1997MNRAS.285..645K

Mon. Not. R. Astron. Soc. 285, 645-650 (1997)

A new class of rapidly pulsating star - II. PB 8783 C. Koen/ D. Kilkenny/ D. O'Donoghue,2 F. Van Wyk1 and R. S. Stobie 1 'South African Astronomical ObseIVatory, PO Box 9, ObseIVatory 7935, Cape, South Africa 2Departrnent of Astronomy, University of Cape Town, Rondebosch 7700, South Africa

Accepted 1996 October 15. Received 1996 October 8; in original fonn 1996 May 9

ABSTRACT

Some 27 h of high-speed photometric monitoring of the sdB + star PB 8783 has revealed it to be a rapidly pulsating star of the EC 14026 class. Six periodicities in the range 120 to 135 s have been tentatively identified in the light curve. The pulsation amplitudes of all the modes are small, mostly below 5 mmag. Spectrograms show the star to be composite, and UBVRIJHK photometry indicates that the companion to the subdwarf is of early-F spectral type. Key words: binaries: general - stars: individual: PB 8783 - stars: oscillations - stars: variables: other.

1 INTRODUCTION In Paper I of this series (Kilkenny et al. 1997), a new class of rapidly pulsating sdB stars, the EC 14026 stars, was announced. To date, more than half a dozen of these stars have been discovered. Observations of four of these are now sufficient for publication (Kilkenny et al. 1997; O'Donoghue et al. 1997; Stobie et al. 1997b). Three of the well-observed EC 14026 stars were drawn from the Edinburgh-Cape (EC) Blue Object Survey (Stobie et al. 1997a), while the fourth, the topic of this paper, is the previously known sub dwarf PB 8783 (Berger & Fringant 1984). In Section 2 our high-speed photometry of the star is described, and the identification of periodicities in the data detailed. All well-studied EC 14026 stars are composite systems, comprising an sdB star and an FIG star, probably on the main sequence. PB 8783 is no exception to this pattern, and in Section 3 further spectroscopic and photometric measurements of it are described, and used to estimate spectral types, gravities and temperatures for the component stars. Section 4 deals with the evidence in favour of the sdB star, rather than the cooler companion, being the source of the oscillations. 2 FREQUENCY ANALYSIS OF THE OBSERVATIONS During an interval of 11 d, 27.2 h of high-speed photometry of the star were obtained. The observing log is given in Table 1. The integration time used was 10 s in all runs except those on JDs 9957 and 9958, when 5 s was used. The observations were made without any filters in the light

beam. The 0.5-m telescope data were acquired with the South African Astronomical Observatory (SAAO) Modular Photometer and a red-sensitive photomultiplier tube; the data obtained with the 0.75-m telescope employed the University of Cape Town (UCT) High-Speed Photometer with a blue-sensitive tube. We will analyse the data obtained on the two telescopes separately. This is necessitated by the fact that the amplitudes of the variations seen in the 0.75-m data are slightly larger than the O.5-m amplitudes. The reason for this lies in the different spectral responses of the photomultiplier tubes in use on the two telescopes. The reader is reminded that the high-speed observations were obtained in white light, i.e. with no filters in the light beam. The Sll photomultiplier tube in the UCT High-Speed Photometer mounted on the 0.75-m telescope has an effective wavelength similar to Johnson B, although with a much broader bandpass. The GaAs tube in the Modular Photometer on the O.5-m telescope, on the other hand, has an even wider wavelength response. In particular, the GaAs tube has substantial sensitivity extending to the near-infrared, whereas the blue-sensitive Sl1 tube cuts off on the redward side of the Johnson V filter. The proportion of the cooler, hence redder, companion star measured by the GaAs tube is thus larger, causing the variations ofthe blue star to appear smaller (see also Section 4 below). The marked difference in amplitudes observed using the different telescopes is the prime reason that we analyse the two sets of observations separately. If the data were to be treated together, the apparent decrease in the amplitude of the variations in the 0.5-m observations (obtained entirely subsequent to the 0.75-m observations) would resemble

© 1997 RAS

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System

1997MNRAS.285..645K

646

C. Koen et al. Table 1. Log of the high-speed photometric observations of PB 8783. Starting Time JD2440000+

Telescope (metres)

Run Length (Hours)

9955.54 9957.54 9958.45 9960.46 9961.49 9963.52 9965.47 9966.56

0.75 0.75 0.75 0.50 0.50 0.50 0.50 0.50

0.88 3.80 4.91 4.82 3.99 3.22 4.46 1.11

amplitude waning due to mode beating. We would thus expect to identify spurious modes with frequencies close to those truly present in the observations, were the data from the two data sets combined. A sample light curve is shown in Fig. 1. The variable amplitude of the pulsations is clearly visible, and hints strongly at interference between different modes. Thus, the total amplitude of variation is large when the different modes are in phase (e.g. panels four and five of Fig. 1) and small when modes are in antiphase (e.g. the first panel of the figure). An amplitude spectrum of the longest run (4.9 h) is plotted in Fig. 2 (top). A highly significant peak at a period of 122.7 s (/=8.152 mHz) is apparent. Also plotted in Fig. 2 are the spectra of the residuals after pre-whitening by the one (middle) and two (bottom) best-fitting sinusoids respectively. The combination of the modes giving rise to the highest peaks in the top two panels of Fig. 2 obviously gives rise to the beating visible in Fig. 1, as noted above. The bottom panel of Fig. 2 shows that there may be periodicities remaining in the data, albeit at amplitudes at the 1 mmag level. We searched all the runs for significant signals at frequencies outside the range shown in Fig. 2. None were found, with amplitudes exceeding ~ 0.7 mmag, in the frequency range 15-50 mHz. Tentative evidence for very lowfrequency signals will be discussed in Section 4. The results of successive rounds of pre-whitening of the 0.75-m telescope data are shown in Fig. 3. The periodogram of all the observations is plotted in the top panel. The remaining panels show the periodograms of the residuals of the data, after non-linear least squares fitting offrom one to five sinusoids to the data. Initial estimates for each new frequency were found from the position of the highest peak in the periodogram of the residuals from fitting the previous set of frequencies. The frequencies and amplitudes listed in Table 2 were obtained from the last round of fitting. For comparison, Table 2 also shows the results of fitting six sinusoids to the somewhat longer set of O.5-m telescope observations. The agreement between the frequencies found in these independent data sets (different times, telescopes, photometers and observers) is encouraging. [Note that the apparently large difference in frequency between the two frequencies near 7.4 mHz (difference 0.012 mHz) and 7.8 mHz (difference 0.022 mHz) can be ascribed to their being respectively 1 and 2 cycle per day alias pairs: the differences are therefore apparent, rather than real]. The one striking discrepancy between the two sets of frequencies is the doublet of 8.151 and 8.153 mHz found in the O.5-m data, corresponding to a singlet at 8.152 mHz in the 0.75-m

-.025 f-

...

." ::.~~~ ,..: ..•. o rr\"""",o:, . ...:::

.025

-..... ~... ... ..." .:. '.... ~ ,: ...,"":. .".!~~......:..: ~-:':'II'I',,?' ••",,:~.......,.,.: •• I

I

r-". .. . .. - . -....... ,

..

##.,. . . . . . . .

•

•

.:

..

. . -.- .:-

••••

~

:-:.

...:

-:- • ..,

;.

~

-.'

yt .... ....

.

"'.

~

-.- .. .

,.«... ",,':" -."-I' ••.:. ....--;,!\ .... ;

...........

l. v) and Sz(Tz, v). These are obviously incapable of reproducing the Balmer jump. The U - B colour (as well as Figs 5 and 6) indicates that there is a significant Balmer jump, so the U flux was excluded from the blackbody fitting procedure. The uvby data were also not included in the blackbody fitting but were rather used as a check on the consistency of the photometry. The results of the fitting revealed a tight correlation between TI and Tz in the sense that for larger values of Tz, larger values of TI are required [along with a concomitant decrease in the scaling factors (ri/dY and (rz/d)Z]. Three sample fits yielded values for (Ti> Tz) of (20 000 K., 5750 K), (24500 K., 6000 K), (36000 K., 6250 K); these sample fits are illustrated in the top panel of Fig. 6. The blackbody curves of the blue and red stars (with appropriate scaling) for these three sample fits are shown in this panel (temperatures appear next to the corresponding curves). The sum of the contributions is also shown as the curves passing through the observed BVRIIHK fluxes (these three curves are indistinguishable except for log l < 3.6). The correlation between the blue and red star temperatures can be understood as arising from the need for the blue star contribution to decrease at short wavelengths (log l < 3.8) as the red star contribution increases. This can only be achieved if the blue star temperature increases and its radius decreases. We note also that the high temperature of the red star precludes the use of the Allard et al. (1994) flux ratio diagrams to identify accurately the properties of the two stars. Particularly, examination of their fig. 8 shows that the sdB and red star flux ratio sequences are virtually parallel for spectral types of the red star earlier than G8, so that small changes in the orientation of the required connecting line between the sequences cause large changes in the identification of the stellar types.

S-I

cm- 2 A-I. The abscissa is wavelength in A.

Log (Wavelength) 3.40

3.80

4.20

o

o

o ~~~~--------------------~

'"

C'I

o

.no

o :J(T)

...........

C

LL

o

3.40

3.80

4.20

Log (Wavelength) Figure 6. Fits of blackbody curves (top panel) or Kurucz model atmospheres (bottom panel) to the UBVRIJHK (filled squares) and Stromgren uvby (filled triangles) measurements of PB 8783. The ordinate is F, in erg S-I cm- 2 Hz- t • The absicssa is loglo of the wavelength in A. Numerical annotation refers to the temperatures of the corresponding curves. See text for more details. © 1997 RAS, MNRAS 685, 645-650

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System

1997MNRAS.285..645K

A new class of rapidly pulsating star - II As mentioned above, it is clear from Figs 5 and 6 that there is a substantial Balmer jump observed in PB 8783. In order to include this extra information in the fitting, Planck functions were replaced by model atmospheres from Kurucz (1992) to represent Sl (Tl> v) and Sz(Tz, v). For the red star logg=4 models were used, while for the blue star logg=5 models were used (a little lower than typical values of logg for sdB stars). The model spectra with temperatures closest to 36 000 and 6250 K are shown in the lower panel of Fig. 6, along with the sum of these two curves. It is evident that the sum of these two curves has far too much flux compared with the observed U and u fluxes. In addition, the sum has a steeper gradient across the optical region than the observed BVRI fluxes: the observed I flux is slightly, but significantly, too large (the errors on the UBVRI fluxes are smaller than the plotted symbols). In order to explain the size of the Balmer jump, it is clear that the blue star temperature must be reduced (so that its Balmer jump increases) or the red star temperature increased (for the same effect). In order to obtain quantitative estimates, a least squares fit to the fluxes was performed for T1 ranging from 20000 to 40 000 K and Tz ranging from 5250 to 8250 K. Because of the extra complexity involved in using those blue photometric fluxes that include strong absorption features, especially the Balmer jump, only the ubyVRIJHK fluxes were used in these fits. As with the blackbody fits, a tight correlation between T1 and Tz was found, with Tz about

649

1000 K hotter than previously to account for the Balmer jump. Only three points on the (Tl> Tz) grid searched gave acceptable solutions: (20000 K, 6500 K), (24 000 K, 6750 K) and (33000 K, 7000 K). The correct point on this one-parameter family of solutions was identified using the strength of the Ca II K line: as the temperature of the red star increases from 6500 to 7000 K, the Ca II K line increases in strength. In addition, the contribution of the red star to the flux at 4000 A increases relative to that of the blue star. A tight constraint on the red star temperature is thus obtained: the strength of the observed Ca II K line (Fig. 5) could only be explained by the (33 000 K, 7000 K) solution. This fit is illustrated in Fig. 7 and is in satisfactory agreement with all observed fluxes. Much hotter temperatures can be ruled out because the slope of the composite spectrum across the BVRI bands is steeper than observed. From examination of the quality of fits obtained by varying the temperatures in the region of the optimal solution, we estimate errors of 200 and 2000 K for Tz and T1 respectively. From the fit of the Kurucz models, the ratio of the fluxes at 5500 A (the effective wavelength of the V band) was found to be 2.81. Mv for main-sequence stars of T z = 7000 K (spectral type Fl) is 3.3 (Lang 1992), implying that, if both components are at the same distance, Mv of the blue star is 4.4. Heber (1986: tables 6 and 8) finds that the sdB star SB 446 has a temperature of 33500 K and Mv = 4.4. The assumption that PB 8783 is a physical binary is therefore

Log (Wavelength) 0

o

3 . 40

3.70

4.00

4.30 T1: 33000 R1: 0.17 G1: 5.7 T2: 7000 R2: 1. 44 816 d:

CO

o

...

0

-~

.

0",," ~

........... ::l

C

LLo o C\.I

o o~__~____~__~__~____;===~=d o

3.40

3.70

4.00

4.30

Log (Wavelength) Figure 7. Best-fitting Kurucz model atmospheres to the ubyVRIJHK fluxes. The UBVRIIHK fluxes appear as filled squares and the Stromgren uvby fluxes as filled triangles. The model for the blue star has a temperature of 33 000 K, that for the red star has a temperature of 7000 K. Other details as for Fig. 6. © 1997 RAS, MNRAS 685, 645-650

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System

1997MNRAS.285..645K

650 C. Koen et al. well justified. In addition, we can assume, in contrast to other studies of sub dwarf binaries (e.g. Allard et al. 1994, Theissen et al. 1995) where the red star is of later type and consequently more luminous than a main-sequence star, that the red star in PB 8783 is on the main sequence. Using the radius of a main-sequence star of the same effective temperature, the distance of the system and the radius of the blue star can be estimated. Assuming further that the blue star has a mass of 0.5 M 0 , typical of sdB stars (Saffer et al. 1994), loggl can be estimated. The resulting values are: '2 = 1.4 ± 0.1 R 0 , 'I = 0.17 ± 0.03 R 0 , d = 800 ± 120 pc, loggl =5.70 ± 0.15. The wavelength range spanned by the spectrum in Fig. 5 is too short for it to serve as an additional check on the derived physical parameters of the stars. However, the values for Tl> T2 and loggl are consistent with those found from the absorption line profile analysis of O'Donoghue et al. (1997). The quoted errors should not be regarded as being statistically independent of each other, as the relevant quantities are correlated. Note that the above analysis neglects interstellar reddening (although at the high galactic latitude of PB 8783 this may not be a serious oversight) and has used models with Population I composition. The parameters derived above should therefore be treated with appropriate caution. 4 WHICH STAR IS THE PULSATOR?

Although it is more likely that the more compact star, i.e. the sdB star, is the source of the oscillations, this should not be taken for granted. One method of establishing which star gives rise to the pulsations is to determine the amplitude as a function of wavelength. PB 8783 was observed on two nights successively in the U and V filters for ~ 10 oscillation cycles in each filter. Interchange of measurements through the two filters, and then interpolation of the mean amplitude, was necessary in order to remove statistically the effect of amplitude variations arising from beating. The result was a ratio of amplitudes Au/Av= 4.2. The decomposition of the fluxes from the red and blue stars derived in the last section yield UI =2.04 x 10- 25 , VI = 1.15 X 10- 25 (blue star); and UI = 1.02 X 10- 25 , V2=3.23 X 10- 25 (red star). This implies that UI/VI = 1.8, whereas U2JV2 = 0.32; it is thus clear that the sdB star is the source of the rapid pulsations. A great many stars of the spectral type of the cooler star in the PB 8783 pair are known to be b Scuti pulsators, with typical periods of a few hours. We therefore examined the low-frequency parts of our amplitude spectra for excess power. There are indeed similar features at frequencies of about 54 and 135 JlHz (periods of 5.1 and 2.1 h) in the periodograms of the 0.5- and 0.75-m data, but these are not so pronounced as to be entirely convincing. The problem is, of course, that the inevitable slow drifts in atmospheric

transparency that affect our data have the same time-scales as b Scuti variations. 5 CONCLUSIONS

The periodicities identified in our observations of PB 8783, ranging from 120 to 135 s, extend the range of periods seen in the BC 14026 stars to below 130 s (cf. Kilkenny et al. 1997, O'Donoghue et al. 1997, Stobie et al. 1997b). Furthermore, it appears that PB 8783 has a particularly rich oscillation spectrum, making the star an attractive proposition for well-constrained pulsation modelling. Being the brightest of the known BC 14026 stars, fairly accurate absolute photometry of PB 8783 was obtained from the U up to the K band. This photometry allowed the determination of the temperatures of the component stars to within quite narrow limits. Furthermore, estimates of the distance of the system and the gravity of the subdwarf star were made which were found to be in accord with parallel studies. ACKNOWLEDGMENTS

The authors are grateful to Fred Marang (SAAO) and to Kristen Larson (Rensselaer Polytechnic) for obtaining some of the photometry discussed in the paper. REFERENCES Allard F., Wesemael F., Fontaine G., Bergeron P., Lamontagne R, 1994,i\J, 107, 1565 Berger J., Fringant A-M., 1984, A&AS, 58, 565 Bessell M. S., 1979, PASP, 91, 589 Fabregat J., Reig P., 1996, PASP, 108,90 Heber U., 1986, A&A, 155, 33 Kilkenny D., 1995, MNRAS, 277, 920 Kilkenny D., Koen c., Q'Donoghue D., Stobie R S., 1997, MNRAS, 285, 640 (Paper I, this issue) Kurucz R L., 1992, in Barbuy B., Renzini A, eds, IAU Symp. No. 149, The Stellar Populations of Galaxies. Kluwer, Dordrecht, p.225 Lang K. R, 1992, Astrophysical Data: Stars and Planets. SpringerVerlag, New York Loumos G. L., Deeming T. J., 1978, Ap&SS, 56, 285 Q'Donoghue D., Lynas-Gray A E., Kilkenny D., Stobie R S., Koen C., 1997, MNRAS, 285, 657 (Paper IV, this issue) Saffer R A, Bergeron P., Koester D., Liebert J., 1994, ApJ, 432, 351 Stobie R S. et al., 1997a, MNRAS, submitted Stobie R S., Kawaler S. D., Kilkenny D., Q'Donoghue D., Koen c., 1997b, MNRAS, 285, 651 (Paper III, this issue) Theissen A, Moehler S., Heber u., Schmidt J. H. K., de Boer K. S., 1995, A&A, 298, 577 Wilson W. J., Schwartz P. R, Neugebauer G., Harvey P. M., Becklin E. E., 1972, ApJ, 177,523

© 1997 RAS, MNRAS 685, 645-650

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System