A&A 406, 1001-1017 (2003)
DOI: 10.1051/0004-6361:20030848
F. Comerón1 - M. Fernández2 - I. Baraffe3 - R. Neuhäuser4,5 - A. A. Kaas6
1 - European Southern Observatory, Karl-Schwarzschild-Strasse 2,
85748 Garching, Germany
2 -
Instituto de Astrofísica de Andalucía, CSIC, Apdo. 3004, 18080 Granada, Spain
3 -
C.R.A.L. (UMR 5574 CNRS), École Normale Supérieure, 69364 Lyon Cedex 07, France
4 -
Max-Planck-Institut für extraterrestrische Physik,
Giessenbachstrasse 1, 85741 Garching bei München, Germany
5 -
Astrophysikalisches Institut, Universität Jena, Schillergässchen
2-3, 07745 Jena, Germany
6 -
Nordic Optical Telescope, Apartado 474, 38700 Santa Cruz de
La Palma, Spain
Received 30 September 2002 / Accepted 28 May 2003
Abstract
A spectroscopic survey of a small area at the
center of the Lupus 3 star-forming core has revealed four new
mid-to-late M-type members, including a M7.5 brown dwarf. One of
the new members, classified as M5, displays prominent forbidden
lines and strong H
emission (
Å), in addition to other permitted lines, and its
luminosity is far below that of other members of the region with
similar or later spectral types. We estimate a mass accretion rate
rate of
yr-1 for this
object, although with uncertainties that do not exclude values as
low as 10-10
yr-1. Based on the
ratio, the detection of HeI, and the CaII infrared
triplet, we argue that most of the H
emission is produced
near the surface of the object, probably in accretion columns or
at the base of jets, rather than in a low-density extended region.
The strong emission-line spectrum superimposed on an unusually
faint photospheric continuum thus seems to be a real, intrinsic
feature rather than a result of the viewing geometry caused by an
edge-on disk blocking the light from the central object. Other
Lupus 3 late-type members also display noticeable underluminosity,
all of them having
Å as a result of the faint underlying continuum. We
tentatively interpret these findings as evidence for the pre-main
sequence evolution of objects with very low (possibly substellar)
initial masses being significantly modified by accretion.
Key words: stars: pre-main sequence - stars: low-mass, brown dwarfs - stars: winds, outflows
The Lupus star-forming complex is one of the belt of low mass star-forming
regions within 200 pc from the Sun, together with
those in Chamaeleon, Rho Ophiuchi, R Coronae Australis, and
Taurus. Physically, it belongs to the larger structure of the
Scorpius-Centaurus OB association (Blaauw 1991), of
which Lupus is currently one of the remaining active star forming
sites together with the more conspicuous one of Rho Ophiuchi. The
generic denomination of Lupus clouds applies to a number of
distinct, loosely connected entities scattered over a large area
of the sky and approximately at the same distance (e.g.
Schwartz 1991). Its large-scale molecular contents as
traced by 12CO
)
has been studied by
Murphy et al. (1986), who estimated a total mass of
from the integrated brightness
temperature, well in agreement with the
obtained in the more recent study of Tachihara et al. (2001) using the same tracer. Tachihara et al.
(1996) studied individual molecular concentrations
at a higher spatial resolution in 13CO
,
deriving individual masses in the range
100-1200
.
Comprehensive surveys of the pre-main sequence population
associated to the Lupus clouds have been carried out by
Schwartz (1977), who extended the list of
H-emitting members previously identified by
Henize (1954) and Thé (1962), and by Wichmann
et al. (1997) and Krautter et al.
(1997). A detailed spectroscopic study of many of the
members identified in Schwartz's work was presented by Hughes et al. (1994). Further studies on individual pre-main
sequence members of the Lupus complex can be found in Reipurth &
Zinnecker (1993) and Nürnberger et al.
(1997). Special attention has been paid to the
Lupus 3 cloud, the one containing the richest stellar aggregate,
where Nakajima et al. (2000) recently identified new
faint, moderately obscured members with estimated masses near the
substellar limit by means of near-infrared imaging. The
large-scale distribution of pre-main sequence stars in the Lupus
clouds and their relationship to the concentrations of molecular
gas has been studied by Tachihara et al. (2001).
Being one of the most nearby active star forming regions, the Lupus clouds are well suited for the detailed study of the individual properties of its pre-main sequence population. On the other hand, the Lupus clouds are in many respects similar, but not quite identical, to other nearby low-mass star forming regions that can be studied to the same level of depth and detail, thus allowing direct comparison between the stellar populations in those different aggregates and their relationship to the interstellar component out of which they are born (e.g. Chen et al. 1997). A particular aspect of interest in this context is the greater ratio of late M-type members with respect to earlier-type stars in the Lupus clouds, as compared to regions in principle similar like the Taurus clouds. This was already noted by Appenzeller et al. (1983) and confirmed by Hughes et al. (1994) and Wichmann et al. (1997). The new low mass members detected by Nakajima et al. (2000) are consistent with this feature of the Lupus population.
In this paper we present the results of a survey of a small area
at the core of the Lupus 3 cloud, as signaled by the Herbig Ae
star HR 5999 and its optical companion HR 6000,
in which we have identified new low mass members by means of their
characteristic H
emission and have obtained good quality
spectra of previously known low-mass members. The H
equivalent widths measured for the members of our sample span a
wide range of values, making them suitable for addressing the
possible effects of accretion activity, as traced by H
emission, on the actual tracks followed by pre-main sequence
objects in the temperature-luminosity diagram. Strong accretion
has been recently suggested to dramatically shift the position of
very low mass stars and brown dwarfs in that diagram (Fernández
& Comerón 2001), and new supporting observational
evidence is discussed in this paper.
The spectroscopic observations presented in this paper were
obtained using the FORS1 visible-light camera and spectrograph at
Antu, Unit Telescope 1 of the ESO Very Large Telescope. Short
images (30 s) of overlapping fields around the HR 5999/HR 6000 pair
through the Bessell I filter were obtained on the night of 29/30
April 2001 in order to provide a reference image for the slitless
spectra extraction. Slitless spectroscopy was then obtained on the
night of 1/2 May 2001 of the two fields immediately East and West
of HR 5999/HR 6000, using an instrumental configuration consisting
of the 300V grism used simultaneously with the Gunn r filter, to
isolate the region around H
between 6120 Å and
6940 Å. The resolution yielded by this instrumental setup, in
the prevailing seeing conditions of 0''8 under which our
observations were taken, is approximately 11 Å per resolution
element. Each of the two fields was imaged for one hour, split in
three exposures with small telescope offsets perpendicular to the
dispersion direction in between.
The slitless spectroscopy images were reduced and combined in near-real time by means of dedicated IRAF scripts that automatically detected stars in the images and extracted their 1D spectra from the slitless spectroscopy image (see Comerón et al. 2000 for a detailed description of this procedure). Emission-line stars with late-type underlying continua were then selected by visual inspection of the extracted spectra for follow-up multi-object spectroscopy (MOS).
The movable-blades facility available in FORS1 allowed us to fully define the instrument setup for the ensuing MOS, including the masks to be used, while the second field was being observed in slitless mode. Thanks to this, the MOS observations were performed immediately after the slitless ones on the same night of 1/2 May 2001. The setup used consisted of the 300I grism yielding a resolution similar to that of the slitless spectra, but now covering a more extended range between approximately 6000 Å and 11 000 Å. The actual range for each object depended on its position in the field. Three exposures were taken through each mask, each of 10 min in duration, at telescope positions slightly offset along the slits.
Besides the images obtained as a reference for the slitless
spectroscopy with the VLT, we obtained
images with DFOSC
at the 1.5 m Danish telescope in La Silla on the nights of 22/23
and 23/24 March 2002 to determine the broad-band spectral energy
distribution of the newly found members. Three dithered images
were obtained of each of the two overlapping fields East and West
of HR 5999/HR 6000. Thanks to the larger field of view of DFOSC,
these images included the fields for which slitless spectroscopy
had been obtained the previous year as well as some of the
surrounding area. The combined images add up to effective exposure
times of 120 s in V, 60 s in
,
and 45 s in i. Photometry
of the stars observed in MOS was performed on the DFOSC images
using the PHOT task under the APPHOT package in IRAF. Given the
uncrowdedness of the field and the lack of prominent nebulosity,
the capabilities of PHOT were deemed sufficient for our purposes.
For the sake of comparison with other studies, we have transformed
the i magnitudes to
magnitudes using Eqs. (23) of Fukugita
et al. (1996).
The broad-band photometry in the visible was complemented with
near-infrared
imaging obtained with SOFI at the New
Technology Telescope (NTT) on 19 August 2001 and 20 July 2002. The
August 2001 data contain J-band imaging for all the objects,
while the July 2002 ones provide
imaging for the central
parts of the FORS1 field. In both cases, dithered images of the
fields East and West of HR 5999/HR 6000 were obtained,
sky-subtracted by median filtering of the individual frames
without correcting for the telescope offsets, and combined by
shift-and-add. The photometry on the reduced frames was performed
using also PHOT, like in the previously described set. A caveat to
keep in mind is that the visible and infrared observations are
non-simultaneous. Therefore, the possibility of mismatches between
the visible and the near-infrared observations due to time
variability, which is discussed in Sect. 3.1, cannot be
excluded.
Before entering in the description and discussion of our
results, it seems convenient to consider in some detail the
distance to the Lupus 3 star forming region, as this is a
necessary ingredient (and an important caveat) in the estimate of
luminosities that we discuss below. Most studies have determined
or adopted distances in the range 130-170 pc on the basis of the
membership in the Scorpius-Centaurus association (see e.g. the
discussion in Murphy et al. 1986, and also
Krautter 1991) or by estimating spectroscopic
parallaxes for early-type surrounding stars that are expected to
belong also to Scopius-Centaurus (Hughes et al. 1993).
A distance as short as 100 pc has been suggested by Knude & Høg
(1998) based on the extinction distribution of
foreground and background stars with distances measured by Hipparcos. The best value currently available for the distance to
the Upper Centaurus-Lupus subgroup of the Scorpius-Centaurus
association is pc, derived by de Zeeuw
et al. (1999) based on Hipparcos trigonometric
parallaxes. Murphy et al. (1986) have suggested that
the Lupus complex, while still physically belonging to
Scorpius-Centaurus, may be in front of it, with the extinction
produced by the cloud being responsible for the lower number of OB stars in that direction.
The trigonometric parallaxes measured by Hipparcos for
HR 5999 and HR 6000 allow in principle a more
direct derivation of the distance to Lupus 3. Somewhat
surprisingly, the derived distances for these stars are 207 pc and
241 pc respectively, the uncertainty being in both cases
approximately pc. This determination is not incompatible
with the normally adopted distance to the bulk of
Scorpius-Centaurus, which is considerably extended in the plane of
the sky: assuming that the depth is of the same order as its
spread in the sky, a distance of
200 pc of Lupus 3 places
it in the far side of the association rather than outside it. We
note that a similarly high distance of
pc to the
Lupus star forming complex, based on a larger set of Hipparcos distances, was derived by Wichmann et al.
(1998).
![]() |
Figure 1:
A composite of two FORS1 I-band images indicating
the positions of previously and newly identified objects. The
field measures
![]() |
Open with DEXTER |
To be consistent with the individual distances to HR 5999 and HR 6000, with that derived by Wichmann et al. (1998), and with the membership in Scorpius-Centaurus, we have opted for the rather eclectic solution of adopting a distance of 200 pc for Lupus 3, corresponding to a distance modulus of 6.5 mag. We should stress the important difference that this implies with respect to distances adopted in previous works: by comparison to Hughes et al. (1994), our distance modulus is 0.8 mag greater than in that work, and it would reach one full magnitude if we adopted the sheer average of the Hipparcos trigonometric parallaxes for HR 5999 and HR 6000. Luminosities may thus change by a factor of two depending on which distance modulus (the one of Hughes et al. or the one adopted in this work) is assumed. It is important to bear in mind that the distance uncertainty is by far the dominating factor of uncertainty in the luminosities derived for individual members of Lupus 3. However, we do not consider that depth effects contribute to relative differences in luminosity among the members. Judging from the extent of the Lupus 3 cloud on the sky from Digitized Sky Survey images, we estimate its physical size to be between 4 and 5 pc along the largest dimension projected on the sky. A size at least twice as large along the line of sight would be required to produce differences in luminosity exceeding the 10% level. We thus consider it highly unlikely that depth yields any significant contribution to the spread in luminosities among the objects discussed here.
Table 1: Positions, photometry, and spectral types of the observed stars.
The review of the spectra obtained in our slitless images
revealed all the previously known T Tauri stars in the observed
region plus four additional, previously unrecognized emission-line
objects. Figure 1 shows the composite of two FORS1 I-band images
of the surveyed field indicating the positions of all the
spectroscopically identified members of the star forming region,
either from Hughes et al. (1994) or from the present
work. It should be noted that a few additional possible faint
members, not plotted there, have been reported by Nakajima et al.
(2000) based on their infrared colors. Given the
constraints due to the number of slits that can be defined in a
single setting of FORS1 and their relative positions, it was not
possible to obtain MOS spectra of all H-emitting stars in
the region, both new and previously known. Instead, our selection
for MOS targets focused on the newly found H
-emitting
objects, and previously known members were observed only if it was
feasible to position a slit on them with the same setting.
Preference was given however to faint objects suspected to have
late-type slitless spectra but no discernible H
emission
rather than to bright known members. However, none of those faint
objects was confirmed as a candidate Lupus member. Other than the
object labelled Par-Lup3-4, none of the candidate faint
members in Nakajima et al. (2000) survey was included
in our spectroscopic sample.
The positional and photometric data of the objects
for which MOS spectra were obtained are given in Table 1 together
with their spectral type (see Sect. 3.2). The nomenclature for
all previously known members refers to Schwartz's
(1977) catalog, although some of them have other
denominations as well; we note in particular
Sz102 = Th 28 (Thé 1962), also known
as Krautter's star. For the new members, we have followed
the same criterion as in Fernández &
Comerón (2001) and thus denominate them
generically as Par-Lup3, where "Par" stands for Paranal. As it
turned out, the MOS setting serendipitously included a fifth
late-type object near the position of one of the four new
H-emitting objects, which we have denominated
Par-Lup3-1/cc 1 following the suggestion of Neuhäuser
et al. (2002) on naming of companion candidates.
Its actual relationship to Lupus 3 is very doubtful, as discussed
below. Unless otherwise indicated, the
photometry is
from our DFOSC observations and the JHK photometry from our July
2002 observations. Unfortunately, several of the Sz stars observed
with FORS1 in MOS as well as Par-Lup3-1 were out of the field of
view in the July 2002. In those cases, the JHK photometry has
been adopted from Hughes et al. (1994) or from the
J-band August 2001 observations for Par-Lup3-4.
It is interesting to compare our photometry with that published
by Hughes et al. (1994) in the visible and the infrared
and with that of Nakajima et al. (2000) in the
infrared, as well as the two epochs in J available for some of
the objects from our own observations. Most objects display at
most the moderate variability characteristic of T Tauri stars with
amplitudes typically <0.3 mag in all the bands. This is the case
of Sz 100, Sz 106, Sz 108B,
Sz 110, Sz 112, and Sz 114. A
comparison between the June 2001 and July 2002 J-band photometry
also shows differences within this range for Par-Lup3-2
and Par-Lup3-3, while only one-epoch data are available
for Par-Lup3-1, as well as for Sz 108A due
either to scattered light from HR 5999/6000 or to saturation in
the infrared. Greater differences are seen however for other
objects indicating large-amplitude variability. For
Sz 102 we obtain V and
magnitudes that are
fainter than those in Hughes et al. (1994) by 0.51 and
0.25 mag respectively, the
magnitude staying approximately
equal. Sz 109 appears distinctly redder in our DFOSC
observations than in the data of Hughes et al. (1994):
while our
magnitude is only 0.05 mag fainter than in that
work, V and
are fainter by 0.91 mag and 0.65 mag,
respectively. The large variability of this object has been also
noted by Nakajima et al. (2000), who found it to be
at least 0.2 mag brighter in J than Hughes et al.
(1994). Also Sz 114 is fainter and redder in
our observations than in those of Hughes et al. (1994)
by 1.04 mag (V), 0.66 mag (
)
and 0.34 mag (
). The
features in the VRI variability may be explained in principle by
the appearance and disappearance (due to rotation or spot
lifetime) of surface regions at a temperature different from the
normal photosphere, possibly hot regions at the bottom of the
accretion columns. Further photometric monitoring is needed to
confirm or disprove this interpretation.
In the infrared, differences above the 0.3 mag level are found
only for Par-Lup3-4 by comparing our July 2002
photometry
with that of Nakajima et al. (2000). This object was
clearly fainter and redder in July 2002, with a difference with
respect to Nakajima's data of 0.71 mag (J), 0.37 mag (H) and
0.12 mag (
). The J magnitude observed in June 2001 is also
brighter than in July 2002 by 0.45 mag. These changes in the
brightness and colors are one of the peculiar features of
Par-Lup3-4, together with others that will be discussed in this
paper.
We find a clear systematic difference between our assigned spectral
types and those of Hughes et al. (1994), in the sense
of ours being 2 spectral subclasses later (we exclude from
this comparison Sz102, whose spectrum is featureless in
the considered range except for the abundant emission lines and
the telluric absorption bands, thus precluding a spectral
classification). Since the spectra themselves are not included in
Hughes et al.'s paper, we cannot reassess their classification to
attempt to identify the origin of the discrepancy. However, we
suspect that the reason may be related to the spectral ranges used
in their work and ours. Our classification is based on a
comparison to the spectral sequence for K-to-late M stars of
Kirkpatrick et al. (1991) and, given the quality
of our spectra, we estimate it to be accurate to within half a
spectral subclass. The comparison makes use of
temperature-sensitive features like the shape of the broad
continuum peak between
7300 Å and
7600 Å, the
change of slope that occurs at
8000 Å, and the height of
the rise between
8000 Å and
8150 Å. For late M
spectra these features provide a sensitive sequence allowing the
accurate determination of spectral subtypes down to at least M8,
also for very young objects (see e.g. Fig. 2 of Comerón et al. 2000). On their side, Hughes et al.
(1994) base their classification on the
5800-7000 Å
interval, which is probably better suited for the classification
of earlier spectra. It is also somewhat more prone to veiling as
the continuum emission due to accretion increases in importance
towards the blue with respect to the underlying photospheric
continuum, which might explain at least in part why the
classification based in that bluer range is systematically
earlier. Since the
6500-9000 Å range has been used in most
recent discussions on the classification of late M- and L-type
objects in star forming regions, and for the sake of consistency
with previous work by us and other authors, we consider it more
appropriate for the current study. Analogous reasons of
self-consistency in the analysis of the data have led us not to
consider the other members in the field that were not observed and
classified by us.
![]() |
Figure 2: All the FORS1 spectra obtained in the observations described in Sect. 2.1. To homogenize the presentation, all the spectra have been normalized to the flux in the 9000-10 000 Å interval. The vertical scale has been further reduced by a factor of 2 in the case of Sz 102. |
Open with DEXTER |
We have checked our spectral classification against qualitative criteria based on indices that measure the strength of temperature-sensitive features. The R4 / R7 index proposed by Prosser et al. (1991), which measures the height of the local peak between 7000 Å and 7068 Å (R7 passband) with respect to the nearby depressed continuum between 6750 Å and 6844 Å (R4 passband), yields an appropriate determination of the spectral type over the whole late-K to mid-M range and can be applied also at later M types, as described by Comerón et al. (2000). For types later than about M6.5, the VO index defined by Kirkpatrick et al. (1995) measuring the depth of the VO feature near 7450 Å provides also a sensitive spectral type indicator. Overall, we find an excellent agreement to within half a spectral subclass for all the objects that we classify as M6 or later. At earlier types, for which only the R4 / R7 index is sensitive, we find a precise agreement for Par-Lup3-4 at spectral type M5. For the other objects, our assigned spectral types are between half and one spectral subclass earlier than those derived from the R4 / R7 vs. spectral type calibration of Prosser et al. (1991). We suspect that the offset may be due to systematic differences between the spectra of the young objects that we are considering here and the evolved field stars that were used in that paper to calibrate the relationship between that index and the spectral type. It is important to note that the R4 / R7 index is defined near the blue edge of the spectral range that we use for classification, and may thus be expected to be more sensitive to veiling. The effect of veiling should be an increase in the value of the index due to the dilution of the peak sampled by R7 and resulting in the assignation of a spectral type earlier than the unveiled one. This effect goes into the direction opposite to the systematic offset that we find in our data, which leads us to conclude that veiling may introduce at most a small bias in our spectral type determination, most probably not exceeding one spectral subclass.
Table 2: Equivalent widths of the main emission lines observed in the spectra of the stars discussed in this paper.
Several features of the very low mass members of this region of
Lupus 3 are already visible at a glance in Fig. 2. M types later
than M5 are abundant, also among objects previously classified by
Hughes et al. (1994), and one of the new objects is
classified as late as M7.5. Secondly, the H
equivalent
widths have large variations as observed in other star forming
regions, but reach rather extreme values with respect to the
underlying continuum in the well-known case of Sz102 and
the newly found Par-Lup3-4. The emission spectrum of
these objects is discussed in further detail below.
The classification is rather uncertain for Par-Lup3-1/cc 1,
the faint object 3'' southwest of the M7.5 object
Par-Lup3-1. The fact that its spectrum is clearly that of
a late-type star, that it is so close along the line of sight to
the latest-type H-emitting member identified in the star
forming region, and that it is much fainter than it, led us to
consider the possibility that Par-Lup3-1/cc 1 might be a
bound substellar companion to Par-Lup3-1 with a spectral
type later than M. However, a comparison to the sequence of late M
and L spectral types published by Kirkpatrick et al. (1999) shows that the features that can be
discerned in the spectrum of Par-Lup3-1/cc 1 are clearly
M-like and exclude the possibility of a very late M or L type. We
rather consider Par-Lup3-1/cc 1 a likely distant mid-M
giant, a possibility that appears reasonable in view of the
relatively low galactic latitude of Lupus 3 and the fact that
Par-Lup3-1 is projected on an area of low dust column
density, as is obvious from the high stellar density in this
direction. Therefore, we will not consider
Par-Lup3-1/cc 1 among the members of Lupus 3 in the rest
of this paper.
Perhaps the most striking feature of the sample of objects considered in this paper is the variety of emission line spectra found and the existence of some rather extreme cases worth considering in some detail.
Table 2 lists the equivalent width of the main lines and, for
comparison, the same information for LS-RCrA 1
(Fernández & Comerón 2001). The similarity of
one of the newly found objects, Par-Lup3-4, with
LS-RCrA 1 is remarkable as illustrated by Fig. 3. The
emission lines have their origins in the different components
associated to the accretion/mass loss processes that accompany the
early stages of stellar evolution and that have been extensively
studied in stars of somewhat higher masses than those considered
here (e.g. Hartigan et al. 1995;
Hirth et al. 1997). Forbidden lines such as those seen
in the spectra of Sz102, Sz106, and
Par-Lup3-4, as well as some amount of H
emission,
are commonly observed in actively accreting young stars. These
lines are thought to have their origin in a magnetically driven
wind stemming from regions in the vicinity of the star (typically
<1 AU), either in the star/disk magnetopause (Shang et al.
2002; see also Kwan & Tademaru 1988; Hirth et al. 1997) or a more extended range in radii (Cabrit et al. 1999). The forbidden-line emission spectrum is
similar to those of the closely related Herbig-Haro objects.
In the commonly accepted magnetospheric accretion model, the
circumstellar disk is magnetically truncated near the surface of
the star (Königl 1991), with material flowing from
the disk towards the star in accretion columns that follow the
magnetic field lines connecting the inner edge of the disk with
high latitude regions of the star. Much of the permitted-line
emission, particularly H,
is thought to have its origin in
these columns, as confirmed by the broad velocity profiles
observed in H
(Muzerolle et al. 1998a,
2000). At the base of these accretion columns, close
to the surface of the star, shock fronts develop creating a high
temperature layer (Königl 1991; Gullbring
1994; Lamzin 1995). Much of the HeI
emission, the broad component of the CaII infrared triplet lines
(and possibly also the narrow component according to Batalha et
al. 1996), and the continuum emission at short
wavelengths veiling the photospheric spectrum, are thought to
arise in the hot spots at the base of the accretion columns. An
additional contribution to the H
emission in the vicinity
of the star may arise from the base of a collimated jet, also near
its surface (Takami et al. 2001, 2003).
![]() |
Figure 3: Spectrum of Par-Lup3-4 with the main emission lines identified. For comparison, the spectrum of LS-RCrA 1 (from Fernández & Comerón 2001) is shown. |
Open with DEXTER |
Many of the detailed diagnostics on the physical conditions
and the structure of the regions where both the outflow and
accretion signatures originate can be derived from high resolution
spectra, as yet unavailable for our sources. Nevertheless, the
lines visible in our low-resolution spectra give already some very
useful directions. The wide range of H
intensities found
in our sample, spanning over two orders of magnitude, hints to a
range of accretion activities. On the other hand, also forbidden
lines associated to outflows display intensities ranging from
undetectable to the extreme case of Sz102, with varying
levels of correlation to the lines associated to accretion.
Most of the CaII infrared triplet emission, if not all of it,
is likely to be due to the mass accretion process. Although the
structure of the lines is unresolved in our spectra, the
equivalent widths of the CaII triplet at 8498 Å, 8542 Å, and
8662 Å that we detect in Sz100, Sz102,
Sz113, Par-Lup3-4, and LS-RCrA 1, are
well above those typically resulting from chromospheric activity
in young late-type stars as well as those of the narrow-band
component possibly related to accretion (Batalha et al.
1996). However, they are within the range observed in
the broad-line component (Hamann & Persson 1992) that
is generally attributed to accretion. This is therefore the most
likely origin also in our objects. In all cases the line ratios
are much closer to 1:1:1 than to the expected ratio 1:9:5 for
optically thin emission, as noted by Fernández & Comerón
(2001) for the case of LS-RCrA 1 and as
usually found among T Tauri stars (Herbig & Soderblom
1980) and sources powering Herbig-Haro outflows
(Reipurth et al. 1986). The absence of any detectable
forbidden [CaII] emission at 7291 and 7324 Å suggests indeed
that the observed CaII emission arises in a high density region
(
cm-3), under conditions found near the
surface of the object, rather than in a lower density outflow.
As argued below, the weakness of the forbidden lines in
Sz 106 suggests that the contribution of the outflow
emission to the H
flux is small. Otherwise, the CaII
infrared triplet is poorly correlated with H
among our
objects.
Muzerolle et al. (1998b) have examined the correlation
between the accretion rates as derived from ultraviolet veiling
and the intensity of different emission lines, finding that the
tightest match occurs for the CaII infrared triplet. For the
correlation between CaII(
)
flux,
,
and the mass loss rate, the linear fit to
their Fig. 9 can be expressed as
![]() |
(1) |
![]() |
(2) |
Table 3:
Estimated accretion rates from the CaII(
)
line.
Table 4: Outflow parameters derived from forbidden lines.
The HeI line, weakly appearing in the spectra of Sz102,
Sz106, Sz110, Sz113, and
Par-Lup3-4, as well as LS-RCrA 1, may be formed
in principle either in the accretion shock (Lamzin et al.
1996) or in the disk wind, provided that sufficiently
high temperatures are produced (Beristain et al.
2001). Besides Sz102, the most prominent
emission in HeI is found in Sz113, a star having also
intense H
and CaII infrared triplet lines but no forbidden
lines indicating outflows. The other objects displaying HeI
emission also have strong H
emission, with the exception
of Sz110 where it is only moderate. Disk winds hot enough
to produce HeI emission may be expected to produce also [OII]
emission at 7319 and 7329 Å, which is observed only in
LS-RCrA 1 and possibly Par-Lup3-4. For these
reasons, we thus favor the base of the accretion columns as the
formation site of the observed HeI emission.
The amplitudes observed in the VRI bands, as well as the rate at which they decrease towards longer wavelengths (see Sect. 3.1), are also expected as a consequence of accretion by the formation of hot spots in the surface. The origin of the near-infrared variability of Par-Lup3-4, although probably related also to accretion, may be somewhat different from that at visible wavelengths and related to changes in the inner regions of the accretion disk (Eiroa et al. 2002). Near-infrared monitoring of other star forming regions (Kaas 1999; Carpenter et al. 2002) has shown that important variations in the K band, typically around 0.3 mag but reaching up to 1 mag for some sources, is common among young stellar objects, with a trend for larger amplitudes to be associated with stronger near-infrared excesses.
Forbidden lines of [OI], [NII], [SII], and [OII] are detected prominently in Sz102 and Par-Lup3-4, and much more faintly in Sz106. Ratios among some of these lines allow useful estimates of physical conditions in the circumstellar medium and mass loss rates through their dependencies with temperatures and densities (e.g. Osterbrock 1989). A difficulty nevertheless arises with the interpretation of our low resolution spectra, since forbidden lines in T Tauri spectra are usually split into two separate components with near-zero and high, generally blueshifted velocity respectively (e.g. Appenzeller et al. 1984). Besides the kinematical differences, both sets of lines sample different physical conditions (Hamann 1994). Unfortunately, due to lack of resolution in the spectra available for the objects discussed here it is not possible to carry out a separate analysis of the low- and high-velocity components. This is an important caveat at the time of interpreting the quantities estimated in this Section.
The ratio [SII](6717)/[SII](
6731),
practically insensitive to the temperature, is often used to
derive the electron density
in the emitting volume. The
values we find for Sz102 and Par-Lup3-4 are very
similar, also to those measured for LS-RCrA 1 (Table 4),
and indicate
near 2000-3000 cm-3. These are indicative
values only, due both to the blend of the low and high velocity
components of those lines as pointed out above, and to the fact
that they (especially the high velocity component) arise from an
extended region (e.g. Cabrit et al. 1999) probing a
wide range of densities.
Keeping the caveats about the components blend in mind, it is
possible to estimate upper limits to the mass loss rates following
the method outlined by Hartigan et al. (1995) based
on the luminosities of the high velocity component of the
[OI]6300 or the [SII]
lines. For this
purpose, we first derive the luminosity in the line from its
measured equivalent width and from the dereddened R0 magnitude
of the object:
![]() |
= | ![]() |
(4) |
![]() |
= | ![]() |
(5) |
It must be stressed however that these are upper limits to the
mass loss rate, since we use the total equivalent width of the
lines given the insufficient resolution to isolate the high
velocity component - or even to establish its existence. Mass loss
rates can be much lower if the low velocity component dominates
the emission. Indications that this may indeed be the case come
from the existing evidence for a correlation between the accretion
(traced by infrared excess due to circumstellar emission or by
veiling) and the mass loss processes (traced by forbidden lines),
as has been discussed by Cabrit et al. (1990) among
other works. In their comparison between outflow rates and
accretion rates derived from spectral veiling, Hartigan et al.
(1995) also find a loose correlation between both,
the accretion rate being typically 10-100 times higher than the
outflow rate. Taken at face value our results seem to depart
considerably from this correspondence, as the mass loss rates
discussed above are typically an order of magnitude greater
than the accretion rates. We take this as an indication that
actual mass loss rates are much smaller than the values given in
Table 4, a hypothesis that could be easily confirmed if
mid-resolution spectroscopy failed to reveal the blueshifted
component on which the mass loss rate estimate is based. As a
matter of fact, applying this derivation to Sz102 results
in a mass loss rate of
yr-1versus an accretion rate 40 times smaller (which would result in
its complete evaporation if maintained for only 1-2 Myr!). In that
case however published medium resolution spectra (Appenzeller et al.
1984) fail to reveal any significant high
velocity component, in consistency with a much lower mass loss
rate.
In addition to H,
some of the upper lines of the Paschen
series of hydrogen are visible in the spectra of Sz102,
Sz106, and Par-Lup3-4 allowing the measurement
of the Paschen decrement. Pa9, Pa10, Pa11, Pa12, and Pa 14 are
clearly visible in Sz 102, while only Pa9, Pa10, and Pa12
can be reliably measured in the other two objects. The ratios
among these lines in Sz 102 and Sz 106 follow
the expected values for recombination lines in a hot gas (
K), which is only weakly dependent on temperature and
density (Brocklehurst 1971). This is not the case
however for Par-Lup3-4, whose Paschen decrement is
noticeably slower: for example, the intensity ratio Pa12 / Pa9 is
0.8, about twice the expected value, and a similar
behavior is seen for Pa10 / Pa9. This was also noticed by
Eislöffel et al. (1990) in their comparison
between T Tauri and AS 353A, the source powering
HH 32, attributing the slower Paschen decrement of the
latter to optically thick emission. The same conclusion may apply
to Par-Lup3-4.
![]() |
Figure 4:
Near-infrared color-color diagram displaying all the
sources for which near-infrared photometry is available. The solid
line represents the locus of unreddened main-sequence stars
down to spectral type M6 (Bessell & Brett 1988), and
the dashed straight line is the classical T Tauri star locus
derived by Meyer et al. (1997). The dotted line is the
reddening vector representing the displacement of a star obscured
by AV = 10 mag using the extinction law of Rieke & Lebofsky
(1985). Filled circles indicate sources with
EW(H![]() |
Open with DEXTER |
The accretion signatures in the visible among T Tauri stars and protostars ("Class II'' and "Class I'' sources respectively; Adams et al. 1987) are usually correlated with near- and mid-infrared excesses indicating the existence of warm circumstellar material or the reprocessing of light from the central object by dust in a circumstellar envelope. This is clearly visible even in a small-size sample like the one discussed in this paper, as can be seen in Fig. 4. In it we plot the near infrared colors of all sources for which such information is available from Table 1, together with the locus occupied by normal main-sequence dwarfs (Bessell & Brett 1988), by classical T Tauri stars (Meyer et al. 1997) and a reddening vector (Rieke & Lebofsky 1985) whose length corresponds to AV = 10 mag and whose origin lies at the position of a dereddened M6 dwarf. Such a diagram provides an easy way of identifying sources with infrared emission of circumstellar origin, which are those lying to the right of the reddening vector as plotted in the figure (Lada & Adams 1992).
The filled circles in Fig. 4 represent the sources with a
H
emission equivalent width exceeding 100 Å.
Par-Lup3-4 is represented twice, corresponding to the
JHK photometry published by Nakajima et al. (2000)
and the one obtained by us in July 2002. With the single exception
of Sz 113, all the filled circles fall in the region
denoting the existence of infrared excess. On the other hand, it
is worth noticing that the JHK photometry of LS-RCrA 1
published by Fernández & Comerón (2001) places
it at virtually the same position as Sz 113, indicating
that the correlation between strong H
emission and
near-infrared excess suggested by Fig. 4 is not strict, and that
strong emission lines relative to the continuum can be observed
also among objects with no traces of near-infrared excess. It is
interesting to notice also that all the stars with large H
equivalent widths, with the exception of the extreme case of
Sz 102, lie close to the unreddened locus of
classical T Tauri stars.
One of the most interesting features of the present sample is
that the two sources with the largest H
equivalent width,
Sz 102 and Par-Lup3-4, are also the faintest
ones at infrared wavelengths. Sz 102 is also the bluest
in
,
which is partly due to the intense emission lines
that dominate over the photospheric emission in the R band
(inducing in the case of Sz 102 a bluening of 0.45 mag in
)
and partly due to intense veiling. The bluening in
due to the emission lines in the R band also affects
noticeably the colors of Sz 106 (0.08 mag),
Sz 113 (0.09 mag), Par-Lup3-4 (0.37 mag), and
LS-RCrA 1 (0.31 mag). The faintness of
Par-Lup3-4 may be partly explained by its late spectral
type, but the sample contains at least 7 objects that are both
later and much brighter, including the latest object of the sample
that we estimate to be 2.5 subclasses later. We note that
Sz113, which we classify as M6, also has strong H
and CaII emission and is over one magnitude fainter than the rest
of the objects of similar spectral type at
and J. These
differences tend to maintain themselves or decrease only slowly as
we move towards longer wavelengths, contrarily to what would be
expected if they were due to different amounts of extinction along
the lines of sight towards each object.
It must be noted that the truly anomalous property of the sources
with large H
equivalent widths is the low luminosity of
the underlying continuum, rather than the intrinsic intensity of
the H
emission. A comparison of the quantity
,
which provides a measure
of the absolute intensity of the H
emission, shows that
Par-Lup3-4 and LS-RCrA 1 display H
intensities similar to those of Sz 109,
Par-Lup3-2, and Par-Lup3-3, which are in the
same spectral type range but have normal luminosities.
Sz 100, Sz 112, and Sz 114, having
spectral types M5-M6, have higher H
fluxes but no signs of
underluminosities.
To better quantify the apparent underluminosity of the sources
with the largest H
equivalent widths, we have followed the
same procedure as outlined by Fernández &
Comerón (2001) to place all the M-type objects of
our sample in a temperature-luminosity diagram enabling comparison
to pre-main sequence evolutionary tracks. The estimate of the
extinction AV towards each source is made by comparing the
(J-H) color to an assumed intrinsic (J-H)0 using the Rieke &
Lebofsky (1985) extinction law, which gives
AV = 9.36 [(J-H) - (J-H)0]. For the intrinsic colors we take
as a base those of Bessell & Brett (1988), to which
we add
mag in order to match the colors
adopted at spectral type M6 by Comerón et al. (2000)
for the Chamaeleon I population. Such offset is intended to
roughly account for the redder intrinsic colors of young, lower
surface gravity objects at a given spectral type, as described by
Comerón et al. (2000). The values of AVcalculated in this way are thus 1.0 mag lower than those that
would be obtained directly using Bessell & Brett's colors with no
further correction. Even lower values, closer to zero, are
obtained by dereddening the positions of the objects in Fig. 4
to the position of the T Tauri star locus, in which
(J-H)0 is higher, although the applicability of this line to
objects as late as the ones discussed here may be questioned. The
results are given in Table 5. We have adopted AV = 0 for those
sources for which we obtain negative values of the extinction. In
general such negative values do not exceed 0.6 mag and may be
attributed to errors in the photometric measurement or in the
assumed intrinsic colors. The exceptions are Sz 109 and
Par-Lup3-2, for which we obtain
AV = -2.7 and -2.4,
respectively. We also adopt AV = 0 for Par-Lup3-1,
consistent with its visible colors and its location in a zone of
low extinction on the background.
The reader is referred to Fernández & Comerón (2001) for a discussion on the temperature scale and bolometric corrections that we also use here. We note that, although systematic differences may exist between the adopted and the actual values of temperatures and bolometric corrections, such differences should affect in a similar manner all the objects under consideration and thus should have virtually no impact on the evaluation of the underluminosity on which we focus our discussion. On the other hand, the choice of the J band as the reference for the derivation of the luminosity reduces the impact of possible errors in the derivation of the extinction, which we obtain using the (J-H) color, and of the possible contribution of circumstellar material to the flux. We have excluded from this discussion Sz102, whose spectrum is likely to be heavily veiled due to accretion, thus making the spectral classification very uncertain. In any case, it is most probably earlier than M and possibly close to the K0 spectral type assigned to it by Hughes et al. (1994).
In the case of Par-Lup3-4, the strong variability in infrared colors adds an extra difficulty to the estimate of extinction and luminosity. The value of AV obtained from the (J-H) color in Table 1 is 5.6, but it reduces to 2.4 if the colors of Nakajima et al. (2000) are adopted. However, the higher derived extinction when the object is fainter and redder turns out to produce a difference of only 0.21 mag in the dereddened magnitude J-AJ, thus making the luminosity estimate little dependent on the chosen set of JH magnitudes.
![]() |
Figure 5:
Temperature-luminosity diagram showing the positions of
the objects listed in Table 2, with the exception of
Sz102. The superimposed curves are the pre-main sequence
evolutionary tracks of Baraffe et al. (1998). As
discussed in Sect. 2.3, a distance of 200 pc has been adopted
for the Lupus 3 star forming cloud. Full circles indicate
H![]() ![]() ![]() ![]() |
Open with DEXTER |
The temperature-luminosity diagram is shown in Fig. 5 where, to
further illustrate the point, we have also included the position
of LS-RCrA 1, taken directly from Fernández &
Comerón (2001). Overplotted in that figure are the
pre-main sequence evolutionary tracks of Baraffe et al.
(1998) that may be used in principle to estimate the
age and mass of each object. All the objects with
Å tend to lie in a small range of ages below 2
Myr (note that the precise value depends on the adopted distance,
as discussed in Sect. 2.3). LS-RCrA 1 and
Par-Lup3-4, the objects with the most intense emission
lines in the diagram, fall significantly below the isochrones
corresponding to the ages around 10 Myr or less that should be
expected for the members of this star forming region. A similar
but less marked trend is seen in Sz106 and
Sz113, the next objects in order of decreasing H
equivalent width. This trend is further illustrated in Table 5,
where we list the masses and ages obtained by using the
evolutionary tracks of Baraffe et al. (1998). These
values are only indicative of the true physical parameters of the
observed objects, given the sources of uncertainty involved in
evolutionary calculations at such early ages (Wuchterl
2001; Baraffe et al. 2002) even
regardless of the effect of accretion. The correlation between
H
equivalent width and derived age (in a direction which
is furthermore the opposite from what would be expected if younger
objects had higher accretion rates) and the derivation of
unrealistic ages for members of a star forming region having
extreme H
equivalent widths obviously calls for a
cautionary use of pre-main sequence evolutionary models when
evaluating the physical parameters of such objects. We thus
consider that the values given in Table 5 are reliable only for
objects having H
equivalent widths below
100 Å.
It should be noted that one of such objects, Par-Lup3-1,
is a bona fide brown dwarf judging from the derived mass and
from its M7.5 spectral type together with its young age.
Table 5: Derived masses and ages using the Baraffe et al. (1998) evolutionary tracks
Deep pointed observations of Lupus 3 in the ROSAT archive obtained with the Position-Sensitive Proportional Counter (PSPC) include all the objects discussed in this paper. Merging together all the existing PSPC datasets (200348p, 200351p, 200358p, 200359p, 200736p, 201209p, 201210p, 201211p, 201213p, 201213p-1, and 201276p) which amount to approximately 15 ksec of integration time, it is thus possible to obtain detections or upper limits for our sources in order to study the correlation between the X-ray luminosity and the bolometric luminosity or other signs of activity. We used the EXSAS source detection routine running under MIDAS to perform local and map maximum likelihood source detection as described in Comerón et al. (2000), and considered association with a visible source if the positional coincidence between the optical and the X-ray position is better than 40''. In addition, Sz 108 (both components unresolved) is detected in a shorter integration pointing of the ROSAT HRI (High Resolution Imager; dataset 201649h) within 3'' from the optical position.
Besides the HRI detection of Sz 108, only Sz 112 is detected in the ROSAT observations. A strong detection is obtained within the PSPC error box of Sz 106, but the nominal X-ray position is closer to HR 5999 than to Sz 106 and we thus consider than an association with the much brighter star HR 5999 is far more likely. For all the other objects only upper limits to the X-ray luminosity are obtained. To convert the observed fluxes or upper limits to X-ray luminosity, we used a conversion factor of 10-11 counts cm2 erg-1 based on the assumption that the X-ray emission can be approximated by a typical T Tauri-like Raymond-Smith spectrum with a plasma temperature of 1 keV, as justified in Comerón et al. (2000), corrected for the approximate value of the extinction derived as described in Sect. 3.4.
Table 6: X-ray luminosities.
The X-ray luminosities and upper limits are given in Table 6,
together with the maximum likelihood parameter, ML, for the two
detected sources (where ML = 5 corresponds to a detection). The upper limits are generally very close to the
normal level of emission from very low mass young stars and to
those of the actually detected objects (
), suggesting that a somewhat deeper integration may add
some more detections.
The relatively high upper limits that we find for Sz 113
and Par-Lup3-4 are due to their low bolometric
luminosities, and would be similar to those found for the other
objects if their bolometric luminosities were replaced by those of
objects with weak H
of the same spectral type in the
region. This may not be true however for Sz 102: if its
luminosity were replaced by that of a
1 Myr-old K-type
star, its non-detection implies
,
or
about one order of magnitude below the expected level.
Par-Lup3-4 and LS-RCrA 1 seem to be so far
the only examples of mid-M type objects with such strong emission
lines relative to an unusually faint photospheric continuum.
Equivalent widths of several hundred Å have been reported in
the last years also for a number of very late-type members of
star-forming regions with no traces of underluminosity (Martín
et al. 2001; Briceño et al. 2002;
Zapatero Osorio et al. 2002; Barrado y Navascués et al. 2002; Luhman et al. 2003). All those
cases correspond to objects with spectral types later than M7,
and inspection of the published spectra shows that the large
equivalent widths are due to the decreasing flux in the local
(pseudo-) continuum at H
rather than to the strength
of H
,
the
ratio being several
times smaller than for Par-Lup3-4 and LS-RCrA 1.
However, it is worth mentioning the recently reported case of
KPNO Tau 12 (Luhman et al. 2003), a M9 young
brown dwarf with strong H
emission as well as detectable
HeI and CaII triplet emission, whose position in the
temperature-luminosity diagram also falls significantly below the
isochrones around which other young members of the Taurus clouds
tend to appear.
The other object with extreme H
emission in our sample (in
this case concerning both relative and absolute intensity of
the emission lines) is Sz 102, whose apparent
underluminosity given its spectral type was already pointed out by
Hughes et al. (1994). These authors noted that the
geometry of its associated outflow (Krautter 1986)
was suggestive of an edge-on disk perpendicular to it. Hughes et al.
thus suggested that the reason for the apparent
underluminosity of Sz 102 could be due to the fact that
we may be seeing it in scattered light, rather than directly. Many
other young stellar objects surrounded by edge-on disks are
currently known (e.g. Burrows et al. 1996; McCaughrean
& O'Dell 1996, Stapelfeldt et al.
1998; Woitas & Leinert 1998, Monin &
Bouvier 2000; Brandner et al. 2000;
Jayawardhana et al. 2002; Grosso et al.
2003) or suspected (Martín 2000).
Can the underluminosity of the Lupus sources discussed here,
LS-RCrA 1, and KPNO Tau 12 be also due to a
particular viewing geometry rather than to the intrinsic
properties of the object? In principle, an edge-on disk may
enhance the intensity of emission lines by blocking the light from
the central object if they formed in a region well detached from
the photosphere, not occulted by the circumstellar disk, so that
even moderate emission-line luminosities may produce large
equivalent widths when superimposed on a photospheric continuum
seen only in scattered light. Evidence for such an extended
emitting region is indeed provided by intense forbidden lines in
both Par-Lup3-4 and LS-RCrA 1, as discussed in
Sect. 3. However, the excitation conditions of the gas in the
extended, low-density regions of the disk wind set a limit to how
much H
flux can be contributed by them. Indeed, as noted
by Fernández & Comerón (2001) the
H
/[SII] ratio in LS-RCrA 1 is 11, far above the
value of 4 that is found in the highest excitation Herbig-Haro
objects. A lower, but still high ratio of 6.7 is measured in
Par-Lup3-4, similar to that of Sz 102. The
ratios are much higher for the other objects displaying both a
certain degree of underluminosity and strong H
emission,
Sz 106 and Sz 113, where forbidden lines are
either faint or not detected at all. Interestingly, the situation
is different for two well-known objects having confirmed edge-on
disks, HH30 (Cohen & Jones 1987) and
HV Tau C (Monin & Bouvier 2000), indicating
that in those two cases the complete emission-line spectrum,
including H
,
can be explained by the disk wind only. On
the opposite, it appears that, for all the objects discussed here,
most or nearly all of the H
emission is contributed by a
high excitation, high density emitting region that must lie closer
to the object such as the accretion columns extending up to the
inner edge of the truncated accretion disk only a few object radii
away, or the basis of a jet.
While H
emission is the most easily observable indicator
of inflow and outflow activity, it is also a rather poor
quantitative indicator of those phenomena due to the broad
diversity of environments in which H
emission is produced.
Stronger constraints are provided by the CaII and HeI lines, whose
origin in accretion near the object surface is much better
established (e.g. Alencar et al. 2001), and whose
origin in outflows would be harder to account for in view of the
high temperatures and mass loss rates needed (Hartmann et al.
1990; Beristain et al. 2001). For the
CaII infrared triplet lines in our objects, their ratios close to
1:1:1 indicate that they are formed near the photosphere of the
central object rather than in an extended low density envelope
(Graham & Heyer 1988). Observational evidence for the
intimate connection between these lines and the accretion process
is provided by the observed correlation between variations in the
red wing of H
and the strength of the CaII broad component
in BP Tau (Gullbring et al. 1996) or by the
correlation between the strengths of CaII and HeI in time series
observations of DF Tau (Johns-Krull et al.
1997), besides the already noted correlation between
veiling and CaII emission investigated by Muzerolle et al.
(1998b) that we used in Sect. 3.3.1 to derive
accretion rates. Also the appearance of HeI emission, needing high
excitation temperatures that can probably happen only at the bases
of the accretion columns, argues for its origin near the surface
of the object.
Since these lines most likely form in the immediate vicinity of the
central object, they should be also
obscured by a possible edge-on disk if that were the cause of the
underluminosity. The prominence of these lines in Sz 113,
Par-Lup3-4, and LS-RCrA 1 thus indicates that
they are intrinsically intense as compared to the underlying
continuum. In other words, our data argue in favor of
intrinsically high
ratios (where
and
are respectively the luminosity of any of the
emission lines formed near the surface of the object and the
bolometric luminosity) for the objects considered here, regardless of whether or not they are occulted by edge-on disks.
The arguments discussed in the previous paragraph in favor of intrinsically high line-to-continuum ratios only show that such ratios may not be due to partial blocking by edge-on disks, but in principle does not rule out the possibility that the overall underluminosity may be due to edge-on disks blocking both the continuum and the line formation region. If this were the cause, normally luminous objects with similar emission line to bolometric luminosity ratios should be much more common (and brighter, hence easier to identify) than the underluminous ones, since the latter would be only the result of a particular viewing geometry. However, the fact that no late-type objects with such extreme emission line spectra and normal luminosity have been identified leads us to consider it highly unlikely that edge-on disks are the cause of the phenomena described here.
Another interesting possibility is that the apparently underluminous objects are actually fully embedded in a thick circumstellar envelope and undergoing the very high accretion and mass outflow rates characteristic of Class I sources, which would naturally produce the large emission-line to continuum ratios. Class I objects with spectral types only slightly earlier than those of Par-Lup3-4 or LS-RCrA 1 and very similar emission-line spectra have been observed in the visible by Kenyon et al. (1998). The observed emission does not come in this case directly from the central star, which is hidden from view by extinctions in the AV = 30-60 mag range. Instead, the star is seen via the light scattered by the walls of cavities in the envelope having a relatively unimpeded line of sight toward the central star. This indirect view of the star through scattered light would in turn explain the apparent underluminosities. Some of the spectra presented by Kenyon et al. (1998), in particular those of HH31 IRS2 and 04158+2805, bear indeed a great resemblance to those of Par-Lup3-4 or LS-RCrA 1, although without the extreme relative emission line strengths that we measure in our sources.
However, a distinctive characteristic of optically visible Class I sources is that they are spatially extended and resolved at the distance of nearby star forming regions. Kenyon et al. (1998) find optical sizes normally in the 5''-10''range for their Taurus sources, which would be easily resolved also in our Lupus images despite of the somewhat larger distance. These are also typical sizes expected from models of the appearance of circumstellar envelopes in the near-infrared (Whitney et al. 1997). However, none of our underluminous sources, including LS-RCrA 1, shows any hints of resolved circumstellar nebulosity. The image quality of our observations of Lupus 3 is not very good (FWHM of 1''4 in both our visible and near-infrared images), but should suffice to resolve nebulosity of the same physical size as that of the Taurus sources. Much more constraining are our near-infrared observations of LS-RCrA 1, as noted in Sect. 3.3.2, which allow to set an upper limit of only 0''38 (60 AU) to the extent of any possible infrared nebula around this object. As noted by Whitney et al. (1997), the existence of extended nebulosity is an unambiguous criterion for the distinction between Class I and Class II sources, and none of our objects can be classified as Class I according to it.
Class I sources also tend to have conspicuous near-infrared
excesses, as confirmed by both observations and models (Whitney et al. 1997). As discussed in Sect. 3.3.3, we do see a
trend among sources with high H
equivalent widths towards
having near-infrared excesses (see also Fig. 4). However,
leaving aside the most extreme case of Sz 102, two of our
sources (Par-Lup3-4 and Sz 106) have colors
significantly bluer than the average of the Whitney et al.'s
sample, and two other (Sz 113 and LS-RCrA 1)
display no signs of infrared excess at all. Moreover, as noted in
Sect. 3.3.3, all the late-type sources with large H
equivalent widths occupy positions in the (J-H), (H-K) diagram
close to the locus of unreddened classical T Tauri stars
(class II objects) as observed by Meyer et al. (1997).
Such lack of strong infrared excess is reproduced in the Class I
object models of Whitney et al. only when the line of sight is
approximately coincident with the axis of the dust-free bipolar
cavities, thus providing a nearly unobscured view of the central
object. However, reproducing the underluminosity that we observe
among our sources does require a significant amount of
obscuration, and therefore most probably redder infrared colors
than the ones that we observe.
Although the available imaging observations do not favor the interpretation of the observed features as due to a possible Class I nature of the objects under discussion, it is interesting to note that a more definitive proof could be provided by polarimetric observations offering the possibility of probing the properties of circumstellar envelopes (Whitney et al. 1997).
A third possibility to be considered is that the evolution of
our sources along their convective tracks may have been
substantially modified by accretion. The impact of sustained
accretion on a limited area of the stellar surface after reaching
the birthline (Stahler 1983) has been studied by
Hartmann et al. (1997; see also Siess et al.
1997). The results for stars with initial masses of
0.3
and above show that the luminosity evolution of a
star is accelerated by accretion, in a way such that the star
appears in the
diagram at the position that a
non-accreting star with the same instantaneous mass, but an older
age, would have. Observationally, the difference between accreting
and non-accreting stars at a given spectral type is that the
former started their post-birthline evolution with a lower mass,
and are at a given time less luminous than the non-accreting ones.
While this is a good qualitative description of what is actually
observed in our sample, the magnitude of the observed effect is
huge (in terms of difference in L at a given
)
compared to the quantitative effects estimated by Hartmann et al.
(1997). Given the reasonable assumptions of that work
regarding the treatment of accretion, it seems very difficult with
such an approach to reach the level of the observed effects.
However, Hartmann et al. limited their study to initial masses
larger than 0.15
.
As Table 5 indicates, the initial
masses at the birthline of the objects that we discuss here may be
far below that value. The effect of accretion may conceivably be
more dramatic for such low initial masses closer to the brown
dwarf regime, even with the moderate (in absolute terms) accretion
rates that we estimate in Sect. 3.3.1. Given the present data
and uncertainty of mass estimates, this conclusion remains
speculative and deserves an extension of the work by Hartmann et al. (1997) to lower masses. This work is under
progress. On the other hand, we must note in this context the
intriguing non-detection in X-rays of the most extreme
emission-line object in the region, Sz 102, which
suggests that its underluminosity in
is corresponded by
an underluminosity in X-ray emission as compared to typical low
mass objects of the same age. Such correspondence can be readily
explained by an edge-on disk or a circumstellar envelope that
should also attenuate the X-ray emission produced by chromospheric
activity in the central object, while it is less straightforward
to explain how accretion-induced underluminosity may also decrease
the level of X-ray emission. An X-ray detection (or a more strict
upper limit) of Par-Lup3-4 or LS-RCrA 1 would be
valuable, as it would decide whether or not Sz 102 may be
regarded as a special case.
What may be the implications of the existence of such low mass
underluminous objects on the study of the mass and age
distributions in star forming regions? As Fig. 5 shows, if our
favored hypothesis of accretion-modified evolution is the correct
one to explain the existence of such underluminous objects,
existing evolutionary tracks that do not include accretion are of
little use for the derivation of the physical parameters of such
objects, in agreement with the cautionary advise of Hartmann et al. (1997) on the interpretation of ages in star
forming regions. These objects are easy to exclude from the
samples used to study the mass function and age distribution of
star forming regions provided that spectroscopy is available. They
may represent however a contamination in samples selected on the
bases of their near-infrared colors for which spectroscopy in the
visible is not available, a frequently encountered case and an
unavoidable one when studying heavily embedded aggregates. The
scarcity of very low mass objects with H
equivalent widths
much above 100 Å, such as Par-Lup3-4 and
LS-RCrA 1, suggests in principle that such contamination
may not be very important. Nevertheless, the possibility that the
effects of strong accretion during early evolution may remain
after accretion has ceased needs to be taken into account. It
should be noted in this context that temperature-luminosity
diagrams of star forming regions often show outliers falling well
below the band defined by the bulk of the members, but whose
spectral characteristics are otherwise undistinguishable from
those of apparently younger members of the same spectral type. The
explanation of such outliers is unclear, and in general it may not
be excluded that they are the result of an earlier episode of star
formation. However, in the light of the results presented here, we
offer the alternative explanation that they may actually be
post-accretion objects.
In this paper we have reported on new slitless and multi-object
spectroscopy observations of the central area of the Lupus 3 star
forming cloud in a search for new members characterized by their
H
emission. The observations have revealed four previously
unidentified members having spectral types between M4.5 and M7.5,
the latter one being a brown dwarf.
The spectra of previously known and new members present a wide
range of H
equivalent widths reaching up to 430 Å. The
new object with the largest H
equivalent width,
Par-Lup3-4, displays a rich emission-line spectrum
indicating both accretion and mass loss, as well as strong
variability in both brightness and color in the near-infrared and
moderate near-infrared excess. The strength of the emission lines
and the late spectral type of the underlying continuum make it
very similar to a previously reported object, LS-RCrA 1,
the only other mid-M type object displaying such intense
emission-line spectra identified so far. Even more remarkably,
Par-Lup3-4 shares with LS-RCrA 1 the property of
being much less luminous than objects of their spectral types
without such prominent emission-line spectra and a few million
years of age. The same seems to apply to Sz 102, although
its featureless spectrum in the range covered here precludes its
classification. Noticeable underluminosity, although not so
extreme, is also found among other objects in our sample with
H
equivalent widths exceeding 100 Å, Sz 106 and
Sz 113. A comparison among the H
fluxes shows
that the reason for the large H
equivalent widths is the
faintness of the underlying continuum, rather than the absolute
intensity of the emission lines, except in the case of
Sz 102 that is also the strongest H
emitter of
our sample.
Based on the correlation between CaII emission and mass accretion
rate, we estimate accretion rates between
and
yr-1 for the objects in our
sample. An upper limit of several times
10-8
yr-1 is found for the mass loss rate of
Par-Lup3-4, although much lower values are possible in
the likely case that the forbidden-line emission is dominated by a
low velocity component.
We have considered different possible causes for the underluminosity
of objects with large H
equivalent widths. The strength of
H
,
the CaII infrared triplet, and HeI with respect to the
forbidden line spectrum leads us to the conclusion that most of
the H
luminosity is probably contributed by high density
regions near the surface of the star, and that the high
is an intrinsic feature of these objects
rather than a geometric effect caused by partial obscuration by an
edge-on disk. We also find it unlikely that these objects are
embedded Class I sources seen through scattered light, due to the
lack of spatially extended nebulosity and to the moderate or
non-existent infrared excesses among them. A third possibility is
that these objects started their post-birthline evolution at very
low, probably substellar masses, and that further accretion has
modified their evolution making them appear at present much older
than they actually are given their mass. This effect is suggested,
although with less dramatic effects, by evolutionary models
incorporating accretion at higher masses, and an extension of
those works to the substellar range may be needed to confirm that
it is an adequate explanation in the cases discussed here. We
speculate that the members of star forming regions sometimes found
with apparent ages well above those of most of the other members
might actually be post-accretion objects.
Two of the previously known objects, Sz 108 and Sz 112, are detected in X-rays with emission levels similar to those observed in low mass stars in other star forming regions. Upper limits available for the other sources discussed in this paper are also consistent with those levels, although only in one case (Sz 102) does the upper limit suggests that the global underluminosity is corresponded by an underluminosity also in X-ray emission.
The detailed modelling of T Tauri stars that high resolution spectroscopy and time monitoring have made possible calls for a similar observational effort focused on objects such as those described here. Also, the distance to regions such as Lupus and R CrA may allow the resolution of circumstellar disks by adaptive optics imaging, and therefore a direct test of the possibility that underluminosity may be caused by edge-on disks. Another attractive observational approach is polarimetry, which offers the possibility of probing the existence and geometry of circumstellar envelopes. If the underluminosity turned out to be real, as we propose in this work, the observations of these objects will provide important constraints to new generations of pre-main sequence evolutionary models including accretion.
Acknowledgements
We are pleased to thank the staff of Paranal Observatory for their support during our observations, especially Drs. Hermann Böhnhardt and Olivier Marco. We are particularly thankful for their acceptance of our request for a change of instrument coordinated with another program of ours scheduled during the same nights, despite of the increased support effort involved. The support received at the La Silla Observatory by the 2.2 m and the NTT Teams, and in particular Drs. Lisa Germany, John Pritchard, Malvina Billères, and Olivier Hainaut, is also gratefully acknowledged. Extensive and constructive comments by an anonymous referee led to substantial improvements in the contents of this paper. IB acknowledges discussions with Lee Hartmann and Kevin Luhman on the edge-on disk hypothesis. MF has been partially supported by the Spanish grant AYA2001-1696. RN wishes to acknowledge financial support from the Bundesministerium für Bildung und Forschung through the Deutsche Zentrum für Luft- und Raumfahrt e.V. (DLR) under grant number 50 OR 0003.