KECK/HIRES SPECTROSCOPY OF V838 MONOCEROTIS IN OCTOBER 2005^*

The eruption of V838 Monocerotis (V838 Mon) was discovered in the beginning of 2002 January. Initially thought to be a nova, the object appeared unusual and enigmatic in its nature. The eruption, as observed in the optical, lasted about three months (Munari et al. 2002b; Kimeswenger et al. 2002; Crause et al. 2003). After developing an A-F supergiant spectrum at the maximum at the beginning of 2002 February, the object evolved to lower effective temperatures and in 2002 April it practically disappeared from the optical, remaining very bright in the infrared. At the same time, a B3V companion to the erupted object was discovered in the optical (Munari et al. 2002a). A detailed analysis of the evolution of the object in the outburst and decline can be found, e.g., in Tylenda (2005).

Several mechanisms have been proposed to explain the eruption of V838 Mon, including an unusual nova (Iben & Tutukov 1992), a late He-shell flash (Lawlor 2005), and a stellar merger (Soker & Tylenda 2003). They have critically been discussed in Tylenda & Soker (2006). These authors conclude that the only mechanism that can satisfactorily account for the observational data is a collision and merger of a low-mass pre-main-sequence star with an ∼8 M_☉ main-sequence star.

In 2006 October/November, the B3V companion significantly faded and a strong Hα emission appeared in the spectrum of V838 Mon (Goranskij 2006; Bond 2006). Late in 2004, an emission-line spectrum, composed mainly of [Fe ii] lines, started to develop (Barsukova et al. 2006), reaching its maximum around the 2006 eclipselike event (Munari et al. 2007).

In the present paper, we present and discuss a high-resolution spectrum of V838 Mon acquired with the Keck I telescope in 2005 October. At that time, the [Fe ii] emission-line spectrum was already well developed and the object was a year before the eclipse of the B3V companion.

A high-resolution spectrum of V838 Mon was obtained on UT 2005 October 13 with the High Resolution Echelle Spectrometer (HIRES; Vogt et al. 1994) attached to the Keck I telescope. We used the C4 decker, which provides a slit with dimensions 1148 × 35. The resultant resolving power was R ≃ 34,000. The HIRES instrument makes use of a 3-chip mosaic as a detector. We obtained two exposures of the object with a total time of 960 s.

Data reduction was performed with the IRAF³ standard procedures (Massey 1997). Data were debiased and flat-fielded using the ccdproc task. Cosmic-ray events were removed from the CCD frames with a method described in van Dokkum (2001). After a careful check of the removal, we are confident that the procedure did not introduce any spurious effects to the spectra. Order tracing and extraction were done using the tasks in the echelle package. The orders corresponding to the bluest part of the spectra were underexposed and we were not able to extract them, even with a pinhole trace as a reference. The wavelength calibration was based on ThAr lamp observations and it is accurate to within 0.003 Å. The two extracted echelle spectra were averaged giving a spectrum with a total wavelength coverage of 3720–7962 Å. In the range there are, however, numerous gaps caused by the inter-chip spaces on the mosaic and the orders that do not overlap.

Using an extracted pinhole frame, we corrected all the orders of the echelle spectrum for a blaze effect. Inspection of the overlapping and blaze-corrected orders revealed that, although correctly flattened, they were systematically tilted in such a way that the long-wavelength end of an order was always higher on a relative intensity scale than the short-wavelength end of the neighboring order. The discrepancy was of 15% of the average intensity of the order. We reduced the tilt by multiplying each order by a linear function (i.e., with 1 in the middle of an order range and 1 ± 0.075 at the edges). All the overlapping orders were then safely merged and, subsequently, we got a spectrum composed of three parts: (i) completely covered wavelength range over 3720–4883 Å, (ii) wide range over 4951–6469 Å with two narrow gaps at 6308.5–6310 Å and 6401–6424.5 Å, and (iii) a part of the spectrum over the range 6544–7962 Å with 10 nonoverlapping orders. The three pieces we call hereafter parts blue, green, and red, respectively.

Since no spectrophotometric standard stars were observed during our observing run at Keck I, we decided to flux calibrate the HIRES spectrum on the basis of the method described in Suzuki et al. (2003), namely, by using a flux-calibrated spectrum of V838 Mon obtained elsewhere with a lower spectral resolution. Such a low-resolution spectrum was acquired on 2005 November 22 with the Grating Spectrograph with SITe mounted at the 1.9 m Radcliffe telescope at the South African Astronomical Observatory (SAAO). We made use of a grating no. 7 with 300 lines mm⁻¹ and a slit with a projected width of 15. The last figure was comparable to seeing conditions during the observation (∼14). The total duration of the exposure was 1000 s. Three spectrophotometric standard stars, i.e., LTT9239, EG21, and LTT3218, were observed during the night. This allowed us to calibrate the low-resolution spectrum in absolute flux units, while the wavelength calibration was performed with ThAr reference lamp spectra. All the data reduction and calibrations were carried out with standard IRAF procedures. The extracted and flux-calibrated spectrum covers the range 3817–7235 Å with a resolving power of R ≃ 1 000.

The flux calibration of the HIRES spectrum was performed as follows. The high- and low-resolution spectra were corrected to the heliocentric rest frame and they were cross-correlated in order to find a relative shift between them in the wavelength domain. We found a shift of 5.5 Å (note that the value is close to the spectral resolution in the SAAO observations), which was eliminated by displacing the low-resolution spectrum. Next, the HIRES spectrum was smoothed to the resolution of the observations at SAAO and the spectra were divided one by the other. Rather than to calibrate each single order of the echelle spectrum, as it was done in Suzuki et al. (2003), we decided to calibrate separately only the three mentioned above parts of the merged spectrum. To each of the pieces, a low-order polynomial was fitted giving a conversion ratio (CR; see Suzuki et al. 2003). Since the SAAO spectrum has a narrower wavelength range than the high resolution one, we extrapolated the CRs to the full ranges of the blue and red parts. Finally, the HIRES spectrum was put on the absolute flux scale by dividing its parts by the appropriate CRs. We note that the flux calibration of our final spectrum is uncertain, particularly in the ranges not covered by the SAAO observations and in the regions where strong atmospheric molecular bands are present (the object was observed at unequal air masses in the different facilities).

Although the low-resolution spectrum was obtained about 6 weeks after the HIRES observations, we are quite confident that the spectrum of V838 Mon did not change significantly during this period in the observed spectral range. A comparison of the spectra smoothed to the same resolution shows that there are no significant discrepancies which would affect the flux calibration. Moreover, from the photometric behavior of the object, illustrated, e.g., on V. Goranskij's Web site,⁴ one can see that the assumption of a constant flux in B and V bands during the period is justified.

The flux-calibrated spectrum was then corrected for the interstellar reddening using the IRAF's task deredden with E_B−V = 0.90 (Tylenda 2005) and with a standard ratio of total to selective extinction R_V = 3.1.

The flux-calibrated and dereddened spectrum of V838 Mon is presented in Figures 1–3. For a comparison, a synthetic spectrum of a B-type star (see Section 6) is overplotted with the blue line. Most of the identified atomic and molecular features are indicated in the figures. Wavelengths are given in the heliocentric rest frame and the units of the flux are 10⁻¹³ erg s⁻¹ cm⁻² Å⁻¹.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Our observations reveal an unusual and very complex spectrum of V838 Mon. It contains contributions from several physically different environments. In its blue part a photospheric spectrum of the B3V companion can clearly be seen, while the green and red parts are dominated by molecular absorptions that severely affect the flux of a very cold supergiant. A particularly striking feature of the presented spectrum is the presence of prominent emissions of [Fe ii]. Finally, numerous, mostly resonance, lines showing P Cygni profiles as well as a few pure absorptions of neutral atoms can easily be found.

In the case of such a complex spectrum, detailed identification of all the features is a difficult task. The identification procedures and their results are described in the following sections. We also present results of basic measurements, e.g., radial velocities of different line systems and molecular bands.

The atomic line identification was based mainly on the Atomic Spectra Database Lines Form⁵ (provided by the National Institute of Standards and Technology (NIST)), the Atomic Line List version 2.04 by van Hoof,⁶ and on the multiplet tables by Moore (1945).

All the identified atomic emission and absorption features are listed in Tables 1 and 2 (respectively) in order of their laboratory wavelengths. Columns 1–3 of the tables contain laboratory wavelength, ion name, and multiplet number from Moore (1945). Column 4 in Table 1 gives integrated (dereddened) fluxes of the emission lines (for which a reliable measurement was possible). In the same column of Table 2, the profile type (pure absorption or P Cygni) is specified. Notes on the features are given in the last column of the tables.

4.1. Emission Lines

4.1.1. Forbidden Lines

A great majority of the emission features seen in the spectrum are forbidden lines of Fe ii. They are remarkably numerous. We have identified 49 lines from 11 different multiplets. All of them arise from levels with energies (above the ground level) between 2.0 and 3.4 eV. Most of them are very strong and often blended with other spectral features. In addition to these forbidden transitions, we have also found two permitted lines of Fe ii at 5018 Å and 5169 Å (a third line from the same multiplet falls in a gap between two chips).

All the unblended [Fe ii] emissions show practically the same but rather unusual shape. It can be well fitted with a theoretical Lorentzian profile, as can be seen from Figure 4. The profile shown in this figure is a mean shape obtained from seven strongest and "clean" [Fe ii] lines. When averaging, the normalized profiles of individual lines were weighted according to the rms² values of the local noise in the spectrum. From 30 [Fe ii] lines, for which we successfully fitted a Lorentzian profile using IRAF's splot, we have obtained a typical full width at half-maximum⁷ (FWHM) of 76.3 ± 17.7 km s⁻¹ (median and standard deviation). All the lines have practically the same radial velocity. Measurements performed with the IRAF's rvidlines task gave a heliocentric radial velocity of V_h = 13.3 ± 0.7 km s⁻¹ (mean and standard deviation).

Zoom In Zoom Out Reset image size

Other forbidden lines seen in our spectrum include those of [O i] 5577 Å, 6300 Å, 6364 Å, and [N i] 5198 Å. These are rather weak features, observed in spectral regions strongly contaminated with molecular bands. Their measurements are therefore uncertain. The measured FWHM and radial velocity of the [O i] lines are 80 ± 30 km s⁻¹ and 40 ± 10 km s⁻¹, respectively. The [N i] line has a radial velocity of about 80 km s⁻¹.

It should be noted that some of the emission lines listed in Table 1 have uncertain identification. Two strong features at about 5041 Å and 5056 Å were identified as Si ii, but other identifications are possible. Three lines recognized as [Ni ii] 3993 Å, 4201 Å, and 4326 Åare weak, which makes their identification uncertain. However, the ionization potential of Si and Ni being similar to that of Fe and low excitation energies of the proposed multiplets make the identification probable.

4.1.2. Balmer Lines

Inside the cores of the photospheric Balmer lines of the B-type companion weak emission features can be found. Their presence is evident when one compares the observed photospheric profiles with a synthetic spectrum of an early B-type star (see Figure 1). The following discussion is limited to Hβ only, as the emission feature is most evident in this line. In order to better characterize the emission, we have subtracted the underlying stellar absorption. The procedure was as follows. To the wings of the absorption line, we fitted a profile in the form

$$\begin{equation} \phi (\lambda)=d\exp \lbrace [-a(\lambda \pm \lambda _c)^b+c]^{-1}\rbrace \end{equation} \tag{ 1 }$$

with a, b, c, and d as free parameters. The central wavelength, λ_c, was taken as the laboratory wavelength of the Hβ line corrected for the radial velocity of the B3V star (see Section 6). Equation (1) is based on an empirical profile given in Chauville et al. (2001). However, for our purpose it was generalized to treat also the local continuum level as a free parameter. The fitting procedure used the least-squares method and it was applied to all the data points in the Hβ wing (i.e., not only to three points, as in Chauville et al. 2001). The core region dominated by emissions was, obviously, excluded from the fit. The extracted emission feature is shown in Figure 5. It can be interpreted as a single emission with an absorption superimposed on it. In this case, fitting Lorentzian profiles gives the emission component with an FWHM of ∼42 km s⁻¹ at a radial velocity of ∼14 km s⁻¹, while the absorption component appears at ∼10 km s⁻¹ and has an FWHM of ∼24 km s⁻¹. Note that the parameters of the emission component in this case are close to those of the [Fe ii] lines. Alternatively, if the observed profile is fitted with two emission components, they have radial velocities of ∼−10 and ∼28.8 km s⁻¹, and FWHMs of ∼30 and ∼24 km s⁻¹, respectively.

Zoom In Zoom Out Reset image size

4.2. P Cygni Features

The spectrum displays a number of strong resonance lines showing P Cygni profiles. The most prominent ones are those of Mn i 5395 Å and 5433 Å, Cr i 5410 Å, Ba i 5535 Å, Ca i 6573 Å, K i 7699 Å, and Rb i 7800 Å (only the Cr i line is not a resonant one). They are shown in Figure 6. Other lines showing P Cygni signatures, although not as notable as in those listed above, are Mg i 4571 Å, Fe i 5110 Å, and possibly also a few weak lines of Cr i.

Zoom In Zoom Out Reset image size

Most of the P Cygni profiles have a deep absorption component extending from ∼50 km s⁻¹ (as a crossover from emission to absorption) to ∼−95 km s⁻¹. The most extreme exception is the K i line, which shows blueshifted absorption down to ∼−145 km s⁻¹. Emission components peak at ∼80 km s⁻¹.

The shapes of the absorption components indicate inhomogeneities in the outflowing matter. Particularly in the profiles of Rb i and Mn i, one can clearly see an additional narrow absorption component (NAC) centered at −82 km s⁻¹ (see panel (b) in Figure 6). The measured FWHM of the narrow feature inside the Rb i 7800 Å line is 19.3 km s⁻¹.

As noted above, the Mg i 4571 Å line shows only weak signatures of a P Cygni profile. Indeed, it is dominated by a slightly asymmetric emission and exhibits only a very weak absorption at approximately the same velocity as the NAC.

The profile of Fe i 5110 Å is also exceptional. Its absorption component is so deep that it removes all the light from the cold star and reaches the flux level of the B3V companion (see the top panel in Figure 2). In addition, it is considerably broader than most of the other P Cygni absorptions and extends from ∼80 km s⁻¹ to ∼−120 km s⁻¹. Most probably, the λ5110 absorption is saturated. Other Fe i lines identified in the spectrum appear as absorptions only, without any signs of emission components (see Section 4.3).

As Table 2 indicates, almost all the abundant alkali metals show P Cygni profiles. Thus, one would expect that it should also be the case of the Li i doublet at 6708 Å. However, in the raw spectrum no clear emission component can be seen close to the clear absorption of Li i. The doublet falls in a region that is dominated by a strong absorption band of TiO and it is difficult to define an underlying molecular baseline. Moreover, the doublet resides very close to the γ (1, 0) band head of TiO and most probably a part of the emission component of Li i is absorbed by the saturated molecular band. A similar situation also occurs for the Rb i 7948 Å line. Its profile is strongly affected by the (B − X) (0, 0) absorption band of VO and the expected emission component cannot be seen on the raw spectrogram.

4.3. Absorption Lines

The complexity of the spectrum makes a direct identification of molecular bands very difficult. For identification purposes, we have performed a series of simulations of molecular spectra. First attempts showed that a model capable of reasonably reproducing the observed molecular features should include, at least, two components, i.e., a stellar photosphere of a cool supergiant and an absorbing outflowing material with an excitation temperature significantly lower than the stellar temperature.

As a model of the stellar cold photosphere, we have used a synthetic spectrum generated from a model atmosphere from the NextGen grid⁸ (Hauschildt et al. 1999). These models include TiO opacities, which is crucial in the case of the analyzed spectrum. We chose a model atmosphere with T_eff = 2400 K, and with the lowest gravity available within the grid, i.e., log g = 0.0. For these parameters there is only a NextGen model with solar metallicity, which is probably close to the true metallicity of V838 Mon (Kipper & Škoda 2007).

Our choice on T_eff is a result of initial modeling of the observed spectrum, where we used atmosphere models for T_eff = 2000, 2200, and 2400 K. With the value of 2400 K, we have obtained most satisfying fits to the observed band profiles from higher rotational levels.

The effective temperature of the cold star can also be determined from photometric measurements. Munari et al. (2007) measured BVR_cI_c magnitudes of V838 Mon on 2005 December 1, while Henden et al. (2006) measured UBVR_cI_c on 2005 December 25 and JHK' magnitudes on December 18. Thus, the photometry was done ∼two months after our spectroscopy. Given very slow evolution of the object at that epoch, we can safely assume that the magnitudes did not change significantly between 2005 October and December. Applying the same procedure to the above photometric results as that used in Tylenda (2005), i.e., fitting a standard supergiant photometric spectrum with E_B−V = 0.9 and taking into account a contribution from a standard B3V star, we have obtained a spectral type of M6–7 and T_eff ≃ 2480 K. Adopting R = 3.1 and a distance of 8 kpc (as in Tylenda 2005), the effective radius and luminosity of V838 Mon are ∼840 R_☉ and ∼2.4 × 10⁴ L_☉, respectively. The resultant fit is shown in Figure 7. When fitting a blackbody to the IJHK magnitudes, a temperature of 2320 K has been obtained and, correspondingly, an effective radius of ∼905 R_☉ and a luminosity of ∼2.2 × 10⁴ L_☉. The blackbody fit (with the contribution from the B3V companion added) is shown as a dashed curve in Figure 7. Thus, the estimates of T_eff obtained from the photometric measurements are well consistent with the effective temperature we used to model the cold photosphere.

Zoom In Zoom Out Reset image size

The photosphere model spectrum was then used as an underlying spectrum absorbed in the outflowing molecular gas. In radiative transfer modeling of absorption in the outflow (scattering was not included), the gas was treated as a plane-parallel homogeneous slab of matter. In this one-dimensional approach, three parameters characterize the outflowing gas, i.e., the column density of an absorbing molecule, excitation temperature (the same for rotational and vibrational components), and a radial velocity of the slab (no velocity gradient was implemented). Finally, to the resultant spectrum a photospheric contribution of a B3V star was added (a synthetic spectrum of T_eff = 18,000 K and log g = 4.0 was used for this purpose, see Section 6).

It should be noted that the above approach was not meant to reproduce the observed spectrum in the whole spectral range. Simulations were performed separately for different narrow spectral regions and usually for one molecule at once. It allowed us to identify and study individual bands dominating in a given spectral range. With this modeling procedure, rough estimates of physical parameters of the outflowing material (excitation temperature, column densities, radial velocity) were also possible in a number of cases.

A list of the identified molecular bands is presented in Table 3. The first column of the table gives observed wavelengths of features. In most cases this is the wavelength of a band head formed by the R-branch, but in a few cases this corresponds to a local minimum of the absorption feature formed by Q- or P-branches; otherwise an empty space is left. Column 2 contains laboratory wavelengths of the features. Columns 3–5 specify the molecule name, the electronic system with a vibrational band identification, and the branch designation, respectively. Column 6 gives a heliocentric velocity of the feature, if a clear band head is observed. References to appropriate molecular data and comments on individual features can be found in the last two columns. Most of the identified molecular features are also indicated on the spectrograms in Figures 1–3.

The spectrum is dominated by strong TiO absorption bands, which are present in all parts of the HIRES spectrum. Prominent are also bands of VO, AlO, ScO, and YO. In the following subsections, we describe in some detail the identified features for each of the molecules.

5.1. TiO

Much effort was devoted to understand electronic transitions of TiO in laboratory (Phillips 1973), in astronomical observations (Ram et al. 1999), as well as in theoretical computations (Schwenke 1998). In the identification procedure, we mainly consulted databases of line positions computed from laboratory measurements (Jørgensen 1994; Plez 1998; Schwenke 1998). These line lists also provide data for bands not measured in laboratory and in these cases discrepancies between different sources are often significant. The lack of accurate spectroscopic data disables a detailed analysis of some TiO bands, e.g., the γ' (3, 0) band close to 5350 Å and higher overtones of γ' bands.

We have identified three electronic systems of TiO in the spectrum of V838 Mon, i.e., α (C³Δ − X³Δ), γ (A³Φ − X³Δ), and γ' (B³Π − X³Δ). The molecular bands of the γ and γ' systems have generally a characteristic triple structure formed by their main branches. In the case of the α transitions, the separation of branches is so small that the three components overlap and the triple structure is not seen. Additionally, in some bands of the γ and γ' systems, relatively strong satellite branches are observed, e.g., the R₃₁-branch of the γ' (1, 0) band at 5827 Å. Finally, we have also identified forbidden transitions of TiO in the spectrum. The latter point is discussed in Section 5.6.

The observed features of TiO arise both in the stellar photosphere and in the outflow. Although the latter component dominates formation of the observed bands, it would be impossible to satisfactorily reproduce a number of the observed features without a contribution from the photosphere. Unfortunately, we have not been able to set valuable constraints on the effective temperature, since there are only a few regions in the spectrum, where the stellar component dominates. Performed simulations show, however, that the selected atmospheric model with T_eff = 2400 K works good in reproducing high excitation features seen in the spectrum (see below).

Nonetheless, the most prominent absorptions of TiO seen in the spectrum of V838 Mon arise in a low excitation environment. A comparison between the observations and spectra computed from our slab model with different excitation temperatures suggests a mean value of T_ex = 500 ± 100 K. Adopting this temperature and comparing the observed spectrum to the simulations of the γ' (2, 0), (1, 0), (0, 0), and γ (2, 0) bands, an estimate of the TiO column density of ∼16.8 dex cm⁻² has been obtained. Assuming that all Ti is locked in TiO and taking a Ti atomic abundance of 10⁻⁷ relative to hydrogen, a column density of hydrogen of ∼23.8 dex cm⁻² has been derived.

A characteristic feature of the observed TiO spectrum is a strong saturation of bands originating from the ground vibrational state. The effect is clearly seen in the γ' (0, 0) and (1, 0) bands. In the γ' (2, 0) band, the saturation is weaker and completely disappears in the (3, 0) band. Similarly, a strong saturation is seen in the α (0, 0), (1, 0), (2, 0) bands, and probably also in the (3, 0) and (4, 0) bands. In the red part of the analyzed spectrum, the γ (0, 0) and (1, 0) bands are saturated. Due to the strong saturation in the γ and γ' systems, the triple structure of the bands can be seen only through weak emission-like features that appear between the saturated components, see, e.g., the γ (0, 0) and (1, 0) bands at 6670 Å and 7100 Å (Figure 3).

Radial velocities of some TiO bands can be determined by comparing rotational features of the observed bands with simulated spectra. Although the rotational structure is completely smoothed in the observed band heads, it is recognizable in those parts of the spectrum, where only absorptions from higher rotational levels are present. Two examples of such rotational features are shown in Figure 8, where a simulated spectrum is overplotted on the observed one. The cold slab had a excitation temperature of 500 K in the simulations. Note however that the contribution from the slab is negligible in the γ(3, 1) and (4, 2) bands. To fit the rotational components, the modeled spectrum has been shifted to V_h = 58 ± 5 km s⁻¹. Since these rotational features coming from high excitation levels are predominantly formed in the atmosphere of the central star, the derived velocity is expected to be close to the radial velocity of the star.

Zoom In Zoom Out Reset image size

Radial velocities can also be determined from the observed positions of band heads. In the case of the TiO bands present in our spectrum, the formation of a band head is often strongly influenced by the presence of satellite bands and a non-negligible isotopic shift in wavelength. An example is presented in panel (a) of Figure 9. This head, belonging to the α (2, 0) band, is formed by the R₂-branch and a contribution from satellite bands is small in this case. However, as can be seen in the rotational profile⁹ overplotted on the spectrum in Figure 9, the head is actually formed by multiple components. The strongest ones belong to the three main TiO isotopomers, i.e., ⁴⁶Ti¹⁶O, ⁴⁷Ti¹⁶O, and ⁴⁸Ti¹⁶O. For solar Ti isotopic ratios, i.e., 9.3, 10.1, 13.4, 13.8 for 48/46, 48/47, 48/49, 48/50, respectively (Cowley 1995), all the isotopic components should significantly contribute to the shape of the α (2, 0) band head. Since they are not distinguishable in the observed smooth head, only a position of the most blueshifted component belonging to the ⁴⁶TiO isotopomer can be measured with a reasonable accuracy. In this way, a velocity of V_h = −67 km s⁻¹ has been inferred.

Zoom In Zoom Out Reset image size

Another example of a TiO complex band head is shown in panel (b) of Figure 9. As seen in the rotational profile, the structure of the γ' (1, 0) band head is even more elaborate than in the previous example. This is due to the presence of multiple heads formed by satellite branches. The observed head is again very smooth and individual components, clearly distinguishable in the overplotted rotational profile, are not seen in the observed spectrum. An estimate of the outflow velocity is very uncertain in this case, but by fitting the rotational profile to the overall absorption profile of the band, a velocity of V_h = −58 ± 10 km s⁻¹ has been derived.

As can be found in Table 3, the most blueshifted molecular component is observed in the α (0, 0) band at 5164.6 Å. A velocity of −125 km s⁻¹ has been derived assuming that the blue edge of the head is formed by the ⁴⁶TiO isotopomer, instead of the most abundant ⁴⁸TiO. A similar velocity was found for the α (1, 0) band at 4952.6 Å.

The derived radial velocities of TiO features are in a range between 58 km s⁻¹ and −125 km s⁻¹. It is reasonable to assume that the molecular absorption arises in an outflow with a velocity gradient. This can explain different velocities of the bands with different excitation requirements, and can also account for the observed smooth structure of the band heads.

5.2. AlO

We based our identification procedure of the AlO bands on an analysis of the B²Σ⁺–X²Σ⁺ system performed in Coxon & Naxakis (1985). Absolute strengths of the electronic transitions ware taken from Partridge et al. (1983).

The blue–green (B²Σ⁺–X²Σ⁺) band system of AlO is evidently present in the analyzed spectrum. The most prominent feature is the saturated (0, 0) band with a main head at 4842 Å, see panel (c) in Figure 9. Three band heads of different branches are close enough to form this relatively sharp head. In this case, a radial velocity can be measured accurately and we have found V_h = −82 ± 6 km s⁻¹.

Higher vibrational overtones, Δv = +1 and Δv = +2, of the B − X system can also be recognized in the spectrum of V838 Mon. The more intense Δv = +1 overtone is seen up to the (6, 5) band around 4754 Å. The radial velocity of the band heads of the first overtone cannot be defined accurately due to the poor quality of the spectrum in the blue range. However, the wavelengths of bands from higher vibrational levels imply a velocity of V_h ≃ 58 km s⁻¹. The Δv = +2 overtone is only hardly seen in the spectrum. The most intense (2, 0) band is contaminated by the He i 4471 Å line of the B3V star.

In the green part of the spectrum, vibrational bands belonging to the Δv = −1 sequence are clearly visible. The bands are shown in Figure 10. The (0, 1) band at 5079 Å is particularly strong. Its band head is saturated and shifted to V_h = −82 km s⁻¹. One can also distinguish other members of the same overtone originating from higher vibrational levels up to v'' = 4. According to our simulations, these bands, to be observed, require a rather high excitation temperature of about 1500 K, which suggests that they originate in, or close to, the atmosphere of the star. However, sharp band heads belonging to transitions with v'' ⩽ 3 are shifted to a velocity V_h = −82 km s⁻¹, which is inconsistent with the stellar velocity. The velocity rather indicates that the absorbing gas is located in outer parts of the outflow. Note that this velocity is almost the same as that of the narrow absorbing component seen in several atomic lines, e.g., Rb i (see Section 4.2).

Zoom In Zoom Out Reset image size

5.3. VO

The spectrum of V838 Mon shows strong vibrational bands of the C⁴Σ⁻– X⁴Σ⁻ and B⁴Π– X⁴Σ⁻ systems of VO. For the purpose of our simulations, a line list for these systems was computed using spectroscopic constants from Cheung et al. (1982). Rotational strength factors were taken from Kovács (1969) with corrections applied after Whiting et al. (1973). In the case of the B − X (0, 0), (1, 0), and (2, 0) bands, a list of energy levels was generated by diagonalization of the perturbed Hamiltonian from Adam et al. (1995) and Cheung et al. (1994). The transformation matrices were next used to compute rotational strengths of individual lines to properly include effects of perturbations. The rotational structure of the (2, 0) band was not analyzed in laboratory. Therefore, spectroscopic parameters for this band were extrapolated from the lower overtones. Electronic strengths were taken from Karlsson et al. (1997). Franck–Condon factors were computed with the LeRoy's codes¹⁰ (Le Roy 2004, 2007).

As can be seen in panel (d) of Figure 9, the unsaturated C − X (0, 0) band has two heads. One is formed by the R₁- and R₄-branches, while the other, redshifted by 1 Å, is formed by R₂ and R₃. Both are present in the observed spectrum and the former one appears at a velocity of −77 km s⁻¹. A regular ladder of rotational lines of the (0, 0) band formed by main P-branches can be recognized starting at 5753 Å (see Figure 2). Fortunately, precise laboratory wavelengths are available for these lines and a detailed modeling of the band was possible. The simulations suggest that this part of the spectrum is formed at a low excitation temperature of ∼300 K. The derived velocity of the absorbing gas in the rotational transitions of the (0, 0) band is V_h = 43 km s⁻¹. From the heads of higher transitions of the C − X system, namely the (1, 0) and (2, 0) bands, velocities of −77 km s⁻¹ and −45 km s⁻¹ were derived, respectively.

Additionally, within the B − X system, the bands (0, 0), (1, 0), and (2, 0) were identified. The (0, 0) and (1, 0) bands, together with their simulated spectra, are presented in Figure 11. Most of the features of these bands are only slightly shifted from their laboratory wavelengths and the measured velocity is V_h = −2 km s⁻¹. A rather uncertain value of ∼350 K was deduced from the simulations as an excitation temperature of the absorbing gas (mainly from the intensities of absorption features at 7910 Å and 7920 Å). Furthermore, in order to reproduce some features seen in the spectra of the bands with higher vibrational numbers, e.g., the absorption feature at 7474 Å belonging to the (2, 1) band, it was necessary to add an additional layer of absorbing VO gas with a rather high excitation temperature, namely, T_ex=1500 K. This hotter component, treated in the simulations as a slab between the stellar photosphere and the cold slab, appears at a velocity of ∼58 km s⁻¹.

Zoom In Zoom Out Reset image size

Our simulations of the C − X (0, 0), B − X (0, 0), and (1, 0) bands give satisfactory fits to the observations when a column density of the VO absorber is of ∼16.4 dex cm⁻². If we assume that all V is locked in VO molecules and taking the solar abundance of V relative to H of 10⁻⁸, this corresponds to a column density of H of ∼24.4 dex cm⁻². The result is consistent with a similar estimate performed from the TiO bands in Section 5.1.

5.4. ScO

The only electronic system of ScO present in the spectrum is the A²Π–X²Σ⁺ system. Absorption features of the (0, 0), (1, 1), (2, 1) bands are clearly seen in the green part of the spectrum. Moreover, one band of ScO, i.e., the (0, 1) band, was found in emission. For the purpose of our simulations, line positions of the A − X (0, 0) band were computed from the Hamiltonian diagonalization parameterized by Stringat et al. (1972) and Adams et al. (1968). Higher vibrational levels were approximated from data in NIST.¹¹ For the positions of transitions of the (1, 0) band, Rice & Field (1986) was consulted. Strengths of rotational lines were found in the approach proposed in Herbig (1974). It should be noted that although the line positions of the (0, 0) band are exact their strengths are only approximate.

Figure 12 shows a region of the observed spectrum covering two subbands of the (0, 0) band. Simulations of this spectral region suggest the presence of at least two different velocity components. The band head formed by the ²Π_3/2^RQ_2G + ^RR_2G (at 6036 Å) and ²Π_1/2^QQ_1G + ^QR_1G (at 6079 Å) branches are shifted to −55 km s⁻¹ (as measured for the lowest part of the profile). The second velocity component at 45 km s⁻¹ was found from the position of the head formed by the ²Π_1/2^RR_1G branch at 6064.17 Å. This latter head is formed at a rather high rotational number, i.e., N = 58, contrary to the heads observed at −55 km s⁻¹, which are formed at N = 28 and N = 17. The relative velocity shift can be understood as a result of decreasing population of high rotational levels with increasing velocity in the observed outflow. A lack of accurate laboratory wavelengths disables an analysis of other absorption bands of ScO.

Zoom In Zoom Out Reset image size

Remarkable is the presence of the A − X (0, 1) ²Π_3/2^RQ_2G and ²Π_1/2^QQ_1G subbands of ScO in emission at 6408 Å and 6458 Å, respectively (see the bottom panel of Figure 8 for the latter subband). The observed velocity of this features agrees well with the velocity of emission components seen in the atomic P Cygni profiles (see Section 4.2). The origin of this molecular emission can be explained by fluorescent pumping of the upper electronic level by absorption in the (0, 0) 6090 Å and 6132 Å subbands, clearly seen in the spectrum of V838 Mon.

5.5. YO

All the features of YO identified in the spectrum of V838 Mon belong to the A²Π–X²Σ⁺ system. A list of the transitions used in our identification procedure was prepared on the basis of data provided in Bernard & Gravina (1983). There is another well-known system of YO, i.e., B²Π − X²Σ⁺ around 4819 Å (Bernard et al. 1979), but it falls into a spectral region of saturated TiO bands and was not found in our spectrum.

We performed a detailed simulation of the A²Π_3/2–X²Σ⁺ transition. The results of this simulations are shown in Figure 13. The upper two plots in Figure 13 are rotational profiles of the transition obtained for two excitation temperatures, 750 K and 500 K. It can be seen that a sharp edge of the (0, 0) band head at 5973 Å is formed with the higher temperature. In the observed spectrum this head is very smooth and resembles more the feature seen in the profile for 500 K, and this is a value we took for further simulations. It should be noted, however, that the observed profile of the band can be influenced by the physical structure of the wind, which is ignored in our simulations; in particular, a velocity gradient in the molecular outflow can smooth the shape of the band heads, and even if the temperature is high, no sharp head would be observed. Nonetheless, to perform full simulations of the A²Π_3/2–X²Σ⁺ transition of YO it was necessary to include the γ (3, 0) band of TiO as an additional important absorber in this spectral range. The green line in Figure 13 shows the TiO contribution to the observed spectrum. The final simulation which includes absorption of YO and TiO with an excitation temperature of 500 K is shown at the bottom of Figure 13. From the overall fit of the (0, 0) band to the full simulation, a velocity of V_h = −8 km s⁻¹ has been derived.

Zoom In Zoom Out Reset image size

Note however that a sharp (1, 1) band head formed by the ^RQ₂₁-branch at 5988 Å appears at higher excitation temperatures, at least 750 K, as can be seen from Figure 13. Such a sharp feature seems to be present in the observed spectrum, suggesting that this absorption feature is formed in gas with a somewhat higher temperature than the one found for the (0, 0) band head. This is confirmed by the measured velocity of the (1, 1) band head, V_h = 58 km s⁻¹, which is the same as that of other high-excitation features found in the spectrum.

5.6. TiO—forbidden transitions

Very interesting from the spectroscopical point of view is the presence of forbidden transitions of TiO, i.e., the c¹Φ–X³Δ (0, 0) band around 4737 Å and the b¹Π–X³Δ (0, 0) band at 6832 Å. The stronger b − X band is shown in detail in Figure 14. The intensities of the bands were predicted in ab initio calculations of rovibrational energy levels of the 13 lowest electronic states of TiO by Schwenke (1998). As shown in Figure 14, our simulation can reproduce the forbidden bands quite well. The best fit indicates a velocity of V_h = −80 km s⁻¹ and an excitation temperature of about 500 K. However, when the column density found for other TiO bands (see Section 5.1) is assumed, oscillator strengths of the b¹Π–X³Δ band must be enhanced by a factor of 70 ± 35 in order to reproduce the observations. To our knowledge this is the first, in astrophysics as well as in laboratory, observation of TiO forbidden bands.

Zoom In Zoom Out Reset image size

A number of broad absorption lines of the Balmer series, from Hβ down to at least H12, as well as a few He i lines, can clearly be seen in the blue part of our spectrum (Figure 1). These features are attributed to the B3V companion. We compared the observed spectrum to a grid of synthetic spectra BSTAR2006, generated with the Synspec program (Lanz & Hubeny 2007). Solar metallicity models, with effective temperatures between 15000 and 30000 K and gravities in the range log g = 2.0–4.75, were considered. A turbulent velocity of 2 km s⁻¹ was adopted. Good fits to the observed spectrum were obtained for T_eff = 18000 ± 2000 K and log g = 4 ± 0.75. The low signal-to-noise ratio in the part dominated by the B3V star and uncertainties in the flux calibration did not allow us to obtain more precise estimates of the star parameters. In any case, the above values agree well with those expected for a B3V star. In the following discussion, we use the synthetic spectrum with T_eff = 18000 K and log g = 4.0. This spectrum, broadened with a rotational velocity of 250 km s⁻¹ (see below) and shifted to a radial velocity found for the hot star (see below), is overplotted on the observed spectrogram of V838 Mon in Figures 1–3.

A comparison of the observed spectrum to the synthetic one shows that the absorption lines of the B3V companion are rotationally broadened. In order to find the rotational velocity, we convolved the synthetic profiles with a Gaussian profile broadened with a grid of rotational velocities (we made use of the rotin3 program¹²). Note that this procedure does not take into account limb darkening nor gravitational darkening, which can be important for profile formation in a rapidly rotating star. Lines usually used to investigate rotational broadening of B-type stars, i.e., He i 4471 Å and Mg ii 4481 Å, are weak, noisy, and affected by emission lines and/or molecular bands in our spectrum. However, we have found that in the green part of the spectrum the He i 5876 Å line is relatively strong and has a "clean" profile. This green spectral region is dominated by the cool photosphere but, fortunately, in the spectral vicinity of the λ5876 line the flux from this component is practically reduced to zero due to a saturated TiO band (see Figure 2) and the line from the B3V component is well seen. Thus, when determining the rotational velocity of the B3V component we relied mainly (but not only) on the profile of He i 5876 Å. Using the T_eff = 18000 K, log g = 4.0 template, we found a projected rotational velocity Vsin i = 250 ± 50 km s⁻¹. This is a rather large value¹³ and indicates that the rotation axis is close to the plane of the sky.

We also attempted to measure the radial velocity of the hot star. We used the Fourier cross-correlation method (Tonry & Davis 1979) as it is implemented in the IRAF's task rvcor. As a template, we used our synthetic spectrum broadened to the rotation velocity of 250 km s⁻¹. The cross-correlation was performed for the region between 3815 Å and 4180 Å, with all emissions and ISM absorptions in this range removed manually. (We performed tests with ranges including Hβ and Hγ, but they did not give reliable results. These spectral regions are too strongly contaminated by atomic emissions and molecular absorptions. Also, including the very noisy shortward part of the spectrum spoils the cross-correlation function and gives unrealistic results.) As a result, we obtained a heliocentric velocity of 37.0 ± 16.8 km s⁻¹ (with a Tonry & Davis 1979 r value of 18.0), or correspondingly 19.6 ± 16.8 km s⁻¹ in the local standard of rest. The error value was taken directly from the output of the rvcor procedure and, in our opinion, it is slightly overestimated.

Several interstellar absorption features can be identified in our spectrum. The strongest ones are those of Ca ii H, K and Na i D₁, D₂. They are shown in Figure 15. Interstellar components can also be found in the deep P Cygni absorption of K i at 7699 Å  (see the upper panel in Figure 6). The Na i and K i interstellar absorptions in the spectrum of V838 Mon have already been studied in a number of papers (e.g., Kipper et al. 2004; Munari et al. 2005). As can be seen from Figure 15, the profiles of the Ca ii lines, not described in the literature so far, have a similar structure to those of Na i. All the four profiles displayed in Figure 15 show three components at 32, 42, and 65 km s⁻¹, similarly as observed in Munari et al. (2005). We note, however, that unlike previous studies, where the interstellar lines were observed in the spectrum of V838 Mon itself, the features presented in Figure 15 are observed in the spectrum of the B3V companion. As can be seen from Figures 1 and 2, the local continuum near the discussed lines comes from the B3V companion. At the wavelengths of the Ca ii lines V838 Mon is too cool to emit any significant radiation, while at the wavelengths of the Na i lines an absorption band of TiO is so strong that it eats up practically all the flux from the cool supergiant. The fact that the interstellar line profiles in the spectrum of the B3V companion are very similar to those in the spectrum of V838 Mon itself is consistent with the idea that both stars are related, at least that they are at a similar distance.

Zoom In Zoom Out Reset image size

We have identified six diffuse interstellar bands (DIBs) in our spectrum, i.e., at λλ4428, 5780, 5797, 6284, 6614, and 7224 (the central wavelengths are taken from Herbig 1995). They all are redshifted to a heliocentric velocity of ∼60 km s⁻¹. It is well known that equivalent widths of the features at 5780 and 5797 Å correlate very well with interstellar extinction. Using conversion formulae from Herbig (1975), we derived E_B−V of 0.95 and 0.90 for the DIBs at 5780 Å and 5797 Å, respectively. Taking the relation from Herbig (1993), which is based on a larger number of data points, gives E_B−V ≃ 1.00 for both features. Given considerable uncertainty of these estimates, due to uncertainties in the equivalent width measurements and the intrinsic scatter in the relations used, our result is consistent with E_B−V ≃ 0.9 obtained in other studies (as summarized in Tylenda (2005)).

We have obtained a spectrum of V838 Mon on 2005 October 13 using the Keck/HIRES instrument. The spectrum of a resolving power of ∼34,000 covered a wavelength range of 3720–7960 Å. The spectrum is very complex and shows numerous spectroscopic features which can be identified as arising from different components or environments. V838 Mon itself, i.e., the object which erupted in 2002, is seen mainly in the green and red parts of the spectrum as a cool continuum with strong, broad, and complex molecular absorptions. Numerous atomic lines showing P Cygni profiles or broad absorption features are also present in this component. The blue part of the spectrum is dominated by a photospheric spectrum of the B3V companion, showing broad absorptions lines of H i and He i. In the blue and green parts of the spectrum, numerous emission lines can be identified. These mostly arise from forbidden transitions in singly ionized iron. Finally, several absorption features of interstellar origin have been identified. These include atomic resonance lines as well as DIBs.

The numerous [Fe ii] emission lines observed in our spectrum evidently arise from rarefied and (partly) ionized matter. All the lines show the same profile, which can be well fitted with a Lorentzian profile. The lines are centered at a heliocentric radial velocity of ∼13 km s⁻¹ and have a typical FWHM of ∼80 km s⁻¹.

In the cores of the photospheric absorption lines of the B3V companion we see weak emission features. The double feature in the Hβ line can be interpreted as a single emission with a narrow absorption superimposed on it. The derived parameters of the emission component are close to those of the [Fe ii] lines. This suggests that the Balmer emission features arise from the same environment as the [Fe ii] emission lines.

The spectrum displays a number of strong atomic, mostly resonance, lines with P Cygni profiles. Evidently, V838 Mon, being almost four years after its outburst, continues losing mass. The emission component typically peaks at a heliocentric velocity of ∼80 km s⁻¹, while the absorption component usually extends from ∼50 km s⁻¹ to ∼−100 km s⁻¹. The absorption component often shows a complex profile. In particular, in a few lines a strong and NAC is clearly seen at a velocity of ∼−80 km s⁻¹. A few atomic lines were observed to show pure absorption profiles.

The spectrum from the cool component is dominated by numerous, often very strong, molecular bands in absorption. Some bands are so deep that the flux from the cool component practically goes to zero. In these cases, we have a sort of spectral windows, through which we can see the B3V companion, even in the red part of the spectrum. All the identified bands are from oxides. They include TiO, VO, AlO, ScO, and YO.

Our analysis shows that the bands are formed in different regions. Some of them, especially those from highly excited states, are formed, at least partially, in the photospheric regions. The temperature in these layers is as high as ∼2400 K. These bands show a heliocentric radial velocity of ∼58 km s⁻¹, which we identify as a radial velocity of V838 Mon itself.

Most of the bands are formed in regions significantly cooler than the photospheric ones. The vibrational temperatures derived from some bands go down to ∼500 K. These are usually very strong bands arising from ground states. The radial velocity of these bands is usually much lower than that of the photospheric regions and goes down to ∼−125 km s⁻¹. Overall, our analysis of the molecular bands shows that they are mostly formed in outflowing matter with a significant velocity gradient.

We have detected bands arising from forbidden transitions in TiO. As far as we know, these are first observations of TiO forbidden transitions in astrophysical objects.

The observed spectrum of the B-type companion, seen primarily in the blue part of our spectrum, can be fitted with a model atmosphere spectrum with solar abundances, an effective temperature of ∼18000 K, and a gravity, log g ≃ 4.0. These values agree well with those expected for a B3V star. The absorption lines are significantly broadened, which we interpret as a rotational broadening with a velocity, V sin i ≃ 250 km s⁻¹. We have derived a heliocentric radial velocity of the B3V companion of 37 ± 17 km s⁻¹.

Several interstellar absorption features are seen in our spectrum. They include lines of Ca ii, Na i, and K i, as well as DIBs. The lines of Ca ii and Na i are observed in the spectrum of the B3V companion. Their profiles are very similar to those observed in the spectrum of V838 Mon during the 2002 outburst. This result is consistent with the idea that V838 Mon and the B3V companion are related objects. From equivalent widths of DIBs we have estimated an interstellar reddening of E_B−V ≃ 0.9–1.0, which agrees well with interstellar extinction values obtained by other authors using other methods.

A detailed analysis and interpretation of the data presented in the present paper, as well as of data from other sources, will be done in a forthcoming paper.

M.S., R.T., and T.K. were supported by the Polish Ministry of Science and Higher Education under grant N203 004 32/0448, for which they are grateful. T.K. thanks T. Tomov for his helpful advises on data reduction and spectrum analysis with IRAF. M.K. was supported by the Foundation for Polish Science through a FOCUS grant and fellowship, by the Polish Ministry of Science and Higher Education through grants N203 005 32/0449 and 1P03D 021 29. M.G. acknowledges support by Polish KBN grant 1P03d 017 27 and SALT International Network grant 76/E-60/SPB/MSN/P-03/DWM 35/2005-2007. This paper uses observations made at the South African Astronomical Observatory (SAAO). The authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of Mauna Kea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain.

Facilities: Keck:I (HIRES), Radcliffe (Grating Spectrograph)