Forum Springer Astron. Astrophys.
Forum Whats New Search Orders

Astron. Astrophys. 328, L37-L39 (1997)

Previous Section Next Section Title Page Table of Contents

4. The radial velocity of [FORMULA] Crt

Smalley et al. (1997) measured the RVs of individual lines, which were then averaged for each spectrum. This procedure was also attempted here. However, the rotational velocity of [FORMULA] km s-1 (Smalley et al. 1997) leads to severe blending which makes identification of the features and assignment of a laboratory wavelength to a certain blend difficult. The result was that in a typical spectrum the RVs of individual lines varied between -10 and +12 km s-1. The mean RV also sensitively depends on the subjective choice of what is considered as outlier not to be taken into account in the average. The mean error of the average RV is about 2-3 km s-1, similar to what Smalley et al. (1997) report for their RVs. Considering the error and the substantial uncertainty of the laboratory wavelengths of the features, it is surprising that Smalley et al. obtained RVs, which - with only one exception out of 7 measurements - are within 0.3 km s-1 around +10.5 km s-1.

We therefore decided to use a different method. Adopting the atmospheric parameters obtained by Smalley et al. (1997), i.e. [FORMULA] K, [FORMULA] and [FORMULA] km s-1, and assuming solar abundances (consistent with the results of Smalley et al.), model spectra for the two wavelength regions were computed, based on Kurucz (1993) atmospheric models and the atomic line list obtained from VALD (Piskunov et al. 1995). The RVs of [FORMULA] Crt are then obtained by cross-correlation of each observed spectrum with the model. The results are given in Table 1.


Table 1. The heliocentric radial velocities of [FORMULA] Crt obtained by cross-correlation of the observed spectra with a model spectrum. The columns give the heliocentric Julian date (HJD) of mid-exposure, the central wavelength of the spectral region observed, the RV from the cross-correlation, and the weight of the RV used in computing the mean RV.

In order to reduce the influence of noise, all spectra were filtered using a Gaussian filter of FWHM=0.03 nm (about 3 pixels). The model spectrum contains the same pattern of blends as the observed spectrum, however, the depths of some features are quite different (mainly the weak features are stronger in the observed spectrum than in the model). For the final RVs in Tab. 1 only those features were taken into account during the cross-correlation whose depths and shapes were similar to those in the observed spectrum. These were 4 strong lines/blends in the 490 nm region and 2 in the 505 nm region; therefore, we introduce a relative weight of 0.5 for all RVs from 505 nm spectra. The first spectrum has a strong contamination by a solar spectrum, which was subtracted during the reduction. The RV from this spectrum, however, is much larger than that from the 490 nm spectrum taken immediately afterwards; the difference is at least twice that met in other cases of two or more spectra taken in the same night (cf. Tab. 1). We therefore decided not to use this RV (weight 0).

By comparing the RVs from spectra taken in the same night, a measure of the RV-error is obtained. Using the weighted standard deviation, the RV error is between 1.1 and 1.5 km s-1 ; this is a factor of 2-3 better than the one obtained from identifying individual lines and averaging, and better by the same factor than the error given by Smalley et al. (1997).

Comparing the RVs in Tab. 1 with each other, there seems to be no systematic: no systematic difference between the two spectral regions and also no systematic behaviour of the RV with time. We therefore computed the weighted mean RV of [FORMULA] Crt:


where [FORMULA] is the standard deviation of an individual RV from the mean. It is not significantly larger than the above estimated error of an individual RV; this is another evidence for constant RV of [FORMULA] Crt during the time span of 8 days.

On the other hand, our mean RV is significantly different from the mean RV reported by Smalley et al. (1997). Of course, before discussing this, we need to look for systematic errors. The analysis of the solar spectra showed that our spectra have no systematic error in the velocity zero point. The main difference is the different methods by which to analyse the spectra. In those cases, where we attempted to identify the individual lines, the mean RV obtained are consistent with the RVs reported in Tab. 1, considering the much larger error of these RVs. If there is a systematic error at all, it is in the opposite direction: the mean RVs from individual lines tended to be smaller than the ones from the cross-correlations. The differences in line depths between the observed and model spectra indicate systematic differences between the spectra and the template used in the cross-correlation; however, if the pattern is similar no systematic error in the position is expected. To check this, also cross-correlations with other features and with the whole spectrum were done. The RVs are always consistent with those in Tab. 1within the errors.

We thus conclude, that the RV of [FORMULA] Crt is different from the one obtained by Smalley et al. (1997) in Mar.-May 1994 and Jan. 1995. Assuming that there is no systematic error in the velocity scale of Smalley et al., the RV of [FORMULA] Crt is clearly variable, within a range exceeding the difference between the two mean velocities, i.e. more than 8 km s-1, thus confirming the results obtained early this century (Campbell & Moore 1928). The reason for the variability may be binarity, but the data obtained so far do not allow to derive a period. The amplitude of the RV variation argues against a very long period, but it must certainly be much longer than 8 days. The period of 7.87 days reported in Sect. 2must be considered spurious caused by the small number of measurements and the extremely sparse sampling. Since the RVs reported by Campbell & Moore (1928) are between -3 and +16 km s-1 and our constant RV is near the minimum and the one given by Smalley et al. (1997) is near the maximum of this range, an unfortunate sampling of an elliptical RV-curve missing the minimum or maximum is also no longer an explanation. Given the considerable brightness of the star, and its chance of being a Sirius-like system, we urge all observers, who have the opportunity, to take spectra of [FORMULA] Crt in order to find the cause for the RV variations.

Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 1997

Online publication: March 26, 1998