THE ASTROPHYSICAL JOURNAL, 460:L49[–]L52, 1996 March 20
© 1996. The American Astronomical Society. All rights reserved. Printed in
U.S.A.
-------------------------------------------------------------------------------
Next: §4. DISCUSSION
Previous: §2. THE 5.52 YEAR SPECTRAL CYCLE
Top: TITLE PAGE
-------------------------------------------------------------------------------

                          §3. PHOTOMETRIC OSCILLATIONS

     The highly homogeneous NIR photometry of Whitelock et al. (1994) is an
important piece of data to study the S Doradus behavior of [&eegr;] Car. The
H-band data reported by these authors, transformed into flux units, are plotted
in Figure 1, together with the epochs of minimum excitation, predicted by the
spectroscopic period. There is a remarkable correlation between the He I Weq
variations and the NIR light curve, the minimum excitation phases corresponding
to the maximum continuum brightness. This behavior characterizes [&eegr;] Car
as a bona fide S Dor variable. In this case, the low-excitation phases would be
caused by the radial expansion and photospheric cooling of the central star
(Lamers 1995). The amplitude of the S Dor variability is smaller in shorter
wavelengths (van Genderen et al. 1994), probably because of the light
absorption and scattering inside the Homunculus. Light-curve peaks, however,
can be seen up to the V band in some shell events.

     At close inspection, it can be noticed that for almost each feature in the
NIR light curve there is a corresponding one in the He I Weq. The long-term
decrease in He I Weq has a corresponding trend in the brightening of the
continuum. This trend could be explained by a secular expansion and
photospheric cooling of the central star, rather than the decrease in the
circumstellar reddening (van Genderen et al. 1994). The H-band maxima culminate
in narrow peaks coinciding with the sharp minima of the He I Weq, which
indicates that the variations are not sinusoidal.

     The NIR light curve gives support to the 5.52 yr period in two ways. The
first is that a shell event predicted to have occurred in 1975.88 coincides
with a photometric peak, observed in 1976.1. The difference in phase is only
0.04, despite the poor sampling in the NIR light curve at that epoch.

     Second, the period derived only on the basis of the NIR light curve is
very similar to the spectroscopic one. In order to derive an independent
period, I averaged the time interval from peak to peak in the Whitelock et al.
(1994) H-band data. The recurrence time is 1971 days, with an uncertainty of 43
days, close to the observational time sampling. This recurrence time is
indistinguishable from the spectroscopic 2014 day cycle (5.52 yr), although it
was derived using different techniques in different time intervals. Whitelock
et al. (1994) obtained a somewhat shorter period (1870 days) by Fourier
analysis of their NIR data. However, in the present case, the method based on
the sharp peaks must be preferred, because the light-curve irregularities can
contaminate the results. I assumed, in this Letter, that the period is best
represented by the spectroscopic data, because it spans a longer time interval.

     If the periodicity is as stable as it seems to be, it could be seen in
other time intervals. Epochs of zero phase (maximum light or minimum
excitation) were predicted for the past centuries and compared with the
observations reported by Innes (1903). The most ancient data are not very
reliable, but the magnitudes after 1811 are referred to a system of known
stars. Before entering in the big fading phase, after 1850, [&eegr;] Car
underwent three great bursts, in 1827, 1838, and 1843, reproduced from Innes
(1903) in Figure 2. According to the 5.52 yr period, peaks should have occurred
in exactly those years and quite close to the predicted zero phases, as seen in
Table 2 and Figure 2. The mean deviation in phase, derived from this set of
epochs, is [&vbm0;]O - C[&vbm0;] = 0.03 and is of the same order of magnitude
as the time sampling. It is unlikely that this coincidence is accidental.

[Image] Fig. 2

     I have surveyed all the historical data, excluding the time interval in
which [&eegr;] Car was faint (1860[–]1948), when the light curve
oscillations disappeared or decreased strongly in amplitude. In the papers of
Innes (1903) and of Polcaro & Viotti (1993) I found additional events in which
the star was noted as bright. The 5.52 yr period predicts epochs of zero phases
not far from some of these events. This is very curious but cannot be used to
give further support to the 5.52 yr period because of the low confidence of the
data. On the other hand, I did not find a single evidence against the proposed
periodicity.

     If the three great bursts of the last century were really connected to the
present ongoing oscillations, the giant eruptions of the LBVs would be due to
the same mechanism that drives the S Doradus variability. This point deserves
further attention, especially through carefully planned observations of the
next minimum excitation phase of [&eegr;] Car.

-------------------------------------------------------------------------------
Next: §4. DISCUSSION
Previous: §2. THE 5.52 YEAR SPECTRAL CYCLE
Top: TITLE PAGE
-------------------------------------------------------------------------------