CARMA CO(J = 2 − 1) OBSERVATIONS OF THE CIRCUMSTELLAR ENVELOPE OF BETELGEUSE

The spectrum for each individual configuration image cube (which are composed of all the appropriate configuration tracks listed in Table 1) along with the multi-configuration image cube can be used to obtain information on the kinematics of the S1 and S2 flows. The spectra corresponding to the C, D, and E configuration image cubes are plotted in Figure 1 for both the high (0.65 km s⁻¹ bin⁻¹) and low (1.3 km s⁻¹ bin⁻¹) spectral resolution data and were obtained by integrating all emission within a circular area of radius 5'' centered on the source. The high and low spectral resolution modes allow two independent sets of spectra to be measured for each observation and thus provide a good check on the data quality. The high-resolution spectra (channel width = 0.65 km s⁻¹) give the best measure of S1 and S2 kinematics and therefore all outflow velocities are derived from these spectra.

Zoom In Zoom Out Reset image size

The E configuration image cube spectrum has a total line width of 29.2 km s⁻¹ and the low spectral resolution profile contains a steep blue wing emission feature between −16.0 km s⁻¹ and −11 km s⁻¹ and a more flat-topped feature between −10.3 km s⁻¹ and −13.2 km s⁻¹. This steep emission wing shows that the turbulence in the flow is less than or equal to the velocity bin size. The blue wing in the high-resolution profile matches the lower resolution profile well but the remainder of the profile looks more complex than the flat-topped feature seen in the lower resolution profile. The profile shape of the CO(J = 2 − 1) line has been well documented by previous single-dish observations (e.g., Knapp et al. 1980; Huggins 1987) and, out of our three individual configuration spectra, we expect the most compact E configuration spectra to resemble these single-dish measurements the closest due to its better sampling of the inner u − v plane and consequent sensitivity to extended structures. This indeed turns out to be the case when we compare our three individual configuration spectra to those previous single-dish profiles. The blue wing emission feature appears again in the D configuration spectrum at the same velocities as those in the E configuration spectrum but the remainder of the profile appears quite different. Between −10.3 km s⁻¹ and +13.2 km s⁻¹ the D configuration spectrum is dominated by a blue wing at ∼ − 10.0 km s⁻¹, a red wing at ∼ + 13.0 km s⁻¹, and an emission feature at ∼0 km s⁻¹.

The line profile has a much lower flux in the high spatial resolution C configuration spectrum due to its lack of sensitivity to extended structure. The blueshifted emission feature located between −16.0 km s⁻¹ and −11.0 km s⁻¹ in the E and D configuration spectra is almost completely spatially filtered by the extended C configuration. This component of the line has previously been associated with the outer S2 flow (Huggins 1987) and as the majority of it has been spatially filtered by our C configuration we expect even less contribution from the S2 flow at lower absolute velocities still. For the redshifted line emission we again expect the majority of the S2 contribution to be spatially filtered, so we conclude that the majority of the emission in the C configuration spectrum emanates from the inner S1 flow. The spectrum is double peaked with the blue- and redshifted wings extending to −9.0 km s⁻¹ and +10.6 km s⁻¹, respectively, and we define these as the outflow velocities of S1. As discussed in Section 2, the C configuration has a resolving out scale of ∼6'' at 1.3 mm and so is not sensitive to angular scales larger than this. If the emission between −9.0 km s⁻¹ and +10.6 km s⁻¹ in the C configuration spectrum appeared as a flat-topped profile then we could conclude that the S1 flow lies within a radius of 3'' from the star. Clearly, however, the lower absolute velocity components of this profile have been spatially filtered so we conclude that the radial extent of the S1 from the star is greater than 3''. If we assume that the S1 flow would produce a top-hat line profile were it not for the resolving out effects of the interferometer, then its integrated line flux is 3.1 × 10⁻¹⁹ W m⁻².

To obtain the most robust value for the S2 outflow velocities we examine the high spectral resolution multi-configuration image cube spectrum, which is composed of all tracks from all three configurations. It is worth stressing that by analyzing the multi-configuration image cube we make the assumption that the physical properties of all three components (i.e., α Ori, S1, and S2) have not changed over the total observation period (i.e., ∼2.5 yr). The profile is found to have a total line width of 28.6 ± 0.7 km s⁻¹, which is in close agreement with previous single-dish observations of the line where values of 30.6  ±  2.5 km s⁻¹ and 28.6 km s⁻¹ were reported by Knapp et al. (1980) and Huggins (1987), respectively. The centroid velocity of the spectrum is −1.1 ± 0.7 km s⁻¹ (v_lsr = 3.7 ± 0.7 km s⁻¹), which is again in close agreement with Knapp et al. (1980) and Huggins (1987) values of v_lsr = 3.0 ± 2.5 km s⁻¹ and v_lsr = 3.7 ± 0.4 km s⁻¹, respectively. The integrated line flux is 1.5 × 10⁻¹⁸ W m⁻² of which approximately 20% emanates from the S1 flow.

The outflow velocities of S2 are −15.4 km s⁻¹ and +13.2 km s⁻¹ which, like the S1 flow, are slightly asymmetric but in the opposite sense. Note that the S1 and S2 outflow velocities are dependent on the adopted radial velocity of Betelgeuse. If, for instance, we instead adopt a radial velocity of 21.9 km s⁻¹ (Famaey et al. 2005) then the S2 outflow velocities become even more asymmetric (−16.6 and +12.0 km s⁻¹) while the S1 outflow becomes less so (−10.2 and +9.4 km s⁻¹). Both S1 and S2 therefore cannot have spherically symmetric outflow velocities regardless of the adopted stellar radial velocity. Adopting a mass of 18 M_☉ and a radius of 950 R_☉ (Harper et al. 2008) then the escape velocity for Betelgeuse is 85 km s⁻¹, which is much greater than the S1 and S2 outflow velocities. This indicates that the majority of the stellar mass-loss mechanism's energy goes into lifting the CO molecules out of the gravitational potential and not into their outflow velocities. These outflow velocities are greater than the adiabatic hydrogen sound speed, which, if we assume that the gas temperature is the same as the excitation temperature, are 1.7 km s⁻¹ and 1 km s⁻¹ for S1 and S2, respectively.

The spectra in Figure 2 are taken from the low-resolution multi-configuration image cube using circular extraction areas ranging in radius from 1'' to 10'' and demonstrate how the line profile changes over these different extraction areas. The most striking change in the line profile is the change in appearance of the extreme blue wing. At small extraction radii where we sample the most compact emission, the feature is weak in comparison with the rest of the line but becomes more dominant as we begin to sample more of the extended emission. This indicates that even the high absolute velocity components of the S2 flow have extended emission and this is why they are almost completely spatially filtered by CARMA's C configuration. The large reduction of flux at −11 km s⁻¹ suggests that there is more material moving toward the observer than at other lower absolute velocities indicating a non-isotropic (or non-spherical) S2 flow. This suggests a more sheet-like (flatter) structure rather than a spherical cap.

Zoom In Zoom Out Reset image size

A subset of the blueshifted velocity channel maps of the low spectral resolution multi-configuration image cube is presented in Figure 3. The first channel map at −17.9 km s⁻¹ shows just the compact unresolved continuum emission with no extended emission present. Between −16.7 km s⁻¹ and −9.0 km s⁻¹, we see evidence for the development of a classical shell signature for the S2 flow. We first sample the highest velocity components where the emission is relatively compact (i.e., between −16.7 km s⁻¹ and −12.9 km s⁻¹) and then sample lower radial velocity components where S2 becomes a faint ring (i.e., between −11.6 km s⁻¹ and −9.0 km s⁻¹). At lower velocities again, these rings disappear into the noise of the maps and possibly extend out beyond the primary beam at zero velocity when the rings should have maximum spatial extent. The emission from the channel maps between −15.3 km s⁻¹ and −11.6 km s⁻¹ corresponds to all the emission in the extreme blue wing component of the multi-configuration image cube line profile discussed in Section 3.1. We can see in Figure 3 that all of this emission is greater than the C configuration resolving out scale, therefore confirming that our C configuration line profile is mainly composed of S1 emission. The shell formation signature of S2 is also apparent in the redshifted velocity channel maps between +7.5 km s⁻¹ and +13.8 km s⁻¹ but the emission appears weaker and the rings fainter therefore indicating that S2 is somewhat fragmented.

Zoom In Zoom Out Reset image size

The multi-configuration maps also show the central compact emission from the S1 flow at velocities between −10.3 km s⁻¹ and +11.3 km s⁻¹. This S1 emission can be seen in the final two maps of Figure 3 as a central slightly elongated emission feature surrounded by the fainter rings of the S2 flow. In the maps where both S1 and S2 are present the emission from S1 appears brighter than the emission from S2. The spatial extent of the S1 flow varies from channel map to channel map but appears to be larger than the 2'' value given by Smith et al. (2009), who observed off-star wind scattered ro-vibration CO lines.

An additional spatially unresolved source is detected in a number of the D configuration image cube maps (both high and low spectral resolutions) and has been previously documented by Harper et al. (2009). The component is present in only five continuous low-resolution channels between ∼ − 4.0 km s⁻¹ and +2.4 km s⁻¹ and is located ∼5'' S–W of α Ori as shown in Figure 4. Its peak emission lies at ∼0 km s⁻¹ and here it approximately equals 60% of the source peak emission. The corresponding channel maps in the E configuration image cube show extended emission out to 8'' in the same S–W direction. This second source does not appear in any of the C configuration channel maps probably due to the lower sensitivity resulting from the smaller HPBW (i.e., the flux is diluted). This discrete second source thus has the effect of adding extra emission to the corresponding multi-configuration image cube maps at the low velocities where it is present.

Zoom In Zoom Out Reset image size

The spatial extent of the S1 and S2 flows around Betelgeuse was not directly determined from either the CO infrared absorption spectra of Bernat et al. (1979) or previous CO single-dish radio observations (Knapp et al. 1980; Huggins 1987; Huggins et al. 1994). Our low spectral resolution multi-configuration image cube has sufficient spatial resolution and signal to noise to make direct estimates of the maximum radius of both flows. The outer S2 flow is not seen in the low absolute velocity channel maps where its spatial extent is maximum and either lies outside of the primary beam or is lost into the noise near the edge of the maps. We derive the maximum outer scale of the S2 flow by looking at the spatial scales of the S2 flow in the higher absolute velocity maps where S2 is present. If we assume that S2 is spherically symmetric with an outer radius R_S2, and is undergoing steady expansion with velocity V_S2, then we can estimate its radius in each velocity channel using the following relation:

$$\begin{equation} {r}_{\rm chan}={R}_{\rm S2} \sin \left[\cos ^{-1}\left(\frac{{v}_{\rm chan}}{{V}_{\rm S2}}\right) \right], \end{equation} \tag{ 1 }$$

where r_chan is the S2 radius in a channel at velocity v_chan.

We use Equation (1) to estimate the maximum projected spatial extent of S2 that occurs at zero velocity. An estimate of the S2 radius per channel (r_chan) was found by creating annuli of increasing radius around the central emission in each relevant line channel map of the multi-configuration image cube, extracting all flux within each annulus and then plotting these fluxes against distance from the star for each channel. The maximum of these resultant curves was then deemed to be the maximum radius of S2 per channel. Figure 5 shows these data overplotted with two model outflows that were created using Equation (1). The blueshifted data points were best fitted by a model outflow of maximum radius 17'' and outflow velocity 17 km s⁻¹, while the redshifted data points were best fitted by a model outflow of maximum radius 16'' and outflow velocity 14 km s⁻¹. It is worth mentioning that this estimate for the spatial extent of S2 is only weakly dependent on our adopted radial velocity value for Betelgeuse and adopting a slightly different value would simply alter S2s outflow velocities. As S2 is not present in our lowest absolute velocity map we are not able to report an estimate of its width.

Zoom In Zoom Out Reset image size

In the left column of Figure 6 we investigate the intensity distribution of CO emission as a function of projected radius, R, for both the S1 and S2 flows. From our discussions in Section 3.1 we can assume that all line emission between −15.4 → −10.3 km s⁻¹ and +12.4 → +13.8 km s⁻¹ emanates solely from the S2 flow. Using the low spectral resolution multi-configuration image cube we integrate the surface brightness over these channels and find that the intensity falloff is proportional to R⁻¹ (Figure 6, top). To investigate the S1 flow intensity distribution around α Ori we integrate the surface brightness over the channels between −9 → +10.6 km s⁻¹. Although these channels contain emission from both S1 and S2, most of the S2 emission here will have larger projected radii and thus the majority of the inner emission should emanate from the S1 flow. Between 05 and 4'' from the star the intensity is again found to be proportional to R⁻¹ (Figure 6, bottom). Such an intensity distribution is expected for an optically thin homogeneous constant velocity outflow with ρ∝1/R². Beyond 4'' the intensity falloff is more rapid and is close to an R⁻² distribution, which may mark the initiation of the current epoch of mass loss.

Zoom In Zoom Out Reset image size

Insight can also be gained into how the intensity varies on different size scales by conducting analysis in the u − v plane and plotting the visibility amplitude of α Ori against u − v distance. The result of this is shown in the right column of Figure 6 where the same channels corresponding to the S1 and S2 flows have been used. The data are azimuthally averaged, and have been binned to produce one data point per kλ. The result for both the S1 and S2 data is a steep drop-off in visibility amplitude over a relatively short u − v distance, signaling that the sources are well resolved. Both sets of visibility data agree with an intensity proportional to (a² + R²)^−1/2, where a is an inner spatial limit. This is because the Hankel transform of this function is q⁻¹e^−2πaq (Bracewell 2000), where q is the u − v distance, and a vertically scaled version of this function is shown to match the visibility data very well in Figure 6. As analysis in both the sky and u − v plane indicate the intensities of both flows are proportional to R⁻¹ we conclude that when azimuthally averaged, both outflows are consistent with an optically thin and quasi-steady flow that is in agreement with Smith et al. (2009; i.e., S1) and Plez & Lambert (2002; i.e., S2).

An exact determination of the maximum spatial extent of the S1 flow is more difficult as we do not see the classical shell formation signature for it as we sample across velocities, like we do for S2. Instead its spatial extent varies over the channel maps with evidence of discrete clumps being present in many of these maps. At 20% of maximum emission in the integrated intensity S1 map (i.e., composed of all channels between −9 km s⁻¹ → +10.6 km s⁻¹) the S1 flow extends out to a mean distance of ∼4'' and is even more extended in the S–W direction due to the presence of the second emission feature in the compact configuration data sets. The HPBW of 09 is not sufficient to determine whether the S1 flow is discrete or an extension of the current wind phase seen in ultraviolet spectra, e.g., Carpenter & Robinson (1997), and centimeter-radio continuum interferometry (Lim et al. 1998; Harper et al. 2001).

In Table 2 we show the derived continuum flux densities for each of the three configuration image cubes and also the multi-configuration image cube. The high spectral resolution (Δv = 0.65 km s⁻¹) image cubes were just wide enough to image the CO line but were too narrow to make accurate estimates of the continuum flux density. Therefore, all continuum flux density estimates are derived from the lower spectral resolution (Δv = 1.3 km s⁻¹) image cubes from which we were able to take accurate measurements at both sides of the line. We fitted elliptical Gaussians to ∼20 continuum channels using CASA's imfit routine, allowing the flux and corresponding uncertainties to be calculated. The source was unresolved in most of these continuum channels.

Betelgeuse is known to show brightness variations at many wavelengths. Goldberg (1984) reports a decrease of half a magnitude in visual brightness over a period of six years. Bookbinder et al. (1987) found stochastic 30%–40% variations in flux density at 6 cm over timescales as short as 10 days to as long as 8 months (i.e., the observational period). A more comprehensive study was carried out by Drake et al. (1992) who observed Betelgeuse with the VLA at centimeter-wavelengths from 1986 to 1990 and found stochastic variability of 22%, 15%, and 21% at 6 cm, 3.6 cm, and 2 cm, respectively, at a variety of different timescales down to less than 1 month. The millimeter-continuum emission that we measure arises mainly from electron–ion and electron–atom bremsstrahlung and possibly dust emission, so it is not unreasonable to also expect variability at millimeter-wavelengths too. The D configuration data were acquired under adverse weather conditions and these data have the highest noise levels out of the three configurations. Its continuum emission measurement is approximately 50% greater than the C and E configuration continuum measurements, which were also acquired approximately two years after the D configuration data. We believe that the continuum emission derived from the multi-configuration image cube is a reasonable estimation of the mean millimeter-continuum flux density over the two-year period and is in reasonably good agreement with the 250 GHz flux density of Altenhoff et al. (1994) who report a mean value of 351 ± 25 mJy for 1986 → 1989.

Bernat et al. (1979) were the first to detect circumstellar absorption lines in CO by looking at the fundamental ro-vibration lines at 4.6 μm. These infrared observations revealed two separate outflows around α Ori; an excited (T_exc = 200 K) S1 flow with an expansion velocity of 9 km s⁻¹ and a less excited (T_exc = 70 K) S2 flow moving with a faster expansion velocity of 16 km s⁻¹. Knapp et al. (1980) were the first to detect emission in the CO(J = 2 − 1) line at 1.3 mm using the 10 m millimeter-wave telescope at OVRO but only detected one component expanding at 15 km s⁻¹. By analyzing the shape of the line profile, they concluded that the S2 radius of 55'' derived by Bernat et al. (1979) was too large and that it lies at a radius of R ⩽ 10''. Since the detection by Knapp, a number of observations at 1.3 mm have been carried out with various beam sizes and all spectra look remarkably similar; that is the profile has a steep extreme blueshifted emission component with the remainder of the profile looking more flat topped and containing a number of less dominant spikes. Huggins (1987) used their single-dish observations (HPBW ∼ 32'') of the CO(J = 2 − 1) line along with excitation and self-shielding models of CO to conclude that the S1 flow makes little contribution to the final emission line. They also identify the extreme blue wing of the line with the S2 flow and predict that it may extend out to a radius of ∼16''. Later, however, Huggins et al. (1994) compared their detected 609 μm ³P₁ → ³P₀ fine structure line of C i with CO data obtained with the IRAM 30 m telescope (HPBWs ∼12'') and find that the expansion velocities in both lines are essentially the same. They conclude that the radial extent of C i is ≲7'' and both the CO and C i are formed in the inner envelope and roughly extend over the same area.

The shape of our multi-configuration line profile for extraction areas of radii 6'' or greater is in good agreement with previous high signal-to-noise single-dish CO(J = 2 − 1) spectra (e.g., Huggins et al. 1994, Figure 1) although the emission spikes in our line profiles are more dominant. Our total line width of 28.6 km s⁻¹ is in good agreement with Huggins (1987) and Huggins et al. (1994), who report line widths of 28.6 km s⁻¹ and 30 km s⁻¹, respectively. The extreme blue wing in both of these spectra is the dominant emission feature of the line and this is also true in our multi-configuration spectra at extraction areas ≳6''. Using the IRAM 30 m telescope, which has an HPBW of only 12'' at 230 GHz, Huggins et al. (1994) produce a similar line profile shape to that presented in Huggins (1987), who have a larger HPBW of ∼30''. From this, one would expect that the majority of the blue wing emission is compact; however, our multi-configuration line profiles suggest otherwise, and show a continuous increase in the blue wing emission as we take larger extraction regions out to 10''. The multi-configuration maps also show a faint ring structure forming at ∼ − 11.6 km s⁻¹ and expanding further out in lower absolute velocity channel maps. This ring emission is fainter than the higher velocity compact emission so we see a drop in flux density in our spectra at the point where these rings form. Therefore, the steepness of the extreme blue wing in our multi-configuration spectrum means that there is merely more CO emitting at higher velocities than at lower velocities, which is indicative of a sheet-like structure moving toward the observer.

The line profiles of higher CO rotational transitions for Betelgeuse have been published in Kemper et al. (2003) and De Beck et al. (2010). De Beck et al. (2010) present high spectral resolution (0.3125 MHz) line profiles for the CO(J = 2 − 1), (J = 3 − 2), and (J = 4 − 3) transitions that were obtained with the James Clerk Maxwell Telescope (JCMT). For the CO(J = 2 − 1) transition the JCMT has an HPBW of ∼20'' and the profile appears similar to our multi-configuration profile over the same flux density extraction area (i.e., Figure 2), with the extreme blue wing component being the dominant feature in both. This feature, which is emission from the S2 flow, becomes a less dominant component of the line profile at the higher CO(J = 3 − 2) and CO(J = 4 − 3) transitions where the JCMT has an HPBW of ∼13'' and 8'', respectively, and does not capture all of the S2 emission, which is shown in Figure 3 to be extended at these velocities. Also, the higher rotational states (J ≈ 10) will be populated more by the higher excitation temperature (∼200 K) S1 flow so these line profiles become dominated by emission from the slower S1 flow. This is confirmed by our narrow SOFIA–GREAT CO(J = 12 − 11) line profile that is presented in the Appendix of this paper, and also by a visual inspection of Herschel-HIFI archival line profiles of the CO(J = 6 − 5), (J = 10 − 9), and (J = 16 − 15) transitions.

The CO 4.6 μm ro-vibration lines have been observed with the Phoenix spectrograph (Hinkle et al. 1998) by Ryde et al. (1999) and Smith et al. (2009) on the 2.1 m telescope at Kitt Peak and on the 8 m Gemini South telescope, respectively. By assuming a Boltzmann population distribution for the ground rotational levels of CO, Ryde et al. (1999) derived a mean excitation temperature of 38⁺⁶_{− 5} K along the line of sight at a projected distance of 4'' north of Betelgeuse. Our CARMA data suggest that the S1 flow extends out to approximately this distance but Ryde et al.'s temperature is not in agreement with either of the line-of-sight S1 or S2 excitation temperatures of 200⁺⁵⁰_{− 10} K and 70 ± 10 K derived by Bernat et al. (1979). This discrepancy may indicate that the excitation is quite non-uniform. Smith et al. (2009) did not derive an excitation temperatures but used their 4.6 μm spectra to reveal extended resonantly scattered CO emission out to ∼2'', a factor of two smaller than our S1 radius. They observe emission over a velocity range of 30 km s⁻¹ but two distinct flows are not detected. Mild (∼20%) density inhomogeneities are reported but overall, their observations are consistent with an optically thin and steady wind which is consistent with our findings.

The S2 flow was first identified in high-resolution K i and Na i absorption spectra by Goldberg et al. (1975) and subsequently re-observed multiple times over the next couple of years (Goldberg 1979). It is interesting to compare these and Bernat et al.'s (1979) CO line-of-sight absorption velocities with those from the CARMA emission spectra obtained at similar spectral resolutions and also to measure, the perhaps co-spatial line broadening of the K i S2 absorption feature.

We have obtained K i 7698.98 Å spectra using the cross-dispersed echelle spectrometers on the Harlan J. Smith 107 inch (2.7 m) reflector at McDonald Observatory. With 2 pixels per resolution element an R = λ/Δλ = 200, 000 and an R = 500, 000 spectrum were obtained on 2007 March 25 and April 13, respectively. The spectra were wavelength calibrated with ThAr lamp lines and the lower resolution spectrum was checked by fitting six symmetric terrestrial O₂ lines in the same order using wavelengths from Babcock & Herzberg (1948). The O₂ lines confirmed that the R = 200,000 calibration was good to better than 0.1 km s⁻¹. Upon cross-correlating the low-and high-resolution spectrum the high-resolution spectrum appeared redshifted by 0.60 km s⁻¹, i.e., one resolution element, for which we do not have an explanation except to note that a similar offset has been reported by Welty et al. (1994). We use the cross-correlation to define the wavelength calibration of the R = 500, 000 spectrum and we adopt a systematic error of σ^sys = 0.2 km s⁻¹.

The high-resolution spectrum is shown in Figure 7 in the adopted stellar center-of-mass rest frame (V_rad = +20.7 km s⁻¹). The S2 feature is deep, well separated from the S1 feature, and very well represented by an absorption model including the hyperfine splitting from the abundant ³⁹K isotope. We adopt the K i 7698.9645 Å line parameters compiled in Morton (2003)⁹ and find a heliocentric S2 absorption velocity of +5.15 ± 0.01(1σ^ran) km s⁻¹ with the uncertainty dominated the absolute wavelength scales. However, the most probable line-of-sight turbulent velocity is much more accurately determined with 0.62 ± 0.02(1σ^ran) km s⁻¹. From their 4.6 μm line profiles Bernat et al. (1979) found a very similar 1 km s⁻¹, these being the sum of thermal and non-thermal motions. These values place tight constraints on the line-of-sight acceleration associated with the S2 flow.

Zoom In Zoom Out Reset image size

The K i spectrum also reveals a slight inflection in the observed line profile at +3.6 km s⁻¹ (heliocentric), which may represent structure in the underlying photospheric profile or additional absorption in which case it has ∼0.1 the column density of S2 ($N_{{\rm K\,{\mathsc i}-S2}}\simeq 1.2\times 10^{11}\ {\rm cm}^{-2}$). The S2 absorption velocity minima can be compared to those obtained by Goldberg (1979, Figure 7) who measured values between 1975 and 1978 of 4.2 ± 0.2 and 5.0 ± 0.2 km s⁻¹ and these differences may result from changes caused by radial velocity changes in the underlying photospheric spectrum. Bernat et al.'s (1979) CO IR absorption observations reveal S2 heliocentric velocities of +4.94 ± 0.30 km s⁻¹ (1979 March 6) and +4.60 ± 0.04 km s⁻¹ (1979 April 14) with turbulent velocities of 4 and 1 km s⁻¹ for the S1 and S2 features, respectively.

In terms of the center-of-mass radial velocity of the star our K i feature implies an outflow velocity of +15.6 km s⁻¹. The blue edge of our CARMA multi-configuration CO profile is estimated to be +15.4 km s⁻¹, which suggests a dynamical association with the CO S2 flow and very close agreement with Bernat et al.'s (1979) CO absorption velocities listed above. Plez & Lambert (2002) have also estimated the radius and velocity of the suspected K i S2 flow using R = 110, 000 resolution long-slit spectra. They found a geometrically thin shell (1'') with velocity of V_S2 = 18 ± 2 km s⁻¹ with a radius of 55'', which is much larger than the field of view of the CARMA spectra. Their long-slit spectra show several smaller partial shells but it is not simple to directly associate the CO emission feature with one or more of these shells especially given the uncertainty in the ionization balances of CO and K i. It is possible that the 55'' shell is associated with the inflexion caused by additional absorption (and low column density) at a velocity slightly higher then S2 observed in our K i profile.