AN HST COS "SNAPSHOT" SPECTRUM OF THE K SUPERGIANT λ Vel (K4Ib-II)^*

The Hubble Space Telescope (HST) SNAPshot program "SNAPing Coronal Iron" (GO Program 11687) aimed to gain further understanding of stellar coronae by searching for highly ionized Fe coronal lines in a variety of representative cool stars. The Cosmic Origins Spectrograph (COS) on board HST was used with its far-ultraviolet medium-resolution grating G130M (R ∼ 20,000) to record the Fe xii λ1349 and the Fe xxi λ1354 coronal lines as diagnostics for environments with temperatures in the range of T ∼ 2 × 10⁶ K and T ∼ 10⁷ K, respectively.

The SNAP targets included the K4 Ib-II supergiant λ Vel (HD 78647). Since λ Vel is not a known coronal X-ray source, the expectation was that the Fe xii and Fe xxi lines would not be detectable. This was confirmed by this observation and λ Vel instead serves as a comparison object to reveal the possible spectral structure in the vicinity of the coronal lines, which might be misinterpreted as hot emission in an otherwise inactive object. An example is the C i λ1354 chromospheric emission, which is blended with Fe xxi in stars for which both features are seen. In this paper, we do not address the presence or absence of coronal plasma but use the spectrum to study the wind and chromosphere rather than a likely very weak corona. The spectrum contains a wealth of diagnostics to investigate λ Vel's outer atmosphere and wind, including thermally and radiatively excited atomic emission, fluorescent atomic and molecular transitions originating from the chromosphere, and overlying absorptions from the outflowing wind. The C ii λ1335, 1336 (UV 1) resonance lines are detected in the observed spectrum and show strong wind-induced P-Cygni profiles.

Only one of the two COS detector segments could be used during the observation due to the extreme brightness of H i Lyα, which would have saturated and possibly damaged detector segment B. With a single detector segment, covering only 140 Å, a remarkable high signal-to-noise spectrum was obtained in a mere 20 minute exposure.

Lambda Vel is a close spectral-type match to ζ Aur A, a star for which the properties of the atmosphere have been spatially probed with eclipse spectra (Wright 1970; Griffin et al. 1990; Baade et al. 1996). IUE spectra revealed that the chromospheric heating of ζ Aur (evolved star) primaries is roughly equivalent to that in single stars of similar spectral-types (Eaton 1992). A comparison of higher signal-to-noise and spectral resolution HST UV eclipse spectra of ζ Aur A with the spectrum of λ Vel shows that both the chromospheric heating and turbulence are closely matched, though the wind acceleration appears slower than single stars of the same spectral type (Harper et al. 2005). Lambda Vel therefore is an important system for detailed study because it is free of the effects of the binarity that obscure the acceleration of the stellar wind in systems like ζ Aur A.

This paper presents an overview and analysis of the λ Vel COS spectrum. Section 2 presents the data that were used in the analysis and the fundamental stellar parameters. Section 3 describes λ Vel's far-ultraviolet spectrum and compares it to similar objects with a higher surface gravity (α Boo) and lower effective temperature (α Ori). Section 4 discusses the line analysis and Section 5 the inferred properties of the chromosphere and wind as inferred from the COS spectrum, with reference to the archival Goddard High Resolution Spectrograph (GHRS) spectra. Finally, Section 6 provides a summary.

The COS observation of λ Vel was a single detector segment, 1350.38 s exposure covering 140 Å (1326.0–1467.0 Å) of the far-ultraviolet spectrum with the PSA/G130M aperture/grating.⁵ The COS instrument and its performance are described by Osterman et al. (2011) and Green et al. (2012). The standard pipeline processed spectrum shows a wealth of details at high signal-to-noise (S/N) and complements previous HST GHRS observations taken in 1994 September, which sampled ⩽40 Å wide regions shortward and longward of the observed COS interval. The COS data provide key information on the continuum flux level over a broader wavelength range in the far-ultraviolet than the more restricted GHRS windows alone and supply a variety of diagnostics of the chromospheric and wind conditions in this object. An overview of the COS spectrum can be found in Figure 1.

Zoom In Zoom Out Reset image size

Measurements of features in the λ Vel spectrum revealed a systematic trend of measured radial velocities (RV) as a function of wavelength, as illustrated in Figure 2, where we show our most reliable measurements of a set of unblended, high S/N atomic (chromospheric) emission lines that are not contaminated by strong overlying wind absorption (e.g., the C ii and Si iv doublets near 1335 Å and 1400 Å are not used here). The properties of the radial velocity trend suggested that it was instrumental rather than astrophysical, a subtle effect not being fully accounted for in the standard STScI pipeline data reduction. We have computed a linear least-squares fit to the trend to enable a correction of the observations.

Zoom In Zoom Out Reset image size

We applied the apparent wavelength trend, defined by the linear fit, to remove the un-physical variation in radial velocities from the measurements, although scatter about the linear fit limits the precision of the correction. We have computed, to characterize this precision, 1σ errors for the slope and intercept of the fit (0.019 km s⁻¹ and 26.6 km s⁻¹, respectively) and for the corrected radial velocities (4.8 km s⁻¹). Our approach does not allow us to determine a correction for the zero point of the velocity scale, but we note that our correction is zero at 1341.48 Å, the crossing point of the linear least-squares fit (using RV_fit = −0.1808 × (λ−1341.48), where RV is in km s⁻¹ and wavelength in Å). We corrected the measured radial velocities by subtracting the fit from the measurements, a process that thus sets the mean of the atomic emission line velocities used to define the correction to zero. The correction allows us to look for more subtle velocity signatures in the COS data, although precision velocity analysis, especially with regard to absolute velocities, remains the provence of the Space Telescope Imaging Spectrograph (STIS), as COS was designed for maximum sensitivity rather than highest resolution or wavelength precision.

Carpenter et al. (1994) assessed basic stellar parameters for λ Vel, finding an effective temperature of 3800−4000 K using VJK colors. The uncertainty was mainly attributable to uncertain reddening but was in agreement with the earlier work of Blackwell & Shallis (1977) that provided an estimate of 3820 K. Carpenter et al. (1994) also derived a mass and radius for λ Vel of 7 M_☉ and 210 R_☉, respectively. Formally, these values implied a luminosity log(L/L_☉) = 3.9 and surface gravity log g = 0.6.

In this paper, we compare the observed spectrum of the K supergiant λ Vel to that of the higher surface gravity K giant α Boo and to the spectrum of the lower effective temperature M supergiant α Ori to illustrate how the components of the spectra change with fundamental stellar parameters.

Alpha Boo (K1.5 III) is one of the brightest stars in the northern sky. Griffin & Lynas-Gray (1999) constructed an empirical energy distribution for the object based on previously published scans. They modeled the energy distribution with Kurucz models (Kurucz 1992) and derived fundamental parameters for α Boo of T_eff = 4290 K and log g = 1.9. More recently, Ramírez & Allende Prieto (2011) undertook a comprehensive analysis of α Boo. Based on model fits of visible to mid-IR spectrophotometric data, they derived T_eff = 4286 K. A mass (1.08 M_☉) and radius (25.4 R_☉) for α Boo followed from isochrone fitting and previously reported angular diameter and distance measurements, respectively; which in turn implied a surface gravity of log g = 1.66. In addition, Ramírez & Allende Prieto (2011) derived [Fe/H] = −0.52 for α Boo, in good agreement with the −0.5 dex first noted by Griffin & Griffin (1967), versus [Fe/H] = +0.23 for λ Vel, a difference of ∼0.75 dex.

Lobel & Dupree (2000) derived T_eff = 3500 K and log g = −0.5 for α Ori based on spectral synthesis calculations at near-infrared wavelengths. These values are in agreement with earlier studies by Carpenter et al. (1994) and Blackwell & Shallis (1977). The accuracy of these basic stellar parameters for α Ori is limited by uncertain values for the mass, radius and distance. Harper et al. (2008) suggested that α Ori is further away than indicated by Hipparcos measurements and used the bolometric flux and angular diameter from Perrin et al. (2004) to obtain T_eff = 3650 K and log g = −0.26. The derived effective temperature of α Ori in various studies appears to converge toward ∼3600 K, while the surface gravity is more uncertain.

The basic stellar parameters of λ Vel and the comparison stars are presented in Table 1.

The far-ultraviolet spectrum of λ Vel is dominated by chromospheric and wind-scattered emission lines, as illustrated in Figure 1. Wind absorption is seen superposed upon the continuum and the strong chromospheric C ii emission lines. For reference as we proceed with our discussion of the components of the λ Vel spectrum, we show in Figures 3 and 4 a comparison of the COS λ Vel spectrum to HST STIS spectra of the K2 giant α Boo (K2 III) and the M2 supergiant α Ori (M2 Iab) to illustrate the change in the nature and character of this spectral region with luminosity and effective temperature, respectively.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Alpha Boo and λ Vel have similar effective temperatures, but α Boo has a considerably higher surface gravity and therefore a less-extended, denser outer atmosphere (see Section 2 and Table 1). Figure 3 compares the COS λ Vel spectrum with a STIS spectrum of α Boo from StarCAT (Ayres 2010), a catalog of high-resolution ultraviolet (UV) spectra of objects classified as "stars," recorded by Space Telescope Imaging Spectrograph (STIS) during its initial seven years of operations (1997–2004).

Lambda Vel and α Ori have low surface gravities and extended outer atmospheres, but α Ori has a cooler effective temperature and a more massive wind, as well as thick circumstellar shells not observed around λ Vel. The definitive far-ultraviolet observation of α Ori until recently was a low-resolution GHRS G140L spectrum (Carpenter et al. 1994), which enabled an initial understanding of this spectral region in the M supergiant as it made clear the existence of a strong far-ultraviolet chromospheric continuum over which a series of strong molecular absorption bands of carbon monoxide from its circumstellar shells was superposed. In α Ori, the chromospheric emission lines characteristics of cool stars are seen superposed on a continuum that is modified by absorption and re-emission in the CO bands. Recently, the ASTRAL: Cool Stars Program (GO 12278) obtained a high-resolution set of STIS observations of α Ori, including this region, which reveals the finer details (Carpenter et al. 2013; Nielsen et al. 2013). Figure 4 compares the COS λ Vel spectrum with the high-resolution STIS ASTRAL spectrum of α Ori.

The various components of the λ Vel spectrum are discussed in the subsequent sections. Line identifications have been made using the Kurucz databases⁶ (Kurucz 1995, 2005) with reference to the NIST Atomic Line Database⁷ and the Ultraviolet Multiplet Tables (Moore 1983 and preceding), the Arcturus UV Spectral Atlas (Hinkle et al. 2005), the IUE studies of Arcturus described in Ayres (1986), Ayres et al. (1986), and the GHRS studies of λ Vel (Carpenter et al. 1999). In addition to wavelength coincidence, we checked to make sure that all of the expected lines, based on intrinsic line strength (e.g., gf value), from a given energy level (lower if absorption, upper if emission) were seen if in the wavelength range of the observation. For the emission lines characterized as fluorescent, we also required that the exciting line and the excitation route be identified for the ID to be considered solid. Lines and excited levels above those populated in plasma with a temperature 20,000 K or lower were not considered unless a radiative excitation process was found to populate the level.

3.1. The Far-ultraviolet Continuum

The strong far-ultraviolet continuum observed in λ Vel is brighter than expected from the photosphere of a K supergiant. The upper panel of Figure 5 compares the observed surface fluxes of λ Vel with two synthetic continua computed with ATLAS9 (Kurucz 1993): one standard model and one in which a temperature minimum (T_min/T_eff = 0.8) is enforced for the uppermost layers. The optical data of Kiehling (1987) are well-matched by the models, but the observed ultraviolet continuum is more than two orders of magnitude above the synthetic spectrum at 2700 Å and more than six orders of magnitude at 1800 Å. The model with the flat temperature minimum matches the observed flux distribution to shorter wavelengths, but even this model falls well below the observations shortward of 2000 Å, implying the presence of a temperature rise above the photosphere. We have further characterized this temperature rise by computing the mean effective radiation temperature corresponding to the observed flux, from the optical down into the far-UV range covered by the COS and GHRS, by setting the surface flux at a given wavelength to πB (where B is the Planck function) and then solving for the corresponding temperature. The lower panel of Figure 5 shows this effective radiation temperature versus wavelength. This confirms the T_min of about 3200 K derived above using the modified Kurucz models and shows the chromospheric temperature rise as one looks to progressively shorter wavelengths, and thus higher altitudes, than included in the modified Kurucz model. The temperature at the longer wavelengths approximates, as expected, the effective temperature of the stellar photosphere. The observed ultraviolet continuum thus must form at chromospheric altitudes. Evidence of a chromospheric continuum in λ Vel was reported by Carpenter et al. (1999) using GHRS data. However, the GHRS wavelength coverage was mainly longward of 1800 Å, with one solitary observation near 1300 Å. The COS spectrum partly fills the gap in the GHRS sampling and confirms the extension of the bright continuum into far-ultraviolet wavelengths.

Zoom In Zoom Out Reset image size

The observed chromospheric continuum emission of λ Vel is reminiscent of the continuum observed in α Ori (Carpenter et al. 1994). However, the absence of the strong circumstellar absorption in the CO absorption bands seen in α Ori, makes the chromospheric continuum in λ Vel more uniform. Alpha Boo (K1.5 III), in contrast, shows a weaker far-ultraviolet continuum. This is consistent with the two supergiants having a greater mass column density (m_col) in their chromospheres than the giant stars, which shifts optical depth unity surfaces to larger radii in the supergiants in accordance with the scaling relation from Ayres (1979): m_col(τ)∝g^−0.5. The temperature of the chromosphere is increasing with height and the continuum flux rises (as the Planck function in the UV is very sensitive to the temperature) as one goes higher in the chromosphere. The chromospheric continuum in λ Vel, as in α Ori, is larger than in the giant star because the optical depth unity surface is at a greater radius and thus presents both a higher temperature and a larger emitting surface.

3.2. The Line Emission Spectrum

3.3. The Absorption Spectrum

We have measured the wavelengths and integrated fluxes of all identified emission features in the COS spectrum and list them in Tables 2 (atomic species) and 3 (molecules). A linear fit is used to quantify the local continuum around and under the emission features so that the continuum could be subtracted from the measurements. The measurements are thus of the net integrated line flux. The line centers and FWHMs were determined by fitting one or more Gaussians to each feature, a single Gaussian to clean, unblended lines and multiple Gaussians in the case of blends. We used the FUSE IDL procedure XGAUSSFIT⁸ to obtain these measurements and intrinsic IDL function GAUSSFIT to estimate the standard errors of those values. The velocity measurements of the emission lines have a mean standard error of 1.03 km s⁻¹, which reflects the accuracy of our ability to fit a line in this data set and is a good estimate of the errors in our relative velocity measurements of individual lines. As noted earlier, the absolute values are more uncertain due to the large COS science aperture, which limits the precision of the absolute velocities to ∼15 km s⁻¹, although relative velocities of groups of lines can be measured to much higher precision by computing means over large sets of lines and/or comparing histograms of the measured velocities of the various groups. Formally, mean velocities of groups of lines have smaller errors by a factor of 1/$\sqrt{N}$, where N = number of lines in the set. However, we choose to compare the distributions of the measured velocities of the different sets of lines by fitting Gaussians to their histograms and comparing the centroids of those fits to best determine their mean relative velocity. The standard errors in the measurements of the FWHM and net integrated flux of each line vary more from line to line, depending on the line strength, but the mean standard errors of the two are ∼2.7 km s⁻¹ and ∼2.0 × 10⁻¹⁵ erg cm⁻² s⁻¹, respectively.

Identifications and measurements of the prominent atomic absorption features are presented in Table 4. These absorption features were also fit with Gaussians, relative to the local continuum, as done for the emission features and with similar accuracy.

The apparent velocities of all the features are relative to the stellar photospheric radial velocity of 17.6 ± 0.3 km s⁻¹ (Gontcharov 2006), which we note is 0.4 km s⁻¹ smaller than used in the original Carpenter et al. (1999) λ Vel study. However, the radial velocities were corrected for the instrumental systematic trend discussed in Section 2, a process that results in the mean velocity of the chromospheric (atomic) emission lines (excluding the wind-impacted C ii and Si iv lines) being set to zero. The angular diameter used to compute the surface fluxes is 11.1 mas (Blackwell & Shallis 1977).

The distributions of the measured velocities for the atomic absorption and emission features are shown in the top panel of Figure 6, while the histograms of the measured velocities of the molecular emission and atomic emission features are given in the lower panel. We note that heavily blended features or features contaminated by overlying wind absorption, e.g., C ii and Si iv near 1335 Å and 1400 Å, are excluded from these samples. It appears that the atomic emission and absorption features belong to two different distributions, with the absorption components blueshifted relative to the emission spectrum by 7.6 km s⁻¹, implying formation in the lower levels of the outflowing wind. The histogram of the RV's of the molecular CO and H₂ emission features define a third population and appear to be blueshifted by 3.3 km s⁻¹relative to the atomic emission features, suggesting formation slightly above the chromospheric regions, in the lowest levels of the outflowing wind, and below the region in which the atomic absorption lines are formed. These values are all well below the terminal velocity of the wind reported by Carpenter et al. (1999) from earlier GHRS data and estimated below from the current COS data set by modeling the C ii P Cyg lines. In order to confirm these conclusions, we have performed a two-sample Kolmogorov–Smirnov (K-S) test on the two pairs of distributions to assess whether: (1) the "Atomic Emission" and "Atomic Absorption" line velocity distributions are from the same or different parent distributions and (2) the atomic and molecular emission line velocity distributions are from the same or different parent distributions. The K-S test run on the first pair gives K-S statistic of 0.404 and a probability of the two distributions being from the same parent distribution as 0.00083, indicating the two are indeed distinct and the velocity difference is meaningful. Similarly, a K-S test on the second pair gives a K-S statistic of 0.617 and a probability of the two distributions being from the same parent as 0.00008, again indicating the two distributions are from different parents and that the velocity difference is meaningful. In both the histograms and the K-S tests, we used the 38 cleanest (unblended, good S/N, uncontaminated by overlying wind absorption) atomic emissions lines, all 55 molecular emission lines, and the 18 cleanest atomic absorption lines.

Zoom In Zoom Out Reset image size

We have used the new COS spectrum, with reference to the archival GHRS spectra, to improve our characterization of the properties of the chromospheric regions and wind of λ Vel. The influence of the stellar wind is seen most clearly in the P Cygni profiles of C ii λλ1334,1335, in the blueshifted Fe ii and Ni ii wind absorption lines in the COS spectrum, and in the self-reversed Fe ii profiles in the near ultraviolet GHRS data. The wind lines studied previously indicate a turbulent velocity that varies with height and ranges from 9–21 km s⁻¹ and a terminal wind speed of ∼30 km s⁻¹ (Carpenter et al. 1999).

The presence of hot plasma in the atmosphere of the star, indicated by previous GHRS observations of Si iii] and C iii] lines at 1900 Å (Carpenter et al. 1999) and FUSE observations of O vi λ1032 Harper et al. (2005), is confirmed by the detection of Si iv λλ1398,1403 in this COS spectrum implying a plasma temperature of ∼8 × 10⁴ K. Both components of the doublet are contaminated by fluorescent H₂ emission making it difficult to measure a reliable flux for this doublet. The suggestion by Ayres (1986) that hot and cold material are intermingled in the chromosphere is supported by the fluorescent molecular lines visible throughout the λ Vel ultraviolet spectrum. Harper et al. (2005) argue that the presence of O vi must indicate magnetic heating, as acoustic heating of significant amounts of material up to 300,000 K is extremely difficult to achieve. We note that Grunhut et al. (2010) reported a 5σ magnetic field measurement for λ Vel, 〈B_z〉 = 1.72 ± 0.33 G.

We have made approximate, Sobolev with Exact Integration (SEI) fits (Lamers et al. 1987) to C ii λλ1334, 1335 (Figure 7) to derive the wind parameters, as demonstrated by Carpenter et al. (1999) using GHRS observations. The C ii line 1335 Å comprises of two albeit unequal lines, so C ii λ1334 is preferred for modeling purposes to estimate wind parameters that are comparable with the results from Carpenter et al. (1999). For this model, we assume the wind velocity is of the form

$$\begin{equation} \omega = \left(1-\frac{1}{x} \right)^{\beta }, \end{equation} \tag{ 1 }$$

where x = R/R_⋆, ω = v(R)/v_∞, and v_∞ is the terminal velocity of the wind, R_⋆ is the stellar photospheric radius, R is the radial height, and β is a constant that specifies the rapidity of the wind acceleration. It is assumed that x > 1, ω ⩽ 1, and R/R_ref > 1. R_ref is the height at which the wind is initiated (1.05 R_⋆ in our calculation) and the altitude from which we integrate (see Harper et al. 1995). The density is determined using conservation of mass and the assumption of a spherically symmetric, isothermal wind. The spectrum around the line is normalized to unity at the peak of the line to permit the SEI modeling, which works in this parameter space (see Carpenter et al. 1999 for further details). The continuum is fitted simultaneously with the line profile. The continuum on the red side of the 1334.5 Å line is depressed by the wind absorption of the 1335.5 Å doublet that lies to the red of the 1334.5 Å line. The terminal velocity (v_∞ ∼ 45 km s⁻¹) and turbulence (v_D ∼ 27 km s⁻¹) from the SEI fit are slightly larger than the values Carpenter et al. (1999) found using the Fe ii, Mg ii, and O i lines in the 1994 GHRS spectra noted above, but the turbulence value matches well that derived from C ii] in the GHRS data. The wind acceleration, however, appears significantly more rapid at this epoch, relative to that measured in 1994 in the GHRS data, with the parameter β from the current fit = 0.35 versus 0.9 in the fit to the 1994 data, confirming the variable nature of the λ Vel wind noted in Carpenter et al. (1999) based on comparisons of the GHRS observations with earlier IUE spectra.

Zoom In Zoom Out Reset image size

The far-ultraviolet spectrum of λ Vel displays a large number of fluorescent transitions that are selectively excited, in most of the cases, by H i Lyα and the O i triplet near 1304 Å. These pumps enable a large number of atomic transitions in S i, Cl i, Fe ii, and Cr ii and molecular lines in H₂ and CO by populating numerous upper levels not collisionally excited. Relative to these atomic lines, molecular CO and H₂ emission features are seen to be blueshifted by 3.3 km s⁻¹, suggesting that they are formed in the atmosphere where the wind acceleration has just begun. Atomic non-fluorescent chromospheric emissions are seen in C i, C ii, Si iv, and possibly N i.

Multiple atomic absorptions attributed to Ni ii and Fe ii lines are visible over the chromospheric continuum and are blueshifted by ∼7.6 km s⁻¹ relative to the chromospheric atomic emission lines (see Figure 6), suggesting formation at somewhat higher altitudes, well into the acceleration zone of the outflowing wind, though still not close to the terminal velocity of the wind at this epoch of ∼45 km s⁻¹. There is no evidence in the λ Vel spectrum of the cold circumstellar CO (A-X) absorption bands seen prominently in the spectrum of α Ori.

The λ Vel spectrum has similarities to that of α Boo (K1.5 III), which is of similar spectral type but considerably higher surface gravity. However, the ultraviolet continuum emission and strong Fe ii fluorescent emission are more reminiscent of the spectrum of the M2 Iab supergiant α Ori, which also has a lower surface gravity than α Boo but a cooler effective temperature than λ Vel.

Our analysis of high S/N, high-resolution UV spectra of cool, evolved stars will continue in a sequence of papers on the HST Treasury Program "ASTRAL: Cool Stars" observations of the M giant γ Cru and M supergiant α Ori, including an overview of these spectra, an investigation of the molecular fluorescence processes active in these stars, and a study of the details of the atomic fluorescence processes whose products are seen in these spectra.

Support for GO-11687 and GO-12278 was provided by NASA through grants from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555.

Facility: HST (STIS, COS, GHRS) - Hubble Space Telescope satellite