The ACS LCID Project – VIII. The short-period Cepheids of Leo A★ Free

Abstract

INTRODUCTION

Classical Cepheids (CC) are the prime distance estimators in the nearby universe thanks to the Leavitt (1908) law, which relates their brightness and period. However, their use is complicated by the existence of other types of Cepheids with different luminosities at a given period: Population II Cepheids (P2C) and Anomalous Cepheids (AC). While the former have periods similar to those of CC (from less than a day to more than 100 d), they are typically fainter at a given period (∼1 to 2 mag) such that recognizing CC and P2C is usually straightforward. On the other hand, AC are only slightly fainter than short-period CC (see e.g. Bono et al. 1997 for a discussion of their formation and evolution). This confusion is even more significant for metal-poor galaxies, in which most of the CC have periods similar to AC (∼0.4-3 d; e.g. IC 1613, Sextans A, Phoenix: Dolphin et al. 2001, 2003; Gallart et al. 2004; Bernard et al. 2010).

An increasing number of studies also suggest that there may be a metallicity effect on the slope and/or zero-point of the period–luminosity (PL) relations (e.g. Sakai et al. 2004; Romaniello et al. 2008; Sandage, Tammann & Reindl 2009; Bono et al. 2010; Storm et al. 2011), which would affect the distance measurements of galaxies for which metallicity is not well constrained. It is therefore essential to obtain complete surveys of Cepheids in a sample of metal-poor galaxies in order to gain a better understanding of their physical and evolutionary properties.

Leo A, a small star-forming dwarf galaxy discovered by Zwicky (1942) and located about 790 kpc away (e.g. Dolphin et al. 2002, hereinafter D02), is among the most metal-poor galaxies of the Local Group (LG). Its current metallicity is A_O = 12+log(O/H) = 7.30 ± 0.05 (i.e. [Fe/H] ∼ −1.5, or ∼3 per cent solar assuming A_{O, ⊙} = 8.87 ± 0.07; Grevesse & Noels 1993) from the spectroscopy of three H ii regions and a planetary nebula (Skillman, Kennicutt & Hodge 1989; van Zee, Skillman & Haynes 2006). Tolstoy et al. (1998) drew attention to the relative lack of an ancient stellar population in Leo A, and this was supported by the discovery of the rarity of RR Lyrae stars by D02.

Leo A has been observed as part of the LCID¹ project, which aims at constraining the star formation histories (SFH) of a sample of isolated LG galaxies based on deep colour–magnitude diagrams (CMD) and time series photometry obtained with the Advanced Camera for Surveys (ACS) onboard the Hubble Space Telescope (HST). The global SFH of Leo A has been presented in Cole et al. (2007), while the analysis of population gradients and spatially resolved SFH will appear in a forthcoming paper (Hidalgo et al., in preparation). Here, we present the search for short-period variable stars, in particular Cepheids, for which these data are particularly well suited.

The paper is organized as follows: the data acquisition, reduction and calibration are summarized in Section 2. We describe the identification of variable stars and their spatial distribution in Section 3. Sections 4 to 6 present the properties of the variable stars and candidate variables detected. The distance estimates based on various tracers and methods are presented in Section 7, and we discuss the implications of our results in Section 8. In Section 9 we summarize our results and present our conclusions.

OBSERVATIONS AND DATA REDUCTION

The observations of Leo A were carried out with the ACS onboard the HST between 2005 December 26 and 2006 January 8. The observed field was slightly offset from the centre of the galaxy, as shown in Fig. 1, to cover a longer radial baseline and study possible stellar population gradients. 16 HST orbits were devoted to this galaxy in two bands (F475W and F814W). Each orbit was split into two ∼1200 s exposures of a given band. This resulted in 16 images per band, for a total exposure time of 19 200 and 19 520 s in F475W and F814W, respectively. The complete observing log is presented in Table 1.

Figure 1.

Sloan Digital Sky Survey image of Leo A. The cross and ellipse indicate the centre and extent of the galaxy (RA = 09^h59^m24|${.\!\!\!\!\!\!^{s}}$|0, Dec. = 30°44^′47^′′, r = 8 arcmin; Vansevičius et al. 2004), while the small and large squares show the location of our field of view (FOV) and that of Dolphin et al. (2002), respectively.

Open in new tabDownload slide

We note that the temporal sampling for this galaxy is not as optimal as was the case in the previous papers of this series. In our study of variable stars in Cetus and Tucana (Bernard et al. 2009, hereinafter Paper I) and IC 1613 (Bernard et al. 2010, hereinafter Paper II), we had 25, 32 and 24 data points per band, respectively, compared to only 16 here. In addition, the observing strategy for these three galaxies provided alternating observations in F475W and F814W, and therefore observations in a given filter that were well separated in time. For Leo A, the paired observations in a given band are only 22 min apart, thus, limiting their utility. While it did not affect our ability to detect candidate variables, some light curves have gaps which could limit the accuracy of the measured mean magnitudes (see Section 5).

The daophot/allframe suite of programs (Stetson 1994) was used to obtain the instrumental photometry of the stars on the individual, non-drizzled images provided by the HST pipeline. Additionally, we used the pixel area maps and data quality masks to correct for the pixel area variation and to flag bad pixels. Standard calibration was carried out as described in Sirianni et al. (2005). We refer the reader to Monelli et al. (2010) for a detailed description of the data reduction and calibration. The final CMD is presented in Fig. 2, where the (F475W + F814W)/2 ∼ V filter combination was chosen for the ordinate axis so that the horizontal branch (HB) appears approximately horizontal.

Figure 2.

CMD of Leo A from our ACS data. The variables are overplotted as labelled in the inset.

Open in new tabDownload slide

The light curves of the variable stars were then converted to Johnson–Cousins BVI magnitudes using the transformations given in Paper I. As we have shown in Paper II from the comparison of the light curves of Cepheids in common between our HST survey and OGLE,² these transformations can be safely used to analyse the properties (e.g. magnitude and amplitude) of the variable stars.

VARIABLE STARS

Identification and period search

The candidate variables were extracted from the photometric catalogues using the variability index of Welch & Stetson (1993); this process yielded ∼700 candidates. A preliminary check of the light curve and position on the CMD, together with a careful inspection of the stacked image, allowed us to discard false detections due to cosmic ray hits, chip defects or stars located under the wings of bright stars. Our final catalogue contains 166 bona fide variables and 33 candidates for which we could not determine a period [e.g. main-sequence variable (MSV) and red giant branch (RGB) variable, eclipsing binaries]. The variables and candidates are shown overplotted on the CMD in Fig. 2 using their intensity-averaged mean magnitudes. The individual F475W and F814W measurements for all the variables and candidates, as well as the transformed BVI magnitudes, are listed in Table 2.

The period search was performed through Fourier analysis (Horne & Baliunas 1986) taking into account the information from both bands simultaneously, then refined interactively by modifying the period until a tighter light curve was obtained. For each variable, data points with error bars larger than 3σ above the mean error bar size were rejected through sigma clipping with five iterations. The periods are given with three significant figures only to reflect the sparse sampling of the light curves. Note that for the same reason, some of the periods may be aliases of the true periods.

The classification of the candidates was based on their light-curve morphology and position in the CMD. We found a large number of Cepheids but very few RR Lyrae stars (156 and 10, respectively), in very good agreement with the study of D02. The small number of RR Lyrae stars – which is not due to incompleteness, our photometry being near 100 per cent complete at the magnitude of the HB – is not surprising considering the SFH of Leo A (Cole et al. 2007; Hidalgo et al., in preparation), which shows a very low star formation rate for the earliest epochs. To obtain the amplitudes and intensity-averaged magnitudes of the variables, we fitted the light curves with a set of templates based on the set of Layden et al. (1999, see Paper I).

Table 3 summarizes the properties of the bona fide variable stars in our final catalogue. The first two columns give the identification number and type of variability, while the next two list the equatorial coordinates (J2000.0). Columns (5) and (6) give the period and logarithm of the period, in days. The intensity-averaged mean magnitudes 〈F475W〉 and 〈F814W〉, and colour 〈F475W〉 − 〈F814W〉 are given in columns (7), (9) and (11). The amplitudes in the F475W and F814W bands measured from the template fits are listed in columns (8) and (10). The last six columns alternately list the intensity-averaged mean magnitudes and amplitudes in the Johnson B, V and I bands. For the candidate variables, we only list the coordinates and approximate magnitudes in Table 4.

Spatial distribution

Fig. 3 presents the spatial distribution of stars in our FOV, where the RR Lyrae stars (open circles) and Cepheids (open and filled stars) are shown. The cross and ellipse represent the centre and Holmberg radius of Leo A (3.5 arcmin, from Vansevičius et al. 2004). Despite their small number, the RR Lyrae stars seem to be roughly uniformly distributed over the field. On the other hand, the Cepheids are mostly located close to the centre of the galaxy, where the main-sequence (MS) stars are strongly concentrated (D02; Hidalgo et al., in preparation). This shows that star formation in the past ∼1 Gyr was confined to the central region of the galaxy.

Figure 3.

Spatial distribution of stars in our observed field, where the variables are overplotted as labelled in the inset. The cross and ellipse represent the centre and Holmberg radius of Leo A (3.5 arcmin; Vansevičius et al. 2004).

Open in new tabDownload slide

We further separated the Cepheids into a bright sample and a faint sample, with a division at W₈₁₄ = 22 (see Section 5). These samples are shown as the filled and open stars, respectively. Interestingly, the Cepheids of the bright sample are even more concentrated close to the centre of the galaxy than the whole sample, all of them being located in the lower half of our FOV. These gradients indicate that the star-forming region has been shrinking with time (see e.g. Hidalgo et al. 2009), at least in the past Gyr or so. This result is in excellent agreement with the SFH as a function of galactocentric radius of Hidalgo et al. (in preparation).

Comparison with previous catalogues

The search for variable stars in Leo A has been carried out by several authors prior to this work. We cross-matched the catalogues from the literature with ours to identify the stars in common and compare the measured properties. These are listed in Table A1. We give their index, type and period measured from our data, together with their identifications and periods from previous catalogues.³

Variable stars in Leo A were first mentioned by Sandage (1986). While the data were not sufficient for a detailed study, Sandage found eight candidates by blinking three plate pairs. However, only the four best candidates are labelled on the finding chart, and no information on periods or magnitudes is given. By comparing his finding charts with ours, we found the following: his ‘definite variable’ (RA = 09:59:19.52, Dec. = +30:44:13.4) is a very bright MS star; one candidate (RA = 09:59:28.88, Dec. = +30:45:42.6) is a blend of two relatively bright RGB stars, and another (RA = 09:59:24.70, Dec. = +30:44:48.1) is close to the tip of the RGB (TRGB); none of these appear to be variable in our data, possibly due to having periods longer than our observational baseline. The fourth candidate is out of our FOV, but corresponds to the D02 Cepheid C2-V67.

The first actual search for variable stars was carried out by Hoessel et al. (1994, hereinafter H94) using data from a variety of telescopes and instruments, from which they discovered 14 suspected variables (including five Cepheids: V5, V8, V9, V10 and V14 in their catalogue). Out of these, V1, V2 and V14 are located outside of our FOV. The stars V3–V6, V8 and V12 are not variables in our data, and V11 is a long-period variable (LPV). We classify the remaining four candidates (V7, V9, V10 and V13) as Cepheids, but the periods from H94 cannot phase our data.

More recently, D02 analysed 28 images (23 in V, 5 in R) taken with the WIYN 3.5 m telescope and found 92 candidates variables (84 Cepheids and 8 RR Lyrae stars). Our catalogue contains 49 of the 50 variables in our FOV. Their C2-V33 appears to be a (non-variable) red clump (RC) star in our data. As shown in Fig. 4, for most of the stars in common we find a very good agreement between their reported periods and ours. For the few variables with discrepant periods, we checked that their value did not give a better light curve. The dotted lines in Fig. 4 show that most of the inadequate periods of D02 are due to the 1 d aliasing usually affecting ground-based data.

Figure 4.

Period difference for the variable stars in common between this and previous works: D02 (open circles), G07 (filled circles) and H94 (open stars). When two periods were listed for a given star, these are joined by vertical lines. The dashed line is the identity line and the dotted lines represent the 1 d aliases of the best period.

Open in new tabDownload slide

The most recent work on variable stars in this galaxy is by the Munich group (Gössl, Snigula & Hopp 2006; Snigula 2006; Gössl 2007, hereinafter G07) using small telescopes with the aim of studying the properties of bright variables in a sample of six northern dwarf irregular galaxies, including Leo A. The work of G07 focused on Cepheids while Snigula (2006) concentrated on LPVs, although some of the stars with insecure classification appear in both. Given the relatively short time baseline of our observations preventing us from detecting long periodicities, most of the previously known stars in our catalogue are in common with G07. In column 7 of Table A1 we use the star indices of G07, where LS, LM and LL refer to short-, medium- and long-period variables, with cut-offs at 1 and 10 d, but also list the indices from Gössl et al. (2006, starting with V-) and from Snigula (2006, starting with LeoA-), when available.

The catalogue of G07 contains 131 candidate variables with R < 23.5 and P < 130 d. Of these, 33 are matches to Cepheids in our catalogue, including 19 with periods in good agreement with ours (see Fig. 4). The remaining ones are either outside of our FOV (45), or not variable in our data (53). We note, however, that all their candidates flagged as ‘best’ and ‘good’ and located in our FOV are indeed Cepheids.

RR LYRAE STARS

10 of the variable stars in our sample are located on the HB and were classified as RR Lyrae stars. From their light curves and periods, we further classify them as seven fundamental-mode (RRab) and three overtone (RRc) pulsators. We find mean periods of 0.636 and 0.366 d for the RRab and RRc, respectively, and a fraction of overtone pulsators of f_c = N_c/(N_ab + N_c) = 0.3. These numbers place Leo A in the Oosterhoff (1939) type II group, whereas the vast majority of LG dwarf galaxies have properties intermediate between those of the type I and type II groups (OoI and OoII, respectively; see Catelan 2009; Smith, Catelan & Clementini 2009). We note, however, that the reliability of these numbers may be affected by the small number of RR Lyrae stars and suboptimal light curves.

In fact, this classification is not as clear from the location of these variables in the period–amplitude diagram. Fig. 5 presents the period–amplitude diagram (top) and the period distribution (bottom) of the RR Lyrae stars. In the top panel, RRab and RRc are shown as the open circles and open triangles, respectively. The solid and dashed curved lines show the loci of OoI and OoII globular clusters from Cacciari et al. (2005). The low-amplitude RRab stars seem to lie on the OoI locus, while the ones with higher amplitude are located between the two groups, suggesting a Oosterhoff intermediate classification instead. The relatively long mean period for the RRab stars may thus be due to the lack of high-amplitude RRab stars.

Figure 5.

Top: period–amplitude diagram for the RR Lyrae stars. The circles and triangles represent RRab and RRc, respectively. The thin dashed line is a fit to the period–amplitude of the RRab; the dotted lines show the 1.5σ limits. The light and dark greyed areas represent the ±1.5σ limits of Cetus and Tucana, respectively, from Paper I. The solid and dashed curved lines show the loci of OoI and OoII globular clusters from Cacciari, Corwin & Carney (2005). Bottom: period histogram for the RR Lyrae stars of the top panel. RRab and RRc are shown as the black dashed and solid grey lines, respectively, while the filled histogram represent the fundamentalized RR Lyrae stars.

Open in new tabDownload slide

The shaded areas in the top panel of Fig. 5 represent the location of RRab stars in Cetus and Tucana from Paper I. The figure shows that the distribution of the RRab of Leo A in period–amplitude space, forming a tight and almost vertical sequence, is very similar to that of Cetus. On the other hand, in Paper II we found that the RRab in IC 1613 were more similar to those in Tucana. Interestingly, both Cetus and Leo A seem to have had a significantly lower SFR at very early epochs than Tucana and IC 1613, which may partly explain the difference.

The very low number of RR Lyrae stars is a remarkable result given the high stellar density in our FOV and completeness at the magnitude of the HB ( ≳ 98 per cent). For comparison, we found 90 or more RR Lyrae stars in roughly similar areas in the other LCID galaxies (Paper I; Paper II). To make this comparison more meaningful, for each of these galaxies we calculated the number of RR Lyrae stars relative to the number of RGB stars within the 2.5 mag below the TRGB. We find 0.01, 0.09, 0.58 and 0.81 for Leo A, IC 1613, Cetus and Tucana, respectively. Assuming RGB and RR Lyrae stars are older than 1.5 and 10 Gyr, respectively, this ratio gives a measure of the fraction of very old stellar population with respect to the intermediate age and old stellar populations of the galaxy. The extremely low value found for Leo A is in very good agreement with the results of the SFH analysis. Cole et al. (2007) and Hidalgo et al. (in preparation) found very little, if any, star formation for ages older than 10 Gyr, and showed that 90 per cent of the star formation had occurred more recently than 8 Gyr ago.

CEPHEIDS

Most of the variable stars in our sample are found in the classical instability strip above the HB. We present their light curves in Fig. 6. Recent work has shown that at the metallicity of the young population of Leo A ([Fe/H] ∼ −1.5), CC and AC can occupy a similar region in the CMD (Fiorentino et al. 2012). CC are young, relatively massive stars ( ≳ 2.2 M_⊙) in the phase of core helium burning. AC, on the other hand, are older and less massive stars on the HB phase of evolution, burning helium in the core in partially degenerate conditions after undergoing the so-called helium flash at the TRGB (e.g. Demarque & Hirshfeld 1975).

Figure 6.

Sample light curves of the Cepheids stars in the F475W (black) and F814W (grey) bands, phased with the period in days given in the lower-right corner of each panel. The open symbols show the bad data points, i.e. with errors larger than 3σ above the mean error of a given star, which were not used in the calculation of the period and mean magnitudes. For a few very bright Cepheids, the light curves have been shifted downward by 1 or 2 mag, as shown by an arrow in the top-left corner of a panel. The complete sample of light curves is available in the online version.

Open in new tabDownload slide

In Fig. 7, we show a comparison of the observed location of the Cepheids in the CMD with the theoretical predictions for both CC and AC with masses 2.0–7.0 M_⊙. The models are based on the BaSTI stellar evolution library (Pietrinferni et al. 2004), and were shifted to the distance and foreground reddening of Leo A. The thick blue line shows the zero-age helium-burning (ZAHeB) locus, which represents the ignition of helium in the core in either degenerate (M ≤ 2.14 M_⊙ at [Fe/H] = −1.5) or non-degenerate (M > 2.14 M_⊙) conditions. The black and blue dashed lines show the theoretical tracks for CC and AC, respectively. They extend from the ZAHeB to the open star symbols representing exhaustion of helium in the core. The figure shows that while all the Cepheids brighter than F814W ∼ 23 are most likely CC, the fainter ones could be a combination of both CC and AC. Note that this result does not change for significantly lower metallicity (i.e. [Fe/H] = −1.8). On the other hand, for [Fe/H] = −1.3 and higher, the AC tracks do not enter the instability strip, leading to a pure CC population.

Figure 7.

CMD showing the observed variable (magenta filled circles) and non-variable (grey dots) stars, where theoretical predictions from the BaSTI library (Pietrinferni et al. 2004) are overplotted. The thick blue line shows the ZAHeB locus for [Fe/H] = −1.5. Evolutionary tracks for both anomalous (blue dashed lines, M = 2.0 and 2.14 M_⊙) and classical (black dashed lines, M = 2.2, 2.4, 2.6, 3.0, 3.5, 4.0, 5.0, and 7.0 M_⊙) Cepheids are shown for the same metallicity. The stellar tracks extend from the ZAHeB to the central helium exhaustion, represented by the open star symbols. The thick red line shows the ZAHeB locus for [Fe/H] = −2.0, which provides a much better fit to the old stellar populations of the HB (see Section 8.1).

Open in new tabDownload slide

Figure 8.

Period–Wesenheit diagram for Cepheids in Leo A. The open circles, filled circles and open triangles represent fundamental, first-overtone and second-overtone pulsators, respectively. A few outliers are labelled and discussed in Section 5. The black dash-dotted and dashed lines are linear fits to the fundamental and FO Cepheids. The grey lines are the same fits to the IC1613 Cepheids (Paper II), shifted to the distance of Leo A.

Open in new tabDownload slide

Fig. 9 shows the Leo A Cepheids in the period–Wesenheit (P-W_I) diagram, together with the OGLE-III Cepheids of the Small and Large Magellanic Clouds (SMC and LMC, respectively; Soszyński et al. 2008a,b, 2010a,b) in the middle and right-hand panels. In all three panels, the Leo A Cepheids are shown with the same symbols as in Fig. 8, and are shown by themselves in the left-hand panel for clarity. The Cepheids of Leo A were shifted according to the TRGB distance determined in Section 7.1, i.e. (m − M)₀ = 24.57. The LMC/SMC apparent magnitudes were converted to absolute magnitudes assuming a distance modulus of 18.515 ± 0.085 to the LMC (Clementini et al. 2003), and a distance offset of 0.51 of the SMC relative to the LMC (Udalski et al. 1999a). The middle panel shows that the bulk of Leo A Cepheids are in relatively good agreement with the SMC Cepheids – even though the significant spread of the SMC Cepheids complicates the comparison – and extend the F, FO, and SO PL relations to shorter periods and fainter magnitudes. This supports their identification as CC.