3 Multicomponent structure of the spectrum

Figure 3: Upper: the spectrum of RW Aur A (average of 1998). Note that the broad emission belongs to Fe II 5991.37Å, and the superimposed narrow absorption to Cr I 5991.86Å. Lower: the spectrum of $\gamma$ Cep (K1III-IV), artificially veiled by a factor of 3

Figure 4: An example of night-to-night variability of the broad emission lines in 1998

Figure 5: A fragment of the average spectrum of RW Aur A (over all seasons) with broad emission lines compared to the spectrum of 41 Cyg (F5II). Note that the Fe I emissions are split by narrow absorptions, while the Fe II, Si II emissions are not

Figure 6: Fluxes of the broad emission lines correlate with colour (fluxes in units of 10-12ergcm-2Å-1)

In this section, we give a brief overview of different spectral features. Several components can be distinguished in the spectrum. The weak absorption lines (WALs) can be identified as the highly veiled photospheric spectrum of a K1-K4V star (average veiling = 3; see Fig. 3). In the spectral classification, we try to avoid the lines originating from low excitation levels, because they can be enhanced by additional absorption in the accreting gas well above the photospheric level (Stout-Batalha et al. 2000). This effect is quite strong in density sensitive lines, like Ba II 6141Å ( $\chi_{\rm exc}=0.7$eV). More details about the accretion enhancement of absorption lines, or shell lines, are given in Sect. 7. The width of the WALs, if interpreted only as rotationally broadened, corresponds on average to $v\,\sin i=30$kms^-1, but was found variable from night to night from 16 to 40kms^-1. Variability in radial velocity of the WALs is discussed in Sect. 6. Hereafter, we refer to these weak absorption lines as the "photospheric spectrum'', although there may be a contribution from the layers above the photosphere (the shell). The spectral region, where the photospheric spectrum is least blended with emission lines, is shown in Fig. 3. In some nights, the veiling was so high that very little of the photospheric spectrum remained visible.

Figure 7: Variability of the inverse P Cyg-profiles of O I 7773Å and He I 5875Å

Figure 8: The two most different spectra show the range of variability in the region of the Na I D and He I lines. Note the strengthening of the Ba II absorption when the accretion components are stronger. The numbers at the blue lines are the excitation potentials of the lower levels in eV; those at the Na I lines indicate velocities relative to D1

Figure 9: The narrow absorptions on top of the broad emissions (upper spectra) turn into narrow emissions (lower spectra) when the veiling is very high

Figure 10: Variability of lines with blue-shifted central absorption components

An outstanding characteristic of the spectrum are the numerous intensive broad emission lines (BELs), most of them belonging to neutral and singly ionised metals. The FWHM of the BELs is 200-280kms^-1for Fe I and Fe II lines, and up to 500kms^-1 for H$\alpha$. The line profiles of the BELs are variable on a time scale of one day. No large variations of the line profiles were noticed during 1-2 hours. An example of the BELs' variability is shown in Fig. 4. The blue and red wings of the BELs at the line base remain symmetrically extended to $\pm 200{-}250$kms^-1, while the intensities of the red and blue parts of the profile can change considerably.

Most of the BELs can easily be identified using the spectrum of the supergiant 41 Cyg (F5II) for comparison (see Fig. 5). Note, that narrow absorptions, similar to the WALs, can be found superimposed on top of the broad Fe I emissions, but are usually absent in the Fe II emissions. As a result, the BELs of Fe II look split into two parts, while the BELs of Fe II and other ions have a more triangular profile. These narrow absorptions on top of emission lines vary in radial velocity and width in correlation with the WALs. One might identify them as the photospheric lines seen through the optically thin emission. Then, the absence of these absorptions on top of the Fe II lines is understandable: the Fe II lines are very weak in the photospheric spectrum of a K dwarf. However, the average profile of these absorptions is systematically broader than that of the WALs.

We measured the equivalent widths of 25 almost unblended, broad emissions of Fe I and Fe II observed in the spectrum with the most intensive lines, and in the spectrum with the least intensive lines. Then, the equivalent widths were converted into fluxes using our photometric data. In both cases the lines can be brought to a curve of growth with $T_{\rm e}=4300\pm500$K and $\log n_{\rm e}=9\pm0.5$. However, the lines can be formed in non-LTE conditions, and these values should be considered as rough estimates.

The fluxes in the emission lines show no clear correlation with the brightness of the star. Instead, there is a good correlation between the line fluxes and the B-V colour (Fig. 6). This correlation is partly due to the contribution of the emission lines to the B and V magnitudes. For the spectrum with the most intensive emissions, we estimated the contribution to the B passband as $0\hbox{$.\!\!^{\rm m}$ }4$, and to the V passband as $0\hbox{$.\!\!^{\rm m}$ }2$. That is, most of the B-V range in Fig. 6 is caused by this effect of the emission lines. The full range of variations in B-V is, however, much larger (see Fig. 1). Other mechanisms, like temperature variations or extinction by circumstellar dust may enter.

The next obvious spectral features are the red-shifted absorption components in many lines. We will refer to these as the accretion components. In some lines, like the O I 7773Å triplet, Na I D₁/D₂, He I D₃, the accretion components are present permanently, though strongly variable in strength. The maximum velocity (extension of the red wing) is about 400kms^-1. Examples of these variations are shown in Fig. 7. Note, that the residual intensity at the bottom of the red-shifted absorption in the oxygen line can be as small as 0.4 of the continuum intensity. In Fig. 8 we show the two most differing spectra in the region of the D₁, D₂ and D₃ lines. In the following analysis we will use as an "accretion parameter'' the equivalent width (EW) of the D₁ absorption between the velocities of +200 and +400kms^-1. There are many other spectral lines, both neutrals and ions, which occasionally show strong accretion components. The spectrum of the accreting gas is described in more detail in Sect. 7. Because of the accretion components, the maximum intensity (or centre of gravity) of the BELs is usually blue-shifted. For example, for He I broad: -36kms^-1; Fe II 5316Å: -28kms^-1; Pa13: -25kms^-1; Fe I 5455Å: -14kms^-1 and Fe I 6191Å: -8kms^-1.

Besides the BELs, there are a few narrow emission lines (NELs) with FWHM of about 40kms^-1. A comprehensive study of the narrow emission lines in spectra of T Tauri stars (including RW Aur) was done by Batalha et al. (1996) with one conclusion being that the lines are formed near the magnetic footpoints of the accretion column. In our spectra, the narrow emission components are clearly visible in He I 5875Å, 6678Å and 7065Å. These lines have both broad and narrow emission components. The line profile of He I can be decomposed into three Gaussians: a broad emission with FWHM  =200-250kms^-1, centred at $-40\ldots-50$kms^-1, a narrow emission with FWHM =35-60kms^-1, centred at about +10kms^-1, and an accretion component with FWHM =150kms^-1, centred at about +250kms^-1. Only a narrow component is present in the He II 4686Å emission, at the average radial velocity of +20kms^-1. Occasionally, weak narrow peaks can be found on top of many other lines in the spectrum which shows the highest veiling (HJD 2450382.5), as shown in Fig. 9. The line D₃ is present in all of our spectra (it falls in the middle of the spectral order), and therefore we use it for the analysis of the NEL correlation with other parameters in Sect. 6.

A blue-shifted absorption component indicating gas outflow (wind) is a typical characteristic of H$\alpha$, H$\beta$, Na I D and the IR triplet of Ca II. Examples of variability in these line profiles are shown in Fig. 10.

And, finally, forbidden lines are always present in the spectra, e.g. [O I] 6300Å, [S II] 6716Å and 6731Å. The line profiles are similar to those published by Hamann (1984).