Photometric Determination of the Mass Accretion Rates of Pre-main-sequence Stars. VI. The Case of LH 95 in the Large Magellanic Cloud^*

With the aim to identify PMS candidates within LH 95, we followed the method described and first tested in De Marchi et al. (2010) and then applied in a series of papers (Beccari et al. 2010, 2015; De Marchi et al. 2011, 2013, 2017; Spezzi et al. 2012; Zeidler et al. 2016). This method relies on the detection of Hα excess emission in low-mass star-forming stars (see, e.g., Königl 1991).

We selected from our photometric catalog in the F555W, F658N, and F814W bands all those stars whose photometric uncertainties δ₅₅₅, δ₆₅₈, and δ₈₁₄ in each individual band do not exceed 0.05 mag. A total of 1294 stars satisfy this condition (gray dots in the color–color diagram of Figure 2) out of 24,515 sources in the complete catalog (black dots in the color–magnitude diagram (CMD) of Figure 3). These stars are typically old MS and do not have appreciable Hα excess; they define the reference with respect to which should look for excess emission in the (m₅₅₅ − m₆₅₈)₀ color at a given (m₅₅₅ − m₈₁₄)₀ color (see the running median represented by the dashed line in Figure 2). For comparison, this reference sequence is in good agreement with the theoretical color relationship in the same filters obtained using the Bessell et al. (1998) model atmospheres for MS stars with effective temperature 3500 K ≤ T_eff ≤ 40,000 K, surface gravity log g = 4.5, and metallicity index [M/H] ≃ −0.5 (Colucci et al. 2012), appropriate for the LMC (dotted line in the same figure). The rms deviation between the model and the data amounts to ∼0.03 mag and is dominated by the systematic departure around (m₅₅₅ − m₈₁₄)₀ ∼ 1.5, most likely due to the coarse sampling of our data. As pointed out by De Marchi et al. (2017), even before correction for reddening, such a color–color diagram provides a robust identification of stars with Hα excess, since in these bands, the reddening vector runs almost parallel to the median photospheric colors of nonaccreting objects, and, moreover, our targets do not have a known patchy nature of interstellar absorption (see Da Rio et al. 2009).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To select the most probable accretors, after the exclusion of the 1294 stars taken as reference, we first selected the targets with δ₅₅₅ < 0.1, δ₆₅₈ < 0.3, and δ₈₁₄ < 0.1 mag, namely, 5155 objects. Then, we retained those whose dereddened (m₅₅₅ − m₈₁₄)₀ color exceeds the local average by at least four times the individual combined photometric uncertainty δ₃ in the three bands F555W, F658N, and F814W, where

$$\begin{eqnarray}&&{\delta }_{3}=\sqrt{\displaystyle \frac{{\delta }_{555}^{2}+{\delta }_{658}^{2}+{\delta }_{814}^{2}}{3}},\end{eqnarray} \tag{ 1 }$$

with δ₅₅₅, δ₆₅₈, and δ₈₁₄ being the photometric uncertainties in each individual band. It should be noted that, for the selected 5155 sources, δ₃ is dominated by the uncertainty on the Hα magnitude, which is, on average, around 0.15 mag, while the median value of the uncertainty in the other two bands is $\langle$δ₅₅₅$\rangle$ ∼ 0.03 and $\langle$δ₈₁₄$\rangle$ ∼ 0.02 mag, respectively. In the end, a total of 247 stars satisfy the condition, and they must be regarded as having a bona fide Hα excess above the 4σ level (circles in Figure 2). This allows us to select the most probable PMS candidates, even when the uncertainty in the F658N band is not negligible. Indeed, as we will show in Section 3.2, our selection in Hα excess emission translates directly into Hα equivalent widths typical of accretors, allowing us to safely remove possible contaminants from our sample, such as chromospherically active stars (see, e.g., White & Basri 2003; Biazzo et al. 2007; Frasca et al. 2008; Beccari et al. 2015). Other classes of objects whose spectra might present Hα emission are interacting binaries, but they are typically very rare in Local Group galaxies (e.g., Dobbie et al. 2014), and their intrinsic colors are bluer with respect to the MS (see, e.g., Beccari et al. 2014 and references therein).

The m₅₅₅ magnitude versus the m₅₅₅ − m₈₁₄ color of the detected sources is shown in Figure 3. From this CMD, the two targets with m₅₅₅ < 22 are bright objects with Hα excess that we exclude from the following analysis, as we are searching for low-mass PMS candidates.⁶ Gouliermis et al. (2007) found in the CMD a pronounced turnoff at V ∼ 22.5 and a red clump (RC) at V ∼ 19 and V − I ∼ 1.2. The RC and old MS population are best matched by a 0.7 Gyr isochrone taken from the Padova-Trieste Stellar Evolution Code (PARSEC; see Bressan et al. 2012). Stars with Hα excess are shown in red, and they define two distinct groups, nicely separated by an 8 Myr old PMS isochrone from the same authors. The theoretical isochrones of Bressan et al. (2012) are shown, respectively, with blue and orange lines in Figure 3, where the distance of 51.4 ± 1.2 kpc (Panagia 1999), corresponding to a distance modulus (m_V − M_V)₀ = 18.55, and a metallicity of Z = 0.007, typical of young LMC stars (e.g., Colucci et al. 2012), were adopted. These isochrones include both the effects of the MW intervening in absorption along the line of sight and the LH 95 mean absorption within the field. According to Fitzpatrick & Savage (1984), the former amounts to ${A}_{555}^{\mathrm{MW}}=0.22$ and $E{({m}_{555}-{m}_{814})}^{\mathrm{MW}}=0.1$, while for the LH 95 field, we considered the following mean extincion values (Da Rio et al. 2009): R₅₅₅ = A₅₅₅/E(m₅₅₅ − m₈₁₄) ≃2.18 and R₈₁₄ = A₈₁₄/E(m₅₅₅ − m₈₁₄) ≃ 1.18.

To determine the presence and extent of possible differential extinction, Da Rio et al. (2009, 2012) analyzed the position of the upper-MS (UMS) stars in the CMD. In particular, after subtracting field stars, they compared the CMD position of the UMS objects with that expected according to grids of evolutionary models. They concluded that there is not a significant level of differential extinction for the UMS stars in the LH 95 field. This conclusion is very relevant to our investigation because young PMS objects and UMS stars share the same spatial distribution (see, e.g., De Marchi et al. 2011, 2013). Those authors also provided the total optical extinction and reddening toward LH 95, namely, ${A}_{555}^{\mathrm{tot}}$ = 0.6 and $E{({m}_{555}-{m}_{814})}^{\mathrm{tot}}\sim 0.3$, respectively; therefore, the reddening within LH 95 is $E{({m}_{555}-{m}_{814})}^{\mathrm{LH}95}\sim 0.2$. Finally, considering the relation A₆₅₈/A₅₅₅ ≃ 0.8 (Rodrigo et al. 2012) and the adopted Galactic and LMC extinction laws, the total extinction in the Hα band toward LH 95 is A₆₅₈ = 0.48.

Looking at the CMD shown in Figure 3, most stars with m₅₅₅ − m₈₁₄ ≳ 0.6 and m₅₅₅ ≳ 22 could be old MS or PMS objects or red giants. This is why it is important to search for PMS objects by analyzing the Hα excess emission as a signature of accretion (and therefore of youth).

As pointed out by De Marchi et al. (2010), the contribution of the Hα line to the m₅₅₅ magnitude is completely negligible; therefore, the magnitude ΔHα corresponding to the excess emission is

$$\begin{eqnarray}&&{\rm{\Delta }}{\rm{H}}\alpha ={\left({m}_{555}-{m}_{658}\right)}^{\mathrm{obs}}-{\left({m}_{555}-{m}_{658}\right)}^{\mathrm{ref}},\end{eqnarray} \tag{ 2 }$$

where the superscript "obs" refers to the observations and "ref" to the reference sequence at each m₅₅₅ − m₈₁₄ color (dashed line in Figure 2). Once ΔHα is derived in this way, the Hα emission-line luminosity L_Hα can be immediately obtained from the photometric ZP, the absolute sensitivity of the instrumental setup (in this case, the F658N of the ACS/WFC), and the distance to the sources. We took the F658N photometric properties of the instrument at the exact observing dates from Ryon (2018), namely, the inverse sensitivity PHOTFLAM = 1.967 × 10⁻¹⁸ erg cm⁻² s⁻¹ Å, and the ZP in VEGAMAG, ZP = 22.383. Considering the rectangular width of the F658N filter RECTW = 74.98 Å and assuming a distance to SN 1987A of 51.4 ± 1.2 kpc (Panagia 1999), we derived a median value of the Hα luminosity of the 245 low-mass objects with Hα excess of ∼1.2 × 10³¹ erg s⁻¹ or ∼0.3 × 10⁻² L_⊙. This value is lower than that measured by De Marchi et al. (2010) in the SN 1987A field (∼10⁻² L_⊙) and that found by De Marchi et al. (2017) in the 30 Doradus Nebula (∼3 × 10⁻² L_⊙). This difference is not surprising because our observations include stars with greater m₅₅₅ − m₈₁₄ colors and therefore lower masses.

The total uncertainty on our L_Hα measurements is typically ∼16% and dominated by the inaccuracy on the Hα magnitude and the uncertainty on the distance and instrumental setup, accounting for ∼5% and ∼3%, respectively (see also De Marchi et al. 2010). Indeed, extinction does not have any influence in the V − H_Hα color excess, and therefore in the L_Hα uncertainty, both because the reddening vector in Figure 2 is substantially parallel to the reference template (dashed line) and because our targets do not seem to have differential reddening, as indicated, for instance, by the relatively compact RC in the CMD.

Following De Marchi et al. (2010), the difference between the observed Hα magnitude (m₆₅₈) and the level of the Hα continuum (${m}_{658}^{{\rm{c}}}$) provides a direct measure of EW_Hα. In particular, since the line width is narrow compared to the width of the filter, the line profile falls completely within the filter bandpass. If we assume that the stars defining the reference template have no Hα absorption features, their m₅₅₅ − m₆₅₈ index would correspond to the color of the pure continuum. Therefore, EW_Hα is given by the following relationship:

$$\begin{eqnarray}\begin{array}{rcl}{\mathrm{EW}}_{{\rm{H}}\alpha } & = & \mathrm{RECTW}\times [1-{10}^{-0.4\times ({m}_{658}-{m}_{658}^{{\rm{c}}})}]\\ & = & \ \mathrm{RECTW}\times [1-{10}^{-0.4\times {\rm{\Delta }}{\rm{H}}\alpha }],\end{array}\end{eqnarray} \tag{ 3 }$$

with RECTW the rectangular width of the F658N filter. As for L_Hα, the statistical uncertainty on EW_Hα, typically ∼6%, is dominated by the uncertainty on the Hα photometry. The validity of this method was also independently tested and then confirmed by Barentsen et al. (2011, 2013), who considered both photometric and spectroscopic observations for T Tauri stars in the Galactic NGC 2264 and IC 1396 young regions. The authors found a strong correlation between Hα equivalent widths derived with both spectroscopic and photometric methods. This already suggests that the possible contribution of veiling due to accretion is greatly reduced by the subtraction of the continuum flux from the band flux.

During the last 30 yr, veiling has been more and more accurately measured, mainly thanks to spectroscopic observations of SFRs in our Galaxy (see, e.g., Hartigan et al. 1995; Manara et al. 2013; Alcalá et al. 2014, to cite a few works), but in general, its effects must also be taken into account for broadband measurements. As discussed in detail in De Marchi et al. (2010), any nebular continuum unrelated to the stellar photosphere (and therefore also the one associated with the accretion luminosity) will add to the intrinsic continuum of the stars, thereby affecting both the observed total level and the slope. This could alter the measured broadband colors of the source, thereby thwarting our attempts to infer the level of the continuum in the Hα band from the observed m₆₅₈ and m₈₁₄ magnitudes and ultimately also affecting the effective temperature and bolometric luminosities that we measure (see Section 3.3.1).

Fortunately, the contribution of the nebular continuum appears to be insignificant for the stars in our sample. To prove this, De Marchi et al. (2010) assumed a fully ionized gas of pure H, considering only bound–free and free–free transitions and ignoring the contribution to the continuum from two-photon emission (Spitzer & Greenstein 1951). Using Osterbrock's (1989, chapter 4) tabulations, they found Hα line intensity and continuum fluxes of the nebular gas for gas electron temperatures in the range 5000–20,000 K (see their Table 2). The purely nebular EW_Hα ranges from 5000 to 9000 Å, or more than 2 orders of magnitude higher than what we measure for our PMS objects (see Figure 4). We can, therefore, safely conclude that the nebular contribution to the continuum is negligible (less than 1%).

Zoom In Zoom Out Reset image size

Furthermore, for gas temperatures in the range 5000–10,000 K, the m₆₅₈ − m₈₁₄ color of the nebular continuum varies from 1.4 to 0.5, spanning a range typical of GK-type stars. Thus, the effects of the nebular continuum on the m₆₅₈ − m₈₁₄ color of PMS stars remain insignificant even for the objects with the highest EW_Hα in our sample. Therefore, for these objects also, the relationships between m₆₅₈ − m₈₁₄ and effective temperature and, in turn, the bolometric luminosity that we will derive in Section 3.3.1 are not affected by the veiling introduced by the additional nebular continuum.

The values of EW_Hα that we obtain in the field of LH 95 are shown in Figure 4 as a function of the dereddened m₅₅₅ − m₈₁₄ color for the 245 low-mass PMS candidates. All selected targets at the 4σ level fall well above the threshold established by White & Basri (2003) to identify probable accretors as a function of spectral type (see also Beccari et al. 2014), thus confirming that our selection criteria of PMS candidates is cautious. Indeed, the sample that we selected must be considered as a lower limit to the number of objects with genuine Hα excess. This means that most probably, we are excluding from the sample the weakly accreting PMS stars, but completeness is not the aim of this work. Instead, we are interested in studying the properties of the mass accretion process in PMS stars, and for this reason, it is important to have a solid sample of candidates.

The values of EW_Hα for our sample range from ∼12 to ∼70 Å, with a median of 29 Å. These values are typical of PMS stars. It should be noted that, because of the width of the specific F658N filter, ΔHα includes small contributions due to the emission of the two forbidden [N ii] lines at 6548 and 6584 Å. De Marchi et al. (2010), following a conservative approach, estimated corrections of ∼0.98, on average, for the ACS F658N filter. This translates into a lower EW_Hα value by ∼0.2–1.4 Å in the range characteristic of our targets, i.e., within the uncertainties of our measurements (see Table 3).

The physical parameters of the PMS candidates identified in Section 3.1, i.e., their effective temperature, bolometric luminosity, mass, and age, were obtained as explained in the following two subsections.

3.3.1. Effective Temperature and Bolometric Luminosity

We derived the effective temperature T_eff from the observed m₅₅₅ − m₈₁₄ color, properly corrected for the total reddening due to both our Galaxy and LH 95, as explained in Section 3.1. The models of Bessell et al. (1998) with 3500 K ≤ T_eff ≤40,000 K, $\mathrm{log}g=4.5$, and metallicity index [M/H] =−0.5 dex were used for the conversion from color to effective temperature, following the work by De Marchi et al. (2010) for the ACS/WFC filters (see that paper for details). Since the models of Bessell et al. (1998) are available only for T_eff >3500 K, for lower temperatures, we decided to consider the T_eff–(V − I_C) calibration by Pecaut & Mamajek (2013),⁷ assuming for simplicity that the calibrated m₅₅₅ and m₈₁₄ magnitudes coincide with V and I_C. The reason for using a different calibration at temperatures lower than those covered by the models of Bessell et al. (1998) is to avoid possibly larger uncertainties in T_eff due to arbitrary extrapolations, as the relationship between T_eff and m₅₅₅ − m₈₁₄ is critical for very cool temperatures. The bolometric luminosity L_⋆ was obtained from the m₅₅₅ magnitude corrected for the interstellar extinction (see Section 3.1), having adopted a distance to LH 95 of 51.4 kpc (Panagia 1999) and a bolometric solar magnitude ${M}_{\mathrm{bol}}^{\odot }$ = 4.74 (Pecaut & Mamajek 2013) and applied at any T_eff the bolometric corrections of the latter authors.

The positions of the PMS candidates in the H-R diagram are displayed in Figure 5, where the ±1σ uncertainties on T_eff and L_⋆ are also shown; in most cases, these errors are within the symbol size. They are mostly due to uncertainties in photometry and distance.⁸ As a reference, we traced the PMS theoretical isochrones of Bressan et al. (2012) for metallicity Z = 0.007, as appropriate for the young populations of the LMC (e.g., Colucci et al. 2012), and for ages of 1, 8, 16, and 32 Myr from right to left (dot-dashed lines). Also shown are the representative PARSEC evolutionary tracks for masses of 0.1, 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.5, 2.0, 3.0, and 4.0 M_⊙ from the same Bressan et al. (2012) models (solid lines). The dashed line in the same figure defines the zero-age main sequence (ZAMS) by Bressan et al. (2012). From the stellar bolometric luminosity and effective temperature, we also derived the stellar radius R_⋆, assuming 5770 K as the effective temperature of the Sun. Typical mean uncertainties on R_⋆ are around 5% and include both uncertainties in T_eff and L_⋆.

Zoom In Zoom Out Reset image size

In the H-R diagram of our PMS candidates, a bimodal distribution in age and T_eff seems to be evident, with a separation around 8 Myr. In particular, stars younger than 8 Myr have a mean T_eff ∼ 3965 K, while older targets have a mean T_eff of ∼4990 K. This apparent bimodality will also be evident in the accretion properties, thus proving that it is not caused by detection limits. This issue will be discussed in the following sections.

3.3.2. Mass and Age

After having identified a population of PMS candidates in Section 3.1 and derived their effective temperature and bolometric luminosity in the previous subsection, it is important for our purposes to determine their mass and age from the H-R diagram (see Figure 5). We followed the approach originally discussed in Romaniello (1998) and most recently refined by De Marchi et al. (2011, 2013). The method, without making assumptions on the properties of the population and the pure basis of the measurement errors, provides the probability distribution for each individual star to have a given value of mass and age, with typical uncertainties of ∼5% and ∼15%, respectively (see De Marchi et al. 2017 for a thorough explanation of the procedure). In particular, we determined the most likely mass M_⋆ of the 245 low-mass PMS candidates of LH 95 by comparing the location of each object on the H-R diagram with theoretical PMS evolutionary tracks. As for the latter, we adopted the already mentioned PARSEC tracks for Z = 0.007 and available down to M_⊙ = 0.09 M_⊙ (Bressan et al. 2012).

Comparing masses computed from different evolutionary tracks is important for the determination of the uncertainty on the mass accretion rate and its relationship with stellar mass and age. In order to assess how differences in the evolutionary models affect our results, in Appendix A, we compare mass, age, and mass accretion rate measurements obtained using the Pisa tracks from Tognelli et al. (2011), available down to M_⋆ = 0.2 M_⊙, with those obtained using the PARSEC tracks. For an extensive discussion of the model-dependent age estimation in clusters, see the recent review by Soderblom et al. (2014).

We determined the ages of individual objects by interpolating between the isochrones in the H-R diagram. As already mentioned in Section 3.1, it is evident from Figure 5 that the PMS candidates appear to be distinct in two populations with a "gap" around 8 Myr. We thus decided to divide the sample of selected PMS candidates into two subsamples depending on their age: from now on, we will indicate as younger PMS candidates those with age t < 8 Myr and older PMS candidates those with t > 8 Myr. With such an age difference, older PMS stars must belong to a previous generation with respect to the younger PMS objects; thus, no spatial relationship (whether a correlation or anticorrelation) should be expected between the two types of objects, as is indeed evident in Figure 1. This will be discussed in more detail in Section 5.1. The younger PMS candidates are about 35% (85/245) of the total sample, while the older PMS candidates represent about 65% (160/245).

The histograms with the mass and age distributions for the 245 low-mass PMS candidates are shown in Figure 6. Different line types correspond to younger (dashed lines) and older (dotted lines) PMS candidates. The mass distributions are peaked at similar values for both younger and older populations, but they show different ranges, with the younger population having a wider range in mass than the older one. Not surprisingly, old PMS stars comprise many low-mass stars, since the more massive objects have already reached the MS. Inspecting the age distribution, a clear separation between younger and older PMS stars is evident. We highlight here that these measurements are not reliable for putting constraints on the shape of the mass function or the exact value of the star formation rate, since we are only considering PMS with Hα excess emission at the 4σ level at the time of the observations. Moreover, we are not taking into account photometric incompleteness, which is unavoidably more severe at lower masses.

Zoom In Zoom Out Reset image size

As already found by De Marchi et al. (2010, 2017) in other regions of the LMC, it is noteworthy that many of the PMS candidates in Figure 5 are close to the MS and would have been missed if no information on their Hα excess had been available. Since we would expect to find very young objects above and to the right of the MS in the CMD, it is indeed customary to identify PMS objects by searching in that area of the CMD. However, this method of identification of PMS stars is not very reliable because of the presence of older populations and possible age spreads in the same field, which prevent the true identification of PMS stars on the basis of the stellar effective temperatures and luminosities alone (see discussion in De Marchi et al. 2010). In fact, these older PMS stars were not detected by Gouliermis et al. (2007) and were considered as field stars by the same authors.

In summary, we find evidence for a series of at least two star formation episodes, which correspond to two distinct stellar populations with different ages. We indeed identify a generation of younger PMS stars with ages ranging from <1 up to ∼7 Myr (median value ∼1 Myr) and a generation of older PMS stars with ages of ∼10–60 Myr (median value ∼50 Myr). Objects of this type are to be expected, also according to the evolutionary tracks. In fact, from the PARSEC tracks at metallicity Z = 0.007, a star with a mass of ∼0.7 M_⊙, i.e., around the peak histogram of our sample (see Figure 6), takes ∼50 Myr to reach the MS.

In the magnetospheric accretion scenario, the accretion luminosity can be determined from the measurement of the reradiated energy from the circumstellar gas ionized and heated by the funnel flows (e.g., Hartmann et al. 1998). The Hα line, and hence its luminosity, generated in this process can be used as a diagnostic to derive the accretion luminosity. From the analysis of a set of L_Hα measurements of a group of T Tauri stars in Taurus-Auriga compiled by Dahm (2008), De Marchi et al. (2010, 2013) found the following L_acc–L_Hα relationship, which we adopt in this work:

$$\begin{eqnarray}&&\mathrm{log}{L}_{\mathrm{acc}}/{L}_{\odot }=\mathrm{log}{L}_{{\rm{H}}\alpha }/{L}_{\odot }+(1.72\pm 0.25),\end{eqnarray} \tag{ 4 }$$

where the ratio L_acc/L_Hα is linear. Recently, Alcalá et al. (2017), using X-shooter spectra of class II objects in the Galactic Lupus SFR, concluded that the empirical relationships they derived between L_acc and the luminosity of several lines from UV to NIR are consistent with being linear.

Taking into account Equation (4), the median value of the accretion luminosity thus obtained for our sample of 245 low-mass PMS candidates is ∼0.17 L_⊙. The statistical uncertainty on L_acc is dominated by the quoted uncertainty of ∼16% on L_Hα mainly associated with the photometric error in the Hα magnitude. There is also a systematic error due to the L_acc–L_Hα relationship, but since we used Equation (4) for all PMS stars, this uncertainty will not prevent the comparison between different targets. Comparable uncertainties in the L_acc–L_Hα relationship were obtained by Alcalá et al. (2017) for the Lupus SFR.

Once L_acc is known, the mass accretion rate ${\dot{M}}_{\mathrm{acc}}$ can be derived from the freefall equation that links the luminosity released in the impact of the accretion flow with the rate of mass accretion according to the following relationship (see, e.g., Hartmann 1998):

$$\begin{eqnarray}&&{\dot{M}}_{\mathrm{acc}}={\left(1-\displaystyle \frac{{R}_{\star }}{{R}_{\mathrm{in}}}\right)}^{-1}\displaystyle \frac{{L}_{\mathrm{acc}}{R}_{\star }}{{{GM}}_{\star }}\,\approx 1.25\displaystyle \frac{{L}_{\mathrm{acc}}{R}_{\star }}{{{GM}}_{\star }},\end{eqnarray} \tag{ 5 }$$

where M_⋆ and R_⋆ are the stellar mass and photospheric radius, respectively; R_in is the inner radius of the accretion disk; and G is the universal gravitational constant. Here R_in corresponds to the distance at which the disk is truncated because of the stellar magnetosphere and from which the disk gas is accreted and channeled by the magnetic field lines; therefore, its value is rather uncertain because it depends on how the accretion disk is coupled with the star. Following Gullbring et al. (1998), we assume R_in = 5 R_⋆ for all PMS stars.

The median value of the distribution of mass accretion rates is ∼7.5 × 10⁻⁹ M_⊙ yr⁻¹, with higher values for the younger PMS candidates (∼5.4 × 10⁻⁸ M_⊙ yr⁻¹) and lower values for the older PMS candidates (∼4.8 × 10⁻⁹ M_⊙ yr⁻¹).

Concerning the statistical errors on ${\dot{M}}_{\mathrm{acc}}$, the first source of uncertainty is L_Hα. With our selection criteria, the typical uncertainty on L_Hα is ∼16% and is dominated by random errors. The other sources of uncertainty for ${\dot{M}}_{\mathrm{acc}}$ are the stellar mass and radius. The uncertainty on R_⋆ is typically ∼5%, including the systematic uncertainty on the distance modulus. As for the mass, its determination is linked to the comparison of the location in the H-R diagram with the evolutionary tracks. When we interpolate through the PMS evolutionary tracks to estimate the mass, the uncertainties on effective temperature and stellar luminosity imply an error of ∼7% on M_⋆. Combining all sources of errors, the statistical uncertainty on ${\dot{M}}_{\mathrm{acc}}$ is ∼18%.⁹

The systematic uncertainty on ${\dot{M}}_{\mathrm{acc}}$ is dominated by the knowledge of the ratio L_acc/L_Hα reported in Equation (4). As already mentioned in Section 4.1, this ratio, even if uncertain by a factor of ∼2 due to the variations of the Hα line intensity, is the same for all stars; therefore, the comparison between different objects is not hampered by this uncertainty, as long as the statistical errors are small (see De Marchi et al. 2010).

As pointed out by De Marchi et al. (2010), other sources of systematic errors on the derived ${\dot{M}}_{\mathrm{acc}}$ are due to theoretical evolutionary tracks and isochrones, reddening, Hα emission generated by processes different from accretion, and contribution of nebular continuum to the photometric colors. Concerning the first source of errors, the main uncertainty on the derived mass and age comes from differences between models computed by different authors or the use of models with metallicities that might not properly describe the stellar population under study. As shown in Appendix A, if we had, for instance, used the Tognelli et al. (2011) PMS tracks instead of those of Bressan et al. (2012) at the same metallicity, we would have obtained similar values of mass and age for the PMS, to within 2% and 6%, with the largest discrepancy for 0.35 M_⊙ ≲ M_⋆ ≲ 0.70 M_⊙ (see Appendix A). Concerning the metallicity, had we used tracks with Z lower by 30%, the masses of our PMS objects would be systematically smaller by about 10% and the ages younger by a negligible amount for the luminosity and temperature ranges typical of our targets. For what concerns the reddening, we followed De Marchi et al. (2010) and concluded that underestimating the E(m₅₅₅ − m₈₁₄) color excess by ∼0.2 mag would lead to a 30% overestimate of R_⋆/M_⋆. This translates into the same overestimate of ${\dot{M}}_{\mathrm{acc}}$, which is smaller than the typical measurement uncertainties in the determination of the mass accretion rate. Finally, the possibility that processes different from accretion (e.g., chromospheric activity, H knots along the line of sight, ionization of H gas from nearby massive stars) or nebular continuum may alter the determination of ${\dot{M}}_{\mathrm{acc}}$ was addressed in detail in De Marchi et al. (2010), with the conclusion that their contribution is negligible. Indeed, the contribution of the background emission was safely removed thanks to the fact that the m₆₅₈ magnitude of each star was determined above the background calculated locally in an annulus of a few arcseconds around the centroid of the star (see also Section 2).

From a first inspection of the spatial distribution of the most probable accreting objects in LH 95 shown in Figure 1, younger low-mass PMS candidates (red diamonds) appear to be clustered around Be stars. In particular, they are nonuniformly distributed in the field of the LH 95 association, but they are concentrated in small clusters around bright massive stars, with a clumpy spatial distribution on the scale of ∼5 pc. Older PMS stars do not seem to form any significant concentration and are uniformly distributed within the region. Gouliermis et al. (2007) suggested the existence of significant substructures of early-type stars containing candidate Herbig Ae/Be stars for LH95. Here we support this scenario, but we also suggest that the subgroups of early-type stars include low-mass young PMS objects, similar to Galactic OB associations, like Orion (Briceño et al. 2005, 2018).

Besides very different ages, as discussed in Section 3.3.2, the two populations of younger and older PMS stars also have considerably different spatial density distributions. We compare these distributions in Figure 7 by means of filled contours. In this figure, we considered the total population of 245 low-mass PMS candidates with well-defined masses and ages. The remarkable feature is the difference in the spatial distribution of younger and older PMS stars, with older objects much more widely distributed and not overlapping with the younger generation.

Zoom In Zoom Out Reset image size

The spatial distribution of the younger and strongly accreting PMS stars in Figure 7 suggests that a recent star formation episode occurred a few Myr ago in regions also including many of the early-type Be stars identified in the LH 95 field. The older and less-accreting PMS stars are instead uniformly distributed without any specific clumping within the field. They might have formed several tens of Myr ago in a more central configuration but later had time to dissipate in a widespread configuration. Their mean age of ∼50 Myr and spatial distribution within ∼40–50 pc at the distance of LH 95 (see also Figure 1) is indeed compatible with a velocity dispersion of a few km s⁻¹, which is typical for young SFRs.

Figure 8 shows the accretion luminosity as a function of the stellar luminosity for both younger and older PMS stars. As already observed in SFRs close to the Sun, L_acc increases with stellar luminosity, with a dispersion appearing to be even smaller for our targets (the recent case of the Lupus SFR by Alcalá et al. 2017 is shown as an example). The accretion luminosity of our PMS candidates mainly falls in the range between 0.2 L_⋆ and ∼L_⋆, with the peak of the accretion luminosity distribution around ∼0.5–0.6 L_⋆, while those of regions in the solar neighborhood, like the Lupus SFR by Alcalá et al. (2017) superimposed in the figure, is typically ≲0.1 L_⋆ (see also, e.g., Muzerolle et al. 1998; White & Hillenbrand 2004; Antoniucci et al. 2011; Caratti o Garatti et al. 2012; Biazzo et al. 2014). In this context, we cannot make a real quantitative comparison between our results and the findings by Alcalá et al. (2017), but it is possible that the differences between the samples could be mainly due to the following reasons: (i) different selection criteria of accreting PMS candidates; (ii) different methods to derive stellar parameters; (iii) different L_acc–L_Hα relationships, which can lead to differences up to ∼0.2–0.3 dex in $\mathrm{log}{L}_{\mathrm{acc}}/{L}_{{\rm{H}}\alpha }$ (for instance, the L_acc/L_Hα ratio in the case of the Alcalá et al. 2017 empirical relationship is not exactly linear, as in our case); (iv) different mass and metallicity ranges, with our targets having M_⋆ = 0.1–1.8 M_⊙ (with a median of ∼0.7 M_⊙) and Z =0.4 Z_⊙, compared to 0.02–2.0 M_⊙ (with a median of ∼0.2 M_⊙) and Z ∼ Z_⊙ for the Alcalá et al. (2017) sample; and (v) other environmental conditions, such as the gas density and contamination.

Zoom In Zoom Out Reset image size

In Figure 9, the L_acc values are plotted in logarithmic scale as a function of the effective temperature of our PMS candidates, together with the sample of Alcalá et al. (2017). The L_acc–T_eff plot appears to be very similar to the H-R diagram shown in Figure 5, with the younger and older PMS candidates well separated in T_eff. In the T_eff range between ∼3.6 and ∼3.7, some Lupus targets seem to have similar log L_acc/L_⊙ values to our PMS stars. This could be an indication of similar accretion properties at a given T_eff range, and therefore stellar mass. Unfortunately, we do not have many objects with very low effective temperatures (in particular with log T_eff < 3.5) to verify the decreasing trend observed in Alcalá et al. (2017) at the very low-mass regime.

Zoom In Zoom Out Reset image size

In Figure 10, the mass accretion rate is shown as a function of the stellar age. In this figure, we divided our sample into lower- and higher-mass targets, where 0.67 M_⊙ is the median mass of all PMS candidates. At first glance, the slope of the $\mathrm{log}{\dot{M}}_{{\rm{acc}}}$–log t relationship appears to be similar for both lower-mass (i.e., ≃−0.72) and higher-mass (i.e., ≃−0.70) regimes and in good agreement with those measured in other Magellanic Cloud (MC) environments (see De Marchi et al. 2011, 2013, 2017). The shaded region in the same figure represents the prediction of viscous disk evolution by Hartmann et al. (1998). These models are able to reproduce the observed decreasing trend of ${\dot{M}}_{\mathrm{acc}}$ with age for low-mass T Tauri stars in SFRs in the solar neighborhood (see dotted region in the same figure and Hartmann et al. 2016 for a recent review of this issue¹⁰ ). The slope of this trend appears to be steeper than those obtained by us for both lower- and higher-mass regimes (≃−1.4 against ≃−0.7). This means that in the PMS candidates of LH 95, ${\dot{M}}_{\mathrm{acc}}$ decreases more slowly with time than what is observed for low-mass T Tauri stars in Galactic SFRs close to the Sun (≲1 kpc; Sicilia-Aguilar et al. 2010). This behavior supports the recent suggestions by De Marchi et al. (2017), according to which, when metallicity is higher, like in the local neighborhood, there are more dust grains in the disk; therefore, the radiation pressure is higher, limiting the accretion process in both its rate and duration, while the mass accretion process seems to last longer at low metallicity. Other authors (e.g., Yasui et al. 2009, 2010, 2016) have concluded that in some low-metallicity environments of the outer Galaxy, the disk lifetimes are shorter than in SFRs in the solar vicinity. However, these works use the dust content of circumstellar disks as a proxy for the total mass of the disks and their lifetimes, while here we directly measure the infall of the probably more abundant gas onto the stars. Therefore, the results of the two studies are not directly comparable. And indeed, our findings are in agreement with optical spectroscopic observations of SFRs in the Galactic anticenter (Cusano et al. 2011; Kalari & Vink 2015), which indicate that a significant fraction of the young stellar objects have preserved their accretion disks, despite the low metallicity. These authors concluded that disk survival may depend not only on metallicity but also on other environmental physical conditions or properties of the central objects. Galli et al. (2015) found the following empirical relationship between disk lifetime and stellar mass: ${t}_{\mathrm{disk}}=4\times {10}^{6}{({M}_{\star }/{M}_{\odot })}^{0.75}$. Such a relationship was found by the authors for stars in the Taurus-Auriga association, i.e., with Z ∼ Z_⊙, but similar results were previously obtained in other solar-metallicity environments (see the case of Lupus in Bertout et al. 2007). If applied to our PMS candidates, neglecting the effects due to different metallicity or binarity (to mention a few), this relationship would imply that disks around stars of ∼0.55 M_⊙ (median mass of our younger PMS stars) survive for t_disk ∼ 2 Myr, similar to the mean age of our younger population (∼1 Myr). A disk dispersal time of the order of 2 Myr is quite at odds with the ages of the older PMS candidates, whose disks have not been totally dissipated even at several tens of Myr. We are therefore led to believe that, unless the older episodes of star formation were much more intense than the most recent one, it is very likely that circumstellar disks live longer in these metal-poor environments.

Zoom In Zoom Out Reset image size

Figure 11 shows ${\dot{M}}_{\mathrm{acc}}$ versus M_⋆ for all PMS candidates and contains several pieces of information in one graph. Younger PMS candidates are marked with diamonds (red for ages younger than 1 Myr and green for 1–8 Myr), while older PMS candidates are represented by filled circles (the targets with ages of 8–16 Myr are shown in blue, while those older than 16 Myr are in black).

Zoom In Zoom Out Reset image size

From first glance, the stars in this plot appear to define a "fan-shaped" area. At a given stellar mass, we notice a wide spread in ${\dot{M}}_{\mathrm{acc}}$ for stars younger than 16 Myr. In particular, this spread in $\mathrm{log}{\dot{M}}_{\mathrm{acc}}$ ranges from ∼1 dex at ∼0.25 M_⊙ up to ∼2 dex at ∼0.67 M_⊙, the mean mass of our targets. Splitting the sample of PMS stars in age bins, it is evident how the mass accretion rate is higher for younger stars, with mean values ranging from ∼2.6 × 10⁻⁷ M_⊙ yr⁻¹ for stars younger than 1 Myr (red diamonds), to ∼3.9 × 10⁻⁸ M_⊙ yr⁻¹ at ∼1–8 Myr (green diamonds), to ∼1.1 × 10⁻⁸ M_⊙ yr⁻¹ for stars with ages of ∼8–16 Myr (blue circles). It is also clear that the slope of the ${\dot{M}}_{\mathrm{acc}}$–M_⋆ relationship changes according to the stellar age, ranging from ∼0.0 for ages younger than 1 Myr, to ∼1.0 between 1 and 8 Myr, up to ∼4 for 8–16 Myr and older stars. Therefore, we conclude that attempting to define a relationship between ${\dot{M}}_{\mathrm{acc}}$ and M_⋆ without taking the age of the star into account can give spurious results and should be avoided.¹¹

Another result we would like to point out concerns the slope of our targets younger than 16 Myr. In Figure 11, we show with dashed lines the trends obtained at given ages (0.25, 0.50, 1, 2, 4, 8, and 16 Myr) considering the relationship between $\mathrm{log}{\dot{M}}_{\mathrm{acc}}$ and ($\mathrm{log}{M}_{\star }$, $\mathrm{log}t$) by De Marchi et al. (2017) and fixing as coefficients related to age and mass those obtained for the mass range 0.5–1.5 M_⊙, i.e., a = −0.59 and b = 0.78 (see their Equation (3)), and as constant c, mainly related to the metallicity, that obtained by the same authors for 30 Doradus in the LMC (i.e., c = −3.67; see their Table 1), which we assume to be at the same metallicity as LH 95. Considering our LH 95 targets with ages younger than 1 Myr and in the same mass range, the slope of the $\mathrm{log}{\dot{M}}_{\mathrm{acc}}$–$\mathrm{log}{M}_{\star }$ (∼1; solid line) is similar to that obtained by De Marchi et al. (2017) for 1 Myr stars in 30 Doradus. The slope of these targets is also similar to that found by Alcalá et al. (2017) for 0.5–1.5 M_⊙ stars in the Galactic Lupus SFR at ∼1–3 Myr (dot-dashed line). This latter qualitative comparison seems to suggest that Galactic and extragalactic SFRs share a similar slope of the $\mathrm{log}{\dot{M}}_{\mathrm{acc}}$–$\mathrm{log}{M}_{\star }$ relation. Moreover, this result also implies that the age is a parameter acting on the objects we can detect, as more massive targets reach the MS faster than lower-mass objects and consequently have lower levels of Hα emission, thus causing the fan shape of Figure 11. This again supports the intercorrelation between mass accretion rate, mass, and age at given surrounding environments.

Assuming that all stars in our sample formed under similar conditions, we can study the simultaneous dependence of ${\dot{M}}_{\mathrm{acc}}$ on both M_⋆ and t through a multivariate least-squares fit of the type ${\dot{M}}_{\mathrm{acc}}\propto {t}^{a}\times {M}_{\star }^{b}$. Adopting this simple relationship in the mass range 0.5–1.5 M_⊙ and for stars younger than 16 Myr, De Marchi et al. (2017) found a ∼ −0.6 and b ∼ 1.3 for ∼300 stars in 30 Doradus in the LMC. If we consider the 54 objects with these characteristics in our LH 95 sample, we find a ∼ −1.1 and b ∼ 1.3. We are indeed cautious about this result because of the relatively few number of targets. Therefore, we now compare the mass accretion properties of the stars in LH 95 with those in several SFRs of our Galaxy and the Magellanic Clouds obtained with the same method. This provides us quantitative information on the effects of the environment during the final stages of the star formation. Following the prescriptions given by De Marchi et al. (2017), it is convenient to use a power-law dependence on mass and age, like

$$\begin{eqnarray}&&\mathrm{log}{\dot{M}}_{\mathrm{acc}}=a\times \mathrm{log}\displaystyle \frac{t}{\mathrm{Myr}}+b\times \mathrm{log}\displaystyle \frac{{M}_{\star }}{{M}_{\odot }}+c,\end{eqnarray} \tag{ 6 }$$

where c reflects environmental effects, such as the metallicity, on the mass accretion rate. These authors studied a homogeneous sample of 1307 objects with 0.5–1.5 M_⊙ younger than 16 Myr in six regions of the MW, LMC, and SMC and analyzed them with the same method as our targets, finding a = −0.59 ± 0.02 and b = 0.78 ± 0.08, respectively. Using the same values of a and b for the PMS candidates in LH 95 with the same restriction in mass and age, we derive c = −3.54. This latter value is consistent with the results for the two clusters in the LMC analyzed by the authors (namely, 30 Doradus and the SN1987A field) and more generally with their relationship c = (−3.69 ± 0.02) − (0.30 ± 0.04) log Z/Z_⊙ obtained for the six clusters in the LMC, SMC, and MW (see their Table 1). In particular, our value is between −3.41, obtained in NGC 346 (a cluster in the SMC with Z ∼ 0.002), and −3.65, found for the MW clusters Trumpler 14 and NGC 3603 with Z ∼ Z_⊙. Even if we do not draw more quantitative conclusions here, our analysis confirms the importance of considering cluster metallicity, besides stellar mass and age, when mass accretion is studied in different environments. Clearly, other physical conditions (like mean gas density or local magnetic field) of the environment might have an effect on the extent and duration of the star formation process in general and the evolution of the mass accretion rate, as suggested by De Marchi et al. (2017).