The star S23 was classified as a B8 star in the early eighties. This classification was confirmed later on based on the observed He I absorption lines in its optical spectrum obtained in 1989. Hence, an effective temperature of about 11 000 K was assigned to S23. We re-observed S23 about 11 years later, in October 2000, with the Fiber-fed Extended Range Optical Spectrograph (FEROS), while FEROS was still attached to the 1.52-m telescope at the European Southern Observatory (ESO) in La Silla (Chile). Surprisingly, our spectra did not show indications for any He I line. This would mean that the spectral type of the star is A0 or later. To determine the effective temperature of S23 at the two observing epochs, we made use of the typical temperature indicator given by the line ratio of the two adjacent lines Mg II (λ4481) and He I (λ4472). From theoretical considerations, this line ratio strongly increases with decreasing effective temperature, as obvious from the dashed line in the following figure. Re-analysing the data from 1989, we confirmed the effective temperature estimate of about 11 000 K, but using our own data, in which the He I (λ4472) line is missing, we could derive only an upper limit for the temperature by assuming that the He I (λ4472) line is just at the detection limit. This delivered a maximum value for the effective temperature of S23 smaller than 9 500 K and would classify it as an A1 star or later.
Even more surprising than the apparent cooling of the star seems to be the simultaneous evolution of its rotation speed given by the projected rotation velocity v sin i as obtained from the width of the Mg II (λ4481) line: while the old spectrum from 1989 delivered a value for v sin i of about 110 km/s, our new observations indicate a much higher rotation speed of 150 km/s. Given the set of stellar parameters of S23, such a high rotation velocity would mean that the star, similar to S65, would rotate with at least 75% of its critical velocity.
Our results are astonishing, since they suggest that within 11 years, the star was spinning up by about 35% while it was at the same time cooling down by more than 1500 K, a scenario which seems not to be physically plausible, for instance with respect to stellar evolution: When a star expands, i.e. cools, it can only decrease its rotation velocity. An increase in rotation speed during the post-main sequence evolution of a star like S23 can only occur during the blue loop, i.e., while the star contracts. But in this case, it would heat up rather than cool down. Hence, stellar evolution cannot be the cause for the changes observed. Instead we need to think of different mechanisms that could explain the peculiarities of S23. In the following, we want to discuss three scenarios which are highly speculative and hence need additional investigation in the future:
Precession of the system
The first possible scenario that could account for the observed peculiarities in S23 might be related to precession of the star with respect to the line of sight. Rapidly rotating stars appear different when seen from different sides. This is because a rapidly rotating star flattens. Hence its poles become much hotter while its equatorial region becomes cool. This effect is known as the gravitational darkening. The difference in temperature between the hot poles and the cool equator thereby increases with increasing rotation speed of the star.The differences in the observed, projected rotation velocities, v sin i, might indicate that the inclination of the system has increased between the years 1989 and 2000. This would mean that we observed the star more from the equator (cooler), while in 1989 it was observed more from the polar region (hotter). Assuming that in 2000 the star was seen almost equator-on and that its rotation was 75% critical, we can compute the resulting surface structure. We find that that star must be flattened (left part of the following figure) and its surface parameters like effective temperature, escape velocity, hydrogen density and mass flux change from the stellar pole to its equator by quite some substantial amounts. These changes are shown as a function of co-latitude (i.e., θ=0 at the pole) in the right part of the following figure.
Based on the different inclination angles under which the star could have been seen in the two epochs of observations, a drop in observed effective temperature in combination with an increase in observed v sin i is thus a possible scenario. Whether such a strong precession indeed occurs, and what might be the reason for such rapid precession movement, is, however, not known. To test the precession scenario, we would thus need to monitor the the star over a period of several years to see whether the He I lines re-appear (i.e., the star "heats-up" again) with a simultaneous decrease of its projected rotation velocity measured from the Mg II lines.
An eclipsing binary scenario
A second possible scenario to explain the strong variations observed in S23 might be that of an eclipsing binary system, in which one component is perhaps a B9 supergiant star and the other one perhaps a A1 supergiant star. The disappearance of the He I lines in our spectra might then be interpreted with the A1 supergiant star occulting more or less the hotter B9 supergiant component, while the spectra from 1989, in which the He I lines were present and strong, indicate that the hotter component was dominating the spectral appearance of the system. To test the hypothesis of S23 being an eclipsing binary, it is certainly useful to perform a high signal-to-noise monitoring of its few available photospheric lines, (i.e., the He I lines and the Mg II line). A further test for the eclipsing binary scenario would be the monitoring of its light curve. Since the hotter component is also expected to be the more luminous one, the system must appear fainter when the B9 star is eclipsed than when the A1 star is eclipsed.A rapidly rotating star with surface inhomogeneities
A third quite interesting scenario is based on the assumption that S23 might have a patchy surface abundance in He. The existence of such inhomogeneous or spotty surface patterns are known especially for some so-called chemically peculiar (CP) stars. These objects at or close to the main-sequence are supposed to be rigidly rotating, and the variations observed in their He I line profiles are interpreted as abundance inhomogeneities in the form of large stellar spots that appear within and again disappear from the line of sight as the star revolves. The occurence of these spots is thereby thought to be linked to the presence of magnetic fields (similar to the spots seen on the surface of the sun).S23 is definitely not a main-sequence star, and to our knowledge, no such abundance or temperature inhomogeneities have been reported for a supergiant so far. In addition, nothing is known about a possible magnetic field on the surface of S23. Nevertheless, such a surface inhomogeneity seems to be an interesting scenario, at least to explain the variations in the He I lines. Whether this inhomogeneity can also explain the different observed values of v sin i, needs to be investigated in more detail.
If S23 possesses He spots, the variation of the He I lines must be correlated to the rotation period of the star. Assuming that S23 is seen edge-on, the derived value of v sin i = 150 km/s delivers the equatorial rotation velocity. With a stellar radius of roughly 60 solar radii, this results in a period of about 20 days. To prove or disprove the scenario of surface inhomogeneities in either He abundance, or temperature, or even both, one would need to monitor the He I and Mg II lines in the spectrum of S23 over at least one rotational period.