P Cygni in a short S Doradus phase. Spectroscopic and photometric ...
EW within
one night (using CCD spectra obtained at the NAO, Bulgaria) and ... is to guide the eye and to emphasise the presence of a slow component in the ...
Spectroscopic and photometric evidences A&A 376, 898-906 (2001)DOI: 10.1051/0004-6361:20010668P Cygni in a short S Doradus phase. Spectroscopic and photometric evidencesN. The UBV photometry reveals a slow (7.4-year), low-amplitude (0.1mag) variation in V in which the star becomes redder when it brightens and vice versa. Simultaneous spectroscopic observations show changes in the Hequivalent width (corrected for the effect of the changing continuum) in positive correlation with the 7.4-year photometric oscillation in the V-band brightness. This result is interpreted as an indication that in P Cygni an increase in the stellar brightness, during the 7.4-year SD phase, is likely accompanied by an increase in the mass-loss rate. The term "LBVs" was coined by Peter Conti in 1984 and comprises the galactic P Cygni type stars, the S Doradus variables and the Hubble-Sandage variables in M31 and M33. The evolutionary status of these objects is still a subject of ongoing debates. In some stellar models, the LBV phase occurs during (or shortly after) the main-sequence phase, produces an enormous mass loss, and prevents evolution to the red (Schaller etal. The last one isa photometric variation of up to 2.5 mag on a time scale of years to decades and even to centuries. A crucial feature of this variability is that the star becomes redder when it brightens and vice versa. In addition, van Genderen (2001) argued that the term "S Doradus variables" should replace "LBV". A definition has been formulated by a small committee at the Workshop "P Cygni 2000: 400 years of progress" (de Groot & Sterken 2001).Historically, the hypergiant PCygni is the first known LBV. The star satisfies the first two criteria for membership in the class. It has undergone at least one SD-eruption (17th century). It has a high luminosity, (Pauldrach & Puls 1990), a high mass-loss rate, (Scuderi etal. Surprisingly, no indication for any SD-phases was known until Markova etal. During this variation, when the star brightens, the effective temperature decreases by about 10% and the radius increases by about 7% leaving the luminosity almost constant. Unfortunately, due to the time limitation of the available data sets Markova etal. For this purpose, we collected and analysed UBV photometry and Hline profiles or equivalent widths obtained over a period of 13.8years. The observational material used is described in Sect.2, where some problems concerning the homogeneity of our sample are discussed and resolved. The results obtained are described in Sect.3, and the physical nature of the variability is discussed in Sect.4. The data cover the period JD2446174 to JD2451102, which corresponds to March 1985 to October 1998. Some of them have already been analysed by Scuderi etal. The distribution of the data by means of their origin is given in Table1, which also lists the accuracy in V and B-V. The observations at the NAO (Markova) were obtained using the coudé spectrograph of the 2-m RCC telescope. They consist of nine photographic spectra (Kodak 103aF) with resolving power and 7 HCCD spectra with and a wavelength coverage of about 110 Åtaken during May-September 1990 and April-October 1998, respectively. For the latter, the detector was an ELECTRON ISD015A CCD with pixels and a pixel size of m. The CCD frames were uniformly reduced by means of a series of modules written in IDL by T. The procedure is entirely standard and consists of background subtraction, cosmic ray hit removal, flat-fielding, wavelength calibration, and continuum normalisation. Pixel-to-pixel sensitivity variation is removed by division by the mean flat field for the relevant night. During the wavelength calibration, the Earth's motion with respect to the heliocentric rest frame is removed. Observations at Ritter Observatory were carried out by members of the observing team with a fiber-fed échelle spectrograph attached to the 1-m telescope and to a liquid-nitrogen-cooled Wright Instruments Ltd. CCD camera, which incorporates an 7701152 EEV CCD05-20-0-202 sensor with pixel dimensions of m. The raw frames were reduced with Ritter Observatory's standard techniques (Mulliss 1996) under Sun/IRAF V2.10.4-p1. Bias subtraction, flat fielding, wavelength calibration, heliocentric correction, and continuum normalisation were carried out with standard IRAF tasks. Table 2:Summary of Hobservations. The spectrograph was equipped with an ST-6 CCD cameras (pixels; observation window 90Å) from June 1994 to September 1995 and with an HPC-1 (pixels; observation window 250Å) afterwards. In both cases the resolution is about 0.24Å/pixel. Wavelength calibration, continuum normalisation, heliocentric correction, and EW measurements were carried out by means of the local KASPEK package (for more details see Leedjärv 1998). The fourth and fifth columns give, respectively, the spectral resolving power, ,and the number of the obtained spectra (one spectrum per night). The resolving power was 7000 with a coverage of 450Å. A signal-to noise ratio of about50 in the continuum was achieved. The EW estimates
include the contribution of both the line core and the emission wings. A relative error of about 4percent in EW was estimated. The EW estimates include only the contribution of the line core. The spectral flux, expressed in continuum units, has been integrated within 60Å around H(Å). The HEW determinations published by Markova etal. The level of the continuum was determined by linear interpolation between the average values of the stellar counts in two bands situated at about 6510Å and at about6617Å. The line flux was then integrated over these limits. The accuracy of the determinations is Å or less. The EW estimates published by Pollmann (1999) cover the period from 1994 to 1999. They originate from amateur observations made with a combination of a 100-mm Maksutov-type reflector (), an objective prism (refracting angle of 30)and a CCD camera (FT800P, pixels). The equivalent width of His determined by spectral flux integration within an interval of 140Å, namely from 6490 to 6630Å. The accuracy of the determinations is 5percent (private communication). The resulting estimates are shown as a function of time in Fig.2 (lower panel). Information concerning the collected HEW measurements and the relevant observations is summarised in Tables2 and3. The total time coverage of the sample is 10years and 6 months, from July 1988 to Jan. The HEW in PCygni was estimated through integration between two continuum points on either side of the line. Because this line has strong emission wings extending to Å (about )from line centre, to obtain reliable EW estimates requires that the two continuum bands be chosen close to but outside the wavelength region6529 to 6595Å. No correction for the contribution of water vapour lines, CII doublet, or NII forbidden lines was made. However, we estimated that in most cases the total effect of blending did not exceed one percent of the HEW. In the extreme case of Ritter Observatory, which is located essentially at sea level, the contribution of water absorption to the Hequivalent width amounts to less than 2%. Table 3:Hequivalent width measurements. The correction coefficient for a given dataset, ,was derived as ,where a and b are the integration limits relevant to the dataset in question (Å and Å). The value obtained for was then applied to all data in the dataset to obtain .Obviously for equivalent widths derived by integration between 6510Å and 6617Å, e.g. We checked that this comparison is valid by studying the variability of the HEW within one night (using CCD spectra obtained at the NAO, Bulgaria) and found that its hourly variations are negligibly small. For example, the rms deviation derived by averaging over 25 spectra obtained within a period of 8hours on 24 June 1994 amounts to 0.5% of the nightly mean. During the next night (total observing time 7hours), the rms deviation averaged over 16 spectra was 0.7% of the EW.In this way, we found that the accuracy of individual values of is better than 9%. The intention of this line is to guide the eye and to emphasise the presence of a slow component in the photometric behaviour of the star. The data confirm the presence of the slow variation in the stellar brightness suggested by Markova etal. Because our sample includes new data, it reveals the presence of two cycles, both incomplete. The reality of the event around JD2448550 must be checked additionally. The variation in V has a time scale of about 7.4years and an amplitude of about 0.1mag, typical of weak-active S Dor variables in a short SD-phase (van Genderen 2001). In addition, the fact that the scatter of the data around the mean light curve exceeds the error in the individual determinations indicates real variations on a shorter time scale (de Groot etal. Upper panel: data from different sources are marked with different symbols: triangles, Taylor etal. The intention of the dashed line is to guide the eye and to emphasise the presence of a slow component in the Hequivalent width variability. Lower panel: data published by Pollmann (1999).Open with DEXTER3.2 Hequivalent width variabilityThe HEW determinations, both original and collected, are shown in Fig.2 as a function of JD. The data (upper panel) clearly indicate the presence of a slow component - we called it the Very-Long Term (VLT) component - in the HEW variability. The solid and the dashed lines represent the pattern of the SD variability obtained as a seventh-order polynomial fit to the V-band data. The data of Pollmann (1999), shown in the lower panel of Fig.2, seem to support this conclusion. In the top panel of Fig.3, the solid line, a seventh-order polynomial fit to the data points, traces out the V-band variability. In the three panels below, the same curve (scaled in an appropriate way to fit the relevant data) is represented by a dashed line. The purpose of these curves is to guide the eye and to make the comparison easier. However, the Hline flux, which can be obtained from the equivalent width by removing the continuum normalisation, may be more physically meaningful than the equivalent width. For the EW observations that were close enough in time to photometric observations to be considered simultaneous, we scaled the EWs to a constant continuum level chosen to correspond to V0 = 4.8mag. Strictly speaking, differences in the continuum flux at Å should be used, but in our case is a good approximation since the colour indices of PCygni do not vary greatly (Fig.1). The corrected EWs from all observers, ,are shown in the bottom panel of Fig.3. Unfortunately, the lack of complete simultaneity between the spectroscopic and photometric observations has led to a noticeable reduction in the number of available EW estimates. In addition, it appears (if it is not an artefact of bad data sampling) that the corrected EW data follow more closely the pattern of the SD variability. This value is probably a bit overestimated since the wind terminal velocity of P Cygni is expected to change slightly and in opposite to the mass-loss rate (Markova etal. Colour information about this variability is not available. Our survey is a natural continuation of the study of Markova etal. We found that P Cygni has experienced a slow variation in apparent magnitude (amplitude of 0.1mag, time scale of 7.4years) with a colour that is redder when the star brightens and vice versa. Most of the time the VLT spectral variability follows the photometric variability through the 7.4-year SD-phase. In addition, we found that the LT variation in the HEW that was noted by Markova etal. Apositive correlation between this variation and changes in the V-band seems to exist. Using the scaling relation for Hequivalent width (optically thick case) derived by Puls etal. This result allows us to suggest that in P Cygni an increase in the stellar brightness, during the 7.4-year SD phase, is likely accompanied by increase in the mass-loss rate. They conclude that it is not impossible that the observations after 1992 represent a cycle nearly twice as long. Thus, we conclude that, during the last two decades, although in quiescence, PCygni experienced at least two low-amplitude SD cycles with several short-term microvariations (de Groot etal. Concerning the location of the origin of variability, two ideas have been announced. The first is that the SD-phases are an atmospheric phenomenon in which a drastic increase in mass-loss rate leads to the formation of a pseudo-photosphere. In the second hypothesis, the variations are, at least partly, due to variations in the underlying stellar radius, i.e. On the other hand, those authors calculated that mass loss and velocity structure variations may in principle cause V-band variations of 0.2 mag if the optical depth of the wind, (at 5555Å), is initially of order unity. In our opinion it seems more likely, at least for PCygni, that the 7.4-year SD oscillations are a mixture of an expanding radius/decreasing temperature and an expanding pseudo-photosphere. PCygni is near the cool side of the bi-stability jump. Pauldrach & Puls (1990) proposed a "feed-back" mechanism, based on the bi-stability, which can provide cyclic variations in the mass-loss rate of PCygni. If both are present simultaneously the shorter one is superimposed on the longer one (van Genderen etal. No SD-phases were found with cycle lengths between 10-20yr. Another important characteristic of the phenomenon is its multi-periodicity/-cyclicity (van Genderen et al. The time scale of the SD oscillations is obviously too long to be explained in terms of pulsation instability like strange-mode oscillations (Kiriakidis etal. Van Genderen (2001) suggested that the long SD-phases can perhaps be identified with the "secular cycles" in Stothers & Chin's models. However, these models do not predict any noticeable shift of the star on the HR diagram and the observed variations in the stellar brightness and colour should be therefore attributed to changes in the opacity of the wind. This fact limits the implication of the models only to LBV members that show increased mass-loss rate during SD-phases. Besides, the models do not predict a sharp physical division between S- and L-SD phases. Concerning the LT (600)variation in the HEW curve it is perhaps interesting to note that Markova etal. So it seems likely that the LT variability is rather due to a large-scale time-dependent structure in the wind, than to variations in the mass-loss rate. AcknowledgementsNM is grateful to Henny Lamers for his valuable comments and suggestions. The constructive remarks of the referee, Dr. Arnout van Genderen, have helped to significantly improve the paper and are very much acknowledged. Technical support at Ritter Observatory is provided by R. This work was supported by NSF to MES (Bulgaria) trough grant F-813/98 (N.M.). M. 1994, MNRAS, 268, L29 In the text NASA ADS van Genderen, A. M. 2001, A&A, 366, 508 In the text NASA ADS van Genderen, A. J. 1999, MNRAS, 303, 116 In the text NASA ADS de Groot, M., & Sterken, C. Ser., 233, 288 In the text de Groot, M., Sterken, C., & van Genderen, A. M. 2001a, in P Cygni 2000: 400 years of progress, ed. Ser., in press In the text de Groot, M., Sterken, C., & van Genderen, A. Ser., 120, 76 In the text Lamers, H. M. 1995, ApJ, 455, 269 In the text NASA ADS Langer, N., Hamann, W. Ser., in press In the text Markova, N., & de Groot, M. E. 1996, MNRAS, 283, L69 In the text NASA ADS Meynet, G., Maeder, A., Schaller, D., & Charbonnel, C. W. 1995, ApJ, 451, L60 In the text Stothers, R. W. 1996, ApJ, 468, 842 In the text NASA ADS Stothers, R. W. 2000, ApJ, 540, 1041 In the text NASA ADS Taylor, M., Nordsieck, R.