EVIDENCE OF V-BAND POLARIMETRIC SEPARATION OF CARBON STARS AT HIGH GALACTIC LATITUDE

A characteristic property of stars evolving from red giant stage to planetary nebulae (PNe) is polarization; observed polarization reaches a maximum during the proto-planetary stage and gradually decreases as the star evolves into a PN (Johnson & Jones 1991). Polarimetric studies of post-asymptotic giant branch (AGB) stars by Trammell et al. (1994) and Parthasarathy & Jain (1993) have helped to identify aspherical morphology shells that originate during mass-loss phase as common features of these objects. Quasi-periodic variations in linear polarization are known to be observed in many late-type variables. However, not much work has been done on polarization of carbon stars; we report here V-band polarimetric measurements of a sample of carbon stars at high Galactic latitude. Polarization of carbon stars are known to show a flatter wavelength dependence than the oxygen-rich stars. Stars with circumstellar materials exhibit a certain amount of linear polarization. The degree of polarization increases with asymmetries present in the geometry of the shells. By modeling the observed polarization, it is possible to study the geometry and structure of the shells. Polarization models of circumstellar grain scattering shows that graphites do not adequately explain the mean polarimetric behavior of carbon stars (Raveendran 1991). Rowan-Robinson & Harris (1983) have shown that graphites also do not adequately explain the IR spectra of circumstellar dust envelopes around carbon-rich objects. It was suggested by Czyzak et al. (1982) that graphite formation in circumstellar envelopes is not a very likely process and its likely that carbon grains exist most likely as amorphous carbon.

While Parthasarathy et al. (2005) find no significant variation in polarization for HD 100746 at U, B, I bands, etc., Raveendran (1991) finds mild wavelength dependence of polarization in the case of a few carbon stars of Mira variables. In particular, the polarization was found to increase systematically toward the red in RT Pup, UU Aur shows a dip in the polarization curve around the V band, and X Vel shows a larger polarization in the R band than in the V band. Multiband polarimetric observations by Serkowski et al. (1975) showed that the interstellar polarization has a peak occurring at a median wavelength λ = 0.545 μm. The maximum effect of interstellar polarization can be felt, therefore, in the yellow spectral region. It is important therefore to undertake follow-up observations on several occasions to study the variations in polarization particularly for those that show significant polarization in the V band. Our present sample includes faint high latitude carbon stars from Christlieb et al. (2001). We have studied a large fraction of this sample based on low- as well as high-resolution spectroscopy (Goswami 2005; Goswami et al. 2006, 2007, 2010b; Goswami & Aoki 2010a). A primary characteristics of these stars is that they provide evidence of nucleosynthesis of the low- and intermediate-mass stars as most of them are polluted by now-extinct AGB stars. The basic parameters of the program stars are listed in Table 1.

Observations were made using the 2 m telescope of IUCAA, Pune using the polarimeter (IMPOL) attached to it. The polarimeter available at the cassegrain focus of the 2 m IUCAA telescope has a rotating half-wave plate (HWP) and Wollaston prism through which light passes before forming a pair of images of an object on the CCD (Ramprakash et al. 1998). The HWP can rotate in several discrete steps, such that its fast axis makes angles (α) with some reference direction (generally celestial north–south). The light which is transmitted out of the Wollaston prism forms two images of any celestial source on the CCD, with the ordinary and extraordinary set of rays. For each object observations were taken at four different positions of the HWP: 0°, 225, 45°, and 675.

Among BVRI polarimetry, the V band is known to show the maximum polarization (Serkowski et al. 1975) and is ideal for the detection of polarization caused due to scattering by dust. Moreover, because of a low photon flux in the B band, the polarization measurements in the B band often have large errors. For these reasons, we have conducted only V-band polarimetry for the program stars. In addition to the program stars we have observed three polarization standard stars and one unpolarized standard star for polarimetric calibration.

Data reduction was carried out using various tasks in IRAF. The task PHOT in APPHOT was used to measure the stellar flux. Zero polarization standard stars were observed to check for any possible instrumental error, which proved to be smaller than 0.1%. For stars which show low degree of polarization, measurement accuracy is an important issue. The measured errors are smaller than 0.1% for majority of the observed stars.

V-band polarimetric measurements for the standard polarized and unpolarized stars are presented in Table 2. Estimates from literature are also listed for a comparison. A close agreement between the estimated and literature values lends support to the reliability of our results. The estimated error in polarization is less than or near about 0.1% for all stars (Table 3) except for the star HD 209621 for which the error is ±0.23. Since polarimetric error in our case is photon noise dominated, in the case of very bright stars this error becomes close to zero. Our program stars include objects brighter as well as fainter than HD 209621. Thus, the error in HD 209621 can be considered to be a typical error value.

The polarization standard stars HD 147084 and HD 160529, both observed on the same night, show an offset of about ∼−10°; the near equality of the offsets is as expected for consistency, and a similar offset was also reported earlier by Sen et al. (2000). The offset is essentially a result of constraints in instrumental mounting, such as constraint in mounting the polarimeter with the axis of fixed HWP making as angle 0° with celestial north–south axis (this will give 0 offset). We have applied an offset of −10° to the estimated position angles of all the stars observed on the same night as that for HD 147084 and HD 160529. The expected final values of position angles (θ°), obtained with the necessary offset corrections are listed in Table 3.

The linear polarization (p%), with the errors in the measurements of the program stars, is presented (Table 3) without any corrections applied to the observed values for interstellar polarization. For stars with zero polarization (⩽0.1%) the position angles are not defined, hence position angles are listed only for stars with non-zero polarization. The interstellar polarization is presumed to be negligible because of their high Galactic latitudes. The interstellar reddening, listed in Table 3, for the program stars is small. For the program stars for which color excess have been known, it is instructive to compare the spatial distribution of the color excess with polarization. In Figure 1 (upper panel), we have plotted the V-band polarimetric estimates of the stars (p%) against interstellar reddening E(B − V) along with the mean interstellar polarization ∼3E(B − V; Parthasarathy et al. 2005). Estimated V-band polarization is found to be larger than the mean interstellar contribution to the polarization in most of the cases indicating that they are intrinsically polarized.

Zoom In Zoom Out Reset image size

Polarization does not seem to have any co-relation w.r.t. metallicity ([Fe/H]) of the stars (Figure 1, middle panel). The observed V-band percentage polarizations are larger for high Galactic latitude objects (Figure 1, lower panel); however there are two outliers (shown with triangles in Figure 1, lower panel) that are high Galactic latitude objects but show weaker polarization. These two objects (HD 100764 and HE 1152−0355 with [Fe/H] ∼ −0.6 and −1.3, respectively) are not so metal-poor compared to the other stars that show larger polarization. It has been observed that the objects clearly fall into two distinct groups; five objects showing polarization below 0.4% and the other five showing polarization above 1.0%. HD 206983 and HD 196944 in our sample do not show significant evidence of linear polarization in the V band. The properties of the individual stars are discussed below.

3.1. HD 206983, HD 196944

3.2. HD 100764, HD 168986, HD 198269, HD 209621, and HE 1152−0355

3.3. HE 1027−2501, HE 1305+0007, HE 1429−0551, LP 625−44, and HE 1523−1155

Our V-band polarimetric observations presented for a sample of 10 carbon-enhanced metal-poor (CEMP) stars show five objects with polarization p% > 1 and the other five with p% < 0.4. While the size of the sample is relatively small, it should be noted that these stars belong to similar group in terms of carbon enhancement and metal deficiency. Thus, a clear separation of these objects into well-separated V-band polarimetric groups is significant. Intrinsic polarization of late-type stars are usually attributed either to the Raleigh scattering of the light emerging from the limb of a star in which case the sources of photometric asymmetry necessary for a net observable polarization are suggested to be non-radial pulsation of the star, variation of temperature over the surface, and the presence of giant convection cells (Harrington 1969; Schwarzschild 1975) or due to scattering by molecules and dust grains in an extended asymmetric circumstellar envelope (Kruszewski et al. 1968; Shawl 1975; Daniel 1978). It seems circumstellar grain scattering is the main mechanism responsible for the observed polarization in the sample of program stars. It would be worthwhile to conduct polarimetric observations for the entire sample of carbon stars from Hamburg ESO survey (Christlieb et al. 2001) for a statistical evaluation of their polarization properties.

Except for HD 100764, our results are the first estimates of polarization for the program stars. While the separation of the stars into two distinct polarization groups is a significant feature, it is not clear at the moment if this is a time-dependent property. While a few of the CEMP stars are confirmed as radial velocity variables and binaries, binarity of others are not yet established. Two important means for establishing variability and binarity of stars are long-term photometric observations as well as long-term spectroscopic radial velocity measurements. Time variation of polarization measurements for these objects would also be an indicator of their variability caused by the morphology of the circumstellar dust clouds that might have originated during AGB mass loss of the companion stars. Thus, the detection of polarization from these objects would lend support to those physical scenarios that consider binarity of the objects to explain their formation mechanism and peculiar abundances of heavy elements.

We thank the staff at IUCAA observatory for time allocation and assistance during the observations. This work made use of the SIMBAD astronomical database, operated at CDS, Strasbourg, France, and the NASA ADS, USA. We thank Prof. A. V. Raveendran for useful discussions. S.S.K. is a JRF in the DST project SR/S2/HEP-09/2007; funding from this project is greatfully acknowledged.