**3. Photometry**

Photometric observations of GR 290 were initiated in the early 1960s by the Italian astronomer Giuliano Romano in the Asiago Observatory [1]. He obtained a light curve with the brightness of a star varying irregularly between 16 m. 7 and 18 m. 1, and classified it as a variable of the Hubble-Sandage type based on the shape of the light curve and GR 290's colour index.

Subsequent photometric investigations of GR 290 were undertaken by Kurtev et al. [6] and later by Zharova et al. [27]. The cumulative light curve derived in the latter work and covering half a century shows that GR 290 exhibits irregular light variations with different amplitudes and time scales [27]. The star shows large and intricate wave-like variations, with duration of the waves amounting to several years. In general, its variability is irregular, with the power spectrum fairly approximated by a red power-law spectrum [28] (i.e., the one dominated by a long timescale variations). Moreover, Kurtev et al. [6] discovered short-timescale variability with amplitude ∼0m. 5, which is also typical an LBV star.

Polcaro et al. [10] used various collections of photographic plates to further extend the historical light curve back to the beginning of the 20th century. The data between 1900 and 1950 sugges<sup>t</sup> that no significant eruption took place during that half century. On the contrary, after 1960, two clear, long-term eruptions are evident (see Figure 2).

New photometric data, collected in Table 3 and shown as magenta dots in Figure 2, confirm the conclusion of Maryeva et al. [22] and Calabresi et al. [29] that the star has reached a long lasting visual minimum phase in 2013, and its brightness has been relatively stable since then.


**Table 3.** New photometric observations of GR 290 acquired by our group since Polcaro et al. [10].

*a* observatories: Loiano: 1.52 m telescope at the Loiano station of the Bologna Astronomical; Observatory-INAF. ARA: 37 cm telescope of the Associazione Romana Astrofili at Frasso Sabino (Rieti); RTT-150: 1.5 m Russian–Turkish telescope. CMO: 2.5 m telescope of the Caucasian Mountain Observatory.

It is generally observed that, during the S Dor cycle, the colour of a typical LBV is bluer at the light minimum than close to the light maximum. In contrast, Polcaro et al. [10] demonstrated that (*B* − *V*) colour of Romano's star is constant over time, within the error bars. There is no clear evidence for a variation of (*B* − *V*) as a function of the visual magnitude, and our new photometry obtained after 2015 confirms this conclusion (see Figure 3). This is consistent with Romano's star being hotter (about 30,000 K) than a typical LBV, with the slope of optical spectrum defined by a Raleigh-Jeans power-law tail.

**Figure 2.** The historical light curve of GR 290 in the B-filter from 1901 to 2019.

**Figure 3.** (**top**) Light curve of GR 290 in the B, V and R filters obtained by us between 2010 and 2019 and partially published in Polcaro et al. [10]. (**bottom**) (*B* − *V*) and (*V* − *R*) colour indices for the same time interval.

In other spectral ranges, the object is much less studied than in optical range. Only a single measurement of its magnitude is available in ultraviolet and infrared ranges, corresponding to different moments of time defined by a mean epoch of the individual survey (Table 4).


**Table 4.** Stellar magnitudes of GR 290 in ultraviolet and infrared range.

### **4. Spectroscopy and Determination of Physical Parameters**

The first description of optical spectrum of Romano's star can be found in the article of Humphreys [33]. The spectrum was obtained in August 1978 at Kitt Peak National Observatory, when brightness of the star was *V* = 18.00 ± 0.02 [33]. Humphreys noted: "Its spectrum shows emission lines of hydrogen and He I. There are no emission lines of Fe II or [Fe II]." and classified the star as a peculiar emission-line object<sup>6</sup> [33].

<sup>6</sup> In Humphreys [33], Romano's star is identified as B 601.

In 1992, T. Szeifert obtained a spectrum of Romano's star right before the historical maximum of its brightness. Szeifert [4] described it as "Few metal lines are visible, although a late B spectral type is most likely" (Figure 4). On the other hand, Sholukhova et al. [34] obtained the next spectrum in August 1994 and classified the star as a WN star candidate. Since 1998, regular observations of GR 290 carried out on the Russian 6m [5,35] and spectra published by Sholukhova et al. [35] indicate that the spectrum of GR 290 has not reverted to a B-type spectrum. Thus, Szeifert's [4] spectrum is unique and corresponds to the coldest and brightest state of the star measured so far.

**Figure 4.** Comparison of normalized optical spectra of GR 290 obtained with Calar Alto/TWIN in October 1992 by Szeifert [4] and with GTC/OSIRIS in September 2018. Spectra are displaced vertically for illustrative purposes.

Studies of GR 290 devoted to its spectral variability show that its spectral type changes between WN11 and WN8 [8,35,36]. Since the beginning of the 2000s, it has made this transition twice [10]. Viotti et al. [37,38] first described an anticorrelation between equivalent width of 4600–4700 Å blend and the brightness. Later, Maryeva and Abolmasov [36] found a correlation of spectral changes and the visual brightness typical for LBVs: the brighter it is, the cooler the spectral type. However, as noted by Humphreys et al. [9], GR 290 does not exhibit S Dor like transitions to the cool state with an optically thick wind, but instead varies between two hot states characterised by WN spectroscopic features. Among all known LBVs, only HD 5980 [39] convincingly shows a hotter spectrum in the minimum of brightness. Other LBV stars showing WN-like spectrum in quiescent "hot" phase usually stop at colder spectral types such as WN11 (for example AG Car [40] and WS 1 [41,42]) or Ofpe/WN9 (for example R 127 [43] and HD 269582 [44]).

As already mentioned, since the autumn of 2013, GR 290 is in a minimum brightness state with *V* = 18.7–18.8 mag. Due to this, it has been challenging to obtain its spectra with good enough quality for wind speeds to be adequately estimated. In summer of 2016, GR 290 was observed with the Optical System for Imaging and low-Intermediate-Resolution Integrated Spectroscopy (OSIRIS) on the *Gran Telescopio Canarias (GTC)* [22]. These observations gave the best spectral resolutions and signal-to-noise ratios ever obtained for this object, and allowed to estimate an average radial velocity (RV) of the object, RV(GR 290) = −163 ± 32 km s<sup>−</sup>1, which is consistent, within the uncertainties, with the heliocentric velocity −179 ± 3 km s<sup>−</sup><sup>1</sup> of M 33 galaxy.

New spectra of GR 290 were obtained with the OSIRIS spectrograph in September 2018 [23]. Detailed analysis of the spectra obtained in 2016 and 2018 did not reveal any changes (Figure 5). As before, the star displays a WN8h spectrum with forbidden nebular lines.

**Figure 5.** Comparison of normalised optical spectra of GR 290 obtained with GTC/OSIRIS in July 2016 (grey thick line) and September 2018 (black dash-dotted line). Spectra are nearly identical.

The large number of acquired spectra allows tracking the quantitative changes of physical parameters of the star over time. To do it, a numerical modeling of GR 290's atmosphere using CMFGEN code [45] was started by Maryeva and Abolmasov [46], who constructed models for two states—the luminosity maximum of 2005 and the minimum of brightness in 2008. Then, Clark et al. [47] estimated the parameters of GR 290 during the moderate luminosity maximum of 2010. Polcaro et al. [10] built nine models for the most representative spectra acquired between 2002 and 2014. The results of calculations from Polcaro et al. [10], Clark et al. [47] and Maryeva et al. [22] are summarised in Table 5, along with the parameters estimated using the spectrum of September 2018. Comparisons of observed spectra with corresponding models are shown in Figure 6.

**Table 5.** Derived properties of Romano's star at the moments corresponding to different acquired spectra. H/He indicates the hydrogen number fraction relative to helium, *f* is the filling factor of the stellar wind. Details of modeling may be found in [10,22,47].


*a*Clark et al. [47] assumed a distance to M 33 of 964 kpc.

Numerical calculations show that the bolometric luminosity of GR 290 is variable, being higher during the phases of greater optical brightness [10,46]. At the same time, the wind structure of GR 290 also varies in correlation with brightness changes—the slow and dense wind at brightness maxima becomes faster and thinner at minima (Figure 7), and the effective temperature<sup>7</sup> of the star increases from 25 kK (with WN11h spectral type) during the maximum of 2005 year to 31–33 kK (WN8h) during the minima.

<sup>7</sup> Effective temperature is defined as a temperature at radius *R*2/3, where the Rosseland optical depth is equal to 2/3.

**Figure 6.** Normalised optical spectra of GR 290 compared with the best-fit CMFGEN models (green line). The model spectra are convolved with a Gaussian instrumental profile. Description of observational data may be found in [10,22,23]. Notice that "September 2006" spectrum was obtained by P. Massey with WIYN 3.5 m telescope [18]. Spectral types are estimated based primarily on relative strengths of N V, N IV, N III, N II and He II *λ*4686 emission lines [48]. Spectra are displaced vertically for illustrative purposes.

Figure 8 shows the positions of the star in the H-R diagram at different times. The object clearly moves well outside the typical LBV instability strip [49,50], deep inside the region of Wolf-Rayet stars, except for a moment of maximum brightness in 2005. On average, GR 290 lays on the 40–50 M evolutionary tracks from the Geneva models [51] with rotation. Using CMFGEN, we found that hydrogen mass fraction in the atmosphere of GR 290 is 35% [22], and used this estimation for determination of current stellar mass and age. According to this tracks, the Romano's star should now be 4.5–5.7 Myr old and should have a mass of 27–38 M.

**Figure 7.** Change of the wind structure and extent over time. The region where *ne* ≥ 10<sup>12</sup> cm<sup>−</sup><sup>3</sup> is shown in dark red, 10<sup>12</sup> ≥ *ne* ≥ 10<sup>11</sup> cm<sup>−</sup><sup>3</sup> in red, and 10<sup>11</sup> ≥ *ne* ≥ 10<sup>10</sup> cm<sup>−</sup><sup>3</sup> in orange. Solid black line shows the radius where Rosseland optical depth (*τ*) is 2/3. Scale in units R is shown at the top.

**Figure 8.** Position of GR 290 in the Hertzsprung-Russell diagram at different times. Numbers correspond to: (1) October 2002; (2) February 2003; (3) January 2005; (4) September 2006; (5) October 2007; (6) December 2008; (7) October 2009; (8) December 2010; (9) August 2014; and (10) July 2016 and September 2018. The hatched strip shows an average line along which GR 290 moved during its recent luminosity cycles. The Geneva tracks [51] for 40 M (dashed line) and 50 M (solid line) with rotation are shown by blue lines, with dark blue part corresponding to the hydrogen burning in stellar core. Triangles mark the positions of late-WN stars from Large Magellanic Cloud (LMC), whose data were taken from Hainich et al. [52]. In addition, the positions of LBV stars P Cygni and HR Car are shown with circles. Data for these objects were taken from the works of Najarro [53] and Groh et al. [54]. Grey solid line is Humphreys-Davidson limit [3], grey dash-dotted line is LBV minimum instability strip as defined in [54].
