A The resolution of X-Shooter

# X-Shooter spectroscopy of young stellar objects in Lupus.

Atmospheric parameters, membership and activity diagnostics12
###### Key Words.:
stars: pre-main sequence, fundamental parameters, chromospheres – accretion – open clusters and associations: Lupus
3

## Abstract

Context:

Aims:A homogeneous determination of basic stellar parameters of young stellar object (YSO) candidates is needed to confirm their pre-main sequence evolutionary stage, membership to star forming regions (SFRs), and to get reliable values of the quantities related to chromospheric activity and accretion.

Methods:We used the code ROTFIT and synthetic BT-Settl spectra for the determination of the atmospheric parameters ( and ), the veiling (), the radial (RV) and projected rotational velocity (), from X-Shooter spectra of 102 YSO candidates (95 of infrared Class II and seven Class III) in the Lupus SFR. The spectral subtraction of inactive templates, rotationally broadened to match the of the targets, enabled us to measure the line fluxes for several diagnostics of both chromospheric activity and accretion, such as H, H, Ca ii and Na i lines.

Results:We have shown that 13 candidates can be rejected as Lupus members based on their discrepant RV with respect to Lupus and/or the very low  values. At least 11 of them are background giants, two of which turned out to be lithium-rich giants. Regarding the members, we found that all Class III sources have H fluxes compatible with a pure chromospheric activity, while objects with disks lie mostly above the boundary between chromospheres and accretion. YSOs with transitional disks displays both high and low H fluxes. We found that the line fluxes per unit surface are tightly correlated with the accretion luminosity () derived from the Balmer continuum excess. This rules out that the relationships between and line luminosities found in previous works are simply due to calibration effects. We also found that the Ca ii-IRT flux ratio, , is always small, indicating an optically thick emission source. The latter can be identified with the accretion shock near the stellar photosphere. The Balmer decrement reaches instead, for several accretors, high values typical of optically thin emission, suggesting that the Balmer emission originates in different parts of the accretion funnels with a smaller optical depth.

Conclusions:

## 1 Introduction

A full characterization of members of young open clusters (OCs) and star forming regions (SFRs) is necessary to understand the mechanisms of star and planet formation, as well as the different phenomena occurring during the early phases of stellar evolution. It is therefore mandatory to derive, as homogeneously as possible, the basic physical parameters of young stellar objects (YSOs), such as effective temperature (), surface gravity (), mass and projected rotational velocity (). These quantities are closely related to the evolutionary stage of the YSO and are also directly involved in the generation of stellar magnetic fields by the dynamo processes. In the context of the magnetospheric accretion model, magnetic fields play a key role in the infall of matter from the disk onto the central star (Hartmann et al. 1994, 2016, and references therein). They are also responsible for the strong activity observed from the photospheres to the chromospheres and coronae of YSOs and late-type stars in general (e.g., Feigelson & Montmerle 1999; Berdyugina 2009, and references therein). Circumstellar disks, stellar rotation, magnetic activity and accretion are all strongly involved in the process of planet formation (e.g., Hawley & Balbus 1991; Brandenburg et al. 1995; Larson 2003, and references therein).

Deriving the basic stellar parameters is particularly hard for objects belonging to very young OCs and SFRs where several phenomena, such as fast rotation, cloud extinction and mass accretion, affect their spectra, making the analysis of their photospheres difficult. In particular, the veiling of the spectra, arising from accretion shocks and, possibly, from disk gas inside the dust sublimation radius, dilutes the photospheric lines and it must be properly taken into account to derive reliable stellar parameters (see, e.g., Frasca et al. 2015, and references therein). Moreover, emission lines, which are numerous and very strong and broad in the spectra of highly accreting objects, need to be masked, leaving only limited spectral ranges suitable for the analysis of photospheric lines.

On the other hand, it is well known that the luminosity of several emission lines of the Balmer series, the He i and Ca ii lines are well correlated with the accretion luminosity (e.g., Herczeg & Hillenbrand 2008; Alcalá et al. 2014, 2017, and references therein), as well as with hydrogen recombination lines in the near-infrared (Muzerolle et al. 1998; Calvet et al. 2004; Alcalá et al. 2017), highlighting the importance of these emission features as accretion diagnostics.

Spectra with both a very wide wavelength coverage and a high or intermediate resolution are particularly helpful, as they offer the possibility to find spectral intervals free from or little affected by these issues, where fully resolved absorption lines can be modeled to consistently derive atmospheric parameters, radial velocity, , and veiling. For this reason, the VLT/X-Shooter spectrograph (Vernet et al. 2011), with its wide wavelength coverage, is an ideal instrument to reach these goals.

Thanks to their relatively close distance to the Sun (150–200 pc, Comerón 2008) and the large numbers of Class II and Class III infrared sources they contain, the dark clouds in Lupus are among the best sites for which a spectroscopic determination of the atmospheric parameters for YSOs can be carried out down to the hydrogen burning mass limit. In addition, extinction to the individual YSOs is relatively low in comparison with many other SFRs.

The selection of candidates and the characterization of some of the confirmed Lupus members, both physical and kinematical, has been performed in recent years (see, e.g., Evans et al. 2009; Comerón et al. 2009, 2013; Galli et al. 2013; Herczeg & Hillenbrand 2014; Bustamante et al. 2015). However, most of these works are based on non-simultaneous optical and IR photometry and low-resolution spectroscopy. Moreover, the atmospheric parameters are derived mostly for the Class III sources. A homogeneous determination of the atmospheric parameters, , and veiling for a large sample of Class II sources in Lupus is still lacking. In this work we aimed at filling-in this gap. Our study is based on X-Shooter spectra of more than one hundred YSO candidates in the Lupus I, II, III, and IV clouds.

The high quality and the intermediate resolution of these spectra allowed us also to investigate the behavior of hydrogen, calcium and sodium emission lines that are diagnostics of both chromospheric and accretion activity.

The paper is organized as follows. In Sect. 2 we present the data sample. The analysis of the spectra for the determination of stellar parameters and the radial velocities is described in Sect. 3. The results are discussed in Sect. 4, where the properties of the targets, the membership to the Lupus SFR, and the line fluxes per unit surface are discussed. The main conclusions are summarized in Sect. 5.

## 2 Data sample

In total, we analyzed X-Shooter spectra of 102 objects, 43 of which were obtained in 2010, 2011, and 2012, during the INAF guaranteed time observations (GTO, Alcalá et al. 2011a). The remaining sources were observed in the framework of the ESO-P95 proposal from March to August 2015 (48 objects) and the ESO-P97 proposal from May to June 2016 (5 objects). In addition, X-Shooter spectra of six YSOs were retrieved from the ESO archive. Our sample comprises sources belonging to the infrared (IR) Class II. However, seven Class III objects were also included. Six of them have been already investigated by Stelzer et al. (2013) and Manara et al. (2013). The sample accounts for roughly half of the entire known population of YSOs in Lupus, but  95 % of disky YSOs. For the criteria of target selection and details about the data reduction the reader is referred to Alcalá et al. (2014, 2017).

We remark that 13 out of the 103 objects studied here were suspected to be non-members based on the appearance of their spectrum (Alcalá et al. 2017). The analysis we performed in this work allowed us to definitely reject them as members.

The target list is given in Table 1 together with the heliocentric Julian Date of observation and some of the parameters derived in this work. The spectra of Sz 68, Sz 74, Sz 83, Sz 102, Sz 105, and the archive data (i.e., IM Lup, Sz 77, EX Lup, GQ Lup, Sz 76, RXJ1556.1-3655) were acquired using the 05/04/04 slits in the UVB/VIS/NIR arms, respectively. All the other targets were instead observed with the 10/09/09 slits in the UVB/VIS/NIR arms, respectively.

For our analysis, particularly for the  measurement, it is important to know the actual spectral resolution. To this purpose we used Th-Ar calibration spectra taken during our observing runs. We found that the resolving power of VIS spectra taken with the 09 slit was in the spectral region used for the  determination ( 9700Å), while in the highest resolution mode (04 slit) it is about 17 000. More details can be found in Appendix A.

## 3 Atmospheric parameters, veiling, radial and projected rotational velocities

The determination of the atmospheric parameters ( and ), the veiling , the projected rotational velocity () and the radial velocity (RV) was accomplished through the code ROTFIT (e.g., Frasca et al. 2006, 2015).

We adopted as templates a grid of synthetic BT-Settl spectra (Allard et al. 2012) with a solar iron abundance and effective temperature in the range 2000–6000 K (in steps of 100 K) and from 5.5 to 0.5 dex (in steps of 0.5 dex). This grid covers the wide range in spectral type (SpT), from early-K to late-M, spanned by our targets and extends also to lower gravities typical of giants.

The code ROTFIT was already applied to X-Shooter spectra of Class III sources in Stelzer et al. (2013). However, the code was improved since then with the inclusion of further spectral segments to be analyzed both in the VIS and UVB spectra. Furthermore, in this version we implemented an accurate search for the minimum of the chi-square () in the space of parameters and the evaluation of the errors based on the 1 confidence levels. The veiling, , was also considered as a free parameter. Details on the application of the code to the X-Shooter spectra are given in Appendix B.

The RV was needed to align the synthetic template spectra with the observed ones. The peak of the cross-correlation function (CCF) was fitted with a Gaussian for a more accurate determination of its center (see the insets in Fig. 19). For each i spectral segment, the error of the value, , was estimated by the fitting procedure curvefit (Bevington 1969), taking into account the CCF noise, which was evaluated as the standard deviation of the CCF values outside the peak. Then, the heliocentric correction of the RV measurements was performed with the IDL procedure helcorr. The final RV for each star, reported in Table 1, is the weighted mean of the values obtained from all analyzed spectral segments, using as weights . The standard error of the weighted mean was adopted as the RV uncertainty. These errors range from 0.7 to 12 km s, with a median value of 2.3 km s. We note that, as expected, the RV errors tend to be larger for the stars with higher and/or with low S/N spectra, or strongly affected by veiling (see Table 1).

### 3.2 Accuracy of Teff and logg determinations and comparisons with the literature

The  errors reported in Table 1 range from about 30 to 250 K, depending on  (and on the S/N), with a median value of 70 K. The only exception is SSTc2dJ161045.4-385455 ( = 514 K), a likely non-member with a low-signal spectrum.

To estimate the external accuracy of , we have compared our values with those derived from the spectral classification performed by Manara et al. (2013) for the Class III objects and by Alcalá et al. (2014, 2017) for the Class II sources, who adopted the temperature scales of Luhman et al. (2003) and Kenyon & Hartmannn (1995) for the M- and K-type stars, respectively. The comparison is shown in Fig. 1. The general agreement of the two  determinations is evident, with a scatter of the residuals 100 K rms. Nevertheless, this comparison displays a systematic offset between the two  determinations, which is more evident in the M-type domain, with the  values derived from SpT being about 40 K higher, on average, than those from ROTFIT. This offset is likely due to the different  scales between the BT-Settl models and the empirical SpT– calibration defined for M-type stars by Luhman et al. (2003). We note that the  values of Luhman et al. (2003) for early M-type stars are systematically higher by  70 K than those reported by Pecaut & Mamajek (2013) for young stars, which are based on the fitting of the optical/IR spectral energy distributions.

As a further check for the accuracy of our determinations we have searched for spectroscopic  measurements in the literature, not based on X-Shooter data. We compared our results with the values reported by Comerón et al. (2013) and Herczeg & Hillenbrand (2014). Both works are based on low resolution (–1000) optical spectra. We have only one star in common with Comerón et al. (2013), namely SSTc2d160836.2-392302, for which the  agreement is excellent (see Fig. 2). Twenty of our targets have been observed by Herczeg & Hillenbrand (2014), who list the spectral type derived from suitable spectral indices. We converted their SpT into  by using the calibration put forward by them. The comparison is displayed in Fig. 2. The agreement is evident, with an average offset of +28 K and a scatter of about 100 K, as in Fig. 1. The latter can be considered as the typical external accuracy of our  values. We note that, despite the lower number of points, the agreement in the M-type domain looks better than the one shown in Fig. 1. This is likely owing to the  scale, based on BT-Settl spectra, adopted both in the present work and in Herczeg & Hillenbrand (2014).

Concerning , the errors are in the range 0.1–0.5 dex, with a median value of 0.21 dex. Only four objects have larger errors (up to 0.9 dex). We did not find other estimates of gravity in the literature, except for the value of =4.0 reported by Comerón et al. (2013) for SSTc2d160836.2-392302, which is consistent with ours (4.040.13).

### 3.3 Projected rotational velocity

The intermediate resolution of X-Shooter allows us to measure  only for relatively fast rotators. To check the minimum  that can be measured in these spectra and to estimate the  errors, we have run Monte Carlo simulations on synthetic spectra with the same resolution and sampling of X-Shooter, following the prescriptions given in Frasca et al. (2015). We found that the minimum detectable  with the 09 slit is about 8 km s. Therefore, we considered as upper limits all  values lower than 8 km s. We also evaluated the upper limit on  for the spectra acquired with the 04 slit, finding a value of 6 km s. Details about the Monte Carlo simulations can be found in Appendix C. The errors of  range from 1 to 15 km s(Table 1), with a median value of about 4 km s.

Determinations of  can be found in the literature only for a handful of the brightest sources in our sample. For Sz 68, Torres et al. (2006) report a value of  km s, which is compatible with our determination of 39.6 km s. Dubath et al. (1996) give only a lower limit of  km s for Sz 121, which is one of the most rapidly rotating stars in our sample ( km s). The same authors report  km s for Sz 108, for which we find an upper limit of 8 km s. The value of 22.4 km s quoted in the Catalog of Stellar Rotational Velocities (Glebocki & Gnacinski 2005) for RY Lup is marginally consistent with ours ( km s). Finally, in the same catalog,  km s is reported for Sz 98, while we find  km s.

### 3.4 Veiling

After several tests made with our code on Class III and X-Shooter archive spectra of non-accreting stars, we realized that values of veiling as large as 0.2 can be found, although no veiling is expected in such cases. This is probably a combined effect of the intermediate spectral resolution, small differences in the continuum setting between target and template spectra, and some trade-off between parameters. For this reason, all values are considered as not significant and have been replaced with an upper limit of 0.2 in Table 1. We measured significant values of veiling () for 12, 19, and 44 YSOs at 970 nm, 710 nm, and 620 nm, respectively. This trend is in line with the enhancement of with decreasing wavelength (see Fig. 22).

Herczeg & Hillenbrand (2014) report values of the veiling at 750 nm for sixteen of our targets. For about a half of them we derived only upper limits, , while for the remaining YSOs the veiling is always rather small. However, the comparison of these data shows that we can measure the veiling at 710 nm whenever Herczeg & Hillenbrand (2014) quote .

Using the mass accretion rates derived by Alcalá et al. (2017) on the same data set, we found that high values of veiling () are measured only for the strongest accretors (, see Fig. 3).

## 4 Results

### 4.1 Membership

The atmospheric parameters, the RV, and the equivalent width of the lithium line at 6707.8 Å () were used to verify the membership of our targets to the Lupus SFR. Here we used the values derived by Biazzo et al. (2016).

The  values derived with ROTFIT allow us to identify in our sample contaminants whose IR colors mimic those of cool very young stars. In Fig. 4 we show the - diagram of our targets, where the stars suspected to be non-members have been highlighted with green asterisks. We also plot the pre-main sequence (PMS) isochrones of Baraffe et al. (2015) and the post-main sequence isochrones from the PARSEC database (Bressan et al. 2012).

All green asterisks at  K have and correspond most likely to giants. Four of these are indeed in the red giant branch (RGB) and possibly close to the red giant clump; a few others, with lower gravity, could be in the AGB phase. The star with the highest temperature, SSTc2dJ161148.7-381758, could be a MS star unrelated to the Lupus SFR, as indicated by its RV= km s. The green asterisk with the largest errors located at =5114 K and =3.4 corresponds to SSTc2dJ161045.4-385455. Its position on the - plane may be compatible with the PMS locus, taking into account the errors, but the value of RV= km s strongly suggests that it is not a member of the Lupus SFR.

The other two indicators of membership, namely the RV and , are plotted in Fig. 5 with the same symbols as in Fig. 4. It is worth noting that, except for two objects (SSTc2dJ160708.6-394723 and SSTc2dJ161045.4-385455), all the stars classified as non-members based on the - diagram and/or discrepant RV, have undetectable Li i lines. The non-member with the highest (364 mÅ) is SSTc2dJ160708.6-394723. For this object we estimate  K and , thus it lies in the RGB region of the - diagram (Fig. 4). Its RV is inconsistent with the Lupus SFR. Likewise, SSTc2dJ161045.4-385455 has a rather large (about 300 mÅ), but, as already mentioned, its RV =  km s and the absence of emission lines rule out its membership to the Lupus SFR. Although with large errors, its atmospheric parameters suggest a G-type giant or subgiant. These two stars are most likely lithium-rich giants. More details are given in Appendix D.

Three other non-members based on the - diagram (Sz 78, Sz 105, and SSTc2dJ161222.7-371328) have instead RVs compatible with Lupus, but no Lithium. These stars have all very low  values of 2.04, 0.76, and 0.99, respectively, thus ruling out a PMS nature.

In conclusion, by using the three indicators, , RV, and , we were able to reject 13 stars as Lupus members. All the remaining 89 targets were considered as members, although a few of them have an RV slightly outside the SFR average derived by us ( km s) and external to the RV distributions for on-cloud and off-cloud sources defined by Galli et al. (2013). Nonetheless, they display a large that reinforces their membership; they could be spectroscopic binaries.

A small group of six Class II sources, namely Sz 69, Lup706, Par-Lup3-4, 2MASS J16085953-3856275, 2MASS J16085373-3914367, and SSTc2dJ154508.9-341734, has instead an RV well consistent with the Lupus SFR but a quite low lithium content ( mÅ). Nevertheless, the spectra of these objects display other features that indicate their membership (c.f. strong and wide emission lines, veiling, etc., Alcalá et al. 2014, 2017). These cases will be discussed in more detail in Biazzo et al. (2016). Finally, the star Sz 94, classified as a Class III source with proper motions and RV compatible with Lupus, shows , making its membership dubious.

### 4.2 Hertzsprung-Russell diagram

In Alcalá et al. (2017) we derived masses using four different evolutionary tracks, but only for the Class II sources. Here we derive masses and ages for both the Class III and Class II sources homogeneously, and using the  values resulting from the ROTFIT analysis.

In Fig. 6, we report the position of our targets in the Hertzsprung-Russell (HR) diagram, where we used the  values derived in the present work and the luminosities calculated by Manara et al. (2013) and Alcalá et al. (2014, 2017), with the same method, for the Class III and Class II sources, respectively. In the same figure we overplot the PMS evolutionary tracks and isochrones by Baraffe et al. (2015).

Most of the targets are located between the isochrones at 1 and 10 Myr, while a few of them (Lup706, Sz106, Par-Lup3-4, Sz123B, Sz133, SSTc2dJ160703.9-391112, and Sz102), displayed with larger open circles in Fig. 6, appear to be subluminous. These are likely objects with rotation axes perpendicular to the line of sight where the edge-on disks reduce the stellar luminosity by extinction and light scattering (Alcalá et al. 2014, 2017). Several of these objects show in their optical spectra an enhancement of outflow tracers like the low-velocity component of the O i 6300 Å line (Bacciotti et al. 2011; Natta et al. 2014; Whelan et al. 2014). This line originates in a more extended volume with respect to the accretion tracers that arise from a much closer region to the stellar surface, and are suppressed by the optically thick edge-on disk. This effect is particularly important in Par-Lup3-4, Sz102 and Sz133 (Bacciotti et al. 2011, Natta et al. 2014, Nisini et al. in prep.). Another object, namely SSTc2dJ160708.6-391408, classified as a flat IR source (Merín et al. 2008) and shown with a green filled square in Fig. 6, appears subluminous because the derived stellar luminosity does not account for reprocessed stellar radiation in the infalling envelope of gas and dust (see Appendix C in Alcalá et al. 2017). This object is further discussed in Appendix D.

The evolutionary tracks and isochrones can be used to estimate the masses and ages of the targets from their location in the HR diagram by minimizing the quantity:

 χ2=(Teff−Tmod)2σ2Teff+(L−Lmod)2σ2L, (1)

where and are the stellar effective temperature and bolometric luminosity, respectively, with errors and , respectively. The effective temperature and stellar luminosity of the evolutionary tracks are denoted with and , respectively. We assigned to each YSO the mass and age corresponding to the closest track and isochrone, which minimize .

The same procedure was used for evaluating masses and ages with the Siess et al. (2000) evolutionary models that do not cover the very low-mass regime but extend to higher masses. For the stars in the mass range 0.1–1.4  we found an excellent agreement between the masses derived with the two sets of tracks, with differences within 0.05  rms. Masses and ages, with the exception of the underluminous sources, are reported in Table 1. For the most massive star, SSTc2dJ160830.7-382827, whose position in the HR diagram is outside the mass range covered by the Baraffe et al. (2015) tracks, we adopted the Siess et al. (2000) models and we found =1.8 and  Myr.

The mass determinations are in very close agreement with those of Alcalá et al. (2017). As stressed in previous works, the age determination suffers from uncertainties due to the error on the distance of the clouds, source variability, extinction and depends on the adopted set of evolutionary tracks (see, e.g., Comerón 2008, and references therein). Therefore, the ages of the individual sources are affected by large uncertainties. However, the median age of the full sample analyzed in the present work is 2 Myr and the average value is 2.5 Myr, in good agreement with previous determinations (e.g., Alcalá et al. 2014, 2017, and references therein).

### 4.3 Activity/accretion diagnostics

The emission lines comprised in the X-Shooter spectra have been used as diagnostics of chromospheric activity and accretion in previous works of our group (e.g., Stelzer et al. 2013; Alcalá et al. 2014). We have shown how the exceptional richness of information of these spectra allows us to probe the different layers of the atmospheres of PMS stars and to define the “noise” that chromospheric emission introduces into mass accretion estimates (Manara et al. 2013). These spectra allowed us to considerably improve the accuracy of the determination of mass accretion with important implications on the  relationship (Alcalá et al. 2014, 2017).

Here we give a closer look at the H, Ca ii K, Ca ii infrared triplet (IRT), and Na i D. These lines, particularly H and Ca ii lines, have been widely used as diagnostics for both chromospheric activity (e.g., Strassmeier et al. 2000; Montes et al. 2001; Martínez-Arnáiz et al. 2011; Frasca et al. 2016, and reference therein) and accretion (e.g., Muzerolle et al. 2003; Mohanty et al. 2005; Costigan et al. 2012; Alcalá et al. 2014) due to their intensity and the high sensitivity of the CCD detectors at these wavelengths. The cores of the Na i D lines are good diagnostics of chromospheric activity (especially in late-K and M-type stars, e.g., Houdebine et al. 2009; Gomes da Silva et al. 2011), mass accretion (e.g., Rigliaco et al. 2012; Alcalá et al. 2014, 2017) and winds (e.g., Natta & Giovanardi 1990; Facchini et al. 2016).

We calculated the equivalent widths (EWs) and line fluxes by using the spectral subtraction method (see, e.g., Frasca & Catalano 1994; Montes et al. 1995) to remove the photospheric flux and to evidence the core emission. Although this correction is negligible for objects with high accretion rates and very strong emission lines, it is absolutely needed in cases where the photospheric profile is only filled-in with emission or the latter is very weak.

Figure 7 shows the result of the spectral subtraction on the H line of SSTc2dJ160830.7-382827, which is a target with a transitional disk and a weak accretor (Alcalá et al. 2017). In this case the H line displays an absorption profile filled-in with emission, which is only revealed by subtracting the inactive template. The latter has been obtained by the interpolation of the best-fitting BT-Settl spectra at the  and  of the target that are quoted in Table 1.

The inactive templates are rotationally broadened, whenever  is larger than the minimum value that can be resolved by the observations (see Sect. 3.3), Doppler-shifted, and resampled on the spectral points of the target spectra before subtraction. The residual H profile integrated over wavelength (hatched green area in Fig. 7) provides us with the net H EW.

An example of the application of the spectral subtraction method to the Ca ii IRT lines is shown in Fig. 8 for the star Sz 103. We note that the subtraction of the inactive template is necessary for a correct measurement of the Ca ii IRT EWs. In particular, the Ca ii 8662 line shows an emission in its core that does not reach the continuum.

The result of the spectral subtraction in the region of Na i D and He i D lines for Sz 97 is shown in Fig. 9. It is clear that, although the underlying photospheric spectrum barely affects the strong He i D line, it must be subtracted from the target spectrum to get the emission EWs in the Na i D line cores. We stress that the X-Shooter spectra were corrected for telluric absorption lines also at these wavelengths, as described by Alcalá et al. (2014), and the interstellar Na i absorption affects significantly only the more distant non-members (giants).

We converted the observed EWs into line fluxes per unit surface by multiplying them for the continuum flux of the synthetic spectrum corresponding to the  and  of the target. To get a more accurate value of the continuum flux, we interpolated in  and  within the grid of BT-Settl spectra. The fluxes of H, H, Ca ii K, Ca ii IRT, and Na i D lines are quoted in Table 2. In this table we do not report any flux value whenever the spectral subtraction does not give an emission residual profile. We treat as upper limits the values for which the error is larger than the flux.

We have already shown in Stelzer et al. (2013) the agreement between line fluxes per unit surface calculated on the basis of model spectra, as in the present paper, and those derived from the observed flux at Earth and the dilution factor, , where and are the stellar radius and distance, respectively.

We also calculated the ratio of the line and bolometric flux, as a further chromospheric/accretion diagnostic4.

For the stars with H line in emission above the local continuum, we also measured the full width at 10% of the line peak (). This quantity is helpful for discriminating between accreting stars and those with pure chromospheric emission. In particular, we defined “candidate accretors” the objects that fulfill the criterion of White & Basri (2003) that is based on both a km s and larger than given thresholds depending on spectral type, and distinguished them in the following plots with half-filled symbols.

In Fig. 10 we plot the H flux, , and the ratio as a function of the effective temperature. In this figure, the boundary between the accreting objects and the chromospherically active stars, as defined by Frasca et al. (2015) with the same criteria for the members of the Gamma Vel cluster and Cha I SFR, is also marked. We note that all Class III stars lie below the aforementioned boundary and all “candidate accretors” (half-filled symbols) are located above it, while the other Class II sources are scattered around, and mostly below, this dividing line. In particular, 17 Class II sources are located below the dividing line. However, this border cannot be considered as a knife edge that separates accretors from chromospheric sources; therefore only the objects lying well below it are noteworthy. For instance, if we select only the YSOs with , where most of the Class III sources lie, we end up with only eight Class II sources, three of which (Lup 607, MY Lup, and SSTc2dJ160830.7-382827) are those considered as “weak accretors” by Alcalá et al. (2017). The other five objects, namely SSTc2dJ160000.6-422158, SSTc2dJ155925.2-423507, Sz 95, SSTc2dJ161029.6-392215, and SSTc2dJ160703.9-391112 have rather low accretion rates according to Alcalá et al. (2017). The low H flux could indicate that these objects have still rather dense dusty disks that give rise to the IR excess, but a low or moderate accretion rate for which the H flux is comparable with that of chromospheric sources.

The presence above the dividing line of some Class II object with a substantial accretion rate (Alcalá et al. 2017) and not satisfying the White & Basri criterion shows that the latter is not 100 % reliable in selecting accretors.

We note that the objects with transitional disks (TD), shown as green open diamonds in Fig. 10, are spread all over the diagram and no clear behavior appears, in agreement with the finding by Alcalá et al. (2017) for the mass accretion rate.

The H and Ca ii IRT fluxes are well correlated, as apparent in Fig. 11 and suggested by the high value of Spearman’s rank correlation coefficients, which range from to with a significance in the range (Press et al. 1992), for the three Ca ii lines. Least-squares regressions provide the following relations:

 logFHα = 2.35(±0.32)+0.83(±0.07)⋅logFCaII8498 logFHα = 2.23(±0.29)+0.84(±0.06)⋅logFCaII8542 logFHα = 2.38(±0.27)+0.82(±0.07)⋅logFCaII8662. (2)

It is worth noting that most of the sources display H fluxes in excess with respect to the mean flux-flux relations found by Stelzer et al. (2013) for Class III objects in Lupus, TWA, and  Ori SFRs. This is particularly evident for the “candidate accretors” (half-filled symbols) and suggests that the accretion produces a larger H luminosity, compared to Ca ii IRT lines, than that originating from chromospheres. This trend can be explained by the fact that strong accretors generally have also outflows or stronger winds, which contribute to the H emission but do not significantly affect Ca ii lines.

A similar behavior is found when plotting the H versus the Ca ii K flux, as displayed in Fig. 12. These data can be fitted by the following linear relation:

 logFHα = 1.04(±0.41)+1.01(±0.07)⋅logFCaIIK (3)

The H and Na i D fluxes display a correlation (, ), as shown in Fig. 13. We remark that is the sum of the fluxes in the two sodium D lines. A least-squares regression provides the following relation:

 logFHα=2.67(±0.38)+0.82(±0.09)⋅logFNaID. (4)

We have also applied the subtraction method to the strongest emission lines in the NIR spectra that are not severely affected by telluric absorption, namely the He i10830 Å, Paschen (Pa and Pa) and Brackett (Br) lines. The He i10830 Å profile often displays excess absorption components that are related to stellar and/or disk winds (e.g., Kurosawa et al. 2011). Therefore, in these cases the spectral subtraction gives rise to extra absorption. We have, therefore, discarded this line, whose study is deferred to a subsequent work.

The fluxes of the three NIR lines, namely the Paschen and lines and the Brackett line, have been measured for most of the targets (Table 3). The fluxes of these lines are also well correlated with the H flux, as shown in Fig. 14. Least-squares regressions provide the following relations:

 logFHα = 2.52(±0.32)+0.80(±0.08)⋅logFPaγ logFHα = 2.00(±0.27)+0.86(±0.07)⋅logFPaβ logFHα = 2.38(±0.39)+0.93(±0.10)⋅logFBrγ. (5)

### 4.4 Line fluxes and accretion

Line fluxes are also very useful for evaluating the mass accretion rate and for comparing it with the one derived from primary indicators, like the Balmer continuum excess. Alcalá et al. (2017) calculated the accretion luminosity, , by modelling the observed flux-calibrated spectra with the sum of a non-accreting (Class III) template and the emission from a slab of hydrogen, which accounts for . The only parameters needed to convert the flux at Earth into luminosity are the distance to the YSO and the extinction.

Mendigutía et al. (2015) suggested that the relations between accretion and line luminosity, , may be the result of the and correlations. Indeed, both and are obtained adopting the same distance and extinction. Therefore, in their hypothesis, the observed trends may be a “calibration” effect rather than physical correlations.

In the present work we have calculated the line fluxes per unit surface based on the continuum fluxes of the model spectra at the  and  of the target. Thus, neither the distance nor the extinction enter in the determination of .

We show in Fig. 15 that line fluxes are highly correlated with the accretion luminosity reported by Alcalá et al. (2017), as also witnessed by the rank-correlation coefficients ranging from 0.59 to 0.80 with significances always smaller than . This excludes a calibration effect and strengthens the validity of the emission lines as accretion diagnostics.

### 4.5 Flux ratios

The ratio of fluxes in two Ca ii-IRT lines, , is sensitive to the conditions of the emitting plasma, particularly to its optical depth. It was shown that for optically thick sources, such as dense chromospheric solar or stellar plages, the ratio is typically in the range 1–2; optically-thin emission sources, like solar prominences seen off-limb, have instead large values of this ratio,  4–9 (see, e.g., Herbig & Soderblom 1980; Landman 1980, and references therein). As apparent in Fig. 16, the values of this flux ratio are low for all Lupus members in our sample, both Class II and III. This suggests that the bulk of emission is originating in optically-thick regions, either chromospheric plages or the impact regions of accretion flows near the YSOs surface, as already suggested, e.g., by Herbig & Soderblom (1980) and Alcalá et al. (2014).

Another indicator of the physical conditions of the emitting matter is the Balmer decrement, which is here defined as the ratio of the H and H fluxes. We calculated the H flux at the stellar surface in the same way as the H flux. The Balmer decrement is displayed in Fig. 17. We note that all Class III sources have low values of , which range from about 1.5 to 3. These values are, on average, slightly larger than those typical of solar plages or pre-flare active regions (1–2), but significantly lower than in optically thin regions such as solar prominences observed off limb (e.g., Tandberg-Hanssen 1967; Landman & Mongillo 1979). This behavior was already observed in stars with a very high chromospheric activity level and explained as the result of different physical conditions in the chromospheres or of a combination of plage-like and prominence-like regions in the observed flux (e.g., Hall & Ramsey 1992; Chester et al. 1994; Stelzer et al. 2013).

Several accretors share the same locus of the Class III sources in this diagram, but some others (about one third) display very high values of the Balmer decrement, which indicates a prevalence of optically thin emission in their Balmer lines, similarly to results from other studies of accretors (e.g., Whelan et al. 2014; Frasca et al. 2015; Antoniucci et al. 2017).

This suggests that Ca ii and Balmer lines carry informations about different effects/regions of the accretion process, the former mainly originating in optically thick regions, like, e.g., the impact shocks produced by the accretion flows near the stellar photosphere. The Balmer lines can be formed instead in different parts of the accretion funnels with a large range of optical depth.

We note that the subluminous YSOs, with the exception of SSTc2dJ160703.9-391112, Sz 106 and Sz 102, display large values of the Balmer decrement. This can be understood if we assume that an edge-on disk is the cause of the under-luminosity. In this framework, the densest, and thus optically thickest, regions of the accretion flows are suppressed by the optically thick edge-on disk. Moreover, Antoniucci et al. (2017) report that the whole Balmer series decrement shows a common shape (type 1 in their classification) in most subluminous objects, which is compatible with an edge-on geometry. The rather low values of Balmer decrement for SSTc2dJ160703.9-391112, Sz 106 and Sz 102 may be the result of the geometry of the star/disk system and/or of the strong mass accretion rate, which gives rise to dense flows at rather large distances from the central star.

## 5 Summary and conclusions

We have presented the results of the analysis of X-Shooter@VLT spectra of 102 YSO candidates, mostly of infrared Class II, in the Lupus star forming region. The application of the code ROTFIT to specific spectral regions at optical and near-infrared wavelengths has allowed us to measure the effective temperature (), surface gravity (), radial velocity (RV) and projected rotational velocity () with median errors of  K,  dex,  km s, and  km s, respectively. By means of Monte Carlo simulations, we have estimated the minimum detectable  as 8 km s with the 09 VIS slit and 6 km s with the 04 VIS slit, respectively. The comparison with the effective temperatures inferred from the spectral classification performed on the same spectra by Manara et al. (2013) and Alcalá et al. (2014, 2017) and with literature data stemming from low-resolution spectroscopy indicates an external accuracy of about 100 K. A similar statistical evaluation cannot be done for the accuracy of  and , because of the very few values of these parameters available in the literature, which are, however, in good agreement with our determinations. We have also derived the veiling at different wavelengths and confirm previous results that the strongest accretors possess the highest veiling values.

We have shown that 13 candidate members of the Lupus SFR must be definitely considered as non-members on the basis of their discrepant RV with respect to the Lupus SFR and/or the very low  values, which suggest that at least 11 of them are background giants in the RGB or AGB phase. Two of them turned out to be lithium-rich giants.

Thanks to the subtraction of inactive template spectra, we have measured the line fluxes corrected for the contribution of the photospheric absorption, even in the objects with the weakest chromospheric or accretion emission. We found that all the Class III sources lie, as expected, in the locus of the - diagram occupied by the objects where chromospheric emission dominates over accretion as defined by Frasca et al. (2015). The Class II sources are mostly above the boundary between chromospheres and accretion, but some of them are also in the domain of the chromospheric sources. However, all the objects fulfilling the White & Basri (2003) criterion, based on the 10% width and EW of the H emission line, lie, correctly, above the aforementioned boundary, i.e. in the domain of accretors according to Frasca et al. (2015). We do not see any particular trend for the YSOs with transitional disks in this diagram. They display both low and high fluxes similarly to the behavior of their mass accretion rate shown by Alcalá et al. (2017).

The strong correlation between our line fluxes per unit surface and the accretion luminosity () derived by Alcalá et al. (2017) from the Balmer continuum excess has allowed us to confirm the validity of emission lines as accretion diagnostics and to exclude that the relations between and line luminosities are only the result of calibration effects.

We have also investigated the relations between the H flux and the fluxes in different emission lines (Ca ii IRT, Ca ii K, Na i D, Pa, Pa, and Br). We note that, with the exception of the Class III sources and some of the weak accretors, all the objects display an H flux in excess to what it is expected from the relations found by Stelzer et al. (2013) for Class III sources. This behavior is likely the result of different physical conditions in the emitting regions of the accretors with respect to chromospheric emitters, which seem to also affect the flux ratios. The latter have been calculated for two Ca ii-IRT lines, , and for the first two members of the Balmer series, , i.e. the Balmer decrement. We found that the Ca ii-IRT flux ratio takes always small values () that indicate an optically thick emission source. The latter can be identified with the accretion shock near the stellar photosphere. The Balmer decrement takes instead, for several accretors, high values typical of optically thin emission, suggesting that the Balmer emission originates in different parts of the accretion funnels with a smaller optical depth.

###### Acknowledgements.
We are very grateful to the referee, Fernando Comerón, for his useful comments and suggestions. We thank the ESO staff, in particular Markus Wittkowski and Giacomo Beccari for their excellent support during phase-2 proposal preparation, and the Paranal staff for their support during the observations. We thank G. Cupani, V. D’Odorico, P. Goldoni and A. Modigliani for their help with the X-Shooter pipeline. Financial support from INAF is also acknowledged. CMF acknowledges an ESA Research Fellowship funding. This research made use of the SIMBAD database, operated at the CDS (Strasbourg, France).