**6. Wind Properties**

The underlying theory responsible for outflows by hot luminous stars has been known for several decades [81], with a metallicity dependence primarily arising from the variation in iron-peak elemental abundances [82,83]. Individual wind properties of O stars or blue supergiants usually rely on spectroscopic fits to H*α*, as parameterised by the wind strength parameter, *Q* [41], from which the mass-loss rate requires knowledge of the physical radius (from *T*eff, log *L*) and measured or adopted wind velocity. Wind velocities of OB stars cannot be measured from optical spectroscopy, so usually spectral type calibrations are adopted based on measured velocities from UV P Cygni profiles of C IV, N V or Si IV [84]. Until recently, high S/N, high quality UV spectroscopy of OB stars in the Large Magellanic Cloud has been in short supply, but the situation has improved via the HST Large Program METAL (GO 14675, PI Julia Roman-Duval) [85] and upcoming ULLYSES initiative 1.

<sup>1</sup> http://www.stsci.edu/stsci-research/research-topics-and-programs/ullyses.

Specifically for the Tarantula, low resolution UV spectroscopy of the R136 star cluster has added a significant number of wind measurements for early O stars [5]. Wind velocities exceed 3000 km s<sup>−</sup><sup>1</sup> at the earliest subtypes (O2–3), reducing to ∼1500 km s<sup>−</sup><sup>1</sup> for late O-types. Mass-loss rates of emission line stars typically rely on an alternative wind scaling relation, namely the transformed radius, *Rt* [86], which also requires knowledge of wind velocities, although these can be estimated from optical spectroscopy for Wolf-Rayet stars. An added complication arises because radiatively-driven winds are known to be inherently unstable, leading to clumped winds. Wolf-Rayet winds have been known to be clumped for 30 years [87,88], but the degree of clumping for O stars via the H*α* diagnostic remains unclear, with conflicting results obtained from UV resonance lines [89] unless the wind comprises a mixture of optically thin and thick clumps within a much lower density inter-clump medium [90].

Figure 6 presents unclumped mass-loss rates of O, Of/WN and Wolf-Rayet stars in the Tarantula Nebula, obtained from VFTS [50,51,53], HST/STIS [55] and other literature results. Uncertainties have been included wherever possible. Since the primary wind diagnostic in the majority of instances presented here is H*α*, it is apparent that uncertainties are large for those stars with weak stellar winds. In addition, mass-loss rates for Of/WN and Wolf-Rayet stars are anticipated to be reduced significantly owing to wind clumping, as indicated with downward arrows. If volume filling factors are ∼10%, mass-loss rates will be reduced by a factor of √10.

Theoretical mass-loss rates [83] for zero-age main sequence stars at LMC composition [60,61] are included in Figure 6. At face value it would appear that the theoretical mass-loss rates of LMC O stars are supported by theory. However, the following should be borne in mind. It is not clear how significantly wind clumping affects the inferred mass-loss rates of normal O stars, although H*α* results for supergiants are likely to be sensitive to wind clumping. In addition, the vast majority of mass-loss rates of VFTS O stars shown here have been inferred by adopting wind velocities from an assumed scaling relation involving escape velocities [91,92], which are themselves dependent upon spectroscopic gravities. Exceptions are HST/STIS results for early-type stars in R136 which are based on measured UV wind velocities, and span dwarfs, giants, supergiants and main sequence WN stars. In order to verify predictions for lower luminosity (log *L*/*L* < 5.5) O stars, more sensitive diagnostics would need to be employed, such as UV P Cygni lines, providing complications such as porosity are accounted for [90].

It is clear from Figure 6 that rates for the highest luminosity main-sequence Of/WN and WN stars significantly exceed theoretical predictions. This discrepancy is partially addressed through wind clumping, but very massive stars close to their Eddington limits are observed to exhibit enhanced mass-loss rates which are not taken into account in standard theoretical predictions [93,94]. Unsurprisingly, classical Wolf-Rayet stars with log(*L*/*L*) = 5.5 − 6 possess the strongest winds amongs<sup>t</sup> early-type stars in 30 Doradus, with clumping-corrected wind densities an order of magnitude higher than O stars with similar luminosities. It is well known that the wind momenta of WR stars, *Mv*˙ ∞, exceeds the momentum provided by their radiation field, *L*/*<sup>c</sup>*, owing to multiple photon absorption and re-emission within their optically thick winds, permitting *Mv*˙ ∞/(*L*/*c*) > 1 [95].

**Figure 6.** Unclumped mass-loss rates of O-type, Of/WN and Wolf-Rayet stars in the Tarantula Nebula (based on results from VFTS [50,51,53], HST/STIS [55] and other surveys [37,56,59] for WR stars). Filled symbols are within NGC 2070, open symbols elsewhere in the Tarantula. Theoretical mass-loss rates for zero age main sequence massive stars at the Large Magellanic Cloud (LMC) metallicity [83] are included (solid line), based on LMC metallicity evolutionary models [60,61].

### **7. Fate of Massive Stars in the Tarantula**

The conventional picture of massive star evolution close to solar metallicity is that those with initial masses of 8–25 *M* will end their lives as red supergiants (RSG), undergo a H-rich core collapse supernova, leaving behind a neutron star remnant, while higher mass counterparts will either circumvent the RSG phase or subsequently proceed to a Wolf-Rayet stage prior to undergoing core collapse, leading to a H-deficient supernova (neutron star remnant) or faint/failed supernova (black hole remnant) [96]. If one considers the global WR vs. RSG population in the LMC, the lower boundary to the luminosity of WR stars is log(*L*/*L*) = 5.3, while the upper luminosity of RSG is log(*L*/*L*) = 5.5 [48], supporting a transition from RSG to WR for higher mass progenitors at ∼25 *M*.

Close binary evolution severely complicates this scenario, since primaries below 25 *M* can be stripped of their hydrogen envelope, leading to a type IIb or Ib/c instead of a H-rich supernova, while secondaries will be rejuvenated, spun-up, with the potential for a core-collapse supernova for secondaries whose initial masses fall below 8 *M*. To date, there are no unambiguous cases of pre-supernova close binaries in the Tarantula hosting stripped (Wolf-Rayet or helium) stars, although it has been suggested that the WN3 binary BAT99-49 elsewhere in the LMC is the product of close binary evolution [59]. Rapid rotation of the bright O giant component of the high mass X-ray binary VFTS 399 is consistent with this evolutionary scenario. The absence of low luminosity Wolf-Rayet stars in the Tarantula does not exclude the binary channel since low-mass stripped stars would be unlikely to exhibit a Wolf-Rayet spectral appearance [67].

Initially very close binaries may merge on the main sequence, prior to following a relatively conventional evolution, albeit with unusually high rotation rates, which would lead to increased luminosities and potentially evolve blueward off the main sequence [60,61]. Extremely rapid rotation in some VFTS OB stars favours close binary evolution or stellar mergers. Very massive stars in the Tarantula up to ∼300 *M* are expected to lead to 30–50 *M* CO cores and black hole fates, unlikely to produce any associated supernova [97], such that a subset of binary VMS are plausible progenitors

of LIGO black hole binary mergers, although their exact fate crucially depends on their mass-loss properties, which remain uncertain.

### **8. Integrated Properties and Comparison With Star-Forming Regions, Near And Far**

The Tarantula Nebula would subtend little more than one arcsec if it were located at a distance of 50 Mpc, and so provides us with a unique opportunity to compare the individual spatially-resolved properties of an intensively star-forming region and its aggregate characteristics. Using integrated H*α* observations of the Tarantula Nebula [98], an age of ∼3.5 Myr would be inferred from a comparison between the inferred H*α* equivalent width of 1100 Å and population synthesis models for a coeval population at LMC metallicity [12]. This is in reasonable agreemen<sup>t</sup> with the typical age of massive stars, albeit failing to reflect the complexity in its star-formation history (recall Figure 3).

An analysis of the integrated UV spectrum of NGC 2070 supports a young (≤3 Myr) starburst episode [99], while the high spatial resolution of HST/STIS has permitted a comparison between the individual and integrated UV spectroscopic appearance of the central R136 cluster [5]. Very massive stars contributed a significant fraction of its far UV continuum flux, and completely dominate the strong, broad He II *λ*1640 emission. The integrated UV spectroscopic appearance of R136 closely resembles some star clusters in star-forming galaxies at Mpc distances, such as NGC5253-5 [100], suggesting the presence of VMS in these other young massive clusters. From comparison with the predictions of standard population synthesis models, both Starburst99 [101] and BPASS [102] models fail to predict any significant emission prior to the conventional Wolf-Rayet phase (Figure 7), owing to the use of inadequate wind theory for VMS [93,94]. Consequently, neither Starburst99 nor BPASS accounts for the powerful winds of very massive stars in R136, and the adopted mass function follows a Salpeter slope, rather than the top heavy IMF identified by [40] for the Tarantula region as a whole.

**Figure 7.** Comparison between observed He II *λ*1640 emission equivalent widths in R136 [5] versus predicted emission from BPASS (v.2.2.1, red) and Starburst99 (blue) population synthesis models (absorption lines are shown as negative values).

An estimate of the cumulative ionizing and mechanical feedback from massive stars within the Tarantula has revealed a major contribution from VMS towards the collective ionizing output and a dominant role from WR stars to the mechanical feedback [12]. However this analysis relied on calibrations and estimates of spectral types for a significant subset of the massive star content, so we are able to provide updates from recent spectroscopic observations (e.g., VLT/MUSE, HST/STIS) and analyses. We present the updated cumulative ionizing output from 1170 massive stars in the Tarantula Nebula in Figure 8, indicating a total Lyman continuum ionizing output of 1.2 × 10<sup>52</sup> ph s<sup>−</sup><sup>1</sup> within 150 pc of R136a. A quarter of the total ionizing radiation originates from the R136a cluster, while

members of NGC 2070 produce three quarters of the global feedback. Ten systems alone, listed in Table 4, collectively contribute a quarter of the ionizing budget of the Tarantula Nebula, comprising main sequence and classical WR stars, plus early O supergiants. Indeed, half of the global ionizing output originates from 40 early O stars, main sequence WN stars and classical WR stars, with 1130 stars contributing the remaining 50%. The collective bolometric luminosity of these stars is 108.4 *L*, of which the 40 UV-bright stars contribute 108.0 *L*.

**Table 4.** Top ten stellar systems contributing to the Lyman continuum output of the Tarantula Nebula, comprising very massive early O stars and WN5 stars, and classical Wolf-Rayet stars, updated from [12].


**Figure 8.** Cumulative ionizing output (10<sup>50</sup> ph/s) from spectroscopically classified early-type stars in the Tarantula, obtained from VFTS [50,51,53], VLT/MUSE [11], HST/STIS [55] and literature results [56], updated from [12]. Specific regions within 30 Dor are indicated from Table 1.

Recalling Table 2, the highest mass stars and evolved high mass stars in other giant H II regions in the Local Group dominate their radiative and mechanical feedback, including NGC 3372 (Carina Nebula) in the Milky Way, N206 in the LMC and NGC 346 in the SMC. By way of example, Smith [39] established that only a handful of early O-type stars and H-rich WN stars contribute the majority of the Lyman continuum flux of the Carina Nebula, while *η* Car, four Wolf-Rayet stars and two early O supergiants completely dominate the stellar mechanical luminosity. The central ionizing cluster of the Galactic NGC 3603 star-forming region is host to a stellar content analogous to R136a, including a number of early O stars, nitrogen-sequence Wolf-Rayet stars [103,104]. Weak main-sequence wind properties of metal-poor massive stars conspire to even fewer massive stars

(HD 5980, Sk 80) dominating the cumulative stellar feedback in NGC 346. Similar conclusions were reached by Ramachandran et al. [105] for the supergiant shell in the wing of the SMC.

Although the Tarantula Nebula is the most extreme giant H II region in the Local Group, how does it rank against star-forming regions of galaxies in the near universe or knots at high redshift? Figure 9 compares the star-formation rate versus size of regions spanning *z* = 0 to 3.4, adapted from [106], indicating that the Tarantula (red square) is forming stars more vigorously than typical low-redshift counterparts, resembling some star-forming regions at high redshift. Indeed, 80% of the cumulative ionizing radiation originates from NGC 2070, such that this region corresponds closely with typical clumps in the lensed galaxy SDSS J1110 + 6459 at *z* = 2.5 (green circles).

**Figure 9.** Comparison between the integrated star-formation rate versus size of the Tarantula (filled red square) and star-forming knots from galaxies spanning a range of redshifts, adapted from [106].

In addition to its unusually high star formation rate, the Tarantula also possesses high ionization parameter nebular properties with respect to star-forming galaxies in the local universe. Figure 10 presents a Baldwin, Philipps and Terlevich [107] (BPT) diagnostic diagram of Sloan Digital Sky Survey (SDSS) galaxies, in which the Tarantula (red square) has been indicated, along with Green Pea galaxies from Micheva et al. [108] which are low-metallicity, intensively star-forming galaxies exhibiting unusually strong [O III] *λ*5007 emission. Steidel et al. [109] showed that *z* = 2–3 star forming galaxies share similar extreme nebular properties, and a subset of Green Pea galaxies have been established as Lyman continuum leakers [110,111]. Focusing again on NGC 2070, this sits amongs<sup>t</sup> the extreme Green Pea galaxies in the BPT diagram.

**Figure 10.** BPT diagram illustrating the similarity in integrated strengths between the Tarantula Nebula (red square), Green Pea (green circles), extreme Green Peas (blue diamonds), Lyman-continuum leaking Green Peak (pink triangles), updated from [108], plus SDSS star-forming galaxies.
