*3.1. LBV's*

The term "Luminous Blue Variable" is unfortunate in three respects: Many unrelated luminous blue stars are also variable, LBV's are often not very blue or not strongly variable, and it has caused extraneous objects to be included in lists of LBVs, often without observed outbursts [12,13]. Thus we should regard the trigram "LBV" as an abstract label, not an acronym. Many examples are described and listed in [12,14,15]. The present review includes this phenomenon because it may provide some guidance to the physics of giant eruptions, and LBV's have been observed far more often than giant eruptions. Note, however, that suspected analogies between those two categories have not been proven and may turn out to be illusory.

LBV's are defined by a particular form of variability like AG Car in Figure 1 [1]. Their hot "quiescent" states are located along a strip in the H-R Diagram (HRD) shown in the figure, sometimes called the S Doradus instability strip [16]. Its upper and lower parts, LBV1 and LBV2, represent different stages of evolution—providing a clue for theory, Section 5.1 below.

Most stars in the strip are not LBV's. Spectroscopic analyses consistently show that genuine LBV's have smaller masses than other stars with similar *L* and *T*eff [15,17–21]. Consequently, for an LBV the Eddington parameter Γ ≡ *<sup>κ</sup>eL*/4*πcGM* = (*L*/*M*)/(*L*/*M*)Edd is close to 0.5 or somewhat larger. This is not surprising for the luminous classical LBV's; a 60 *M* star, for example, attains Γ > 0.4 before the end of central hydrogen burning [12]. But Γ ∼ 0.5 is remarkable in zone LBV2, where most stars have Γ ∼ 0.2. Evidently each lower-luminosity LBV has lost much of its initial mass.

Since the LBV2 stars have luminosities below the upper boundary in Figure 1, they can evolve across the HRD. Hence we can explain their low masses by supposing that they have already passed through a cool supergiant stage where mass loss was very large [1]. After returning to the blue side of the HRD, they now have large *L*/*M* ratios which cause them to be LBV's. This surmise is confirmed, or at least very strongly supported, by the fact that LBV1's are generally associated with O-type stars but LBV2's are not [12,13,22]. This fact implies that LBV1's are younger than LBV2's, as expected in the evolved-LBV2 scenario. Classical LBV's (LBV1's) are somewhat more than 3 million years old, near or slightly after the end of core hydrogen burning. LBV2 stars are post-RSG's near or after the end of core helium burning (Figure 1 in [12]).

The above account may seem inconsistent, because LBV's are said to have rapid mass loss but the low masses of LBV2's are ascribed to a different evolutionary stage. This semi-paradox arises because the two types play very different roles in this story. LBV1 outbursts probably cause enough mass loss to shape the appearance of the upper H-R Diagram. LBV2 events do not, but they are pertinent because they sugges<sup>t</sup> a connection between LBV variability and the Eddington parameter Γ as noted above. They also give a strong hint that LBV instability occurs in the outer layers, Section 5.1 below.

Figure 1 shows a well-known classical LBV, AG Carinae. It currently has *L* ≈ 1.5 × 10<sup>6</sup> *L* and *M* ∼ 40 to 70 *M*, with *T*eff ∼ 16000 to 25000 K at times when a major LBV event is not underway [15,23–25]. The initial mass was probably above 85 *M* and rotation is non-negligible [25]. In the years 1990–1994, AG Car's photosphere temporarily expanded by an order of magnitude with only a modest change in luminosity [23]. The apparent temperature consequently declined to about 8500 K, shifting much of the luminosity to visual wavelengths. Meanwhile its mass-loss rate increased by a factor of 5 to 10, peaking above 10−<sup>4</sup> *M* y<sup>−</sup>1. (The estimated amount depends on assumptions about the wind's inhomogeneity.) Outflow speeds varied in the range 100–300 km s<sup>−</sup>1. Then, in 1995–1999 the photosphere contracted back to roughly twice its pre-1990 size. The event timescale, about 5 years, was more than 100× longer than the star's dynamical timescale. Perhaps 5 years was a thermal timescale for a particular range of outer layers. AG Car's 1990–1999 event in Figure 1 represents the classic form of high-luminosity LBV event, except that it only partially returned to its pre-1990 state.

Like many other LBV's [26], AG Car has a circumstellar nebula [27–30]. The nebular mass is said to be 5–20 *M*, ejected thousands of years ago and expanding rather slowly. Either the ejecta from multiple events have piled up there, or the star had one or more giant eruptions larger than any LBV events that have been observed in recent times. (The circumstellar material is almost certainly not due to mass loss in a red supergiant stage of evolution, since AG Car is too luminous to become a RSG—see Figure 1. The same statement applies to various other LBV's that have circumstellar ejecta.)

Figure 1 includes another LBV, R 71, to show that rules can be broken. It had an outburst in the 1970's [31], but a later event starting around 2005 was extraordinary [32,33]. Unlike normal LBV events, the luminosity of R 71 substantially increased while the temperature fell definitely below 7000 K. At minimum temperature it exhibited pulsation on a dynamical timescale (cf. comments in [34]). The mass-loss rate rose well above 10−<sup>4</sup> *M* y<sup>−</sup>1, high for its luminosity. Since *L* is poorly known due to an uncertain amount of interstellar extinction, this object may be either a classical LBV1 or an LBV2.

Note that the empirical limit in Figure 1 does not coincide with the LBV instability strip. The instability strip might extend to the boundary, but this detail should warn us that evolution through the LBV1 stage may involve some unrecognized tricks. Each of the following scenarios would be consistent with available data.


Resourceful theorists can devise other possibilities. This review concerns the nature of mass-loss episodes, not the resulting evolutionary tracks. The latter depend on multiple parameters which are very poorly known.

Concerning LBV photospheres, see Section 4.5 below.
