published paper: MNRAS, Vol. 289, p. 979-985
Ivan K. Baldry, Melinda M. Taylor, Timothy R. Bedding, Andrew J. Booth
Chatterton Astronomy Dept, School of Physics, University of Sydney, NSW 2006, Australia
We present Ha spectra of the 35.5 day period Cepheid l Carinae throughout several cycles during 1994 and 1995. A weak Ha emission feature is present at nearly all phases, which is unusual for classical Cepheids. This emission appears both redward and blueward of the absorption feature at different phases.
Cepheids - line: profiles - stars: individual (l Carinae).
l Carinae is the brightest classical Cepheid in the southern sky. It will be a primary target for the Sydney University Stellar Interferometer (SUSI) as part of a programme for measuring the angular diameter of Cepheids (Booth, Davis & Shobbrook 1995), and much work is being done to improve the general understanding and the radial velocity curve of this star (Taylor et al. 1997, hereafter TABC). Here we present spectra showing what appears to be continuous Ha emission throughout the whole pulsation cycle in l Carinae. In long-period Cepheids the Ha line has often shown two components in absorption but only sometimes shown an emission feature (Rodgers & Bell 1968; Wallerstein 1972, 1983). We believe that the true shape and flux of the emission in l Carinae is largely obscured by upper atmospheric and circumstellar absorption as occurs with Mg II emission (Böhm-Vitense & Love 1994). The origin of the emission feature may be from a shock front in the atmosphere of the star or from the upper atmosphere (we avoid the term chromosphere, see discussion by Sasselov & Lester 1994).
From the early work on the Ha line in Cepheids (Wallerstein 1972), it appears that in stars with periods less than about 13 days the Ha line behaves more or less like other absorption lines. Anomalous Ha profiles have been observed in many long-period Cepheids (Grenfell & Wallerstein 1969, Schmidt 1970) and it is generally agreed that the photospheric line is largely obscured by upper atmospheric absorption and emission in these stars. However Schmidt (1970) argued that the wings of Ha could be used as an effective temperature indicator as long as the measurements of the width of the line are made far enough away from the upper atmospheric core. Our original aim was to measure the equivalent width of Ha in l Carinae as a function of phase but this has not been possible with our data set.
l Carinae (HR 3884, HD 84810) has < V > = 3.7, full range DV = 0.7, < B-V > = 1.2 and a spectral type around G5Iab. We adopt a period of 35.54434 days and a zero-phase (maximum light) Julian date of 2447880.81 (Shobbrook 1992). The period is known to vary but not enough to affect the results in this paper.
We use photometric phases referenced to maximum light. In order to convert to a more physically related phase it is best to define zero phase to be minimum radius, which occurs at photometric phase 0.93 in this star (i.e. add 0.07 to all the phases quoted in this paper).
We obtained high-resolution optical spectra of l Carinae at Mount Stromlo Observatory, with the 1.9m telescope and coudé echelle spectrograph, between February 1994 and April 1995. A single spectrum contained about 45 orders, each approximately 100Å long, projected on to a 2K Tek CCD. The dispersion was ~0.05Å per pixel and the full width at half maximum of the instrumental profile was ~0.10Å . Moderately full phase coverage of l Carinae was obtained (the largest gap in the coverage is 0.085 of a cycle between phases 0.609 and 0.694). Table 1 provides a list of the dates and phases of the 33 spectra analysed for this paper.
Table 1: List of the 33 spectra used in this paper.
The data were reduced using the FIGARO software package (Shortridge 1993). For the purposes of this paper only the two orders containing Ha were reduced. Other work has been done on the radial velocities of metal lines by TABC using more of the orders.
The reduction began with bias subtraction followed by extraction of orders. Corrections using the flat fields were not applied to the data as they were not useful either for tracing the shape of the order or for correcting pixel-to-pixel variations on the CCD. This was because the two orders around Ha were in a vignetted part of the CCD, such that the flat field and the star profiles were significantly different and there were not enough photons in the combined flat field images to reduce the photon noise below the level of the pixel-to-pixel variations on the CCD. Once the orders were extracted, the white scattered light level was removed by using an estimate obtained from levels either side of the star light. Cosmic rays were removed by interpolating across that part of the spectrum. The spectra from the 29th March to the 10th April 1995 contained some high-frequency pattern noise which was reduced significantly by smoothing using a three-point triangle. This smoothing was applied to all 33 spectra to ensure consistency. In most cases no pattern noise was left, an exception being the data from the 1st April (see Figure 1, phase 54.244). Finally a continuum was fitted to each spectrum and wavelengths were calibrated using arc spectra. The continuum fitting was the most critical part of the data reduction and was done by eye, aiming to get a good fit over the region of the order 13Å either side of the Ha core at 6563Å. It is not possible to analyse the depth of the wings of Ha using this set of data because of the difficulty of obtaining a good continuum fit over a region much larger than 26Å. In fact the wings will have been removed by our continuum fitting procedure.
The Ha profiles are plotted in Figures 1-4 over the range 6550Å to 6576Å (helio-centric wavelength) and are normalised so that the continuum is at approximately 1.0. Each plot is offset by 0.5 from the previous one. The spectra are plotted from top to bottom in each figure in order of phase, but note that the observations span several cycles. Only 20 of the spectra are shown in the figures because during some parts of the cycle the Ha profiles were not changing significantly from plot to plot.
Figure 1: Plots of the Ha profile in l Carinae when the emission is redward of the central absorption.
Figure 2: Plots of the Ha profile in l Carinae during the change from redward to blueward emission.
Figure 3: Plots of the Ha profile in l Carinae when the emission is blueward of the central absorption.
Figure 4: Plots of the Ha profile in l Carinae during the change from blueward to redward emission.
There is some repetition in the phase coverage from cycle to cycle, for instance the data from the 28th June 1994 and the 8th April 1995 are almost at the same phase. There are four pairs of spectra with a phase difference of less than 0.010 (43.892 & 46.900, 46.449 & 54.446, 46.760 & 54.753, 46.787 & 54.778). There are slight differences between the profiles from cycle 47 and from 55 but we suggest they are caused by the changing strength and position of the terrestrial atmospheric lines. We see no significant cycle to cycle variations in the Ha profile in l Carinae during 1994 and 1995. TABC discuss variations from cycle to cycle in the radial velocities of metal lines.
Throughout the whole pulsation cycle, there is a dominant absorption component (6562.9Å) whose velocity (+4 km/s) appears to be constant. This means that it does not partake in the pulsation and we attribute this component to a circumstellar shell; this was also seen by Rodgers & Bell (1968). Mass loss (Deasy 1988) from the star is the probable origin for the circumstellar material. There must also be at least one other absorption component and an emission component to explain the complexity of the profile at different phases. The radial velocities of the non-constant components are difficult to measure due to blending effects.
From phase 0.1 (just after maximum light) to phase 0.4 there is an obvious red-shifted emission component of Ha . The spectrum from phase 0.107 also shows more possible absorption components at 6561.8Å and at 6563.7Å. The latter feature is possibly caused by a weak terrestrial line (rest wavelength 6563.521Å). Alternatively this profile could be caused by a very weak emission feature not related to the redward emission. These weak features have disappeared by phase 0.244, leaving a P-Cygni-like profile. From phase 0.4 we see a weakening of the redward emission and by phase 0.500 (minimum light) it has disappeared. At this point there may be a very weak blueward emission component but this is not convincing due to the possibility of the line Si I (rest wav. 6560.6Å) becoming stronger and creating this profile. However the blueward emission grows stronger (see Figure 3) and by phase 0.844 it is clearly present. Figure 4 shows the sudden decrease in the blueward emission from phase 0.892 to 0.996. Then from maximum light onwards the cycle starts again with an increase in the redward emission.
During the phases around maximum light (see Figure 4) there is evidence of absorption line doubling, as has been seen in other Cepheids (Grenfell & Wallerstein 1969; Wallerstein 1972, 1983) but there is also possible contamination with terrestrial lines. The weak nature of this effect in l Carinae is consistent with significant absorption line doubling from the upper atmosphere, which has been mostly obscured by circumstellar absorption.
Ha emission is seen in other yellow super-giants (Sowell 1990; Mallik 1993) and not just in long-period Cepheids. Sowell (1990) observed Ha emission in 13 out of 40 stars in his survey and several other stars had possible weak emission features. Mass flows and circumstellar shells are invoked to explain some of the distorted Ha profiles in these stars. For photometrically variable super-giants there are also pulsation and shock mechanisms capable of producing asymmetrical profiles. This makes it difficult to determine the cause of the observed Ha emission in Cepheids because one cannot easily distinguish between a generic super-giant phenomenon, where the emission is constant but the profile changes due to blending effects with an absorption component, and a pulsation phenomenon where the emission actually changes throughout the cycle.
Barrell (1978) discovered a surprisingly high frequency and strength of Ha emission components in beat Cepheids, which are classical Cepheids having more than one mode of pulsation. She suggested that the emission may indicate the existence of an acoustically heated upper atmosphere surrounding beat Cepheids. She (Barrell 1981) later made measurements of the effective temperatures of some beat Cepheids as a function of phase by using the width of the Ha line. This was possible because the emission in the beat Cepheids was not present most of the time and if present was not usually strong enough to affect measurements using the wings of the Ha line.
There have been several studies of Ha behaviour in classical Cepheids (Bell & Rodgers 1967; Grenfell & Wallerstein 1969; Jacobsen & Wallerstein 1982; Rodgers & Bell 1968; Schmidt 1970; Wallerstein 1972, 1979, 1983; Wallerstein et al. 1992). In a few cases Ha emission has been observed but only on the red wing and only for a fraction of the pulsation cycle. Wallerstein (1972) suggested that a small red emission feature in T Mon is due to broad emission which is largely self-absorbed, leaving a small sharp red wing. The strong emission could be caused by shock heating.
Gillet et al. (1994) have observed BL Her, a Population II Cepheid. They observed a small sharp blue wing in Ha emission in phases 0.82 and 0.86 and absorption line doubling before and after maximum light. This is similar to what might be happening in l Carinae between phases 0.8 and 0.2 except that the circumstellar material is obscuring the effect. Fokin & Gillet (1994) model the Ha profile in BL Her in terms of shock wave propagation in an extended atmosphere. We have to be careful about comparing BL Her and l Carinae because of the difference in periods, 1.3 days and 35.5 days respectively.
Now we compare previous work on l Carinae with our own observations. Rodgers & Bell (1968) made a study of the Ha line in l Carinae and presented profiles at 18 different phases, they used the same telescope and spectrograph 32-inch (81 cm) camera at Mt. Stromlo that we used, but with a lower dispersion (10Å /mm compared to 2Å /mm). Their data cover two sections of the pulsation cycle, phases 0.94-0.08 and 0.49-0.73. In the first section, their profiles are similar to ours and show the weak red-shifted emission starting between phases 0.0 and 0.08 (see Figure 4) but there is no evidence for any residual blue-shifted emission at phase 0.94 in their data. In the other section, there are more discrepancies between the two sets of observations. Again, there is no evidence for a blue-shifted emission in their data but they do not have any observations between phases 0.85 and 0.9 when we observed the blue-shifted emission to be strongest. The differences between our observations and theirs may be due to the dispersion, the data recording device (CCD vs photographic plate) or to an intrinsic change in the star.
Böhm-Vitense & Love (1994) made a study of l Carinae in the UV using the International Ultraviolet Explorer (IUE). They studied the emission-line fluxes of several spectral lines (C II, C IV, Mg II, O I) as a function of pulsational phase. It is interesting to compare the Mg II emission with the Ha emission. The Mg II h and k line profiles consist of a strong emission, which is effectively split into red and blue wings by a strong absorption component, producing a double hump appearance. The absorption component is stationary and is attributed to circumstellar gas. The overall flux of the emission changes throughout the pulsational cycle. Between phases 0.8 and 0.9 a very steep increase in flux occurs which they suppose is due to an outward-propagating shock. As well as the overall flux of the Mg II emission changing, the relative strength of the two wings also changes, but this variation is 0.1-0.15 cycles out of phase with our observed Ha emission shifts (see Table 2).
Table 2: Summary of some spectral features in l Carinae.
Table 2 shows that the Ha emission is approximately half a cycle out of phase with the motion of the photosphere as seen from the metal absorption lines. This implies that the emitting gas is travelling in the opposite direction to the photosphere. One possible explanation is that we are seeing moving gas in a shockfront on the far side of the star, this is possible if the radius of the emitting shock is much larger than the photosphere. In this model, the emission from the shockfront on the near side would be obscured by another absorption component, moving nearly in phase with the photosphere, other than the central component due to the circumstellar gas. Another explanation is that there is fairly stable emission from the upper atmosphere. This would arise from the limb and because of projection effects would have almost no Doppler amplitude. An absorption component moving back and forward obscures part of the emission feature creating the red and blue wings. In either case, the Ha absorption velocity (not the circumstellar component) is in the opposite direction to the emission. This means that the Ha absorption is red-shifted between phases 0.6-0.95 and blue-shifted from 0.0-0.5 which gives it a phase lag of ~0.1 behind the metal line velocity. This is consistent with this Ha absorption component being formed higher up in the atmosphere than the metal lines. Further theoretical work may indicate the origin of the Ha emission in l Carinae.
Many thanks to Peter. R. Wood and H. M. Schmidt for the initial set of observations, Michael. S. Bessell for his help and advice in setting up the instrument and to the Director of Mount Stromlo and Siding Springs Observatories for the time on the telescope. We are grateful to Dimitar D. Sasselov, Michael Albrow and Peter Cottrell for helpful comments on this paper. We acknowledge the Harvard ADS Abstract Service as an invaluable aid in finding relevant papers. This work was carried out while the first two authors were in receipt of an Australian Postgraduate Award and a UWPRA scholarship respectively, and was also supported by funds from the Australian Research Council.