**6. Other Issues**

This narrative has omitted stellar rotation even though it is probably important. Rotation would greatly lengthen the narrative, and, more important, would expand the number of free parameters. A traditional exploration strategy makes sense: (1) Begin with simple non-rotating models, (2) learn whether the known processes can account for eruptions without rotation, and then (3) explore the effects of angular momentum. This topic has not ye<sup>t</sup> reached stage 3. In view of the multiple parameters required for a distribution of angular momentum, this approach is particularly justified for expensive 3-D simulations (Section 5.3 above). Apart from *η* Car as noted below [58] and the morphology of LBV ejecta-nebulae [26], there is little observational evidence concerning angular momentum in radiation-driven eruptions.

The same attitude is even more justified for eruption scenarios that require interactions of binary or multiple stars, particularly merger events. As noted many years ago, speculations in that vein allow theorists to "ascend into free-parameter heaven" [131]. Generically they require either small orbits or unusual orbit parameters. Such models are credible for lower-luminosity events that are not discussed in this review (e.g., red transients), because moderate-luminosity star systems are very numerous. The observed lower-luminosity outbursts can be explained by supposing that a tiny fraction of stars experience mergers and other exotic interactions. Stars with *L* > 105.5 *L*, however, are scarce; so we should not see the observed number of LBV's and giant eruptors if unusual circumstances are required. It is true that most massive stars have companions, but only a small fraction of them are close enough for major interactions [132]. Equally important, *there is no evident need* for eruption models of that type. The HRD upper limit in Figure 1 applies to practically all stars above 50 *M*, not just those with close companions. The LBV instability strip becomes much harder to explain if we suppose that it depends on multi-parameter interacting binaries [1,13]. And, perhaps most important, the single-star processes in Section 5 appear sufficiently promising until proven otherwise. In summary: Binary and multiple-system phenomena certainly deserve attention, but they have not ye<sup>t</sup> earned a well-defined place in the giant eruption puzzle.

Binarity does play a role for our best-observed supernova impostor, *η* Car, but it probably did not provide the basic instability mechanism. This object merits additional paragraphs here because so much is known about it, especially regarding some potentially instructive abnormalities. For instance, consider the hot secondary star's high orbital eccentricity, ≈ 0.85, with a periastron distance only about 3× or 4× larger than the primary star's radius [64]. Tidal effects are significant during about 3% of the 5.5-year orbital period, and may have triggered the Great Eruption as noted in Section 3.3. But this is not a straightforward idea! When we take the Eddington factor Γ into account, the companion star's maximum tidal effect is of the order of 10% as strong as effective gravity at the star's surface [58]. The iron opacity peak region is less perturbed because it has a smaller radius, and the core region is practically unaffected. Hence the periastron tidal-trigger conjecture requires an instability that began fairly near the surface—the geyser concept again. Moreover, the eruption did not begin suddenly; instead the star's brightness began to rise and fluctuate years earlier [63,133]. Later the mass outflow persisted long after tidal forces became negligible. Since the tidal maximum at periastron had a duration comparable to the star's dynamical timescale, it was neither an adiabatic nor an impulsive perturbation. Nonetheless the trigger concept has undeniable appeal. One can easily imagine a star expanding due to evolution, until it encountered a radius limit enforced by its companion. This differs from a familiar Roche lobe story in two respects: it was close to the Eddington Limit, and the tidal force made itself felt only for a few weeks near each periastron.

Two other points should be noted about *η* Car's periastron passages. First, after a sufficiently long time, tidal friction should cause the star's outer layers to rotate synchronously with the orbital rate at periastron, like the planet Mercury. The surface rotation period would then be roughly 90 days. In fact the X-rays show a quasi-period of that length [134]. Second, why is the orbit so eccentric? Its period would be only about 130 days if it were circular with *r* = the present-day periastron distance. If the orbit was circular a few thousand years ago, then the simplest explanation for large has two or three parts: (1) Most of the eruptive mass loss must have occurred near periastron, in order to eccentrify the orbit. (2) Several giant eruptions like 1830–1860 were necessary in order to attain ≈ 0.85. (3) However, since that value is very high, some additional factor was probably needed—e.g., asymmetric mass flows. See [58] and references therein.

Another of *η* Car's oddities concerns its equatorial skirt of ejecta. It is manifestly not a rotating disk, but instead appears to consist of radial spikes of ejecta [57,135]. Velocities and proper motions indicate that they formed at about the same time as the Homunculus lobes.

Various authors have speculated that *η* Car's giant eruption was a merger event, entailing a former third star [136–140]. Their scenarios employ at least 8 adjustable parameters, plus qualitative assumptions that are not emphasized, in order to account for 5 or fewer observed quantities. There is no evident need to postulate a third object; the primary star appears well suited to the single-star ideas listed in Section 5 above. (For instance, it is near the Eddington Limit without any reference to companion objects, and probably has a substantial iron opacity peak region.) The most detailed merger model [139] predicted too low a helium abundance, its stated quiescent brightness was far too low, and it was vague about the ejecta morphology. Exotic models can be interesting, but there is no reason to guess that they are necessary for this object. The single-star processes, modified by the known companion star, intuitively seem very promising for *η* Car and have not ye<sup>t</sup> been analyzed in sufficient detail.

High-velocity material associated with *η* Car has been interpreted as evidence for either a blast wave or an merger event [77,140]. Some outlying ejecta have Doppler velocities of 1000–3000 km s<sup>−</sup><sup>1</sup> [26], and light-echo spectra of the Great Eruption may show velocities as fast as 10,000 km s<sup>−</sup><sup>1</sup> [77]. However, other interpretations appear more likely according to the "maximum simplicity" criterion. Judging from H*α* images of the outer ejecta, the high-speed mass and kinetic energy are probably less than 10−<sup>5</sup> *M* and 10<sup>44</sup> ergs, and possibly much less. These amounts are substantially smaller than the mass and thermal energy of the star's opacity peak region, for instance. If an eruptive instability begins suddenly, a small amount of leading material may be ejected to very high speeds, analogous to the acceleration of a SN blast wave as it moves through a negative density gradient. Indeed an acceleration feature like that can be seen in Figure 2 of [34]. The standard super-Eddington flow becomes established after the initial transient burst. This explanation may be wrong, but it as well-developed as the exotic interpretations, and more credible because it fits the other characteristics of *η* Car's ejecta [57]. Moreover, the very-high-velocity line wings in the light echo spectra are so faint that they may be either instrumental artifacts or features caused by Thomson scattering in dense locales of the outflow.

As emphasized in Section 4 above, the brightness and spectrum of a radiation-driven eruption do not tell us much about the star and its structure. However, the post-eruption behavior may give some useful information. At any given time during the event, the entire configuration is close to dynamical equilbrium (including flow processes) but far from thermal and rotational equilibrium. This remains true after the event subsides, leaving a star with a peculiar thermal structure. It should then recover—i.e., find a new equilibrium state—in a few thermal timescales. This process has been observed in *η* Car, and the record is interesting in two respects: it has taken longer than the expected 50 years, and it has been quite unsteady [58]. Major changes occurred at 50-year intervals [58,63], and the spectrum has evolved more rapidly during the past 20 years [141,142]. This temporal structure surely depends on the star's thermal and rotational structure. A preliminary assessment of the recovery problem was reported in [143], but multiple 3-D simulations are needed.

As mentioned near the end of Section 5.2, stellar activity and turbulent MHD may occur in the outer layers of LBV's and/or related stars. This would not be terribly surprising, since one can write the Schwarzschild criterion in a form that looks much like the Eddington Limit. The point is that MHD waves, or similar processes, may assist the outward acceleration forces, and might even produce violent instabilities.

Finally, a point in Section 4.4 merits repetition because it affects this entire topic: a radiation-driven outflow is difficult to calculate. If one writes 1-D analytic equations for radiative transfer and acceleration, they give nonsensical results because a real outflow automatically becomes inhomogeneous. Acceleration and radiation leakage depend on the sizes, spacing, and even the shapes of the granules. These effects are too intricate to calculate ab initio for every model or sub-model. Therefore it might be valuable, and certainly would be interesting, to have some sort of general prescription based on many specialized 3-D simulations. As a first step, those simulations could include only Thomson scattering. What factors determine the characteristic size scales and time scales and density distributions? Cf. [83,84,117].

**Funding:** This research received no external funding, and was supported primarily by photons.

**Acknowledgments:** I am grateful to R.M. Humphreys, J. Guzik, I. Appenzeller, C. de Jager, M. Schwarzschild, E.E. Salpeter, and A.S. Eddington for indicating good points of view for this topic.

**Conflicts of Interest:** The author declares no conflicts of interest.
