next previous
Up: A representative sample of


Subsections

4 Spectral classification

4.1 General methodology

The observed spectrum of Be stars is a composite of the photospheric absorption spectrum and the spectrum produced by the envelope, i.e., an additional continuum component on which emission and absorption lines can be superimposed. For most Be stars, the contribution of the envelope to the continuum in the classical "classification region''($\lambda \lambda$ 3900-4900 Å) is not very important (Dachs et al. 1989). However, the envelope can still contribute emission in the lines of H I, He I and several metallic ions. Weak emission can result in the "in-filling'' of photospheric lines, while stronger emission results in the appearance of emission lines well above the continuum level. With sufficient spectral resolution and a high enough $v \sin i$ the emission lines appear double peaked. A well-developed "shell'' spectrum, with a large number of metallic absorption lines can completely veil the photospheric absorption spectrum (see the spectrum of BD +02$^{\circ}$3815 in Fig. 3).

Since the relative strengths of several He I lines intervene in most classification criteria for the MK system in the spectral range of interest, the spectral classification of Be stars has always been considered particularly complicated. In many spectra, the in-filling of He I lines affects the main classification criteria. When Fe II emission is present, several lines which are used as classification criteria can be veiled (such as the Si II $\lambda \lambda$ 4128-4130 Å doublet).

The high resolutions obtainable at high signal-to-noise ratio with modern CCD cameras improve the situation, since they allow us to disentangle lines that were blended at the resolutions formerly used for spectral classification. On the other hand, the improved resolution means that the traditional criteria are not always applicable. Our spectra have a much higher resolution than the 63 Å mm-1 plates used by Walborn (1971) to define the grid for early-type B stars. Given that the only acceptable methodological procedure in the MK scheme is the comparison of spectra (Morgan & Keenan 1973), all this results in a strong dependence on the choice of standard stars. Unfortunately, the standard stars available for observation are limited by the position of the observatory and time of the year. Our standard stars, taken from the lists of Walborn (1973) and Jaschek & Gómez (1998) are listed in Table 3. We point out that Jaschek & Gómez (1998) give HD 23338 (19 Tau) as a B6V standard and HD 196867 ($\alpha$ Del) as B9V, while Morgan & Keenan (1973) give them as B6IV and B9IV respectively. At our resolution, neither of the two spectra can be justifiably classified as main sequence objects and we endorse the subgiant classification. Indeed HD 23338 looks remarkably similar to HD 23302 (17 Tau), which is given by Jaschek & Gómez (1998) and Lesh (1968) as the primary B6III standard.

For classification purposes, we compared the spectra of the Be stars with those of the standard stars in the interval $\lambda \lambda$ 3940-4750 Å both at full instrumental resolution and binned to 1.2 Å/pixel to mimic the resolution of photographic plates. The comparison was done both "by eye'' and using the measured equivalent widths of the relevant features. All the spectra have been classified in this scheme without previous knowledge of spectral classifications existent in the literature. The derived spectral types are listed in Table 4. Representative spectral and luminosity sequences are shown in Figs. 1 and 2, and some peculiar spectra from the sample are shown in Fig. 3. We find that the accuracy that can be obtained in the classification depends in part on the spectral type of the object, as described below.


  
Table 4: Measured Spectral Type and $v \sin i$ for the Be star sample

\begin{tabular}
{llr}
\hline
OBJECT & Spec. type & $v \sin i$\\  
\hline 
CD $-$...
 ...& B7V & $229 \pm 10$\\  
BD +58 02320 & B2V & $243 \pm 20$\\ \hline\end{tabular}

  
\begin{figure}
\setlength {\unitlength}{1.0in}
 
\begin{picture}
(6.0,5.0)(0,0)
\put(-0.8,-0.8){\epsfbox[0 0 2 2]{h1180f1.ps}}\end{picture}\end{figure} Figure 1: Spectral sequence for main sequence stars. Note the progressive decline of the HeI spectrum from a maximum at B1-B2 and the increase of MgII $\lambda$4481 Å with spectral type. We note that the spectrum of BD +37 675 shows MgII $\lambda$4481 Å $\simeq$ He I $\lambda$4471 Å, which defines B8. However, since HeI $\lambda \lambda$4711, 4009, 4121 Å and CII $\lambda$4267 Å are still visible, we have preferred a B7V spectral type, assuming that HeI $\lambda$4471 Å is partially filled-in. If this is not the case, an intermediate spectral type (B7.5V) would seem necessary
  
\begin{figure}
\setlength {\unitlength}{1.0in}
 
\begin{picture}
(6.0,3.0)(0,0)
\put(-0.8,-0.8){\epsfbox[0 0 2 2]{h1180f2.ps}}\end{picture}\end{figure} Figure 2: Luminosity Sequence near B1. Note the increase in the metallic (mainly O II) spectrum with luminosity class, not so obvious in BD +27 850 because of the later spectral type. The emission veiling in BD +56 473 is too strong to allow an exact classification, even though the strong O II + C III near $\lambda$4650 Å seems to favour the giant classification

  
\begin{figure}
\setlength {\unitlength}{1.0in}
 
\begin{picture}
(6.0,3.0)(0,0)
\put(-0.8,-0.8){\epsfbox[0 0 2 2]{h1180f3.ps}}\end{picture}\end{figure} Figure 3: Some peculiar spectra in our sample. BD +56$^{\circ}$484 (B1Ve) has a very strong emission spectrum, with abundant Fe II features. For late types, Fe II is only seen in shell spectra, like that of BD +43$^{\circ}$1048 (B6IIIshell). BD +02$^{\circ}$3815 (B7-8shell) shows a fully developed shell spectrum. The earliest spectrum in our sample is that of BD +28$^{\circ}$3598 (O9II), where no evidence of emission is detectable. Finally, BD -12$^{\circ}$5132 (BN0.2IIIe) is N-enhanced, showing strong N II $\lambda \lambda$ 3995, 4044, 4242, 4631 Å among others, while C II $\lambda \lambda$ 4076, 4650 Å are absent

4.2 Early B stars

For stars earlier than B3, the classification can be adequately performed using as main indicators the Si IV and Si III lines. The strength of these lines and of the O II spectrum is very sensitive to temperature and luminosity variations. Moreover, the emission spectrum does not generally extend shortwards of $\sim \lambda 4200$ Å. As a consequence, most of our determinations in this spectral range are very secure. Even though the spectral grid is finer than at later types, we believe that most objects are correctly classified to the sub-subtype, i.e., a spectrum classified as B0.7III is certainly within the range B0.5III-B1III and both B0.5III and B1III look inadequate classifications. The luminosity classification is also secure in this range, where we are able to differentiate clearly between giants and main sequence objects.

4.3 Late B stars

For stars later than B5, all the classification criteria available are strongly affected by the presence of an emission continuum. This has resulted in our determinations for this spectral range being slightly less secure than for earlier spectral types. However, with few exceptions, we have been able to assign a spectral type to the correct subtype. This means that we feel that a star classified as B6V would be inadequately classified as B5 or B7. The luminosity classification is slightly less certain. For this reason, we have resorted to using two extra criteria, namely, the number of Balmer lines that could be resolved in the spectrum as it approaches the Balmer discontinuity and the full width half maximum of H$\theta$ (3797 Å), which, among the standard stars, correlates strongly with luminosity class at a given spectral type and is not generally affected by emission in the Be stars. Overall, the three methods do not show strong discrepancies and our luminosity classification can be considered secure, at least to the point of discriminating between giant and main sequence stars.

Lesh (1968) defines B8V by the condition Mg II $\lambda$4481 Å $\simeq$ He I $\lambda$4471 Å. However, as can be seen in Fig. 4, the spectrum of HD 214923 ($\zeta$ Peg), given by Jaschek & Gómez (1998) as B8V standard, shows the Mg II line to be clearly stronger than the He I line. Therefore, this object must be of a later spectral type. Comparison with HD 196867 ($\alpha$ Del) shows that this object is not later than B9V. We have taken the spectral type of this object to be B9V, though we believe that B8.5V could be an adequate interpolation. No object in our sample is so late as HD 214923 and therefore we assign to BD +55$^{\circ}$2411, the only object in the sample with Mg II $\lambda$4481 Å clearly stronger than He I $\lambda$4471 Å a spectral type B8.5V.

In the spectral region B7-B8, where the ratio between Mg II $\lambda$4481 Å and He I $\lambda$4471 Å is the only classification criteria, in-filling can strongly affect the derived spectra. For that reason, we have also used the strength of the Si II doublet and of the whole He I spectrum as additional information. Moreover, at this resolution, C II $\lambda$4267 Å is visible in main sequence stars up to spectral type B7V. Fe II $\lambda$4232 Å starts to be visible in the B8V spectra, but we have not used it as a classification indicator since it can also be a weak shell line.

  
\begin{figure}
\setlength {\unitlength}{1.0in}
 
\begin{picture}
(6.0,3.0)(0,0)
\put(-0.8,-0.8){\epsfbox[0 0 2 2]{h1180f4.ps}}\end{picture}\end{figure} Figure 4: Spectral sequence for late B-type stars. In BD +50$^{\circ}$825 (B7Ve), He I $\lambda$4471 Å is still stronger than Mg II $\lambda$4481 Å, but in BD +19$^{\circ}$578, the two lines have about the same strength, making the object B8V. He I $\lambda \lambda$ 4009, 4121 Å are not visible in this spectrum. Both BD +55$^{\circ}$2411 and HD 214923 have Mg II $\lambda$4481 Å stronger than He I $\lambda$4471 Å, which makes them later, even though HD 214923 is given as B8V standard. Comparison of HD 214923 with HD 196867 (B9IV) and HD 222661 (B9.5V) shows that it is not later than B9V. BD +55$^{\circ}$2411 is earlier than HD 214923, since it shows stronger He I $\lambda \lambda$ 4026, 4387 Å and no sign of Fe II absorption. Therefore a spectral type B8.5V seems justified. This spectrum shows no sign of emission over the whole classification range

4.4 Shell stars

For the purpose of spectral classification, we have only marked as "shell'' stars those showing narrow absorption Fe II lines, either on top of emission lines or blanketing the continuum. Several other stars show double-peaked emission split by an absorption core in some Balmer lines, but this emission is still inside the photospheric absorption feature and does not reach the continuum (e.g., BD +37 675 and BD +42 1376 in Fig. 1). The shell definition is not applicable since no iron lines are visible (see Hanuschik 1995). Exceptions could be BD +47 857 and BD +50 3430 which seem to show absorption cores in some Fe II emission lines and therefore could be shell stars. We will revisit this question in future papers when we discuss the emission line spectra of the objects.

4.5 Distribution of spectral types within the sample

  
\begin{figure}
\includegraphics [width=14.1cm]{h1180f5.ps}\end{figure} Figure 5: The distribution of our sample by spectral subclass for luminosity classes III, IV and V and the shell stars (solid area). Also plotted on each histogram (hollow area) is the overall distribution neglecting luminosity class. The data have been binned into bins n containing objects in the range range B(n-1).5 to B(n).4. Therefore a B0.2 object will appear in the B0 bin, however a B0.5 object appears in the B1 bin
In Fig. 5 we plot histograms showing the total number of objects of each spectral type and luminosity class in our sample. For the purposes of the histograms and the discussion that follows we have classified the two early objects we give luminosity class II as III, and the objects we classify as "III-V'' as IV. Apart from the very earliest objects, which are all giants, the distributions of objects between the three luminosity classes with respect to spectral type are similar. This is probably a selection effect in that there are very few early-type stars, so most are far away and our magnitude limited sample only selects the more luminous giants. From Fig. 6 we note that these very early objects are near the magnitude limit of our sample, thus supporting this interpretation.

  
\begin{figure}
\includegraphics [width=7.3cm]{h1180f6.ps}\end{figure} Figure 6: Apparent B magnitude versus B spectral subclass for the sample. Luminosity class is indicated by the symbols, where a circle indicates III, a cross IV and a square V. Objects with a shell spectrum are indicated by the triangles
Considering the sample as a whole it is interesting to note the majority (34 out of 58) objects are dwarfs, with only 13 out of 58 unambiguously classified as giants. Recall that in Sect. 2 our selection criteria were designed to give an equal number of these objects. This implies that in spectral types given by Jaschek & Egret (1982) many objects that are classified as luminosity class III should in fact be classified as IV or V. In this case it is unlikely that this is a bias created by our selection of objects from the catalogue, as our magnitude limit would be expected to select the more luminous giants over the dwarfs.


next previous
Up: A representative sample of

Copyright The European Southern Observatory (ESO)