A&A 618, A17 (2018) Astronomy https://doi.org/10.1051/0004-6361/201833484 & c ESO 2018 Astrophysics

How common is LBV S Doradus variability at low metallicity??,?? V. M. Kalari1, J. S. Vink2, P. L. Dufton3, and M. Fraser4

1 Departamento de Astronomía, Universidad de Chile, Casilla 36-D, Santiago, Chile e-mail: [email protected] 2 Armagh Observatory, College Hill, BT61 9DG Armagh, UK 3 Astrophysics Research Centre, School of Mathematics and Physics, Queen’s University Belfast, BT7 1NN Belfast, UK 4 School of Physics, O’Brien Centre for Science North, University College Dublin, Belfield, Dublin 4, Ireland

Received 23 May 2018 / Accepted 21 June 2018

ABSTRACT

It remains unclear whether massive star evolution is facilitated by mass loss through stellar winds only or whether episodic mass loss during an eruptive luminous blue variable (LBV) phase is also significant. LBVs exhibit unique photometric and spectroscopic variability (termed S Doradus variables). This may have tremendous implications for our understanding of the first stars, gravitational wave events, and supernovae. A key question here is whether all evolved massive stars passing through the blue supergiant phase are dormant S Doradus variables transforming during a brief period or whether LBVs are truly unique objects. By investigating the OGLE light curves of 64 B supergiants (Bsgs) in the Small Magellanic Cloud (SMC) on a timescale of three years with a cadence of one night, the incidence of S Doradus variables amongst the Bsgs population is investigated. From our sample, we find just one Bsg, AzV 261, that displays the photometric behaviour characteristic of S Doradus variables. We obtain and study a new VLT X-shooter spectrum of AzV 261 in order to investigate whether the object has changed its effective temperature over the last decade. We do not find any effective temperature variations indicating that the object is unlikely to be a LBV S Doradus variable. As there is only one previous bona fide S Doradus variable known to be present in the SMC (R 40), we find the maximum duration of the LBV phase in the SMC to be at most a few 103 yr or more likely that canonical Bsgs, and S Doradus LBVs are intrinsically different objects. We discuss the implications for massive star evolution in low-metallicity environments, characteristic of the early Universe. Key words. stars: variables: S Doradus – stars: evolution – Magellanic Clouds

1. Introduction Physically, LBVs can be understood as variables within the evolved massive blue supergiant zoo. B supergiants (Bsgs) are Massive stars are important producers of radiative and kinetic evolved massive stars (>15 M ) which are thought to be in a energy in the Milky Way, whilst massive stars in low-metallicity transitional stage from the main sequence towards the evolved environments may provide us with key insights into their ion- Wolf–Rayet stage (although as noted in Vink et al. 2010 it is izing input in the early Universe. However, the evolution and unclear whether Bsgs are hydrogen-burning main sequence stars fate of the most massive stars in the Universe, especially at low with a large overshooting parameter, or helium-core burning metal content, is highly uncertain (Langer 2012). A key ques- stars). Most Bsgs display intrinsic photometric variability that tion is whether the bulk of the mass is lost either in stellar winds are characterized by their amplitude and timescales as one of (Chiosi & Maeder 1986), in luminous blue variable (LBV) type three types: eruptions, or binary interactions (Smith & Tombleson 2015), and 1. α-Cyg variables, which display (likely periodic) micro- how mass loss varies with metal content (Vink et al. 2001; Puls variations of .0.1m on timescales from days to months, now et al. 2008). commonly thought to be due to non-radial gravity mode pul- If low-metallicity massive stars indeed lose substantial sations (Lefever et al. 2007); amounts of mass in the LBV phase, this could be pivotal for 2. S Doradus variables (or luminous blue variables; LBVs), our understanding of pristine Pop III stars (Vink & de Koter which display aperiodic variations of &0.1m (van Genderen 2005; Smith & Owocki 2006), as it is usually assumed that these 2001) on timescales from years to decades encompass- objects keep their initial masses intact (Woosley & Heger 2015). ing a wide variety of light curves in amplitude and time, The mass-loss process at low metallicity is key for understand- with a clear distinction made in colour-brightness changes. ing seed black hole masses in the early Universe, the “heavy” Although no commonly accepted paradigm exists for the black holes recently identified from the gravitational wave event cause of the variability it must consist of radius inflation GW 150914 (Abbott et al. 2016), and supernovae and super- to explain the observed changes in optical brightness and novae impostors at low metallicity (Trundle et al. 2008). spectral type, rather than dust obscuration events. Almost all S Doradus variables also display α-Cyg variations, whilst the

? opposite is not the case; Based on observations collected at the European Organisation for 3. η Car variables (or eruptive LBVs; or supernova impostors), Astronomical Research in the Southern Hemisphere under ESO pro- m gramme 299.D-5054(A). which show changes of >1–2 within days indicative of an ?? The full Table1 is only available at the CDS via anonymous ftp eruptive event. These events are extremely rare (around one to cdsarc.u-strasbg.fr (130.79.128.5) or via http://cdsarc. per century in our Galaxy), with the most famous example u-strasbg.fr/viz-bin/qcat?J/A+A/618/A17 being η Car (de Vaucouleurs & Eggen 1952).

Article published by EDP Sciences A17, page 1 of 46 A&A 618, A17 (2018)

Classically, most S Doradus variables have been found serendip- itously either through identification of their peculiar variability 25 (Humphreys et al. 2016), or more recently, from their compact shells by infrared imaging (Gvaramadze et al. 2012). LBVs are rare; ∼20 bona fide LBVs are currently known in the Milky 20 Way and the 1/2 Z Large Magellanic Cloud (LMC), and only one bona fide member, R 40, is known in the 1/5 Z Small 15 Magellanic Cloud (SMC). It is not clear, however, whether the SMC star HD 5980 is an LBV or a Wolf–Rayet star Count (Humphreys et al. 2016; Smith & Tombleson 2015; Vink 2012; 10 Humphreys & Davidson 1994). Additionally, while it is thought that α-Cyg variables form 5 a large fraction (if not most) of all Bsgs (Lefever et al. 2007), it remains unclear how common S Doradus and eruptive LBV 0 variability are among Bsgs. In other words, do all Bsgs display B0 B1 B2 B3 B4 B5 B6 B8 B9 S Doradus variability, albeit at smaller amplitudes or varying Spectral type baselines? For many decades given their extreme rarity (even for massive Fig. 1. Recovered spectral types of Bsgs in our sample (red) and the stars), LBVs were thought to represent a brief evolutionary phase total number of Bsgs in the SMC (solid line) within the OGLE II survey (∼104 yr; Lamers et al. 1998) during a massive star’s life between area. the hydrogen-burning main sequence, and the core helium burn- ing Wolf–Rayet phase. As LBVs lie close to the Humphreys– S Doradus variability among the Bsgs population in order to Davidson instability limit (Humphreys & Davidson 1994), they establish whether most Bsgs are in fact dormant S Doradus are linked to an instability which allows massive stars to lose LBVs or if S Doradus LBVs are intrinsically different to Bsgs. mass, preventing them from becoming red supergiants (RSGs). This paper is organized as follows: Sect.2 describes the anal- Within this period of extreme mass loss the brightness increases ysis of the light curves of known B Sgs. Section3 describes the in line with an inflation of the envelope, resulting in a temperature identification of the sole potential S Doradus variable, AzV 261. decrease (Gräfener et al. 2012; Vink 2015). This picture is corrob- Section4 presents a discussion of the results. Our conclusions orated by their light curves, which show S Dor-type decadal vari- are given in Sect.5. ations (0.5–2 mag) interspersed by micro-variations (of the order of 0.1–0.2 mag). Although, it has traditionally been suggested that S Doradus 2. Analysis of periodicity of B Sgs in the SMC LBVs are transitional objects (Humphreys & Davidson 1994; 2.1. Bsgs sample Lamers et al. 1998), in the last decade, they have been proposed as direct core-collapse supernovae SNe progenitors (Kotak & We selected Bsgs with multi-epoch photometric data obtained Vink 2006; Gal-Yam & Leonard 2009; Groh et al. 2013; Smith during the second phase of the OGLE II were selected for anal- 2014). Additionally, searches for LBVs in nearby galaxies (e.g. ysis. The OGLE II survey obtained multi-epoch I-band imag- Massey et al. 2007) have found a vast number of LBV candi- ing of the SMC. The photometry covers a period of 3 yr span- dates without the distinctive photometric variability, and have ning approximately from 1 June 1997 to 31 November 2000. suggested that there are large number of Bsgs that are not at a cadence of approximately every night. The data cover only eruptive LBVs but probably dormant S Doradus LBV variables. the periods (for approximately 8 months each year) across the This would suggest a much longer lifetime for the LBV phase baseline in which the SMC is visible from the Warsaw telescope (∼105 yr), comprising a significant fraction of the He-burning situated at the Las Campanas Observatory in Chile. The median phase. However, a number of the LBV candidates found by seeing of the imaging is around 100, with a resulting median ran- Massey et al.(2007) have been shown to be sgB[e] stars (a small dom photometric uncertainty of ∼0.007 mag. Bad data points are subset of Bsgs whose evolutionary status remains unclear; see identified and removed according to the difference imaging cri- Miroshnichenko 2006). It has been suggested that sgB[e] and terion. The reader is referred to the OGLE II survey publication LBVs do not evolve into one another (see Humphreys et al. (Szymanski 2005; Udalski et al. 1997) for further details of the 2017b). data. In order to properly unravel the evolutionary status of the The stellar classifications given by Bonanos et al.(2010) LBVs at low metallicity, we need to increase the number of con- were used to compile a list of 179 known SMC Bsgs. In the firmed LBVs in low-metallicity galaxies. The key question is case of multiple varying classifications, we chose the most recent whether S Doradus LBVs are truly special Bsg objects, or if all classification based on spectroscopy. A cross-match with a toler- Bsgs might actually be dormant LBVs. In the latter case the total ance of 100 (the median seeing of the photometry) was performed number of LBVs might be an order of magnitude larger than with the OGLE II I-band imaging. Over the total field of view of the number of bona fide LBVs would suggest (see Massey et al. the OGLE II survey, there are 110 known Bsgs, of which 69 Bsgs 2007) and hence their lifetimes would increase to 105 yr or more, had multi-epoch I-band photometry (20 Bsgs had OGLE II pho- with relevant consequences for their evolutionary state. tometry, but not light curves, while 21 did not have OGLE II In this study, we present a systematic search for LBVs using photometry even though they fall within the OGLE II field of long-baseline (3 yr) medium-cadence (nightly) light curves of view). The recovered 69 Bsgs cover the entire B spectral type, known Bsgs in the 1/5 Z local group dwarf galaxy Small and are split based on their luminosity classification into 16 BIa, Magellanic Cloud (SMC). The data allow us to accurately quan- 24 BIab, and 24 BIb Bsgs, and 5 with no luminosity classifica- tify the incidence of S Doradus variability among known B Sgs. tion. The five Bsgs without luminosity classification were dis- The main aim of the study is to quantify the incidence of carded from the final catalogue. The total number of identified

A17, page 2 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

Table 1. Properties of Bsgs having photometry in the OGLE II survey.

Name SpT Right ascension Declination IIerr Period Significance Aperiodic J2000 J2000 (mag) (mag) (days) (>95%) (>0.1 mag) 2dFS0153 B9Ib 0:36:37.00 −73:34:48.7 13.665 0.006 12.35 No No 2dFS0240 B9Ib 0:38:44.83 −73:35:51.4 14.309 0.007 9.07 No No AzV356 B0Iw 1:04:35.47 −72:04:43.2 13.649 0.007 1245.78 No No AzV376 B1-2Ib 1:05:07.30 −72:24:36.7 13.419 0.007 1245.78 No No AzV351 B1-2Ibe 1:04:21.38 −72:09:14.6 13.412 0.007 414.99 No No AzV394 B1-5Iab 1:06:01.52 −72:23:28.1 13.567 0.007 1245.78 No No AzV329 B1.5I 1:03:22.99 −72:02:35.5 14.028 0.007 1245.78 No No AzV340 B1Ia 1:03:49.43 −72:07:29.4 12.702 0.006 1245.78 No No AzV339 B2.5Iab 1:03:44.18 −72:52:25.8 12.721 0.007 1245.78 No No AzV342 B2.5Iab 1:03:56.07 −72:02:47.4 13.099 0.007 1245.78 No No

Notes. Only the first ten lines are shown here to display the form and content. The full table is available at the CDS.

Bsgs, and the total known in the OGLE II survey area are shown 2dFS0153 as a fraction in Fig.1 as a function of spectral type. The recovery 13.64 fraction is ∼70% until spectral type B6. For binomial statistics, the recovered sample size can predict simple binomial proper- ties at the 99% confidence level with a 10% margin of error. 13.66 Although the selection criteria are unbiased, OGLE II photom- etry saturates around V ∼ 12. Our final sample is thus not uni- 13.68 (mag)

formly distributed in either spectral type or brightness, but does I contain a sufficiently representative sample of Bsgs for our anal- ysis. A list of the Bsgs and their photometric properties is given 13.70 in Table1.

2.2. Analysis of time series data 13.72

S Doradus variables present multiple types of photometric vari- 2450600 2450800 2451000 2451200 2451400 2451600 2451800 ability (van Genderen 2001), presumably due to an inflating Julian Date radius and decreasing photospheric temperature. The photo- Fig. 2. Representative light curve of the Bsg 2dFS0153 from the OGLE metric variability is between 0.1 (below which most micro- survey, where the measured OGLE I-band magnitude is plotted against variations are generally caused by the α-Cyg g-mode pulsations) the Julian Date. Light curves of all 64 Bsgs in our sample are shown in and 2.5 mag (above which most variations are due to eruptive Fig. A.1. events, falling into the eruptive phase of LBVs) on timescales of years to decades. Light curves within this broad magnitude periodic variability is comparatively straightforward. To do so, and time span show diverse characteristics, yet few distinctive we employed the Lomb–Scargle periodogram, which is a statis- features thus excluding them from most other variables. tical tool often used to identify periodic variations in unevenly Primarily, the light curve varies on timescales greater than a spaced observations. A power spectrum for each Bsg light curve year (not considering α Cyg variations), with a systematic trend was constructed. The period range was 0.5–1000 d, with a fre- in brightness. Additionally, a decrease in brightness is accom- quency grid spacing of 0.0037 d−1. The significance of the iden- panied by a bluer colour and vice versa. While the former rules tified peaks was tested by a Monte Carlo simulation (see e.g. out short period variability, the latter precludes variability due VanderPlas 2018; Frescura et al. 2007). To do so, 1000 simu- to dust obscuration events, and is considered a genuine increase lated light curves were generated using a bootstrapping tech- in stellar radius. Considering the periodicity of S Doradus vari- nique, where random data pairs are replaced from the original ables, there are multiple cycles/periodicities present in the light light curves. The bootstrapping technique gives us a significance curve with pronounced beats, although overall the light curve level for the periods detected. Each Bsg showing a peak in the increases/decreases aperiodically. power spectrum greater than the 99% level can be identified as a In this section, we search the light curves of the Bsgs to iden- periodic variable following the methods outlined in VanderPlas tify stars with multiple periodicities, and/or irregular photomet- (2018), and shown in Figs.3 and B.1. ric variability above 0.1m. To place this into context, we also When searching for periodicity, we attempt to differenti- compare our light curves with those of known S Doradus vari- ate between short and intermediate period variables. Bsgs hav- ables in the Magellanic Clouds. The light curves of the Bsgs are ing short periodic variations (essentially with periods less than presented in Figs.2 and A.1. We discuss our results below. 2 days), are generally α Cyg variables (unless the amplitude of their variations exceeds a few tenths of magnitude). Nearly 2.2.1. Searching for periodic variability all Bsgs show periods of 1–2 days with varying amplitudes of 0.05–0.1 mag. However, given that our sampling frequency is In Figs.2 and A.1, we plot the light curves of 64 Bsgs and search about the same, we cannot conclusively identify such variations for photometric variability at periodic and aperiodic timescales as being strictly periodic due to the star. We note that periodic which may have been caused by outbursts. Identification of variations around 0.5 and 1 d are in most cases almost certainly

A17, page 3 of 46 A&A 618, A17 (2018)

0.10 2dFS0153 1400 0.08 1200 0.08

0.06 1000

0.06 800

0.04 m (mag) δ 0.04 600 Lo,mb-Scargle Power

0.02 400 0.02 200

0.00 2dFS0153 0.00 0 0 200 400 600 800 1000 0 200 400 600 800 1000 1200 Period (days) δt (days)

Fig. 3. Representative Lomb–Scargle periodogram of Bsg 2dFS0153. Fig. 4. Two-dimensional histogram (δm–δt) of the Bsg 2dFS1053 from The period in days is plotted against the Lomb–Scargle power. The the OGLE II data, where the colour gradient marks the density of the dashed red and blue lines are the 99% and 95% significance levels iden- points (see the colourbar). The dashed line marks the 10σ level of the tified using the bootstrap Monte Carlo simulations. Periodic variables random photometric uncertainty, while the solid line differentiates vari- are those whose peak in the power spectrum is greater than the 99% sig- ations above the 0.1m level. The δm–δt two-dimensional histograms of nificance level. Lomb–Scargle periodograms of the remaining 63 Bsgs the remaining Bsgs are included in Fig. C.1. are displayed in Fig. B.1. characterize variability in galaxies. Overall, different timescales caused by sampling below the sampling frequency, representing are sampled to varying levels in the plots; however, we are only the nightly half-cadence and cadence levels. Therefore, although interested in aperiodic/periodic variations greater than 0.1m on in many cases multiple frequencies in this time regime are found, timescales greater than a few days. To aid in identification we we cannot isolate them as being strictly stellar (and thereby α compute the δm–δt plots of known S Doradus variables, R 40 and Cyg) with our data. Hence, any periodic variables below 1d are R 110, in the Magellanic Clouds1, and compare them with our discarded from the final sample. In total, we identified five can- Bsgs sample. From the δm–δt plots of the known S Doradus vari- didate periodic variable Bsgs in our sample. Split into groups ables shown in Fig.5, we identify only one BIb star as displaying based on luminosity class, 1 out 24 is BIabs, 1 out of 16 is BIas, aperiodic variability above the significance threshold: AzV 261. and 3 out of 24 are BIbs. In all instances, the period is greater than the Nyquist frequency at the 3σ level (1% level, due to ran- dom variations). We did not attempt to smooth out any long term 3. AzV 261: a candidate S Doradus variable variability in the light curve. Only in the case of AzV 261 were Searching systematically through archival light curves from the multiple periods above the 3σ level found. OGLE database, we found that AzV 261 in the SMC is an excellent S Doradus candidate. In Fig.6a, we show the I-band 2.2.2. Searching for aperiodic variability light curve of the star spanning a decade (1998–2009). The (V−I) colour behaviour is also shown (the additional time series Searching for aperiodic variability is not straightforward com- data come from OGLE III data of Kourniotis et al. 2014). From pared to periodic variability as there are no commonly accepted Fig.6, AzV 261 first displays micro-variations of ∼0.2 mag on metrics for massive stars (see e.g. Labadie-Bartz et al. 2017; timescales of days. In addition, the brightness increased by Findeisen et al. 2015). We employed the δm–δt plots which are 0.53 mag from 13 November 2000 to 20 October 2008 (the suggested by Mahabal et al.(2011) and Findeisen et al.(2015) OGLE-III data are until 2009). Using the Period04 software as a better indicator of variability in a series of tests on sim- (Lenz & Breger 2005), we attempted to characterize further ulated light curves, although the correlation with any period is the periodicity and micro-variations. In the Lomb–Scargle peri- low. The δm–δt plots are only used to identify variability, and odogram, AzV 261 shows four significant peaks (see Fig. A.2 further checks were done by eye as at times some features can- and Table1). However, this light curve has a long-term trend not be compared directly, as suggested by Labadie-Bartz et al. of increasing brightness. We fit a fourth-order polynomial to (2017). remove this trend, and attempt using Period04 to fit periods The δm–δt plot presents the incidence of variability at a par- to replicate the light curve of AzV 261. However, even with ticular timescale in a light curve. It is defined as the sum of all three periods, the residuals of the light curve are significant observations at magnitude m and time t of a light curve, by com- (see Fig.6b) indicating that the observed possibly semi-periodic puting the differences, variations are micro-varitaions (see van Genderen 2001; Lamers et al. 1998). Based on its light curve, AzV 261 is the single likely ∆m = |m − m | (i > j) i j i j (1) LBV candidate in our sample. ∆ti j = |ti − t j| (i > j) In addition, the colour information from Fig.6c strongly supports the LBV candidacy: the (V−I) colour becomes red- and i > j such that each pair is only considered once. By pre- der with time, corresponding to an actual temperature decrease. senting the differences between magnitude and time as a den- sity histogram, we can directly observe how much variability is observed on what timescales (see Figs.4 and C.1). We note that 1 Data are from the American Association of Variable Star Observers the δm–δt plot is closely related to the structure function used to International Database, at http://www.aavso.org

A17, page 4 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

Fig. 5. Two-dimensional histogram (δm–δt) of known S Doradus variables R 40, R 110, S Doradus on (panel a), (panel b) and (panel c) respectively. The candidate S Doradus variable in our Bsgs sample, AzV 261 is shown in (panel d). Symbols are same as Fig.4.

This is consistent with the picture of a brightness increase as In addition, in Massey et al.(1995) the authors obtained slit the temperature decreases due to radius inflation, while the spectra in 13 December 1992 using the Cerro Tololo Interamer- object’s bolometric luminosity stays roughly constant. There- ican Observatory 4m telescope in the blue-air Schmidt cam- fore, AzV 261 is a prime S Dor-type LBV candidate, the only one era. The final data cover the 3900–4800 Å range, at a resolu- among 64 known Bsgs identified by our systematic search. Its tion of 3 Å pixel−1. Based on these spectra the stellar classifica- light curve displays micro-variations of the order of 0.2 mag on tion is noted by the authors of that paper as O8.5Ie, however timescales of a few days to months, with an overall increase over a re-classification of the stellar spectrum (show in in Fig.7) eight years of 0.53 mag in brightness. The increase in bright- points towards an early B2 supergiant classification (Massey, ness corresponds to a redder colour and therefore a cooler tem- priv. comm.), agreeing with the classification by Evans et al. perature. All these features are unique to S Doradus variables (2004). Overall, photometric variability suggests S Doradus can- (van Genderen 2001; Lamers et al. 1998). Although the ampli- didacy for AzV 261; however, the spectral features themselves tude in overall change in brightness is on the lower end of most do not really agree with this notion. So, we must take and anal- confirmed LBVs (which are of the order of 0.5–0.1 mag). Over- yse further spectra to find out whether the effective temperature all, the light curve characteristics of AzV 261 point towards a changes between epochs. S Doradus classification. In the literature, spectroscopic classification of the object taken when its magnitude was fainter (in 30 September 1999) 3.1. Spectral analysis of AzV 261 from Evans et al.(2004) suggests a B2Ib(e) classification. Evans 3.1.1. Medium-resolution spectroscopy et al.(2004) used the 2dF multi-fibre spectrograph mounted on the Anglo-Australian 3.9 m telescope to obtain low-resolution We obtained medium-resolution spectroscopy with the X-shooter spectra over the 3900–4800 Å region with a spectral element slit spectrograph mounted on the UT2 telescope of the Very resolution R ∼ 1500 (for spectral classification), along with Large Telescope (VLT) in Paranal, Chile (see Vernet et al. 2011 Hα spectra (R ∼ 2500) in 30 September 1999. The blue spec- for details) on 2017 December 12. The X-shooter spectrograph trum is shown in Fig.7, while the H α spectrum is plotted in simultaneously observes the 300–2500 nm wavelength range Fig. 12. in medium-resolution (R ∼ 7000) slit spectra using three arms

A17, page 5 of 46 A&A 618, A17 (2018)

13.0 S Dor 13.2 3.5 H– Hδ– Hγ– Hβ–

13.4 –Hζ

13.6 3.0 (mag) AzV 261 I 13.8 14.0 2.5 (a) 1992-Dec-13

0.15 2.0 )

I 1999-Sep-30

− 0.10 V

( 1.5

0.05 Normalized flux + Constant (b) 1.0 0.00 2451000 2452000 2453000 2454000 2455000 Julian Date 2017-Dec-12 0.5 4000 4200 4400 4600 4800

0.3 Angstroms (A)˚

0.2 (a) Fig. 7. Normalized X-shooter spectra of AzV 261 in the MK classifica- 0.1

(mag) tion wavelength range, with key spectral lines marked. Also shown are I 0.0 the archival spectra from Massey et al.(1995) taken in 1992, and from 0.1 − Evans et al.(2004) taken in 1999. A constant is applied for visualisation 0.2 Normalized − purposes. 0.3 − 0.4 − purpose of this study, slit nodding was not used; the infrared 0.20 observations are thus not usable, due to sky contamination, and 0.15 (b) 0.10 were discarded. The spectra were observed in moderate seeing 00 00 00 0.05 conditions (∼1 ) using a slit of of 0.8 and 0.9 in the UVB and 0.00 VIS arm, respectively. The achieved R is 6200 and 7500 in the 0.05 Resdiuals (mag) − 0.10 UVB and VIS arm, respectively. Spectra were reduced using the − 0.15 − X-shooter pipeline, with sky subtraction performed using the opti- 0.20 − 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 mal extraction routine. The final sky-subtracted spectra are given Julian Date +2.45 106 × in Fig.7.

3.1.2. Radial and rotational velocities

2454700 Initial stellar radial velocity estimates were obtained from the 13.80 i 2454550 Doppler shifts of the He spectrum; the strong Balmer lines were not considered as they had significant emission in their 13.85 2454400 cores. The observed He i lines were fitted using a simple Gaus- 2454250 sian plus a low-order polynomial to represent the continuum. 13.90 As discussed below, there is evidence of some stellar rotational 2454100 V broadening, whilst the diffuse lines (due to 1P − 1D and 3P − 3D

13.95 2453950 Julian Date transitions) exhibited significant intrinsic broadening. However,

2453800 visual inspection showed that the fits appeared to identify the line 14.00 centre. For transitions between triplet levels, the fine structure 2453650 multiplet transitions covered a wavelength range corresponding −1 14.05 2453500 to ∼12–15 km s and in these cases the midpoint wavelength

2453350 was adopted. Several lines were not considered as they were 0.06 0.08 0.10 0.12 0.14 0.16 either very weak (e.g. 4169 Å) or had emission in their wings (V I) − (e.g. 5876 and 6678 Å). The radial velocity estimates are sum- −1 Fig. 6. Top: I-band light curve of AzV 261 spanning from 1998 to marized in Table2 and have a mean value of 167 ± 11 km s . 2009. Also shown is the scaled historical light curve of the classic LBV It was also possible to identify features due to the ions, star S Doradus for comparison (crosses). Panel b:(V−I) colour plotted C ii, Mg II , O ii, and Si iii. However these lines were weak and against the Julian date for AzV 261. Middle:(panel a) light curve with often blended; additionally, several appeared to be affected by the overall trend removed by a fourth-degree polynomial, with the best- emission near to the line cores (see Figs.9–11). Radial veloc- fit predicted light curve (the three peaks are given in Table1). Panel b: ity estimates were found from four metal lines using the same residuals between the fit period and the observed light curve. Bottom: V methodology as for the He i spectrum; they are summarized in − vs. (V I) colour-magnitude diagram of AzV 261 over time, with time Table2. They lead to a radial velocity estimate of 168 ± 5 km s−1, given by the colourbar. The behaviour of a brightness increase corre- sponding to a temperature decrease is seen here. in excellent agreement with that found from the He i spectrum. The intrinsically narrow metal line spectra showed a signifi- for ultraviolet (UVB; 300–560 nm), visual (VIS; 560–1020 nm) cant amount of broadening, which could be due to either rotation and infrared (1020–2500 nm) wavelength ranges. For the or to macro-turbulence. Also, the cores of the He i lines were not

A17, page 6 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

Table 2. Radial velocities (vr) and projected rotational velocity (ve sin i) 1995; Hubeny et al. 1998; Lanz & Hubeny 2007). They cover −1 estimates in km s from the He i and metal line spectra. a range of effective temperatures, 10 000 K ≤ Teff ≤ 35 000 K in steps of either 1500 K or 2000 K. Logarithmic gravities (in cm s−2) range from 4.5 dex down to the Eddington limit Species Wavelength vr ve sin i FT PF in steps of 0.25 dex, and micro-turbulences are from 0 to 30 km s−1 in steps of 5 km s−1. As discussed in Ryans et al. He i 3926.54 165 116 101 (2003) and Dufton et al.(2005), equivalent widths and line He i 4009.26 191 88 122 profiles interpolated within these grids are in good agreement He i 4026.27 157 97 113 with those calculated explicitly at the relevant atmospheric He i 4120.90 159 78 91 parameters. He i 4143.75 169 103 105 He i 4387.93 169 76 93 These non-LTE codes adopt the “classical” stationary He i 4437.55 179 – – model atmosphere assumptions, i.e. plane-parallel geometry, He i 4471.58 158 88 88 hydrostatic equilibrium, and optical spectrum that is unaffected He i 4713.24 175 96 60 by winds. The grids assumed a normal helium-to-hydrogen ratio He i 4921.93 151 – – (0.1 by number of atoms) and have been calculated for a range He i 5047.74 168 – – of metallicities; the ratio for an SMC metallicity is used here. C ii 4267.15 166 86 87 However, tests using the Galactic (7.5 dex) or SMC (6.9 dex) Si iii 4552.62 162 106 97 grids lead to only small changes in the atmospheric parame- Si iii 4567.84 174 – – ters (primarily to compensate for the different amounts of line Si iii 4574.76 172 – – blanketing). For further information on the tlusty grids see Ryans et al.(2003) and Dufton et al.(2005). Obviously, a hydro- Notes. The adopted laboratory wavelengths are also listed (see text for static code like tlusty cannot be used to derive wind param- details). eters, and can also not be employed to obtain meaningful fits to wind affected hydrogen lines such as Hα. For these mod- elling types an alternative code such as cmfgen or fastwind sharp, again implying additional broadening mechanisms. We (e.g. Massey et al. 2013) would be needed. However, in the first have attempted to estimate the contribution of rotational broad- instance, we are mostly interested in deriving reliable effective ening using a Fourier transform (FT) methodology (Simón-Díaz temperatures, and comparisons between tlusty and fastwind & Herrero 2007), which has been widely used for early-type have generally shown good agreement between the two codes stars (Dufton et al. 2006, 2013; Lefever et al. 2007; Markova for effetive temperature estimation (e.g. Trundle et al. 2007). & Puls 2008; Simón-Díaz et al. 2010, 2017; Fraser et al. 2010; Initial estimates for the effective temperature and surface Simón-Díaz & Herrero 2014). It relies on the convolution the- gravity could be obtained from fitting the neutral helium orem (Gray 2005), namely that the Fourier transform of con- and hydrogen spectrum. There is a loci of possible solutions volved functions is proportional to the product of their individ- (see e.g. Dufton et al. 1999 for further details), which range ual Fourier transforms. Then the first minimum in the Fourier from an effective temperature and surface gravity (in cm s−2) transform for an observed spectral line, is assumed to be the first of T ' 16 000 K and log g ' 2.40 dex to T ' 23 000 K and zero in the Fourier transform of the rotational broadening pro- eff eff log g ' 3.05. For lower effective temperatures the theoretical He i file. Further details on the implementation of this methodology becomes too weak, whilst for higher values the He ii spectra can be found in for example Simón-Díaz & Herrero(2007) and would be observable. Dufton et al.(2013), whilst the projected rotational estimates, v sin i, are listed in Table2. This range can be constrained using the observed metal lines e and assuming that AzV261 has SMC metal abundances simi- Projected rotational velocities were also estimated from the ii He i spectrum by fitting the profiles of the observed lines, using lar to those found in H regions (see e.g. Kurt & Dufour 1998; a Gaussian function (representing the instrumental profile) with Garnett 1999; Carlos Reyes et al. 2015; Toribio San Cipriano et al. a rotational broadening function. Particularly for the diffuse 2017) and for other hot stars (see e.g. Venn 1999; Rolleston et helium lines, the intrinsic broadening may be significant, whilst al. 2003; Hunter et al. 2007; Trundle et al. 2007), as summarized macro-turbulence and micro-turbulence (Simón-Díaz & Herrero in Dufton et al.(2005). In Figs.9–11, observed and theoretical 2014; Simón-Díaz et al. 2010, 2017) may also be significant. spectraare presentedfor thewavelength region3900–4700 Å. The These estimates (also listed in Table2) should thus be consid- wavelength region 4200–4300 Å is omitted as the only feature ered as upper limits. This is confirmed by the PF estimates for observable was the C ii 4267 Å doublet, which is subject to com- the He i lines being on average 16% larger than their FT equiva- plex non-LTE effects (see Sigut 1996; Nieva & Przybilla 2006). lents, but they provide a useful check on the FT estimates. The theoretical spectra have been convolved with a Gaussian func- Convincing minima in the FT of the metal lines were only tion to allow for instrumental broadening and with a rotational −1 found in two cases with the ve sin i estimates being consistent broadening function with ve sin i = 95 km s . A micro-turbulence with those from the He i spectrum. Additionally, no convincing of 10 km s−1 has been adopted, which is similar to that found in minima could be found for the He i lines at 4437 and 5047 Å, analyses of Magellanic Cloud targets (Hunter et al. 2007; Trundle whilst the line at 4921 Å was affected by emission in its red wing. et al. 2007; McEvoy et al. 2015) with similar effective tempera- Combining all the available FT estimates yields a mean value of tures and gravities to the locus of estimates found for AzV 261. −1 T g ve sin i = 93 ± 13 km s . Spectra for three sets of atmospheric parameters ( eff, log ) are shown: (16 000, 2.45), (18 000, 2.65), and (18 000, 2.80). 3.1.3. Atmospheric modelling Overall the best agreement between observation and theory appears to be for the model with an effective temperature of We have employed model-atmosphere grids calculated with the 18 000 K. Below we discuss the spectra of the different ionic tlusty and synspec codes (Hubeny 1988; Hubeny & Lanz species in more detail.

A17, page 7 of 46 A&A 618, A17 (2018)

H I, He I. As discussed above all three models provide a sat- isfactory fit to the helium spectrum and to the wings of the Balmer lines. The theoretical spectra for the He i line at 3926 Å are too weak, possibly due to the adoption of a semi-classical quasi-static approximation for the line broadening (Dimitrijevic & Sahal-Brechot 1984). Additionally, the theoretical profiles for the He i line at 4713 Å appear to be too strong, although the observed profiles may exhibit emission in its wings. The lack of significant He ii absorption in all the theoretical spectra (e.g. at 4541 and 4686 Å) is also consistent with the observations. C II. For the gravities considered here, the C ii spectrum has a maximum strength at Teff ' 20 000 K. As such it shows little variation in the theoretical spectrum, with the doublet at 4267 Å being in good agreement with observation. As discussed above, the doublet at 4267 Å is difficult to model, and indeed all our the- oretical spectra are stronger than observed, which is consistent with previous investigations using this model grid (Trundle et al. 2007; Hunter et al. 2007). N II. The theoretical spectra adopt a nitrogen abundance of 7.1 dex, which is higher than that found in H ii regions (e.g. 6.55 dex; Carlos Reyes et al. 2015) or the lowest stellar abun- dances (e.g. 6.66 dex; Rolleston et al. 2003). It was chosen to be representative of the typical nitrogen enrichment observed in SMC B-type stars (Trundle et al. 2007; Hunter et al. 2007) and as such it provides no constraint on the atmospheric parameters. However, the reasonable agreement between theory and observa- tion for the N ii line at 3995 Å argues against any large nitrogen enhancement. Fig. 8. Observed and theoretical profiles for the N ii lines at 3995 O II. There is a rich O ii spectrum with some of the stronger and 4630 Å. The theoretical profiles have been convolved with a rota- features identified in Figs.1–3; additionally, most of the tional broadening function and are for atmospheric parameters of unidentified absorption features are also due to this ion. The the- Teff = 18 000 K and log g = 2.65. The adopted nitrogen abundances are oretical line strengths are sensitive to the adopted effective tem- 6.7 (green), 6.9 (blue), and 7.1 (red) dex. Also shown is a theoretical perature and the model with T = 18 000 K provides the best plot for Teff = 16 000 K and log g= 2.45 and a nitrogen abundance of eff 7.1 dex (dotted red). overall agreement with observation. Mg II. The doublet at 4481 Å is too strong in all three theoretical spectrum. However, the observed profile appears to be affected by a least-squares fitting of a linear polynomial to continuum by emission in both wings making any comparison problematic. regions at ±3–5 Å from the line centre. Si II, Si III. The Si ii doublet at 4128–4130 Å appears to be very There is marginal evidence for an observed N ii line at weak in the observed spectrum. The spectra for the two hotter 3995 Å and this would imply a nitrogen abundance of ∼6.7 dex. models appear consistent with the observations. The Si iii triplet Alternatively, a conservative upper limit for the nitrogen abun- transitions between 4552 and 4574 Å are best fitted by the theo- dance would appear to be 7.1 dex. The N ii at 4630 Å is not retical spectrum with Teff = 18 000 K, although the other spectra seen again, implying a similar upper limit on the nitrogen also provide acceptable fits. abundance. Adopting a baseline nitrogen abundance of 6.6 dex In summary the model with Teff = 18 000 K provides the best (Carlos Reyes et al. 2015; Rolleston et al. 2003; Trundle et al. overall fit to the spectrum, and we estimate a stochastic error of 2007; Hunter et al. 2007) would then imply a nitrogen enhance- ±2000 K. This error would translate into an error of ∼0.2 dex in ment of ≤0.5 dex Uncertainties due to possible errors in the the gravity. Combining this in quadrature with a fitting uncer- atmospheric parameters are dominated by the uncertainty due to tainty of ∼0.2 dex leads to an error in the gravity of ±0.3 dex. the effective temperature. This is illustrated in Fig.8 where a the- Hence our best estimate of the atmospheric parameters and their oretical profile is shown for an effective temperature of 16 000 K errors are Teff = 18 000 ± 2000 K and log g = 2.65 ± 0.3 dex. (and log g = 2.45 dex). Adopting this lower effective temper- As discussed above, the strength of the N ii spectrum is close ature would increase the abundance limits by approximately to or below the limit for detection. As nitrogen is important 0.4 dex. for constraining the evolutionary status of early-type stars, we The estimation of the atmospheric parameters depends par- have attempted to set an upper limit on its abundance. Several ticularly on the assumption of baseline abundances for helium, N ii lines can be seen in B-type spectra, especially for targets oxygen, and silicon. It is unlikely that the silicon abundance has showing significant nitrogen enrichments (see e.g. Dufton et al. been effected by nucleosynthetic process since the star formed. 2018). In Fig.8 we show the two intrinsically strongest fea- However, the other two elements could be affected by the mix- tures at 3995 Å and 4630 Å that are not significantly affected by ing of material processed by the CNO bi-cycle to the surface. blending (Dufton et al. 2018), together with theoretical profiles, Adopting a nitrogen enhancement of 0.5 dex, the SMC models which are for our adopted atmospheric parameters and have been of Brott et al.(2011) imply changes in the helium and oxy- convolved with a rotational broadening functions and are for gen abundances of less than <0.01 and <0.04 dex for masses nitrogen abundances between 6.7 and 7.1 dex. The observational between 15 M and 60 M . Increasing the nitrogen enhance- and theoretical profiles were normalized in an identical fashion ment to 0.9 dex (i.e. allowing for possible errors in the effective

A17, page 8 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

Fig. 9. Observed and theoretical profiles for selected wavelength Fig. 10. Observed and theoretical profiles for selected wavelength regions. The theoretical profiles have been convolved with a rota- regions (see Fig.9 for details). tional broadening function and are for atmospheric parameters of Teff = 18 000 K and log g = 2.65 (red); Teff = 16 000 K and log g = 2.45 (blue); and Teff = 20 000 K and log g = 2.80 (green). Selected absorption lines are marked, whilst the narrow features at 3933 and 3968 Å are due to interstellar Ca ii. temperature) would lead to changes of <0.02 and <0.09 dex. This implies that our use of a normal SMC abundance should not be a major source of error.

3.1.4. Hα line profile The Hα line strength and profile in Bsgs is a key indicator of mass loss via stellar winds. The rate of mass loss is still under debate, with differences in measured mass-loss rates varying by Fig. 11. Observed and theoretical profiles for selected wavelength up to factors of 10 from different indicators (e.g. the Hα line regions (see Fig.9 for details). vs. ultraviolet P Cygni lines), which occurs because clumping in the winds significantly affects the profile shapes of emission lines, including Hα. In addition, stellar pulsations in Bsgs may suggests that the star is hotter than the bi-stability jump of be linked to periods of enhanced mass loss. 21 000 K, as absorption appears at and below the jump (Petrov The Hα emission line strength (Fig. 12) changes from et al. 2014). −22.4 ± 1.2 Å in the 1999 Evans et al. spectra to −38.5 ± 0.3 Å in 2017 X-shooter spectra. It is clear that AzV 261 displays a vari- 3.2. Circumstellar dust able stellar wind, although the model fits to observations may not result in accurate mass-loss rates due to the choice of models Most S Doradus LBV variables show evidence of excess free- (e.g. TLUSTY). If the sky subtraction for the multi-fibre spec- free emission (Humphreys et al. 2017a,b; de Wit et al. 2014). In troscopy of Evans et al.(2004) is poor, the measured equiva- contrast with evolved supergiant B[e] (sgB[e]), and warm hyper- lent width would be higher from their spectrum. Their equivalent giant stars, they do not display evidence of warm dust. Compared width estimate is in fact lower, implying that uncertainties in the to the other Hα emitting objects in the Bsgs zoo, S Doradus vari- sky subtraction cannot explain the difference. We do not fit more ables therefore have excess free-free emission resembling that of sophisticated model atmospheres as we are merely interested in classical Be (cBe) stars (which are main sequence fast rotators). observing a change in spectral type (rather than wind variabil- We can therefore use the infrared colour–colour diagrams ity) in order to determine whether AzV 261 is a S Doradus vari- to distinguish AzV 261 from other objects in the Bsgs zoo. able. Regarding the absence of P Cyg absorption, this actually In Fig. 13a, we plot the near-infrared (J−H) vs. (H−Ks)

A17, page 9 of 46 A&A 618, A17 (2018)

1.0 characteristic temperature variability seen in most S Doradus LBV variables, which remains puzzling. Finally, we note that AzV 261 displays an extremely low luminosity for a S Doradus 0.8 –2017-Dec-12 LBV variable (of logarithm of luminosity log L = 4.9 ± 0.1 L ), 1 which implies a stellar mass of 15.9±3 M according to the mod- 0.6 els of Brott et al.(2011). This places AzV 261 at the low-mass end of known S Doradus LBV variables. In Fig. 15, we plot the position of AzV 261 and the Bsgs 0.4 studied here in the Hertzsprung–Russell diagram with respect

Normalized flux to stellar tracks from Brott et al.(2011), along with known 0.2 S Doradus variables in the SMC, LMC, and the Milky Way. The

–1999-Sep-30 effective temperatures and bolometric corrections (to calculate the luminosities of the Bsgs) were estimated using the spectral 0.0 type-effective temperature scale of Trundle et al.(2007), and 6500 6520 6540 6560 6580 6600 6620 6640 computed following the procedure described in McEvoy et al. Angstroms (A)˚ (2015). The data for the luminosities and effective temperatures of the literature S Doradus variables here were taken from Smith Fig. 12. Normalized Hα line profile from the X shooter spectra taken et al.(2018), who present revised distances and luminosities for in 2017 December (red) and archival spectra from Evans et al.(2004) taken in 1999 September shown as solid blue lines. The emission line known galactic S Doradus variables using recent Gaia DR2 par- equivalent width increased by −16.1 Å within this period. We note that allaxes. Their results identify a population of low-luminosity if the sky subtraction from the Evans et al.(2004) fibre spectra is poor, confirmed S Doradus variables in the galaxy which do not have this difference will only be magnified. extragalactic counterparts (e.g. W 243 and HD 168607). The authors of that paper suggest that a systematic search for new colour–colour diagram showing the position of known S Doradus variables in external galaxies might well uncover S Doradus variables, sgB[e], warm hypergiants, cBe stars (taken new low-luminosity LBVs, such as those shown here. The exis- from Humphreys et al. 2017a), and Bsgs in our sample. Also tence of these low-luminosity S Doradus LBV variables in the shown is the main sequence locus at SMC metallicity taken from Milky Way and in the SMC suggest the existence of a differ- Kalari et al.(2018). In addition, we also show the data in the mid- ent instability mechanism, with important implications for SNe infrared (from Humphreys et al. 2017a) for the same stars in the (see Smith et al. 2018). However, we consider it unlikely that [3.6 µm]–[4.5 µm] vs. [5.8 µm]–[8.0 µm] mid-infrared colour– AzV 261 is a member of this group since it does not show the colour diagram. From these diagrams, AzV 261 exhibits modest characteristic temperature variations seen in other LBVs. free-free emission at levels similar to other known S Doradus LBVs at the higher end of those exhibited by cBe stars. In 4. Discussion: S Doradus variables among the Bsg Fig. 14, we plot the spectral energy distribution of AzV 261 along with the TLUSTY model atmosphere for the best-fit stel- population lar parameters. The best-fit model for the infrared excess is The key question we wish to address is whether Bsgs are a black body of Tb = 2550 K. The position of AzV 261 in fundamentally dormant LBVs or if they represent separate stel- infrared colour–colour diagrams, and its spectral energy distri- lar populations, which would make bona fide S Doradus LBVs bution point towards free-free emission. “special objects”. Since the spectroscopic analysis of the surface gravity pre- Although individual LBVs and LBV candidates have been cludes the possibility that AzV 261 is a cBe star, it is most likely discovered at metallicities that are potentially even lower than a S Doradus variable based on its photometric variability and that of the SMC (Drissen et al. 2001; Izotov & Thuan 2008; infrared colours. However, the absence of spectroscopic variabil- ity is puzzling. Based on the absence of spectral non-variability Pustilnik et al. 2008; Herrero et al. 2010; Bomans & Weis 2011), AzV 261 is not a LBV S Doradus variable. in this systematic study of the SMC we found just one S Doradus candidate out of a total of 64 Bsgs. If we assume that all Bsgs pass through the S Doradus phase 3.3. AzV 261 within the Bsgs zoo in their lifetimes, we can set an upper limit to the duration of Based on the observed properties of AzV 261, we attempt to the S Doradus phase. Assuming that the mean duration between classify AzV 261 within the Bsgs zoo (assuming that it may be the end of the H-burning lifetime of a massive star and the onset a S Doradus LBV). Bsgs are diverse, but we attempt to classify of the Wolf–Rayet phase is ∼105 yr, and that all Bsgs become AzV 261 between the main types, namely (i) Bsgs, (ii) sgB[e], S Doradus variables, we expect the duration of the S Doradus and (iii) S Doradus LBV variables. In addition, we include the phase at SMC metallicities to be ∼103 yr. This is much lower (iv) cBe classification, although we note that cBe stars are main than the number suggested if Bsgs were truly dormant S Doradus sequence stars. The main classifying properties of AzV 261 are variables (Massey 2006; Smith 2015) where their predicted (i) a supergiant luminosity classification, (ii) no effective temper- evolutionary lifetime is ∼105 yr, but also lower than that deter- ature variations, (iii) aperiodic increase/decrease in brightness mined in the standard picture of massive star evolution ∼104 yr interspersed with periodic micro-variations, (iv) single-peaked (Humphreys & Davidson 1994; Conti 1976). Hα line emission, and (v) infrared excess resembling free-free The low incidence of LBVs in the SMC seems to suggest that emission due to cold circumstellar disc/ejecta. Bsgs and S Doradus LBVs are truly different from one another, In Table3, we list the properties of the di fferent Bsg sub- and that not all Bsgs become S Doradus variables. This may sug- types, together with those for AzV 261. In general they agree gest that the search for Hα P Cygni profiles, as suggested by best with an S Doradus classification apart from the lack of Massey et al.(2007), might not provide the answer to the total

A17, page 10 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

1.2 1.2

1.0 1.0

0.8 0.8 m] µ

) Dusty 5

. 0.6

H 0.6 [4 − − J m] (

µ 0.4 0.4 [3.6

0.2 Neb. cont. 0.2

0.0 0.0

0.2 0.25 0.00 0.25 0.50 0.75 1.00 1.25 1.50 − 0.0 0.5 1.0 1.5 2.0 − (H Ks) [5.8µm] [8.0µm] − − Fig. 13. Left: near-infrared (J−H) vs. (H−Ks) colour–colour diagram of known sgB[e] (open squares), warm hypergiants (open triangles), cBe stars (open circles), known S Doradus variables (blue dots), BSgs (small dots), and AzV 261 (red asterisk). Data taken from Humphreys et al. (2017a). The dashed red line is the main sequence locus at SMC metallicity. Right: mid-infrared [3.6 µm]–[4.5 µm] vs. [5.8 µm]–[8.0 µm] colour– colour diagram. Symbols are as in the left panel. The position of AzV 261 in both diagrams suggests the presence of free–free emission. The LBVs redward of colours [5.8 µm]–[8.0 µm] > 1.0 are contaminated by polycyclic aromatic hydrocarbon emission from nebulosity, with the regions of nebular contamination and dusty stars indicated following Humphreys et al.(2017a).

12 Justham et al.(2014) considered stellar merging to be a likely − evolutionary path to produce LBV supernovae. The small number of confirmed LBVs in the SMC repre- 13 − sents a puzzle; enhanced LBV mass loss seems to be required if canonical stationary wind mass loss from normal stars is ) 1

− reduced at the lower metal content of the SMC (Mokiem et al.

˚ A 14 2

− − 2007). This raises the question of whether the known population

cms of evolved Wolf–Rayet stars in the SMC have been produced 1 − 15 by rotationally induced chemical homogeneous evolution (Brott − et al. 2011; but see Vink & Harries 2017) or by binary interac- (ergs s λ

F tions (Schootemeijer & Langer 2018; but see Shenar et al. 2016). 16

log − Finally, Vink(2018) recently performed radiative transfer computations with a dynamic version of the Monte Carlo method 17 to study the bi-stability jump of LBVs as a function of metallic- − ity. He found the bi-stability jump to be smaller at low metallic- ity. These results may be related to the low number of S Doradus 1.0 0.5 0.0 0.5 1.0 1.5 variables in the SMC that we find here. Metal-poor galaxies are − − log λ (µm) generally lower mass and less luminous than their metal-rich counterparts, resulting in a smaller absolute population of mas- Fig. 14. Spectral energy distribution of AzV 261 covering photome- try from the literature from U-8.0 µm. Overplotted as a solid line is sive stars (although recent evidence for a top-heavy stellar initial the TLUSTY model atmosphere for the best-fit stellar parmeters of mass function at low metallicities compared to a Salpeter mass Teff = 18 000 K, log g = 2.5. The infrared-excess, corresponding to pos- function suggests that the relative fraction of massive stars may sible free-free emission of cold dust is best fit using a black-body model be higher; e.g. Kalari et al. 2018; Schneider et al. 2018). There- having Tb = 2550 K (shown as the dashed line). The dotted line is of fore, short-lived phases such as LBVs might become harder constant frequency flux density normalized to the 3.6 µm photometry to to observe serendipitously, thus the systematic approach like make the free-free contribution clearer. the one presented here carries important weight in defining whether the incidence and characteristics of LBV S Doradus duration of the LBV phase in massive star evolution. However variables do indeed vary by metallicity, how this influences our it might still be possible that related searches at high spectral view of stellar evolution in the early Universe, and how LBVs resolution, allowing the presence of multiple-troughed P Cygni evolve. absorption components to be revealed, may become an efficient tool for detecting extra-galactic LBVs, as suggested by Groh & 5. Conclusions Vink(2011). Additional evidence for an intrinsic difference between Bsgs In this paper, we presented a systematic search for S Doradus and S Doradus variables comes from the complete absence of variables among the Bsgs population in the 1/5 Z SMC. We −1 rapidly rotating Bsgs with ve sin i & 200 km s (see Vink et al. conducted our search using light curves spread across three years 2010), whilst several S Doradus variables have been found to be (with a cadence of one night) of 64 Bsgs from a total known Bsgs rotating rather rapidly (Groh et al. 2006, 2009). For this reason population of 110 in the SMC within our field of view.

A17, page 11 of 46 A&A 618, A17 (2018)

Table 3. Properties of AzV 261 among the Bsg zoo and cBe stars.

Bsgs sgB[e] S Doradus LBV cBe Luminosity I-III I-III I-III IV-V

Light curve Quasi-periodic Irregular Periodic micro-variations Periodic +Irregular increase/decrease

Spectral type changes Stable Stable Changes aperiodic Stable

Hα emission Variable emission Yes+ forbidden line emission Variable emission Double-peaked variable

Infrared colours None Hot dust+disc Cold dust/free-free emission Free-free emission

Notes. The properties of AzV 261 are highlighted in bold face.

η Car

6.5

R127 MWC 930 R143 6.0 60 M AG Car

S Dor 50 M P Cyg G24.73+0.69 HR Car WRA 751 R71 R110 R 40

) 5.5 R85

40 M HD 160529 L ( 30 M HD 168607 ∗

L 5.0 AzV 261 log 20 M W 243 4.5 15 M

12 M 4.0

8 M 3.5 40000 30000 20000 10000 Teff (K) Fig. 15. Position of AzV 261 on the Hertzsprung–Russell diagram marked by the red asterisk, along the Bsgs studied here as grey squares. The positions of confirmed LBVs from the literature in the galaxy (black circles), LMC (blue circles), and the SMC (red circle) are also shown. Cyan circles mark the position of known LBVs in M31 and M33 (from Humphreys et al. 2017b). The dashed green lines are the stellar tracks from Brott et al.(2011) with the corresponding initial stellar mass indicated, and the solid line is the zero-age main sequence isochrone.

Our motivation was to identify whether all Bsgs display dor- and N. Morrell for collecting a spectrum from Las Campanas. The authors thank mant LBV behaviour at differing timescales or amplitudes, or ESO for providing Directors Discretionary time to perform the X-shooter obser- whether S Doradus LBV variables are truly unique objects. From vations. V.M.K. acknowledges support from the FONDECYT-CONICYT grant No. 3160117. M.F. is supported by a Royal Society – Science Foundation Ireland examining the light curves of the 64 SMC Bsgs, only one Bsg, University Research Fellowship. We acknowledge with thanks the variable star AzV 261 demonstrated S Doradus-like photometric variability. observations from the AAVSO International Database contributed by observers AzV 261 did not display variation in the spectral type. There- worldwide and used in this research. fore, AzV 261 is not a LBV S Doradus variable. Our findings imply that the duration of the LBV phase is ∼103 yr, or more References likely that S Doradus LBVs and Bsgs are likely intrinsically dif- Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2016, ApJ, 833, L1 ferent objects. Bomans, D. J., & Weis, K. 2011, Bull. Soc. R. Sci. Liège, 80, 341 Bonanos, A. Z., Lennon, D. J., Köhlinger, F., et al. 2010, AJ, 140, 416 Acknowledgements. The authors thank the referee, R. Humphreys, for insight- Brott, I., de Mink, S. E., Cantiello, M., et al. 2011, A&A, 530, A115 ful comments. We are grateful to C.J. Evans, and P. Massey for kindly providing Carlos Reyes, R. E., Reyes Navarro, F. A., Meléndez, J., Steiner, J., & Elizalde, their spectra of AzV 261. The authors also thank L. Wyrzykowski and the OGLE F. 2015, Rev. Mex. Astron. Astrofis., 51, 135 consortium for providing their data. V.M.K. also thanks G. Blanc, K. Boutsia, Chiosi, C., & Maeder, A. 1986, ARA&A, 24, 329

A17, page 12 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

Conti, P. S. 1976, Mem. Soc. R. Sci. Liège, 9, 193 Lanz, T., & Hubeny, I. 2007, ApJS, 169, 83 de Vaucouleurs, G., & Eggen, O. J. 1952, PASP, 64, 185 Lefever, K., Puls, J., & Aerts, C. 2007, A&A, 463, 1093 de Wit, W. J., Oudmaijer, R. D., & Vink, J. S. 2014, Adv. Astron., 2014, 270848 Lenz, P., & Breger, M. 2005, Commun. Asteroseismol., 146, 53 Dimitrijevic, M. S., & Sahal-Brechot, S. 1984, J. Quant. Spectr. Rad. Transf., 31, Mahabal, A. A., Djorgovski, S. G., Drake, A. J., et al. 2011, Bull. Astron. Soc. 301 India, 39, 387 Drissen, L., Crowther, P. A., Smith, L. J., et al. 2001, ApJ, 546, 484 Markova, N., & Puls, J. 2008, A&A, 478, 823 Dufton, P. L., Smartt, S. J., & Hambly, N. C. 1999, A&AS, 139, 231 Massey, P. 2006, ApJ, 638, L93 Dufton, P. L., Ryans, R. S. I., Trundle, C., et al. 2005, A&A, 434, 1125 Massey, P., Lang, C. C., Degioia-Eastwood, K., & Garmany, C. D. 1995, ApJ, Dufton, P. L., Ryans, R. S. I., Simón-Díaz, S., Trundle, C., & Lennon, D. J. 2006, 438, 188 A&A, 451, 603 Massey, P., McNeill, R. T., Olsen, K. A. G., et al. 2007, AJ, 134, 2474 Dufton, P. L., Langer, N., Dunstall, P. R., et al. 2013, A&A, 550, A109 Massey, P., Neugent, K. F., Hillier, D. J., & Puls, J. 2013, ApJ, 768, 6 Dufton, P. L., Thompson, A., & Crowther, P. A. 2018, A&A, 615, A101 McEvoy, C. M., Dufton, P. L., Evans, C. J., et al. 2015, A&A, 575, A70 Evans, C. J., Howarth, I. D., Irwin, M. J., Burnley, A. W., & Harries, T. J. 2004, Miroshnichenko, A. S. 2006, Stars B[e] Phenom., 355, 13 MNRAS, 353, 601 Mokiem, M. R., de Koter, A., Vink, J. S., et al. 2007, A&A, 473, 603 Findeisen, K., Cody, A. M., & Hillenbrand, L. 2015, ApJ, 798, 89 Nieva, M. F., & Przybilla, N. 2006, ApJ, 639, L39 Fraser, M., Dufton, P. L., Hunter, I., & Ryans, R. S. I. 2010, MNRAS, Petrov, B., Vink, J. S., & Gräfener, G. 2014, A&A, 565, A62 404, 1306 Puls, J., Vink, J. S., & Najarro, F. 2008, A&ARv, 16, 209 Frescura, F. A. M., Engelbrecht, C. A., & Frank, B. S. 2007, ApJ, submitted, Pustilnik, S. A., Tepliakova, A. L., Kniazev, A. Y., & Burenkov, A. N. 2008, DOI: 10.1016/j.icarus.2012.06.015 MNRAS, 388, L24 Gal-Yam, A., & Leonard, D. C. 2009, Nature, 458, 865 Rolleston, W. R. J., Venn, K., Tolstoy, E., & Dufton, P. L. 2003, A&A, 400, Garnett, D. R. 1999, New Views Magellanic Clouds, 190, 266 21 Gräfener, G., Owocki, S. P., & Vink, J. S. 2012, A&A, 538, A40 Ryans, R. S. I., Dufton, P. L., Mooney, C. J., et al. 2003, A&A, 401, 1119 Gray, D. F. 2005, in The Observation and Analysis of Stellar Photospheres, 3rd Schneider, F. R. N., Sana, H., Evans, C. J., et al. 2018, Science, 359, 69 edn. (Cambridge, UK: Cambridge University Press) Schootemeijer, A., & Langer, N. 2018, A&A, 611, A75 Groh, J. H., & Vink, J. S. 2011, A&A, 531, L10 Shenar, T., Hainich, R., Todt, H., et al. 2016, A&A, 591, A22 Groh, J. H., Hillier, D. J., & Damineli, A. 2006, ApJ, 638, L33 Sigut, T. A. A. 1996, ApJ, 473, 452 Groh, J. H., Damineli, A., Hillier, D. J., et al. 2009, ApJ, 705, L25 Simón-Díaz, S., & Herrero, A. 2007, A&A, 468, 1063 Groh, J. H., Meynet, G., & Ekström, S. 2013, A&A, 550, L7 Simón-Díaz, S., & Herrero, A. 2014, A&A, 562, A135 Gvaramadze, V. V., Kniazev, A. Y., Miroshnichenko, A. S., et al. 2012, MNRAS, Simón-Díaz, S., Herrero, A., Uytterhoeven, K., et al. 2010, ApJ, 720, L174 421, 3325 Simón-Díaz, S., Godart, M., Castro, N., et al. 2017, A&A, 597, A22 Herrero, A., Garcia, M., Uytterhoeven, K., et al. 2010, A&A, 513, A70 Smith, N. 2014, ARA&A, 52, 487 Hubeny, I. 1988, Comput. Phys. Commun., 52, 103 Smith, N. 2015, Very Massive Stars Local Univ., 412, 227 Hubeny, I., & Lanz, T. 1995, ApJ, 439, 875 Smith, N., & Owocki, S. P. 2006, ApJ, 645, L45 Hubeny, I., Heap, S. R., & Lanz, T. 1998, Prop. Hot Luminous Stars, 131, Smith, N., & Tombleson, R. 2015, MNRAS, 447, 598 108 Smith, N., Aghakhanloo, M., & Murphy, J. W. 2018, ArXiv e-prints Humphreys, R. M., & Davidson, K. 1994, PASP, 106, 1025 [arXiv:1805.03298] Humphreys, R. M., Weis, K., Davidson, K., & Gordon, M. S. 2016, ApJ, 825, Szymanski, M. K. 2005, Acta Astron., 55, 43 64 Toribio San Cipriano, L., Domínguez-Guzmán, G., Esteban, C., et al. 2017, Humphreys, R. M., Gordon, M. S., Martin, J. C., Weis, K., & Hahn, D. 2017a, MNRAS, 467, 3759 ApJ, 836, 64 Trundle, C., Dufton, P. L., Hunter, I., et al. 2007, A&A, 471, 625 Humphreys, R. M., Davidson, K., Hahn, D., Martin, J. C., & Weis, K. 2017b, Trundle, C., Kotak, R., Vink, J. S., & Meikle, W. P. S. 2008, A&A, 483, L47 ApJ, 844, 40 Udalski, A., Kubiak, M., & Szymanski, M. 1997, Acta Astron., 47, 319 Hunter, I., Dufton, P. L., Smartt, S. J., et al. 2007, A&A, 466, 277 VanderPlas, J. T. 2018, ApJS, 236, 16 Izotov, Y. I., & Thuan, T. X. 2008, ApJ, 687, 133 van Genderen, A. M. 2001, A&A, 366, 508 Justham, S., Podsiadlowski, P., & Vink, J. S. 2014, ApJ, 796, 121 Venn, K. A. 1999, ApJ, 518, 405 Kalari, V. M., Carraro, G., Evans, C. J., & Rubio, M. 2018, ApJ, 857, 132 Vernet, J., Dekker, H., D’Odorico, S., et al. 2011, A&A, 536, A105 Kotak, R., & Vink, J. S. 2006, A&A, 460, L5 Vink, J. S. 2012, Eta Carinae Supernova Impostors, 384, 221 Kourniotis, M., Bonanos, A. Z., Soszynski,´ I., et al. 2014, A&A, 562, A125 Vink, J. S. 2015, Very Massive Stars Local Univ., 412 Kurt, C. M., & Dufour, R. J. 1998, Rev. Mex. Astron. Astrofis. Conf. Ser., 7, Vink, J. S. 2018, A&A, submitted [arXiv:1808.06612] 202 Vink, J. S., & Harries, T. J. 2017, A&A, 603, A120 Labadie-Bartz, J., Pepper, J., McSwain, M. V., et al. 2017, AJ, 153, 252 Vink, J. S., & de Koter, A. 2005, A&A, 442, 587 Lamers, H. J. G. L. M., Bastiaanse, M. V., Aerts, C., & Spoon, H. W. W. 1998, Vink, J. S., de Koter, A., & Lamers, H. J. G. L. M. 2001, A&A, 369, 574 A&A, 335, 605 Vink, J. S., Brott, I., Gräfener, G., et al. 2010, A&A, 512, L7 Langer, N. 2012, ARA&A, 50, 107 Woosley, S. E., & Heger, A. 2015, Very Massive Stars Local Univ., 412, 199

A17, page 13 of 46 A&A 618, A17 (2018)

Appendix A: Light curves of Bsgs that have data from the OGLE II survey

2dFS0240 14.06 2dFS1217 14.250

14.08 14.275

14.10 14.300

14.12 14.325 (mag) (mag) I I 14.14 14.350

14.16 14.375

14.18 14.400

14.20 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

14.48 2dFS2215 2dFS5057

12.75

14.50

12.80 14.52

(mag) (mag) 12.85 I 14.54 I

14.56 12.90

14.58 12.95

2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450400 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

13.52 12.96 2dFS5095 AzV103

12.98 13.54

13.00

13.02 13.56 (mag) (mag)

I 13.04 I 13.58 13.06

13.08 13.60

13.10

2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450400 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

Fig. A.1. Light curves of 63 Bsgs that have data from the OGLE II survey. The Julian date is plotted against the measured I-band magnitude.

A17, page 14 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

12.200 AzV104 AzV151

12.225 13.35

12.250

13.40 12.275

12.300 (mag) 13.45 (mag) I I 12.325

13.50 12.350

12.375 13.55 12.400 2450400 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450400 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

14.00 AzV155 12.70 AzV173

12.72 14.02

12.74 14.04 12.76 (mag) (mag) I I 14.06 12.78

12.80 14.08

12.82 14.10 12.84 2450400 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

AzV175 AzV18

13.64 12.25

13.66 12.30

13.68 (mag) (mag)

I I 12.35 13.70

12.40 13.72

13.74 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

Fig. A.1. continued.

A17, page 15 of 46 A&A 618, A17 (2018)

AzV200 AzV210 12.64 11.90

12.66 11.95

12.68

12.00 (mag) (mag) 12.70 I I

12.05 12.72

12.10 12.74

12.76 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

AzV214 12.725 AzV215 13.28

12.750 13.30 12.775 13.32 12.800

13.34 12.825 (mag) (mag) I I

13.36 12.850

13.38 12.875

13.40 12.900

12.925 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

13.40 AzV218 AzV221

13.68

13.45

13.70

13.50

(mag) 13.72 (mag) I I

13.55

13.74

13.60

13.76

2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

Fig. A.1. continued.

A17, page 16 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

AzV225 AzV22 13.94 12.05 13.96

13.98 12.10

14.00 (mag) (mag)

I I 12.15 14.02

14.04 12.20

14.06

12.25 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

AzV235 AzV237 12.14 12.32

12.16 12.34

12.18 12.36 12.20 (mag) (mag)

I I 12.38 12.22

12.24 12.40

12.26 12.42

12.28 12.44 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

AzV241 AzV242 12.075 14.10 12.100

12.125 14.15

12.150 (mag) (mag)

I 14.20 I 12.175

12.200

14.25 12.225

12.250

14.30 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

Fig. A.1. continued.

A17, page 17 of 46 A&A 618, A17 (2018)

13.025 13.32 AzV252 AzV253

13.34 13.050

13.36 13.075 13.38 13.100 13.40 (mag) (mag)

I 13.125 I 13.42

13.150 13.44

13.175 13.46

13.48 13.200 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

12.68 AzV257 AzV260 13.32

12.70

12.72 13.34

12.74

13.36 12.76 (mag) (mag) I I

12.78 13.38 12.80

12.82 13.40

12.84 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

13.7 AzV261 AzV263 12.75

13.8 12.80

13.9 12.85 (mag) (mag)

I I 12.90 14.0

12.95

14.1 13.00

2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

Fig. A.1. continued.

A17, page 18 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

AzV266 AzV268 13.26

12.65 13.28

13.30 12.70 13.32 (mag) (mag)

I I 13.34 12.75

13.36

12.80 13.38

13.40

12.85 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

AzV271 AzV292 13.48 13.02

13.50 13.04

13.52 13.06

13.54 13.08 (mag) (mag) I 13.56 I 13.10

13.58 13.12

13.60 13.14

13.62 13.16

2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

AzV314 11.94 AzV31

13.02 11.96

13.04 11.98

12.00 13.06 12.02 (mag) (mag) I I 13.08 12.04

13.10 12.06

12.08 13.12 12.10 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

Fig. A.1. continued.

A17, page 19 of 46 A&A 618, A17 (2018)

AzV324 AzV337 12.60

12.90 12.65

12.95 12.70

(mag) 13.00 (mag) 12.75 I I

12.80 13.05

12.85

13.10 12.90 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

12.66 AzV339 12.625 AzV340

12.68 12.650

12.70 12.675

12.72 12.700

(mag) 12.74 (mag) 12.725 I I

12.750 12.76

12.775 12.78

12.800 12.80 12.825 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

AzV342 AzV343 12.975 13.06

13.000 13.08

13.025 13.10 13.050 (mag) (mag)

I 13.12 I 13.075

13.14 13.100

13.16 13.125

13.150 13.18 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

Fig. A.1. continued.

A17, page 20 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

AzV351 AzV371 13.550 13.375

13.575 13.400

13.600 13.425

13.450 13.625 (mag) (mag) I I 13.475 13.650

13.500 13.675

13.525 13.700

13.550 13.725 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

12.70 12.10 AzV372 AzV373

12.72 12.15

12.74

12.76 12.20 (mag) (mag) I I 12.78 12.25

12.80

12.30 12.82

2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

AzV374 AzV376 13.20 13.350

13.375 13.22

13.400

13.24 13.425 (mag) (mag)

I 13.26 I 13.450

13.475 13.28 13.500

13.30 13.525

2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

Fig. A.1. continued.

A17, page 21 of 46 A&A 618, A17 (2018)

AzV379 AzV394

13.18 13.54

13.20 13.56

13.22 13.58 (mag) (mag) I I 13.24 13.60

13.26 13.62

13.28 13.64 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

AzV404 AzV412 12.22 12.90

12.24

12.92 12.26

12.28 12.94 (mag) (mag)

I 12.30 I 12.96

12.32

12.98 12.34

12.36 13.00

2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

AzV420 AzV438 13.25 12.78

13.30 12.80

12.82 13.35 (mag) (mag) I I 12.84 13.40

12.86

13.45

12.88 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

Fig. A.1. continued.

A17, page 22 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

12.55 AzV445 AzV91 12.425

12.60 12.450

12.65 12.475

12.500

(mag) 12.70 (mag) I I 12.525

12.75 12.550

12.80 12.575

12.600 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450400 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

14.28 AzV93 NGC330-001 12.81 14.29

14.30 12.82

14.31 12.83

14.32 12.84 (mag) (mag) I I 14.33 12.85

14.34 12.86

14.35 12.87

14.36 12.88 2450400 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

NGC330-002 NGC330-004 12.94 13.26

12.95 13.28

12.96 13.30

12.97 13.32

(mag) 12.98 (mag) I I 13.34 12.99

13.36 13.00

13.01 13.38

13.02 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

Fig. A.1. continued.

A17, page 23 of 46 A&A 618, A17 (2018)

13.28 NGC330-005 NGC330-010 13.92

13.30 13.94

13.32 13.96

13.34 (mag) (mag)

I I 13.98

13.36 14.00

13.38 14.02

13.40 2450600 2450800 2451000 2451200 2451400 2451600 2451800 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date Julian Date

14.225 NGC330-019

14.250

14.275

14.300

14.325 (mag) I 14.350

14.375

14.400

14.425 2450600 2450800 2451000 2451200 2451400 2451600 2451800 Julian Date

Fig. A.1. continued.

A17, page 24 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

Appendix B: Lomb–Scargle periodogram of Bsgs that have OGLE II data

0.07 0.08

0.06

0.06 0.05

0.04 0.04 0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power 0.02 0.02

0.01

0.00 2dFS0240 0.00 2dFS1217 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.08 0.08

0.06 0.06

0.04 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.00 2dFS2215 0.00 2dFS5057 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.08 0.10 0.07

0.06 0.08

0.05 0.06 0.04

0.03 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02 0.01

0.00 2dFS5095 0.00 AzV103 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

Fig. B.1. Lomb–Scargle periodogram of 64 Bsgs. The period in days is plotted against the power. The red and blue lines respectively represent the 99% and 95% significance calculated from the Monte Carlo bootstrap simulations.

A17, page 25 of 46 A&A 618, A17 (2018)

0.08 0.12 0.07

0.10 0.06

0.05 0.08

0.04 0.06

0.03

Lo,mb-Scargle Power Lo,mb-Scargle Power 0.04 0.02

0.02 0.01

0.00 AzV104 0.00 AzV151 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.10 0.08

0.08

0.06

0.06

0.04 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.00 AzV155 0.00 AzV173 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.08 0.08

0.06 0.06

0.04 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.00 AzV175 0.00 AzV18 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

Fig. B.1. continued.

A17, page 26 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

0.08 0.10

0.08 0.06

0.06 0.04

0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.00 AzV200 0.00 AzV210 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.08 0.08

0.06 0.06

0.04 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.00 AzV214 0.00 AzV215 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.08 0.08 0.07

0.06 0.06 0.05

0.04 0.04

0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.01

0.00 AzV218 0.00 AzV221 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

Fig. B.1. continued.

A17, page 27 of 46 A&A 618, A17 (2018)

0.08

0.07 0.08

0.06

0.06 0.05

0.04 0.04 0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power 0.02 0.02

0.01

0.00 AzV225 0.00 AzV22 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.200

0.12 0.175

0.150 0.10

0.125 0.08

0.100 0.06 0.075

Lo,mb-Scargle Power Lo,mb-Scargle Power 0.04 0.050

0.02 0.025

0.000 AzV235 0.00 AzV237 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.175 0.08 0.150

0.125 0.06

0.100

0.04 0.075 Lo,mb-Scargle Power Lo,mb-Scargle Power 0.050 0.02

0.025

0.000 AzV241 0.00 AzV242 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

Fig. B.1. continued.

A17, page 28 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

0.08 0.07

0.06

0.06 0.05

0.04 0.04 0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power 0.02 0.02

0.01

0.00 AzV252 0.00 AzV253 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.14 0.08

0.07 0.12

0.06 0.10

0.05 0.08 0.04 0.06 0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power 0.04 0.02

0.02 0.01

0.00 AzV257 0.00 AzV260 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.07 0.4

0.06

0.3 0.05

0.04

0.2 0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power 0.02 0.1

0.01

0.0 AzV261 0.00 AzV263 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

Fig. B.1. continued.

A17, page 29 of 46 A&A 618, A17 (2018)

0.08 0.30

0.25 0.06

0.20

0.04 0.15

Lo,mb-Scargle Power Lo,mb-Scargle Power 0.10 0.02

0.05

0.00 AzV266 0.00 AzV268 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.08 0.08 0.07

0.06 0.06

0.05

0.04 0.04

0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.01

0.00 AzV271 0.00 AzV292 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.08 0.08

0.07

0.06 0.06

0.05

0.04 0.04

0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.01

0.00 AzV314 0.00 AzV31 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

Fig. B.1. continued.

A17, page 30 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

0.07 0.08 0.06

0.05 0.06

0.04

0.04 0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power 0.02 0.02 0.01

0.00 AzV324 0.00 AzV337 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.10 0.08

0.08

0.06

0.06

0.04 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.00 AzV339 0.00 AzV340 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.10

0.08

0.08

0.06 0.06

0.04 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.00 AzV342 0.00 AzV343 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

Fig. B.1. continued.

A17, page 31 of 46 A&A 618, A17 (2018)

0.10 0.07

0.06 0.08

0.05

0.06 0.04

0.04 0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power 0.02 0.02 0.01

0.00 AzV351 0.00 AzV371 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.08 0.10

0.08 0.06

0.06

0.04 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.00 AzV372 0.00 AzV373 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.08 0.07 0.07 0.06 0.06 0.05 0.05 0.04 0.04

0.03 0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power 0.02 0.02

0.01 0.01

0.00 AzV374 0.00 AzV376 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

Fig. B.1. continued.

A17, page 32 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

0.10 0.07

0.06 0.08

0.05

0.06 0.04

0.03 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power 0.02 0.02 0.01

0.00 AzV379 0.00 AzV394 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.10 0.08

0.08

0.06

0.06

0.04 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.00 AzV404 0.00 AzV412 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.08 0.10

0.07

0.08 0.06

0.05 0.06

0.04

0.04 0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power 0.02 0.02 0.01

0.00 AzV420 0.00 AzV438 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

Fig. B.1. continued.

A17, page 33 of 46 A&A 618, A17 (2018)

0.08 0.08 0.07

0.06 0.06

0.05

0.04 0.04

0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.01

0.00 AzV445 0.00 AzV91 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.25 0.10

0.20 0.08

0.15 0.06

0.10 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.05 0.02

0.00 AzV93 0.00 NGC330-001 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.10 0.08

0.07 0.08

0.06

0.05 0.06

0.04 0.04 0.03 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.01

0.00 NGC330-002 0.00 NGC330-004 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

Fig. B.1. continued.

A17, page 34 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

0.10 0.08

0.08

0.06

0.06

0.04 0.04 Lo,mb-Scargle Power Lo,mb-Scargle Power

0.02 0.02

0.00 NGC330-005 0.00 NGC330-010 0 200 400 600 800 1000 0 200 400 600 800 1000 Period (days) Period (days)

0.08

0.07

0.06

0.05

0.04

0.03 Lo,mb-Scargle Power 0.02

0.01

0.00 NGC330-019 0 200 400 600 800 1000 Period (days)

Fig. B.1. continued.

A17, page 35 of 46 A&A 618, A17 (2018)

Appendix C: δm–δt two-dimensional histogram of Bsgs with OGLE II data

0.12 1400 2dFS0240 1750 2dFS1217 0.14

1200 1500 0.10 0.12

1000 1250 0.10 0.08

800 0.08 1000 0.06 m (mag) m (mag) δ δ 600 0.06 750 0.04 0.04 500 400

0.02 0.02 250 200

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

2dFS2215 2dFS5057 0.10 2500 1000 0.200

0.175 0.08 2000 800 0.150

0.06 0.125 1500 600

m (mag) m (mag) 0.100 δ δ

0.04 400 1000 0.075

0.050 0.02 200 500 0.025

0.00 0 0.000 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 1400 δt (days) δt (days)

2dFS5095 1000 0.10 AzV103 0.12 500

0.08 0.10 800 400

0.08 600 0.06 300

m (mag) 0.06 m (mag) δ δ 400 0.04 200 0.04

200 0.02 100 0.02

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

Fig. C.1. Two-dimensional histogram of the δm–δt plot of 64 Bsgs. The δm in mag is plotted against the δt in days, and the colour is indicative of the density of the histogram following the colourbar.

A17, page 36 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

AzV104 AzV151 700 0.200 0.16 1750

0.175 0.14 600 1500

0.150 0.12 500 1250 0.125 0.10 400 1000

m (mag) 0.100 m (mag) 0.08 δ δ 300 750 0.075 0.06

500 200 0.050 0.04

100 0.025 250 0.02

0.000 0 0.00 0 0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 δt (days) δt (days)

0.12 0.10 AzV155 AzV173 800 1750

0.10 700 1500 0.08 600 1250 0.08 500 0.06 1000 0.06 400 m (mag) m (mag) δ δ 750 0.04 300 0.04 500 200 0.02 0.02 250 100

0.00 0 0.00 0 0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 δt (days) δt (days)

0.10 AzV175 AzV18 0.175 1000 800

0.150 0.08 800 600 0.125

0.06 0.100 600 m (mag) m (mag)

δ 400 δ 0.075 0.04 400

0.050 200 0.02 200 0.025

0.00 0 0.000 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

Fig. C.1. continued.

A17, page 37 of 46 A&A 618, A17 (2018)

AzV200 700 0.10 AzV210 800

700 0.20 600 0.08 600 500

0.15 500 0.06 400 400 m (mag) m (mag) δ 0.10 300 δ 0.04 300

200 200 0.05 0.02 100 100

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

0.12 AzV214 0.175 AzV215

0.150 0.10 800 800

0.125 0.08 600 600 0.100 0.06 m (mag) m (mag) δ 400 δ 0.075 400

0.04 0.050 200 200 0.02 0.025

0.00 0 0.000 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

0.10 AzV218 0.200 AzV221 2000 700 0.175 1750 0.08 600 0.150 1500 500 0.125 0.06 1250 400 0.100 1000 m (mag) m (mag) δ δ 0.04 300 0.075 750

200 0.050 500 0.02 100 0.025 250

0.00 0 0.000 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

Fig. C.1. continued.

A17, page 38 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

AzV225 0.200 AzV22 0.12 1400 800 0.175 0.10 1200 0.150 1000 600 0.08 0.125 800 0.06 0.100 m (mag) m (mag) 400 δ δ 600 0.075 0.04 400 0.050 200 0.02 200 0.025

0.00 0 0.000 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

AzV235 AzV237 800 0.12 0.10 500 700 0.10 0.08 600 400 0.08 500 0.06 300 400

m (mag) 0.06 m (mag) δ δ

200 0.04 300 0.04 200 0.02 0.02 100 100

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

900 AzV241 AzV242 0.16 0.175 800 600 0.14 0.150 700 500 0.12 0.125 600 0.10 400 500 0.100 0.08 m (mag) 400 m (mag) 300 δ δ 0.075 300 0.06 200 0.050 200 0.04 100 0.025 100 0.02

0.000 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

Fig. C.1. continued.

A17, page 39 of 46 A&A 618, A17 (2018)

1000 AzV252 AzV253 0.14 1600 0.12

800 1400 0.12 0.10 1200 0.10 600 0.08 1000 0.08

m (mag) m (mag) 800

δ δ 0.06 0.06 400 600 0.04 0.04 400 200 0.02 0.02 200

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

AzV257 0.10 AzV260 0.12 1000 1000

0.10 0.08 800 800

0.08 0.06 600 600

m (mag) 0.06 m (mag) δ δ

400 0.04 400 0.04

200 0.02 200 0.02

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

3000 AzV261 AzV263 2000 0.25 0.40

1750 2500 0.35 0.20 1500 0.30 2000 1250 0.25 0.15 1500 1000

m (mag) 0.20 m (mag) δ δ 0.10 750 0.15 1000

0.10 500 0.05 500 0.05 250

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

Fig. C.1. continued.

A17, page 40 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

1400 0.200 AzV266 AzV268 0.12 1000 0.175 1200 0.10 0.150 1000 800

0.125 0.08 800 600 0.100 0.06 m (mag) m (mag)

δ 600 δ

0.075 400 0.04 400 0.050 200 200 0.02 0.025

0.000 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

AzV271 AzV292 1600 0.12 1400 0.12

1400 0.10 1200 0.10 1200 1000 0.08 0.08 1000 800

m (mag) 0.06 m (mag) 0.06 800 δ δ 600 600 0.04 0.04 400 400

0.02 0.02 200 200

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

AzV314 AzV31 0.10 1000 0.12 1000

0.08 0.10 800 800

0.08 0.06 600 600 m (mag) m (mag)

δ δ 0.06 0.04 400 400 0.04

0.02 200 200 0.02

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

Fig. C.1. continued.

A17, page 41 of 46 A&A 618, A17 (2018)

AzV324 AzV337 2000 1600 0.20 0.25 1750 1400

1500 1200 0.20 0.15

1000 1250 0.15 1000

m (mag) 800 m (mag)

δ 0.10 δ

600 0.10 750

500 0.05 400 0.05 200 250

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

AzV339 AzV340 0.12 500 0.16

800 0.14 0.10 400 0.12 0.08 600 300 0.10

0.06

m (mag) m (mag) 0.08 δ δ 400 200 0.06 0.04

0.04 200 100 0.02 0.02

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

0.10 AzV342 1000 0.16 AzV343 1200

0.14 1000 0.08 800 0.12

800 0.10 0.06 600

0.08 600 m (mag) m (mag) δ δ 0.04 400 0.06 400

0.04 0.02 200 200 0.02

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

Fig. C.1. continued.

A17, page 42 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

0.16 AzV351 0.16 AzV371 1600 1400 0.14 0.14 1400 1200 0.12 0.12 1200 1000 0.10 0.10 1000 800 0.08 0.08

m (mag) 800 m (mag) δ δ 600 0.06 600 0.06

400 0.04 400 0.04

0.02 200 0.02 200

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

AzV372 AzV373 500 700 0.10 0.175

600 0.150 400 0.08 500 0.125 300 0.06 400 0.100 m (mag) m (mag) δ δ 300 0.075 200 0.04

200 0.050

0.02 100 100 0.025

0.00 0 0.000 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

0.10 AzV374 0.175 AzV376 1400 1400 0.150 1200 0.08 1200

0.125 1000 1000 0.06 0.100 800 800 m (mag) m (mag) δ δ 0.075 600 0.04 600

0.050 400 400 0.02 200 0.025 200

0.00 0 0.000 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

Fig. C.1. continued.

A17, page 43 of 46 A&A 618, A17 (2018)

0.10 AzV379 0.10 AzV394 1000 1000

0.08 0.08 800 800

0.06 0.06 600 600 m (mag) m (mag) δ δ 0.04 0.04 400 400

0.02 200 0.02 200

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

0.14 AzV404 0.10 AzV412 1200 500 0.12 1000 0.08 0.10 400 800 0.08 0.06 300 600 m (mag) m (mag)

δ 0.06 δ 0.04 200 400 0.04

100 0.02 0.02 200

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

1400 AzV420 AzV438 0.200 0.10 1600 1200 0.175 1400 0.08 0.150 1000 1200

0.125 1000 0.06 800

0.100

m (mag) 800 m (mag) δ δ 600 0.04 0.075 600 400 0.050 400 0.02 200 0.025 200

0.000 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

Fig. C.1. continued.

A17, page 44 of 46 V. M. Kalari et al.: How common is LBV S Doradus variability at low metallicity?

0.25 AzV445 1600 0.16 AzV91 1400

1400 0.14 0.20 1200 1200 0.12 1000

0.15 1000 0.10 800 800 0.08 m (mag) m (mag) δ δ 0.10 600 600 0.06

400 400 0.04 0.05

200 0.02 200

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 1400 δt (days) δt (days)

0.10 AzV93 0.10 NGC330-001 1600 800

0.08 1400 0.08

1200 600

0.06 1000 0.06

m (mag) 800 m (mag) δ δ 400 0.04 0.04 600

400 200 0.02 0.02

200

0.00 0 0.00 0 0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 δt (days) δt (days)

0.10 NGC330-002 NGC330-004 700 800 0.10

600 700 0.08 0.08 600 500

0.06 500 400 0.06

m (mag) m (mag) 400 δ δ 0.04 300 0.04 300

200 200 0.02 0.02 100 100

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

Fig. C.1. continued.

A17, page 45 of 46 A&A 618, A17 (2018)

0.10 0.10 NGC330-005 NGC330-010 1200

800 0.08 0.08 1000

600 800 0.06 0.06

600 m (mag) m (mag) δ δ 0.04 400 0.04

400

0.02 200 0.02 200

0.00 0 0.00 0 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 δt (days) δt (days)

NGC330-019 0.175 2000 0.150

0.125 1500

0.100 m (mag)

δ 1000 0.075

0.050 500

0.025

0.000 0 0 200 400 600 800 1000 1200 δt (days)

Fig. C.1. continued.

A17, page 46 of 46