence of methane in the spectrum of V838 Mon. We have confirmed the absence of methane from a 3-μm spectrum obtained on 2002 December 17, which shows no absorption in the strong fundamental ν2 band.
The variations in strength of the H2O bands, noted above, are suggestive of changes in the effective temperature of the star and, indeed, corresponding changes in the effective temperature are also suggested by the evolution of the optical spectrum (e.g. Munari et al. 2002; Crause et al. 2003; Rushton et al., in preparation); these changes are no doubt connected with the eruptive behaviour of V838 Mon. We also note that it is almost certainly the case that the great depth of the molecular bands in the IR spectrum of V838 Mon could be caused by non-LTE effects; however, detailed discussion of this topic is beyond the scope of this paper.
The first overtone (δυ= 2) bands of CO (2.3 μm and longer), which were present both in absorption and emission in earlier spectra of V838 Mon (Geballe et al. 2002a,b,c; Banerjee & Ashok 2002), are now strongly in absorption and are deeper than in any previously observed star of any luminosity class. The second overtone (δυ= 3) bands at 1.55–1.63 μm were not present (but see Banerjee & Ashok 2002); however, the situation is complicated by the steepness of the 1.6-μm shoulder of the H2O band and by the presence of prominent features at 1.646 μm and 1.684 μm, identified by Bernstein et al. (2003) as AlO A–X (2,0). There are also similar shaped features at 1.226 μm and 1.242 μm, also identified with AlO A–X (4,0) by Bernstein et al., although TiO λ1.242 μm may also contribute (see below and Fig. 2). The 1.226-μm feature is present in the spectrum of the peculiar evolved star IRAS 18530+0817 (Walker et al. 1997) and also in the spectrum of the M7 III giant R Cas (Joyce et al. 1998).
The 0.8–2.5 μm spectrum also displays several TiO and VO bands, indicating that, whatever V838 Mon is, it was (as of 2002 October) oxygen-rich. Our data show the Δυ=−1 sequence of the 1Π−1Δ band system of TiO, together with the Δυ=+1 sequence of the 1Π−1Σ system in the wavelength range of 0.97–1.05 μm. The Δυ= 0 sequence of the TiO ε system (bandhead at 0.84 μm), present in V838 Mon, has depth comparable with that seen in the very late (M10) asymptotic giant branch (AGB) Large Magellanic Cloud star IRAS 04509−6922 (van Loon et al. 1998); this feature usually disappears in the spectra of brown dwarfs later than L5, consistent with the lack of methane in V838 Mon. The φ system of TiO (Δυ= 0 and Δυ=−1 bands), with bandheads at 1.104 μm and 1.242 μm respectively, are also present, as is the VO A–X Δυ= 0 system at 1.046 μm. We note that we do not see any FeH features at 0.9868 μm, which also weaken considerably with decreasing temperature, essentially disappearing at L8 in brown dwarf stars (McLean et al. 2000); the FeH features at 1.194, 1.21, 1.237 μm are also absent in our data.
The 4p2P°− 3d2D K i lines at 1.1690, 1.1770 μm are just discernible, although the presence or otherwise of the K i 1.2432, 1.2522 μm (4p2P°− 5s2S) transitions is obscured by the TiO feature at 1.242 μm. The Rb i line at 1.323 μm and the Cs i line at 0.8963 μm, also common in the spectra of brown dwarfs (McLean et al. 2000), are present in the spectrum of V838 Mon. There is a feature at 1.282 μm; this is almost certainly not Pβ, because the corresponding Pα, which should be stronger, is either extremely weak or absent.
The IR spectrum of V838 Mon is characteristic of an extremely cool photosphere; it indicates a spectral type considerably later than the very latest of M stars, and hence an effective temperature Teff that is certainly below ∼2300 K. However, the great depths of the CO and H2O bands point to the presence of gas that is much cooler even than this (≲1300 K). The absence of methane at these temperatures illustrates the sensitivity to pressure of the competition for carbon between CO and CH4 (Allard et al. 1997). In brown dwarfs, the transition from carbon in CO to carbon in CH4 takes place in the temperature range of ∼1500–1200 K (Leggett et al. 2002). The determination of the effective temperature of V838 Mon must await detailed analysis, and here we make the working assumption that it is ≲2300 K.
Much of the decline in the visual light of V838 Mon must have been due to emission from the star shifting out of the optical and into the IR as it cooled (i.e. to bolometric correction effects), rather than the formation of an optically thick dust shell (see Kimeswenger et al. 2002). However, we expect that species such as C and Fe will surely condense into dust grains in a photosphere this cool. If we take the bolometric luminosity L from Munari et al. (2002) and an effective temperature of ≲2300 K, then the stellar radius R is
V838 Mon has clearly expanded in size and cooled considerably since its discovery in 2002 January, inconsistent with the canonical post-main-sequence evolution of a normal single star. We also deduce that V838 Mon is almost certainly not a final thermal-pulse event, like Sakurai's object (Evans & Smalley 2002). Not only is the IR development of these objects distinct, the C-rich environment of Sakurai's object contrasts with the O- and H-rich nature of V838 Mon, and they must therefore represent unconnected phases of stellar evolution (for the IR evolution of Sakurai's object, see Geballe et al. 2002e). Comparisons have been made between V838 Mon and a luminous red variable in the Andromeda galaxy, M31 RV (Mould et al. 1990). The latter object had an effective temperature of 4000 K and a radius of 2000 R⊙ at maximum light, and an apparent 1000 K ‘dust shell’ with a radius of ∼8000 R⊙, some 70 d after discovery (Mould et al. 1990); bearing in mind that this object was not as well observed as V838 Mon, the similarity in their behaviour is indeed suggestive.
Models to account for systems like V838 Mon and M31 RV include main-sequence star mergers (Soker & Tylenda 2003) and thermonuclear explosions in cataclysmic variable systems (Iben & Tutukov 1992). Neither is without its difficulties, and both interpret the reddening and the cooling in terms of obscuration by dust; however, we have shown here that the extreme redness and coolness are primarily stellar in origin, rather than dust-related. In the stellar merger scenario, the basic features can be explained in terms of the merger of a binary system consisting of two main sequence stars with masses ≃1.5 M⊙ and ≃0.2 M⊙. In the thermonuclear scenario (Iben & Tutukov 1992), the eruption arises in a binary with orbital period ∼80 min and which consists of a ≃0.6 M⊙ white dwarf and a ≃0.3 M⊙ star; in this case, one might expect abundance anomalies in the ejected material.
In order to make further progress in understanding the nature of V838 Mon, there is a need to undertake a rigorous spectral synthesis for such objects (equivalent to those already being undertaken for brown dwarfs; Allard et al. 1997) – not only to determine properly the photospheric parameters, but also to determine elemental abundances, particularly of those species arising in the later phases of stellar evolution or in a thermonuclear event. Furthermore, the role of grains in the atmosphere in determining the observed properties needs to be included. In brown dwarf stars, grains may have a substantial effect on the opacity of the atmosphere and, in some cases, may sink below the photosphere or gather in discrete clouds (Allard et al. 1996); however, whether this also occurs in the environment of an object like V838 Mon remains to be seen. A further key factor — and which our observations do not address — is whether the material giving rise to the light echo is associated with a previous evolutionary phase of the star (Kimeswenger et al. 2002; Soker & Tylenda 2003), or whether it is interstellar and therefore completely unconnected; identifying the nature and origin of this material is crucial. We are continuing to monitor this remarkable system and further observations will be described elsewhere.
We thank the referee for helpful comments on an earlier version of this paper. The United Kingdom Infrared Telescope is operated by the Joint Astronomy Centre, on behalf of the UK Particle Physics and Astronomy Research Council. TRG is supported by the Gemini Observatory, which is operated by the Association of Universities for Research in Astronomy, Inc., on behalf of the international Gemini partnership of Argentina, Australia, Brazil, Canada, Chile, the United Kingdom and the United States of America. MTR is supported by a PPARC studentship.