The He I $\lambda$10830 Å line as a probe of winds and accretion
in young stars in Lupus and Upper Scorpius Thanks: This work is based on observations made with ESO telescopes at the La Silla Paranal Observatory under programme ID 084.C-0269, 084.C-1095, 086.C-0173, 087.C-0244, 089.C-0143, 085.C-0238, 085.C-0764, 089.C-0840, 090.C-0253, 093.C-0658, 094.C-0913, 095.C-0134, 095.C-0378, 097.C-0349, 097.C-0378, and 0101.C-0866, and archive data of programmes 085.C-0764 and 093.C-0506.

Our sample consists of a total of 117 young stellar objects - 82 Class II sources are located in the Lupus star forming region (d $\simeq$ 160 pc; Alcalá et al. 2014; Alcalá et al. 2017) and a further 35 sources in Upper Scorpius (d $\simeq$ 145 pc; Manara et al. 2020). The Lupus sample is younger (1 – 3 Myr; Comerón 2008) compared to the Upper Scorpius sample (5 – 10 Myr; Luhman & Mamajek 2012), allowing us to investigate the evolution of the He I $\lambda$10830 Å profile for different ages. The targets in our sample have spectral types between K0 and M8.5, with stellar masses in the range of $M_{*}$$\simeq$ 0.1 - 1.6 $M_{\odot}$(see Figures 1 and 2), allowing us to investigate the properties of the He I $\lambda$10830 Å line in relation to varying stellar properties such as spectral type, accretion luminosity, among others. We include information on the spectral types, disk inclination, and accretion properties for the individual targets in Tables A.1 and A.2, however the sample is described in depth in Alcalá et al. (2014); Alcalá et al. (2017) and Manara et al. (2020).

All the observations included in this work have been obtained with the ESO/VLT X-Shooter spectrograph (Vernet et al. 2011). This medium-resolution and high-sensitivity instrument simultaneously covers the wavelength range between $\sim$ 300 nm and $\sim$ 2500 nm. The spectra are divided in three arms which cover three different wavelength ranges. The helium line analysed here is located in the NIR arm, with resolution R $\sim$10500-5300 depending on the slit width. The spectra used in this analysis have been reduced with the procedure described by Alcalá et al. (2014), and the observational details are listed in Alcalá et al. (2014); Alcalá et al. (2017) and Manara et al. (2020) depending on the targets, as discussed in the following section.

The data were initially reduced with the X-Shooter pipeline (Modigliani et al. 2010) using the ESO Reflex workflow. This process is described in detail in previous works where the same data reduction steps are taken, for example Alcalá et al. (2017) and Manara et al. (2020). The telluric correction procedures for both regions are also described in these papers. Telluric correction for the Lupus sample was performed using two separate methods on the NIR and VIS arms (see Appendix A in Alcalá et al. 2014, for detail). The telluric correction for Upper Sco sources was carried out using molecfit (Smette et al. 2015; Kausch et al. 2015) as described by Manara et al. (2020). Radial velocity (from Frasca et al. 2017; Manara et al. 2020) and heliocentric velocity corrections are applied to both the NIR and VIS spectra. To correct for wavelength shifts between the two arms, we used a 1D Gaussian fit to measure the centroid of the Lithium 670.78 nm absorption line and the Pa$\delta$ 1004.94 nm line (observed in both the NIR and VIS spectra). We first fitted the Li 670.78 nm line with a 1D Gaussian, finding median Lithium shifts of $\approx$-0.5 km s^-1  and -0.4 km s^-1  for Lupus and Upper Sco. The Li shift is applied to both NIR and VIS spectra, such that the lithium line is centred on 0 km s^-1. We then corrected for wavelength shifts between the NIR and VIS spectra using the Pa$\delta$ line present in both arms. We found median Pa$\delta$ shifts of $\approx$ 9 km s^-1  and -2.96 km s^-1  for the Lupus and Upper Sco samples (respectively), which is then applied to the NIR spectrum. In a few cases where the signal-to-noise was too low for the lines to be properly fitted, we applied the median shifts listed above to these spectra. We find typical 1$\sigma$ errors on the Li and Pa$\delta$ velocity corrections of 2.96 km s^-1and 2.24 km s^-1, respectively. Adding these in quadrature, the combined error due to the velocity corrections is 3.7 km s^-1and thus do not introduce large uncertainties in the observed maximum velocities in the He I $\lambda$10830 Å line (see Section 4). We describe the process of aligning the NIR and VIS spectra in more detail in Appendix B.

We have gathered further information on our sample for the analyses presented later in this paper. We use the stellar mass and luminosities, and accretion properties ($\dot{M}_{\rm acc}$ and L_acc) from Alcalá et al. (2019) for the Lupus sources, and from Manara et al. (2020) for Upper Scorpius sources, with distances for the individual sources from DR2 of the Gaia Collaboration (2018). The disk inclinations are known for 103 (88%) of the 117 sources in our sample (85% in Lupus, 94% in Upper Sco) and were measured from ALMA continuum observations for the Lupus sample by Tazzari et al. (2017), Ansdell et al. (2018), and Yen et al. (2018) and for Upper Scorpius by Barenfeld et al. (2017). We note that these are measurements of the outer disk inclinations, whereas the He I $\lambda$10830 Å line traces gas close to the star. Since it is possible that there is a misalignment between the inner and outer disk (Bohn et al. 2022, e.g.) this possible issue will be considered when the results are analysed.

We examined the He I $\lambda$10830 Å line profile in all of our targets using scipy routines (Virtanen et al. 2020) in Python 3. We located the positions of all minima and maxima in the spectra between $\pm$ 400 km s^-1with the find_peaks routine, considering only those extrema with absolute intensities larger than twice the rms to avoid measuring noise. Using the location (in velocity) of the extrema and their height relative to the continuum (i.e. absorption or emission), we can distinguish between various profile types, described in further detail in Section 3.2. We used this initial information on the extrema and profile types to provide initial guesses on the peak parameters (amplitude, centroid in km s^-1and standard deviation in km s^-1) for Gaussian fitting. Depending on the profile type, between one to four 1D Gaussians were fitted to the He I $\lambda$10830 Å profiles using scipy’s curve_fit routine, by adding the suitable number of 1D Gaussian functions defined as:

where a is the amplitude, $\mu$ is the centroid and $\sigma$ is the standard deviation of the Gaussian peak.

We measured the $\chi^{2}$ to identify poor fits (i.e. fits with $\chi^{2}$ > 0.5) which we fitted again manually using new initial guesses for the peak parameters and profile type. We also visually checked each fit to ensure the profile is well-fitted (see Figures 18, 19 and 20 for each fitted profile). Out of the 82 sources in Lupus, 30 sources could not be automatically fitted and required manual inputs for a good fit. These profiles generally were categorised into the low-velocity absorption or miscellaneous profile types. Additionally, some of the profile types for these sources were identified incorrectly, therefore they required manual inputs and bounds to accurately fit the He I $\lambda$10830 Å line profile. These sources are listed in Table 1. Only one He I $\lambda$10830 Å profile in Lupus (2MASSJ16085373-3914367) could not be fitted due to low signal-to-noise and/or bad pixels. In Upper Scorpius, all sources could be fitted using a combination of 1D Gaussians, with 15 sources requiring manual inputs to be accurately fitted - these are listed in Table 2.

We identified eight distinct profile types in the data. The conceptual line profiles are presented in the left column of Figure 3 and were created by adding between one to four one-dimensional Gaussians to the continuum, which is set at zero for all profiles. The profile types are as follows:

1. (a)

   Pure emission profiles are characterised by a single peak in emission. There are no sub-continuum absorption features present in these profiles.

2. (b)

   Pure absorption profiles are characterised by a single absorption feature. There are no other absorption features, or emission observed in these profiles.

3. (c)

   P-Cygni profiles are characterised by the presence of a sub-continuum blue-shifted absorption feature, and emission centred near 0 km s^-1or slightly red-shifted. There is no red-shifted absorption feature observed.

4. (d)

   Inverse P-Cygni profiles are characterised by the presence of a sub-continuum red-shifted absorption feature, and emission centred at low velocities. There is no blue-shifted absorption feature observed. The red-shifted absorption feature is generally wider than the absorption observed in P-Cygni profile types.

5. (e)

   Combination profiles are characterised by the presence of two absorption features, one is red-shifted and the other is blue-shifted, and emission centred near $\sim$ 0 km s^-1. These profiles are a combination of the P-Cygni and Inverse P-Cygni profiles, with similar profile characteristics in both the blue- and red-shifted absorption features.

6. (f)

   Low velocity absorption (LVA) profiles are characterised by an overlapping emission feature and absorption feature, both centred near $\sim$ 0 km s^-1. These profiles are distinguished from a P-Cygni or Inverse P-Cygni profile due to the presence of emission on both sides of the absoprtion feature.

7. (g)

   Double absorption (DA) profiles are characterised by the presence of two sub-continuum absorption features generally within $\pm$ 200 km s^-1. There is no emission observed in these profiles.

8. (h)

   Miscellaneous profile types are also observed in the data and are those which cannot be accurately described by any of the above profile types.

Figure 4 shows a sketch of an example protostellar system with accretion flows along which material moves from the inner disk onto the star, and with a jet and disk wind present. The dark grey dashed arrows show different viewing angles which result in different He I $\lambda$10830 Å line profiles. In the scenario where the source is viewed along a line-of-sight passing through an accretion flow, a red-shifted sub-continuum absorption feature is observed in the line profile. Emission-only profiles may be observed if the source is sub-luminous and/or close to edge-on resulting in the observation of only jet/wind emission. Close to face-on, the emission is absorbed by the higher velocity components of a jet/outflow which results in wider blue-shifted absorption profiles at larger velocities. Meanwhile, for closer to edge-on observations, the observer’s line-of-sight passes through the low velocity parts of the outflow (possibly a disk wind), thus producing a narrower absorption feature at less blue-shifted velocities. In some cases, the line-of-sight may pass through both a jet or wind, and an accretion flow resulting in two absorption features in the line profile (see inset in Figure 4). In this figure, we do not include absorption-only profiles, double absorption profiles or low-velocity absorption profiles since they occur less frequently in our sample.

We fitted the He I $\lambda$10830 Å profiles in the Lupus and Upper Scorpius samples as described in Section 3.1. Figure 5 shows the percentage of targets characterised into each profile type for both regions. Information on the profile type and centroid velocities of the emission and absorption features are listed in Tables 1 - 2. We also compare the proportion of profile types observed in our sample to those of a sample of 38 targets located in the Taurus star forming region from Edwards et al. (2006). In Lupus (middle, Figure 5) we find that the most common profile type are inverse P Cygni profiles (27 of the 82 targets, i.e. 33%), followed by combination type profiles (13 targets or 16%). However in Upper Scorpius (right, Figure 5) the combination profiles (11 of the 35 targets, 31%) and inverse P Cygni profiles are the most common types (12 targets, 34%), followed by P Cygni profiles (five targets, 14%). In Upper Scorpius we do not observe any pure emission profiles. In Taurus, a large portion of the He I $\lambda$10830 Å profiles show a P Cygni profile (19 of 38 targets, 50%), followed by the combination profile type (15 targets, 39%). There are no single absorption or double absorption profiles observed in the Taurus sample. We stress that the Taurus sample in Edwards et al. (2006) (38 objects in total) is much smaller than our sample in Lupus (82 objects), but is comparable in size to our sample in Upper Sco (35 objects). The Taurus sample however is predominantly K7 and M0 type stars, compared to a larger range of spectral types from K0 to M8.5 in our sample. We believe these differences may contribute to the varying proportions of each profile type observed in each region. Further, we note that the resolution of the Taurus observations is higher than the resolution of our observations. The resolution with NIRSPEC used for the Taurus sample is R $\simeq$ 25000 (Edwards et al. 2006), while the X-Shooter resolution is $\simeq$ 10000. In velocity, this is roughly equal to 12 km s^-1and 30 km s^-1for NIRSPEC and X-Shooter respectively. However, the widths of the features in our sample are typically larger than this resolution, thus we do not believe these variations are due to an instrumental effect.

In Figure 6 we present scatter plots and histograms of the maximum absorption velocities for the red- and blue-shifted absorption features. The maximum absorption velocity is the highest velocity observed in the absorption feature, i.e. the most blue-shifted velocity in absorption for blue absorption features and the most red-shifted velocity in red absorption features. We calculated the maximum velocities using the fitted Gaussian velocity centroids ($\mu$ or v_cen) and standard deviations ($\sigma$) i.e. v_max = v_cen $\pm$ 3$\sigma$. We display the typical 3$\sigma$ fitting errors for the red- and blue-shifted maximum velocities in panels (b) and (d), respectively, which are obtained from the covariance matrix of the Gaussian fit from scipy curve$\_$fit.

The majority of red-shifted velocities appear to cluster at velocities of $\simeq$200 km s^-1, and increase with stellar mass such that they are compatible with free-fall velocities for stellar radii between R_⋆ = 1 and 2 R_⊙, assuming the radius for magnetospheric accretion is equal to 5 R_⋆ (Hartmann et al. 2016). The free-fall velocities are calculated with the following simplified equation from Hartmann et al. (2016):

where G is the universal gravitational constant, R_M is the magnetospheric radius from which material falls inwards along accretion flows, and M_0.5 and R₂ are the stellar mass and radius in units of 0.5 M_⊙ and 2 R_⊙. The stellar radii for each source in our sample were calculated using the stellar luminosity and effective temperatures found in Alcalá et al. (2019) for the Lupus sample and Manara et al. (2020) for the Upper Scorpius sample. The typical R_⋆ for our targets is $\sim$1.3 R_⊙ for Lupus and $\sim$1.05 R_⊙ for Upper Scorpius. We calculated the radius at which the infalling material reaches free-fall velocities (R_ff) by rearranging Equation 1 in Fischer et al. (2008) so that:

becomes

where R_M is set to 5 R_⋆ following the assumptions of Hartmann et al. (2016) (see above) and where v_esc is the escape velocity defined as:

For sources in Lupus, R_ff is approximately 1.6 R_⋆ while in Upper Scorpius R_ff $\sim$ 1.3 R_⋆ (see Figure 7). This typically corresponds to $\sim$ 1.4–2 R_⊙ in line with the curves in Figure 6 (see Section 5.2 for further discussion).

The blue-shifted velocities in our sample tend to appear at lower velocities than their red-shifted counterparts, but are not as confined to specific velocities. Further, they generally reach escape velocities from the star. We note that the escape velocities calculated for our sample are from the star, and that if the material is launched in a disk wind, the escape velocities from the disk are smaller.

For each source in Lupus and Upper Scorpius, in Figure 8, we plot histograms of the difference between the observed maximum red-shifted velocities and free-fall velocity, and between the maximum blue-shifted velocities and escape velocity from the star. The red-shifted maximum velocities appear to be similar to the free-fall velocities, with a mean value of -80.6 km s^-1. Meanwhile, the blue-shifted velocities are generally larger than the escape velocity with many of the values of v_max-v_esc between 100-200 km s^-1(and a mean value of 157 km s^-1), but their distribution is flatter spanning a large value of mostly positive velocity differences.

In Figures 9 and 10 we present the maximum absorption velocities and the FWHM of the absorption features with respect to the inclination of the source for both the Lupus and Upper Scorpius regions. In Figure 9, we observe the maximum blue-shifted velocity trending to lower velocities at higher inclinations (panel (a)). This is not seen in their red-shifted counterparts, as the red-shifted velocities remain $\geqslant$ 150 km s^-1 (panel (b)), even for highly inclined targets. In combination profiles (downward triangles), most blue absorption features appear to have a maximum velocity of $\simeq$ 150 - 200 km s^-1, while a second cluster of velocities closer to 350 km s^-1  which are generally the red-shifted absorption feature for these profiles. Given that the velocities for combination profiles appear across all inclinations, we believe that they are not due only to a particular viewing angle. Red-shifted absorption features tend to occur at higher velocities than the blue absorption features, and they also appear generally wider (see Figure 10). This may suggest that we can observe a wider range of accretion velocities than velocities in the wind, hinting at a dependence on the source inclination. Red-shifted velocities, however, are not related to the source inclination, like the blue-shifted features as evident in the p-values displayed in each panel. In Figure 10 we find that the red-shifted absorption features are wider (typical FWHM $\simeq$ 100 km s^-1) than the blue-shifted features (typical FWHM $\simeq$ 50–100 km s^-1). This is noticeable even in the combination profiles (downwards triangles). However, we do not find a correlation between the width of the absorption features and the source inclination.

We observe 12 sources in Lupus without sub-continuum absorption features. Emission in the He I $\lambda$10830 Å line arises due to scattering and in situ emission. Pure emission profiles may arise due to a stellar wind, launched from high latitudes on the star and observed at typical inclinations of $\simeq$ 70^∘ (Kwan et al. 2007). Of the 12 stars with emission-only profiles, six targets have known jets. These are: Par-Lup3-4 (Whelan et al. 2014); Sz102 (Murphy et al. 2021); Sz69, Sz73, Sz123A and Sz123B which show both a high-velocity component (HVC) and low velocity component (LVC) in the [OI]$\lambda$6300Å  line (Nisini et al. 2018). The HVC is generally attributed to high-velocity jets which can produce either emission-only profiles (for edge-on sources) or P Cygni profiles (closer to pole-on) - see Section 5.2. A further four sources show signatures of a LVC, possibly a disk wind - these sources are Sz88A, Sz106, Sz133 and RX1556.1-3655 (Nisini et al. 2018).

Five sources (Par-Lup3-4, Sz102, Sz106, Sz123B, Sz133) are sub-luminous (sl) sources, which have a stellar luminosity much lower (by at least a factor ten) than other sources of the same spectral type (Alcalá et al. 2014; Nisini et al. 2018). Only three of the sub-luminous sources (Sz102, Sz123B and Sz133) have known inclinations, all of which are greater than 40^∘. Sub-luminous sources are generally expected to be observed edge-on (Hughes et al. 1994) but it is possible that the disk inclinations reported for these sources are underestimated, or that there is a misalignment between the inner and outer disk in some of these sources (Bohn et al. 2022) as previously mentioned. Jet morphology and proper motions of the Sz102 jet, for example, indicate that the jet is in the plane of the sky (Murphy et al. 2021), i.e. the disk should be edge-on, thus revealing a possible misalignment between the inner and outer disk. Additionally, there are two other sub-luminous sources in the Lupus sample - these are Lup706 (an inverse P Cygni profile) and SSTc2dJ160703.9-391112 (absorption-only profile), so the lower luminosity is not unique to sources showing the emission-only profiles. Three of the sources with pure emission profiles exist in binary systems - these are Sz88A, and Sz123A and Sz123B (Alcalá et al. 2014). The inclination of the remaining source (J16085953-385627) with a pure emission profile is not known and it doesn’t appear to be associated with any known jet.

We searched for a correlation between the centroid velocity of the emission profiles and the source inclination (see Figure 11). Generally, the centroid velocities were blue-shifted by up to 100 km s^-1, and those at higher velocities are narrower than the emission profiles closer to 0 km s^-1(see Figure 12). The fact that we tend to observe emission-only profiles for sources with known jets and at high inclinations lends support to the models of Kwan et al. (2007). They note that when viewing the He I $\lambda$10830 Å line for sources at higher inclinations, a stellar wind, which appears as a P Cygni profile for lower inclinations, may become an emission-only profile as the blue-shifted absorption feature becomes weaker and eventually disappears (see Figure 4). However, we note that for the 70 remaining sources in Lupus, 48 of the sources (68%) have a LVC and/or HVC signature in [OI] (Nisini et al. 2018). HVC signatures in [OI] are observed in sources with P Cygni profiles in the He I $\lambda$10830 Å line at similar rates as the emission-only profiles - seven out of 11 P Cygni profiles in Lupus were observed with a HVC in [OI], compared to six of 12 emission-only profiles. Only ten sources with other profile types, however, are observed with a HVC in [OI]. Of the sources plotted in Figure 11, only one is not observed with either a LVC or HVC in OI - this source is Lup607 which has an inclination of 0^∘. The inclinations of the other two sources without a known jet or outflow (J16085953-3856275 and SSTc2dJ160708.6-391408) are not known. Further, in Figure 11, we plot the sub-luminous sources without known inclinations at 90^∘ (in line with the assumption that sub-luminous sources are often viewed edge-on). These sources show centroid velocities that are blue-shifted to velocities up to -80 km s^-1, in line with the apparent correlation of centroid velocity and source inclination for the emission profiles. However, if the inclinations for the two sub-luminous sources located at $\sim$ 40-60^∘ are underestimated, the correlation between centroid velocity and inclination is less evident if these sources are closer to edge-on ($>$ 70^∘). It is also possible that disk occultation may result in the emission profile becoming blue-shifted, which could explain a trend in the centroid velocity with respect to source inclination.

In a number of emission-only sources (e.g. Sz73, Sz88A, Sz102 and SSTc2dJ160708.6-391408), we observe asymmetric line profiles as is often seen in forbidden emission lines (for example, by Nisini et al. 2018; Banzatti et al. 2019). While the majority of our targets with emission-only profiles are associated with known jets, these profiles can still be produced in other scenarios. For example, in Figures 18 and 19, where we show all Lupus profiles arranged by profile type, we see a number of asymmetric emission profiles (e.g. Par-Lup3-4, Sz73, Sz106, Sz133 and RX1556.1-3655), which appear to have low-velocity absorption features that are not sub-continuum, and thus not fitted by our fitting routine. This is also the case in the Taurus sample studied by Edwards et al. (2006), where four targets show asymmetric emission profiles indicating possible red and/or blue absorption that is not sub-continuum. Overall, we find that sources showing an emission-only profile tend to be associated with known jets and outflows, and appear at high source inclinations.

We observe absorption features in many spectra from both regions in our sample. In the following sections we discuss observed trends in velocities of the wind and accretion features, and investigate if the measured absorption velocities have any dependence on the source inclination and/or accretion properties.

In Figure 5, the proportion of each profile type in both star-forming regions of our sample is presented in comparison to that of the Taurus sample of Edwards et al. (2006). Given that the ages of Taurus and Lupus are similar (Luhman et al. 2010; Comerón 2008, $<$ 3 Myr), we would expect the proportions of each profile type in these two regions to be similar. However, we note large differences in the proportions of the three most common profile types (P Cygni, inverse P Cygni, and combination profiles) between the Taurus and Lupus regions. As noted previously, this is not thought to be due to instrumental effects. It is also important to note that the classification of profile types in the He I $\lambda$10830 Å line varies through the literature. For example, Edwards et al. (2006) define only five distinct profile types, while Thanathibodee et al. (2022) identifies six profile types, compared to the eight types we observe. It could be that the double absorption profiles observed in our sample, for example, are in fact combination profiles with no emission feature. In this case, the number of combination profiles in Lupus becomes 16 ($\simeq$ 20%) compared to 15 ($\simeq$ 39%) in Taurus. This still does not explain the large difference in the number of P Cygni and inverse P Cygni profiles in each region. However, we note that certain regions of the Lupus and Taurus clouds have varying ages and this may suggest that younger sources with outflow phenomena are more common in Taurus.

Between the Lupus and Upper Sco regions from our sample, we find two notable differences in the proportions of each of the three most common profile types (P Cygni, inverse P Cygni, and combination profiles). Firstly, the emission-only profiles disappear entirely in our Upper Sco sample. In Section 5.1 we suggest that the emission-only profiles are the result of jets or winds. If this is indeed the case, this suggests that the jet disappears early on, before the age of the Upper Sco sources. In Figure 16, we present the Gaussian area, in units of km s^-1 $\times$ normalised flux units, of the emission features in all emission-only, P Cygni, inverse P Cygni, and combination profiles in both regions. We compare the Gaussian area of the emission features to the ratio of the accretion luminosity to the stellar luminosity for Lupus (black circles) and Upper Sco (green squares). In general, the size of the emission features in Upper Sco is smaller than in Lupus, and only becomes comparable at high L_acc/L_⋆. In both regions there is a strong trend of the emission with L_acc/L_⋆ suggesting that at least some of the emission originates from the accretion shocks on the stellar surface. Further, this trend supports the results of Kwan et al. (2007), who suggest that stars with higher accretion rates are more likely to exhibit stellar wind signatures.

The second difference between the Lupus and Upper Sco regions in Figure 5 is the significant increase in the proportion of combination profiles, up to $\approx$ 30% in Upper Sco. This suggests that the sources in Upper Sco are still actively accreting, while the blue-shifted features in the combination profiles means that winds and outflows are also present. In Figure 15, we also distinguish the ratio of the Gaussian areas of the absorption features in combination profiles between both regions. We find that the majority of the Lupus combination profiles (seven of 10) have a larger blue-shifted feature, while combination profiles in Upper Sco (seven of 9) tend to be dominated by the red-shifted feature, indicating that the winds detected in the Lupus sample are stronger than those in Upper Sco. This also indicates a difference in the accretion in both regions. However, Manara et al. (2020) found similar values for the mass accretion rates in Lupus and Upper Sco, so this is not likely to be the reason for the increased proportion of combination profiles. We note that the absolute number of combination profiles in Lupus and Upper Sco is similar (12 and 11, respectively) so we cannot rule out the possibility that this difference between the two regions is due to the small sample size.