Spectroscopic events and the 5.5-year cycle in Eta Carinae

A persistent enigma

Often it is claimed, or at least hinted, that Eta's 5.5-year cycle has already been explained by a binary star model. In fact, however, the observed phenomena continue to represent a genuine astrophysical mystery.

Yes, Eta Carinae is most likely a binary system; but binary "models" proposed to explain specific observed effects have often been mutually incompatible with others that focus on other details. Here we put the word "models" in quote marks because so far they've been qualitative and, frankly, vague. No one has yet proposed a scenario which is (1) specific, (2) quantitative, (3) applicable to most of the observed effects altogether, and (4) self-consistent. Moreover, the existence of a companion star remains unproven and single-star "models" are not very difficult to imagine.

We say this not to diminish the published ideas, but rather to emphasize that this remains very much an unsolved problem, and an interesting one. From the point of view of serious astrophysical theory, it's still relatively unexplored territory, and may turn out to have broad implications for the stability of the most massive stars.

Some popular articles and even professional papers since 1996 have given misleading impressions of the problem and its history. We hope the following account helps to put the situation in context.


The 5.5-year cycle makes a good case study of a belated scientific discovery, one that should have occurred decades earlier. Considering that Eta Car was one of the most notorious objects in the sky -- bright, unique, unstable and mysterious -- southern-hemisphere astronomers didn't observe it often enough.* Images, spectroscopy, and even photometry were frustratingly sparse before 1980.

(*For all they knew, this monster could have gone supernova at any time. If it had, in most years of the twentieth century there would have been no detailed observations just before the disaster! -- Even though, as mentioned above, Eta Car was easy to observe and known to be unstable. Most famous astronomical objects are not monitored spectroscopically, and many aren't monitored at all. Someday one of them will do something totally shocking, and astronomers will be embarrassed to admit afterward that they have little or no data leading up to the event. In a sense this has already happened with Eta Car, more than once.)

By the mid-1980s we knew that the spectrum had briefly changed in 1948, 1964, and 1981.The best early discussion of these changes was a 1984 paper by Zanella et al., who suspected a 17-year recurrence period. That paper, [1] in the list of refs. later on this page, deserves special mention in any discussion of this topic.

Timeline graphic

On each occasion the high-excitation [NeIII], [FeIII], and HeI emission lines abruptly disappeared while certain P Cyg absorption features became stronger. Our 1981 IUE data showed a fading of the bizarre FeII 2507 Å "pseudo-laser" emission too, as Viotti later noticed. In each case the spectrum gradually returned to its normal state after a few months. HST observations in the 1990s showed that the [NeIII], [FeIII], and 2507 Å lines originate in ejected gas a few light-days out from the star while the P Cyg absorption occurs in the stellar wind -- see refs. [2].

Zanella et al. [1] conjectured that each "spectroscopic event" was basically a mass ejection which temporarily quashed the star's hard UV radiation, thereby eliminating most high-excitation ion species. Whatever their precise cause, these events are clues to the nature of Eta Car in at least two different ways: They must represent variations in the physical parameters, almost as though we were able to do experiments, while the very existence of the instability may be a clue to the star's famous large-scale instability.

Another such event occurred in late 1986, but, frankly, there were few Eta Car afficionados in those days and we didn't give it much attention. Eta's location in the far southern sky obviously made the object difficult for most astronomers. Then Augusto Damineli in Brazil began the first persistent long-term program in recent decades to monitor Eta's spectrum. For practical reasons he concentrated on just a few red emission lines, including HeI features that are good indicators of a spectrosocopic event. A recurrence in 1992 made the periodicity obvious:

Timeline graphic

Taking the 1948 event into account, Damineli announced a period of about 5.5 years [3]. Whitelock et al. had already recognized a five-or-six year cycle in near-infrared photometry, representing continuum emission in the dense stellar wind [4].

Zanella et al. had been on the right track; their 17-year period was really three 5.5-year cycles, but they had no way of knowing this since astronomers had missed the events of 1970 and 1975. (Damineli found retrospective evidence that an event had indeed occurred in late 1975 [3].) Zanella et al. may have been right about something else as well: Today, the best conjectures to explain an event include a phenomenon very much like their proposed shell ejection (see, e.g., refs. [7-9]).

The 5.5-year cycle was significant in two ways: It offered a clue to the phenomenon, while, equally important, it predicted when the next event would occur:

Timeline graphic

An article in Sky & Telescope magazine, January 1998 issue, reflects some of the views we had then [5]. Written in August-September 1997, three months before the predicted date, its guesses proved to be fairly accurate.

Most readers of this page won't be surprised that (1) at least two different telescope allocation committees expressed disbelief, but (2) the predicted event did occur on schedule. The most dramatic indicator turned out to be an X-ray monitoring project led by Mike Corcoran. The X-ray flux rose tremulously during most of 1997, peaked, and abruptly crashed almost to zero between mid-November and mid-December, see refs. [6-8]. (The peak and the crash exactly followed a reasoned hunch expressed in ref. [5].)

RXTE data plot

Ground-based observations in late 1997 and early 1998 showed the "expected" behavior plus new details. Unfortunately we couldn't obtain any HST/STIS spectroscopy until six weeks into the event, but the results have been among the most valuable clues to what happened even though they began much later than we had hoped. Many aspects of the 1997-98 event were reported at a remarkable meeting in July 1998 [7].

Today there's no doubt that Eta's spectroscopic events recur with a period close to 5.54 years, and that several of them passed unnoticed between 1950 and 1980. But what are they? What's the physics?

Timeline graphic

The 1997-98 event lacked HST coverage before and during the most critical time interval. HST data are essential for two reasons, namely high spatial resolution and ultraviolet coverage. Since no ground-based data could separate the star itself from surrounding ejecta, while the UV is physically critical, the 1997-98 event was indecisive. Thus began our project to observe Eta Carinae with the HST/STIS from early 1998 through the next spectroscopic event in 2003. Since a wealth of extra information automatically accrues independent of the event cycle, we frankly expect this project to be a scientific platinum mine in the long run, with diverse applications hinted elsewhere in this www site.

Causes and implications of the cycle

((( A quasi-theoretical assessment and other material will be provided here later. The spectroscopic events are not understood. Merely saying "it's a binary system" explains little, and might prove to be untrue anyway, contrary to many published assertions. - - - 20 Feb 2003 - - - )))

Spectrum of Eta Car in March 1998 and February 1999

This is a double tracing of the spectrum of Eta Car, the star itself, in March 1998 and February 1999, observed with the HST/STIS/CCD, probably the only GIF file you've encountered lately with dimensions 26731 x 467. Each spectrum covers the wavelength range from 1700 to 10000Å.

These represent STIS data on the star itself, Flambda/lambda plotted on a logarithmic wavelength scale with no attempt to correct for interstellar and circumstellar extinction. Since we can't display the entire wavelength range in one file, it's divided into four parts.

Note how wide these GIF files are, and try scrolling horizontally along each. If you're interested, you can splice all four together to make a really wide GIF file, 26731 x 467, about 160 kB. Printed on a strip of paper at about the same scale as a typical computer display, this spectrum would be roughly 10 m long. The relative spectral resolution is typically 1/7000. At wavelengths longer than 2700 Å most wiggles big enough to notice here are real features, not noise.

We hope it's obvious that such coverage of a complex spectrum is unique among HST data. However, the stellar spectrum isn't nearly as complex as the narrow-line spectra of nearby gas blobs ejected by Eta! For various reasons we don't show them yet.

Printable pdf file with the same tracings divided into many parts (several pages, 2 MB)

References cited above:

  1. R. Zanella, B. Wolf, and O. Stahl (1984) Astron. Astrophys. 137, 79.
  2. Davidson et al. (1995) Astron. J. 109, 1784, and (1997) Astron. J. 113,335.
  3. A. Damineli (1996) Astrophy. J. Letters 460, L49.
  4. Whitelock et al. (1994) Monthly Notices Roy. Astr. Soc. 270, 364.
  5. K. Davidson (1998) Sky & Telescope 95, #1, 36.
  6. Corcoran et al. (1998) Astrophys. J. 494, 381.
  7. See many authors in a proceedings book: Eta Carinae at the Millennium, ASP Conf. Ser. 179
    (ed. J.A. Morse, R.M. Humphreys, and A. Damineli).
  8. Davidson (2002), in The High Energy Universe at Sharp Focus, ASP Conf. Ser. 262, p. 267.
  9. Smith et al. (2003) Astrophys. J. 586, 432.

©2003,2004 The University of Minnesota