4. Discussion and conclusions
Comparing the identifications lists of Beals (1950), de Groot (1969), Ozemre (1978), Markova (1994) and Markova and Zamanov (1995), as well as the spectral atlas published by Stahl et al. (1993) we concluded that more than 70% of the pure emission lines in P Cygni's spectrum (wavelength range from 3500 to 6800Å) appear to be of recent origin. This finding is very important and needs special attention since the appearance of so many new lines in the spectrum would be a remarkable phenomenon in the star's history as it might reflect some changes in P Cygni's atmosphere. But, are these lines really new or is their appearance a result of the more favorable signal-to-noise ratio of the present spectra? A later inspection of the spectra studied by de Groot with the benefit of hindsight showed that some of the forbidden lines (Israelian & de Groot 1992) were already present as early as 1942. One expects any change in the strength of P Cygni's emission lines to be caused or accompanied by a change in the star's Teff. Lamers et al. (1983) argued convincingly that around 1980 P Cygni had Teff =19300 700 K. On the basis of a later photographic study (Lamers & de Groot 1992) it was found that d(log Teff)/dt = -0.027 0.004 per century. This leads to the following temperatures at the following epochs: 1930 (Beals 1950), 19911 K; 1942 (de Groot 1969; Israelian & de Groot 1992), 19762 K; 1990 (this article), 19182 K. Thus, in 60 years Teff has decreased by 729 K, equal to the uncertainty in the determination of Teff. It is, therefore, no surprise that any indication of a change in the emission line strengths can only be marginal, in complete agreement with our finding. A very good look at the older spectra has to be taken before we can be absolutely sure that the photographic spectrum today is really much richer in emission lines than 60 years ago.
Most of the permitted pure emission lines appear to be symmetric with a sharp maximum, blue shifted with respect to Vsys. The similarity in the line-parameters of the FeIII and NII points to a possible formation of both in the same wind layers, where the outflow velocity V 90 km/s. The width of the SiII pure emission lines is considerably grater than that of FeIII and NII and corresponds to an outflow velocity of about 230 km/s, that is, in fact, the terminal velocity of the wind (see Sec. 3.2). Obviously, the Si II emitting region is different from that of FeIII and NII lines. A typical -velocity law with =4, first proposed by Barlow and Cohen (1977), with an initial velocity V0 =4x10-4 V and V =240 km/s (hereafter BC-velocity law) locates the FeIII/NII and SiII emitting regions, respectively, at R 5R and R 100R . All of the observed FeIII and NII transitions are from higher excited levels, which seem overpopulated. At distances R 5R , where these lines probably form, the wind temperature and density are higher than 10x1010 cm-3 and 14000 K, respectively (Drew 1985). In these circumstances fluorescence is a possible excitation mechanism for energy levels between 23.14 and 23.57 eV. In fact, we propose (like Wolf and Stahl 1985) that these levels could be pumped by two UV transitions in HeI atoms at energies of 23.09 eV ( 537Å) and 23.74 eV ( 522Å). In addition, considering the coincidence of the upper levels of FeIII multiplets 113 and 114 with the lower levels of multiplets 118 and 119 one could think about possible cascade transitions between these levels. In the case of the NII levels at energies little above 25 eV dielectronic recombination seems to be more probable. But the problem is that even at the base of the wind the temperature is too low to make this mechanism effective. With respect to the excitation mechanism of the SiII pure emission lines there are two possibilities: fluorescence - pumping of electrons by Lyman and Lyman to levels 42 D and 52 S, respectively - or radiative recombination, both followed by cascade transitions. If these lines are really formed at a distance of about 100R , where the wind temperature and density are below, respectively, 104 K and 5x106 cm-3 (see below), the last possibility seems more probable. In any case a detailed NLTE analysis is necessary to understand the nature of P Cygni's permitted emission-line spectrum
Mean values of -18 2 and 231 3 km/s are derived for Vc and HWZI of the [FeII] lines, respectively. The shape of the profiles - flat-topped - points to a formation of optically thin lines in a constant-velocity outflow. The excellent agreement between HWZI of the lines and V , as determined in Sect. 3.2, confirms the assumption of Stahl et al. (1991) that the [FeII] lines are probably formed in a region expanding at terminal wind velocity.
On our spectra we found evidence for the existence of the red [NII] lines in addition to the yellow one. The shape as well as the width of the profiles appears to be similar and points to a formation in a region with an outflow velocity of about 200 km/s. This region is obviously different from that of [FeII]. Assuming that these three [NII] lines form in the same wind layers, an estimate of the electron density and temperature of the matter can be obtained from their flux ratio. The theoretical intensity of the 6584 line is about 3 times that of the 6548 line. From this we calculate a [NII] ( 6548+ 6584) / [NII] 5755 ratio of 1.50. According to Sobolev's (1975) relation between the above line ratio, Te and ne, a flux ratio between 1 and 2 corresponds to ne =5x106 to 5x107 cm-3 for Te in the range 5700 to 104 K, where the value of 5700 K corresponds to the high-density limit of Sobolev's formula. As Pauldrach and Puls (1990) have shown, the linear velocity law of Waters and Wesselius (1986) as well as the BC-velocity law yield the same density structure at distances larger than 3R . From the mass-loss rate of P Cygni, its stellar radius and its terminal velocity of, respectively, M=1.5x10-5 M /yr, R =76 R (Lamers, 1989) and V = 240 km/s (present study) we derive an electron density between 5x106 and 5x107 cm-3 at distances from 30 to 110 stellar radii. (The matter is assumed to be fully ionized hydrogen.) At these distances the linear velocity law of Waters and Wesselius (1986) gives V=V implying a flat-topped profile with HWZI= 240 km/s for an optically thin line formed there; this is contradicted by the observations. At the same distances the BC-velocity law gives a velocity of 220 to 230 km/s (for V =240 km/s). The observed width of the [NII] profiles, however, is below these values. Obviously, a velocity law flatter than one with =4, but yielding the same density structure, would give better results. Such a velocity law was obtained by Pauldrach and Puls (1990) in the framework of their best model. Its terminal velocity, however, is quite low (195 km/s compared with our value of 240 km/s). In addition, we estimate that the adopted stellar parameters in combination with a BC-velocity law locate the region of [FeII] line-formation at R 110R , where the density drops below 5x106 cm-3.
Adopting for P Cygni a distance of 1.8 kpc, we obtain an angular radius smaller than 1.5 arcsec for the [NII] emitting region. This is in agreement with Stahl (1989) who determined the angular distribution of emission in the [NII] 6584 line, corrected for the average flux from the continuum on either side of the line, and concluded that beyond an angular radius of 1.5 arcsec no evidence for excess [NII] 6584 emission exists. But both results (ours and Stahl's) are in contradiction with Johnson et al. (1992) who did not find any [NII] emission in their on-star spectrum. Differences in the slit position angles used could possibly cause such a discrepancy. This points to a possible asymmetry in the emitting [NII] region close to the star.
© European Southern Observatory (ESO) 1997
Online publication: April 8, 1998