Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.arcetri.astro.it/~lt/preprints/s235/s235.ps.gz Äàòà èçìåíåíèÿ: Tue Sep 11 16:56:43 2007 Äàòà èíäåêñèðîâàíèÿ: Sat Dec 22 06:53:41 2007 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: m 35

A&A manuscript no.
(will be inserted by hand later)
Your thesaurus codes are:
09 (09.08.1; 09.13.2; 09.10.1; 08.06.2; 13.09.6; 09.09.1 S235)
ASTRONOMY
AND
ASTROPHYSICS
28.5.1996
Star formation in the S235A­B complex
Marcello Felli 1 , Leonardo Testi 2 , Riccardo Valdettaro 1 , Jun­Jie Wang 1;3
1 Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, I­50125 Firenze, Italy
2 Dipartimento di Astronomia e Scienza dello Spazio, Universit`a degli Studi di Firenze, Largo E. Fermi 5, I­50125 Firenze, Italy
3 Beijing Astronomical Observatory, Chinese Academy of Sciences, Beijing 100080, P.R. China
May 28, 1996
Abstract. We present near infrared broad band (J,H and
K) and narrow band (H 2 S(1) 1 ! 0 and Brfl) images,
and high resolution molecular observations (C 34 S(2­1), (3­
2), (5­4) and 13 CO(2­1)) around the highly variable H 2 O
maser located between the S235 A and B optical nebulosi­
ties. These observations are part of an on­going search for
the sources of excitation of H 2 O masers in regions of star
formation or, alternatively, for the earlieast evolutionary
phases of massive stars.
We confirm the presence of a highly obscured stellar
cluster between S235 A and B and, from the colour­colour
analysis, we show that the cluster contains many sources
with infrared excess, which are believed to be Young Stel­
lar Objects (YSOs) in an early evolutionary stage.
Diffuse Brfl emission is found mainly in the vicinity
of S235 A, and unresolved Brfl emission is found coinci­
dent with S235 B. Hot molecular hydrogen emission is dis­
tributed around the S235 A nebula, especially in a belt­
like region to the south of S235 A, at the edge of the HII
region.
The driving source of the H 2 O maser does not appear
to be either the YSO inside S235 A or S235 B, but is iden­
tified with a faint near infrared member of the cluster, with
a large (H--K) colour excess, located near the position of
the maser. A hot dust envelope around an early type star
may be the source of the near IR emission. This identifi­
cation is supported by the coincidence of the maser and
the near IR source with the center of a high density and
compact molecular core observed in C 34 S and 13 CO. The
lack of radio continuum emission from the area around the
maser suggests that the star powering the maser and re­
sponsible for the near IR emission must be in a very early
evolutionary stage, highly obscured even at K band and
surrounded by an envelope with such a high density that
any radio continuum emission is strongly self­absorbed. In
any case the evolutionary status of such a star is much ear­
lier than those of the exciting stars of S235 A and S235 B.
Strong variability of the maser emission and large velocity
Send offprint requests to: M. Felli
differences of the maser features with respect to the molec­
ular cloud velocity imply the presence of highly collimated,
energetic and short duration jet activity in this YSO. The
more evolved members of the cluster S235 A and S235 B
lie on the sides of the molecular core, suggesting that star
formation in the cluster is not coeval but proceeds from
the outside towards the core of the molecular cloud.
Key words: HII regions ­ molecules ­ jets and outflows ­
star formation ­ infrared sources: stars ­ individuals: S235
1. Introduction
A reference working model for the cradle of newly born
luminous stars has been around for many years. Since the
star itself is not visible, all the attention is focused on
the cradle itself, i.e.: UC HII regions, hot molecular cores,
cool dust envelopes responsible for the FIR emission, hot
dust cocoons bright in the near IR, bipolar outflows, H 2
jets, masers, etc. However, many of these features usually
occur on widely different scales. Consequently, the corre­
lation among them may often be confused, and misleading
conclusions can be reached, because of the lack of match­
ing (high) resolutions.
Different and complementary approaches have been
used to search for the earliest phases of massive stars. Sev­
eral pieces of evidence suggest that H 2 O masers are one
of the best indicators for selecting the target fields and
that the study of their association with near IR sources,
UC HII regions and hot molecular cores are the neces­
sary follow­ups to unveal the earliest YSOs. In fact, even
though originally H 2 O masers were discovered in diffuse
HII regions, it has now become clear from surveys of IRAS
selected sources (Palla et al. 1993) that H 2 O masers not
associated with diffuse HII regions may be the majority,
up to 80%, suggesting that the maser emission occurs in
an earlier phase, much before the onset of an (UC) HII
region (Codella et al 1994, Codella & Felli 1995). Also,

2 M. Felli et al.
on simple energetic arguments (Elitzur et al. 1989), H 2 O
masers must be very close to the stellar source of energy
(from 10 14 to 10 16 cm). Consequently, VLA maser po­
sitions with a precision better than 0:1 00 offer the most
accurate locations to be searched.
The new scenario that emerges from these studies is
that in the very early stages bounded ionized winds or
UC HII regions maybe present around the star and close
to the H 2 O maser, but so small and optically thick that
they are undetectable in the radio continuum. These early
stages are usually highly obscured also at K band and
maybe observable in the near IR thanks to the emission
of hot dust near the star, or in molecular lines sensitive
to high densities (Cesaroni et al. 1994). As the ionized re­
gion expands, the HII region becomes detectable, the H 2 O
maser disappears and the K band dust emission strongly
decreases (Testi et al. 1994; Hunter et al. 1995; Tofani et
al. 1995; Felli et al. 1996).
In the present paper we test these ideas on the S235 A
­ B star forming complex. S235 is the most prominent of
a group of optical nebulosities which lie toward the an­
ticenter of our Galaxy. The area around S235 contains
traces of both advanced evolutionary stages (S235 itself)
and less evolved stages such as three small optical nebu­
losities: S235 A and S235 B about 10 0 south of S235 and 1 0
apart from each other, and S235 C, 3 0 further to the south.
S235 is connected to an extended molecular cloud at ­20
km s \Gamma1 , while S235 A, S235 B and S235 C are embedded
in a smaller molecular cloud at ­17 km s \Gamma1 , elongated in
the north­south direction and with a maximum extension
of ¸ 10 0 . The region has been mapped in several molecular
lines, but with a resolution of the same order as the mutual
separation between S235 A and S235 B, so that it is dif­
ficult to make precise association of molecular peaks with
either of the two nebulosities (Evans & Blair 1981; Ho et
al. 1981; Sandell et al. 1983; Lafon et al. 1983; Stutzki et
al. 1984; Nakano & Yoshida, 1986 [hereafter NY]; Wilking
et al. 1989; Snell et al. 1990; Plume et al. 1992). S235 A
and S235 B are also strong near IR sources, known as IRS3
and IRS4 (Evans & Blair 1981; Evans et al. 1981). Both
are associated with Hff emission (Krassner et al. 1982),
but only S235 A has corresponding radio emission (Israel
& Felli 1978), which suggests a classical HII region na­
ture. S235 B, despite many attempts, remains undetected
in the radio continuum. Its nature is still a puzzle, expe­
cially considering that it has strong Brff and Brfl emission
(Krassner et al. 1982; Thompson et al. 1983). A highly
self­absorbed UC HII region or a stellar wind have been
put forward as possible explanations, but are somehow in
contradiction with its bright aspect in Hff, which would
exclude a strong local obscuration. S235 C is also detected
in the radio continuum, has a partial shell morphology and
is associated with a HH object.
In between S235 A and S235 B a highly variable H 2 O
maser has been detected (see e.g. Persi et al. 1994, To­
fani et al. 1995). No radio continuum emission from a
point source (Ÿ 1 00 ) down to 0:3 mJy at 8:4 GHz has been
found in a region ¸ 10 00 in radius around the maser (To­
fani et al. 1995). Three possibile sources of excitation of
the maser have been considered: 1) the S235 A HII re­
gion, 2) the S235 B YSO and 3) a so far undetected YSO
closer to the maser (Tofani et al. 1995). While the first one
is improbable based on simple morphological arguments
(basically the large distance of the maser from the outer
boundary of the HII region), the remaining two (as well
as the true nature of S235 B) demand more observations.
We present here arcsec resolution near IR images, high
resolution molecular observations and an extended patrol
of the H 2 O maser emission which can improve our under­
standing of this star forming region and possibly reveal
the true location of a new YSO in a very early phase.
A distance of 1.8 kpc will be assumed, following NY,
for the molecular cloud and the stellar cluster .
2. Observations
2.1. J H K broad band images
The near infrared observations were carried out on
February 12, 1994 with the 1.5 m TIRGO 1 telescope at
the Gornergrat Observatory and with the Arcetri near­
infrared camera ARNICA (Lisi et al. 1993; Hunt et al.
1996a). The camera is based on a NICMOS3 256\Theta256
pixel HgCdTe array detector developed by Rockwell Inter­
national. The scale on the detector is 0:95 00 =pixel, hence
the field of a single frame covers about 4 0 \Theta 4 0 . For each fil­
ter thirteen offset frames were taken, all of which contain
the S235 A­B region. The standard star AS 10 from the
ARNICA list (Hunt et al. 1996b) was observed for photo­
metric calibration.
All the observed images were reduced by using the AR­
NICA (Hunt et al. 1994) and IRAF software packages. Sky
subtraction and flat fielding were performed using the me­
dian average image of the thirteen frames. We combined
all the images and produced a 7 0 \Theta 7 0 mosaic image. Aper­
ture photometry was performed using the APPHOT rou­
tines of IRAF. The limiting magnitudes in our mosaic im­
age are 17.3, 16.2 and 16.0 mag (5oe in 4 00 aperture) in J,
H, K bands, respectively.
One of the sources (S235 B?, in the following we shall
indicate with ? the stars within the nebulosities, to dis­
tinguish them from the diffuse emission) was so bright at
K that it was saturated in our images; for this reason a
shorter integration image was taken at K on the February
16, 1994. This image had a much lower sensitivity and was
used only to determine the K magnitude of S235 B?.
Careful astrometry of the near infrared images was
made with the Digitized Sky Survey of the Space Tele­
scope Science Institute (Testi 1993). The astrometric cal­
ibration error is less than one arcsecond.
1 The TIRGO telescope is operated by the C.A.I.S.M.I.--
C.N.R., Firenze, Italy

M. Felli et al. 3
20 s
25 s
30 s
35 s
40 s
45 s
5 37 50
h m s
+35 37'
o
38'
39'
40'
41'
42'
43'
Fig. 1. The full mosaic image in K­band centered on the
S235 A­B region. Axes are Right Ascension and Declination
at 1950.0 equinox.
In Fig. 1 the full mosaic image in K­band centered on
the maser position is presented. Three diffuse nebulosities
can be seen: S235 A and S235 B in the center and S235 C
in the southern part. From the figure we can also distinctly
see that there is a stellar cluster in the center of the region,
which is highly obscured optically.
2.2. Narrow band images
Narrow band images were obtained at the TIRGO tele­
scope with ARNICA on the November 8, 1995, at the
H 2 S(1) 1 ! 0 and at the Brfl wavelengths. The plate scale
of the instrument is the same as for the broad band obser­
vations, but the useful field of view is ¸ 2 0 \Theta 2 0 (Gennari
& Vanzi 1994). The data reduction has been performed
in the same way as for the broad band data, but, due to
the smaller field of view, the final mosaics cover only a
¸ 3 0 \Theta 3 0 region around the water maser position.
To calibrate the images and to subtract the contin­
uum emission, the K band image was used. The narrow
band images were first convolved with a gaussian of proper
width in order to match the K band PSF, then flux cal­
ibration was obtained assuming that a set of stars does
not have detectable line emission, hence the flux density
at the wavelength of the narrow band images should be al­
most the same as that at K. After calibration, the K band
image was subtracted from the narrow band images. In
Table 1 the central rest frequencies, the bandwidth of the
filters, and the noise level of the final images are reported.
Due to non perfect PSF matching and saturation of
S235 B?, the continuum subtraction performed as ex­
plained above was not satisfactory in removing strong
point sources from the images. For this reason we tried
also to subtract the two calibrated narrow band images
from each other. Comparing this result with that obtained
subracting the K band image, we found that this method
was more efficient in removing the point sources. In Fig. 9
the narrow band images obtained with this method are
presented.
Table 1. Central rest wavelength, bandwidth and noise in the
final images for the narrow band observations.
Name – (a) \Delta– (a) oe
(¯m) (¯m) (erg cm \Gamma2 s \Gamma1 arcsec \Gamma2 )
H2 S(1) 1 ! 0 2:1249 0:0235 2 \Theta 10 \Gamma16
Brfl 2:1647 0:0224 2 \Theta 10 \Gamma16
(a) Gennari & Vanzi 1994
2.3. Molecular line observations
With a resolution of 33 00 NY found that the position of
the maximum intensity of CS(J=1--0) is located ¸30 00
northeast of S235 B, while the 12 CO emission has a more
elongated shape, extending approximately in the north­
east/southwest direction from the H 2 O maser to S235 B.
They also reported the detection of a 12 CO molecular out­
flow centered on S235 B.
The region between S235 A and S235 B was observed
with a better resolution and sensitivity by Cesaroni, Felli
and Walmsely (1996) as part of a larger survey in several
molecular lines of a selected list of H 2 O masers without
nearby radio continuum emission. Here we report the re­
sults that are relevant to the present discussion.
The observations were made with 30 m IRAM radiote­
lescope of Pico Veleta. The parameters of the observed
lines are reported in Table 2. The pointing was checked
every hour and it was found to be better than 5 00 .
Table 2. List of observed molecules, rest frequencies, forward
efficiency, main beam efficiency and half power beamwidth
Molecule š F eff B eff HPBW
(GHz) ( 00 )
C 34 S (J=2--1) 96.412984 0.92 0.72 25
C 34 S (J=3--2) 144.617141 0.90 0.55 16
C 34 S (J=5--4) 241.016172 0.86 0.37 10
13 CO(J=2--1) 220.398686 0.86 0.41 11
The three C 34 S transitions were observed simultane­
oulsy, while the 13 CO(J=2--1) line was observed in a sepa­

4 M. Felli et al.
rate setup. The alignement between the different receivers
was checked through continuum cross scans on Jupiter,
at the three frequencies of observation, and found to be
accurate to within 4 00 .
The data were calibrated by using the standard chop­
per wheel technique (Kutner and Ulich 1981). The ob­
served line intensities are expressed as main beam bright­
ness temperature (TMB ), corresponding to a temperature
scale corrected for forward scattering and spillover losses,
but not for the coupling of the source to the antenna beam.
The main beam brightness temperature is given in terms
of the effective beam efficiency, B eff , and the antenna for­
ward efficiency, F eff , using the formula TMB = F eff T \Lambda
a =B eff ,
where T \Lambda
a is the antenna tempearture outside the atmo­
sphere corrected for rear spillover and resistive scattering.
The front­end receivers employ SIS mixers having sys­
tem temperature, after correction for atmosphere and tele­
scope efficiency, of 340 K at 96 GHz, 470 K at 144 GHz
and 1500--2200 K between 220--240 GHz. Our spectrom­
eter was a filter bank consisting of 256 channels with 25
MHz bandwidth for 13 CO(2--1) and two filter banks of
512 channels (one of this split into two parts of 256 chan­
nels each) with 512 MHz bandwith for the C 34 S lines.
The corresponding velocity resolutions are 0.14 km s \Gamma1
for 13 CO(2--1) and 0.12, 0.08 and 1.2 km s \Gamma1 for the three
C 34 S lines, respectively.
The integration time was 2 minutes in the total power
mode. The molecular cloud was mapped at offsets of 12 00
with respect to the H 2 O maser position, over an incom­
plete grid of 5\Theta5 positions.
3. Results
3.1. The molecular clump
Figure 2 shows a contour map of C 34 S(J=3--2) superim­
posed on our K band image. The center of the C 34 S molec­
ular core is unresolved and the peak coincides with the
H 2 O maser. S235 B and S235 A are located at the edges
of the molecular core, ¸ 23 00 and ¸ 33 00 away, respectively.
The peak main beam brightness temperature is 4.3 K and
the peak velocity ­17.1 km s \Gamma1 .
Figure 3 shows the profiles of 13 CO(J = 2 \Gamma 1) around
the H 2 O maser position. The peak main beam brightness
temperature (43 K) occurs 12 00 to the north of the maser,
at a velocity of ­17.1 km s \Gamma1 . The velocity integrated in­
tensity distribution (not shown) is extended and similar
to that found by NY in 12 CO, i.e. peaking between the
maser and S235 B. From Fig. 3 we clearly see that a new
component at velocity ¸ ­15 km s \Gamma1 becomes prominent
in the south­west part of the map, close to the position of
S235 B.
As far as the CO outflow reported by NY, the blue­
shifted lobe, from ­24 to ­20 km s \Gamma1 , can be barely seen
also in the 13 CO(J = 2 \Gamma 1) data (see Fig. 4) and it is lo­
cated between the H 2 O maser and S235 B (as also in NY).
3 0 s
3 1 s
3 2 s
3 3 s
3 4 s
5 3 7 3 5
h m s
4 0 "
+ 3 5 4 0 0 0 "
o
2 0 "
4 0 "
+ 3 5 4 1 0 0 "
o
Fig. 2. Contour map of
integrated C 34 S (J=3--2) emission overlayed on the K band
gray­scale image. The cross marks the position of the H2O
maser (ff(1950) = 05 h 37 m 31:86 s , ffi(1950) = 35 ffi 40 0 17:77 00 , To­
fani et al. 1995). Contour levels are from 2:9 to 9:9 Kkm s \Gamma1
in steps of 1 Kkm s \Gamma1 , the thicker contour represents the half
power level and it is at 5.4 Kkm s \Gamma1 . The HPBW is 16 00 . Also
indicated are the three infrared sources (M1, M2, and M3) close
to the maser.
Fig. 3. 13 CO(J=2--1) line profiles over the observed region.
The center of each spectrum is offset in steps of 12 00 in ff; ffi
with respect to the H2O maser position. The velocity of each
box is from ­28 to ­6 km s \Gamma1 , the intensity scale from ­5.3 to
44.5 K

M. Felli et al. 5
However, we do not detect the red­shifted lobe observed
in 12 CO from ­13 to ­6 km s \Gamma1 (which is also weaker in
NY). The red wing of the 12 CO profile may be affected by
a superposition of different velocity components south of
S235 B and has been questioned by Snell et al. (1990) and
by NY themselves. In conclusion, we confirm the presence
of a blue­shifted lobe (at velocities up to 7 km s \Gamma1 from
the rest velocity of the cloud) between the H 2 O maser
and S235 B, but we cannot support S235 B as the unique
source of the outflow. With the present result the possibil­
ity that the origin of the blue outflow coincides with the
H 2 O maser should also be retained.
Fig. 4. Map of the 13 CO(J=2--1) blue lobe. The velocity in­
terval is from ­24 to ­20 km s \Gamma1 . The contour unit is 2,3,4,5 K
km s \Gamma1 . The position of the H2O maser (open triangle), of the
three infrared sources M1, M2 and M3, and of S235 B are also
indicated.
The comparison between the very high peak 13 CO(J =
2 \Gamma 1) brightness temperature (43 K) with the peak tem­
perature in the 13 CO(J = 1 \Gamma 0) line (8 K) and with the
peak temperature in the 12 CO(J = 1 \Gamma 0) line (25 K)
observed with lower resolution by NY can give us some
indications on the molecular core density.
In the hypothesis of a constant temperature cloud we
shall consider two possibilities: 1) the 13 CO(J = 2 \Gamma 1) line
is optically thick, 2) the 13 CO(J = 2 \Gamma 1) line is optically
thin. In the first case the (1­0) and (2­1) 13 CO tempera­
tures can be reconciled only by a beam filling factor effect
and implies a very small source size. A beam filling factor
¸ 1=5 is necessary, which is compatible with the ratio of
the two beam areas,
\Gamma 12
33
\Delta 2
= 1
7:5 , and implies a source size
less than 10 00 .
In the second case, since the (2 \Gamma 1) optical depth is
higher than the (1 \Gamma 0), an excitation temperature much
higher than 43 K must be invoked.
The comparison between 13 CO and 12 CO tempera­
tures also requires a very large optical depth, even after
correcting the 12 CO observations for the source size.
However, if we allow for a temperature gradient in the
cloud, an intrinsic 12 CO temperature lower than the ob­
served 13 CO is not unplausible. In fact, we can hypothesize
a cooler surrounding envelope at 25 K which is optically
thick at 12 CO and optically thin at 13 CO and a hotter in­
ner core at a temperature –43 K which is optically thick
at 13 CO.
We can estimate a lower limit for the column density
and the extinction, from our 13 CO (2 \Gamma 1) data using op­
tically thin approximation:
N( 13 CO) = 7:3 \Theta 10 13 Ü 13
`
\Deltav
km s \Gamma1
'
T ex
1 \Gamma e \Gamma 10:58
Tex
(from Wootten et al. 1983). From the measurements of
Evans & Blair (1981) of the 12 CO(2 \Gamma 1) and 13 CO(2 \Gamma 1)
lines we can estimate Ü 13 for that transition to be ap­
proximately equal to 1. With T ex – 43 K, and \Deltav =
2:5 km s \Gamma1 , we obtain N( 13 CO) ? 4 \Theta 10 16 cm \Gamma2 . As­
suming the standard abundance of 13 CO, N(H 2 ) = 5 \Theta
10 5 N( 13 CO), then the total column density becomes
greater than 2 \Theta 10 22 cm \Gamma2 . Using the standard relation
N(H 2 ) = 10 21 AV cm \Gamma2 mag \Gamma1 (Bertoldi & McKee, 1992),
this implies a visual extinction ? 20 magnitudes (thus the
extimate of 9 magnitudes made by Tokunaga & Thomp­
son 1979 should be considered as a lower limit). With a
size of the clump of the order of 10 00 the mean density in
the molecular clump is n(H 2 ) AE 10 5 cm \Gamma3 .
The molecular hydrogen density can be also obtained
with a completely independent method from the three
C 34 S line intensities by using the statistical equilibrium
model of Cesaroni et al (1991). In this model the statis­
tical equilibrium equations for the level populations are
solved in the Sobolev approximation, with a constant ve­
locity gradient assumed equal to the ratio between the line
width and the source size. The input parameters to the
model are the kinetic temperature, the H 2 density and the
C 34 S abundance, and are varied until a best fit with the
observed intensities (1.5, 4.4 and 2 K for the (2--1), (3--2)
and (5--4) transitions, respectively) is obtained. In Fig. 5
the profiles of the three C 34 S lines at the position of the
peak (corresponding to the position of the H 2 O maser) are
shown. Due to the low signal to noise of the observations,
only a crude estimate of the parameters can be obtained.
We obtain a resonable agreement between the model and
the observed data for T– 50 K, n(H 2 ) ¸ 10 6 \Sigma 40% cm \Gamma3 ,
and X(C 34 S) ¸ 2 \Xi 4 \Theta 10 \Gamma10 .
In summary, all the indications point to a very high
density (n(H 2 ) ¸ 10 6 cm \Gamma3 ), small and very hot (T kin –
50 K) core , probably hotter than the surrounding molec­
ular gas, located at the position of the maser.

6 M. Felli et al.
Fig. 5. Profiles of the C 34 S lines observed at the position of
the H2O maser.
The mass of the hot core, assuming a constant density
of 6 \Theta 10 5 cm \Gamma3 (compatible with the two estimates given
above) within a spherical source with a diameter of 35 00 ,
derived as the diameter of the circle with an area equal to
that inside the 50% peak intensity contour of the C 34 S(3--
2) (see Fig, 2) and deconvolved for a beam FWHM of 16 00 ,
is of the order of 450 M fi .
3.2. H 2 O maser variability
H 2 O maser emission at 22 GHz from this region has been
known for a long time and is highly variable: in 1974 only
one component at ­22.3 km s \Gamma1 (Lo et al. 1975) is present,
in 1976 a high­velocity (­4 km s \Gamma1 ) and a low­velocity (­59
km s \Gamma1 ) feature appear (Blair et al. 1978), in 1978­79 a
group of components at ­60 km s \Gamma1 and two components
(at ­17 and 4 km s \Gamma1 , respectively) are present, which
steadily decrease and completely disappear in 1979 (Ro­
driguez et al. 1980), and, finally, only one broad compo­
nent between ­61 and ­57 km s \Gamma1 is found in 1983 (Henkel
et al. 1986). In our Medicina 2 data the source has been
observed since 1989 (see Persi et al. 1994 for a recent sum­
mary of the results).
The light curves of the four velocity components ob­
served at Medicina since 1986, including more recent ob­
servations after Persi et al. (1994), are reported in Fig. 6.
The two brightest ones (partly blended) are at ­58 and ­60
km s \Gamma1 , as in the 1978­79 and 1983 spectra. The ­60 km
s \Gamma1 component reached a peak of 111 Jy on 25/01/1993.
The components at 0 and ­70 km s \Gamma1 are much weaker
and only occasionally rise above the mean noise of ¸5 Jy
2 The Medicina radiotelescope is operated by the Istituto di
Radioastronomia C.N.R., Bologna, Italy
Fig. 6. Light curves of the 4 velocity components found in the
H2O spectra in the Medicina observations. Day 0 corresponds
to 23/jun/1987.
(3 oe). After 1995 the maser is quiescent (i.e. below 3 Jy).
The peak maser luminosity is 1.8 10 \Gamma5 L fi , and during
quiescent period is less than 2 10 \Gamma7 L fi .
Accurate maser position estimates were made with the
VLA by Tofani et al. (1995), at the time of the peak of
the ­60 km s \Gamma1 component. The ­70, ­60 and ­58 km s \Gamma1
components were not spatially resolved. The position of
the 0 km s \Gamma1 component remains unsettled because it was
quiescent at the time of the VLA observations. It should
also be noted that the velocity difference between the ­60
km s \Gamma1 maser and the peak of the molecular emission (¸ ­
17 km s \Gamma1 ) is quite large (43 km s \Gamma1 ), much larger than the
10 km s \Gamma1 half width dispersion of the velocity difference
over a large sample of CO outflows associated with H 2 O
masers (Felli et al., 1992; Codella and Felli, 1995; Anglada
et al., 1996). Only in the 1978­79 observations there were
components at the same velocity as the molecular cloud.
3.3. The S235 A­B stellar cluster
Figure 1 clearly shows that the S235 A and B nebulosi­
ties host stellar sources (S235 A? and B?) that are the
brightest members of a stellar cluster visible at K band.
The cluster was found by Hodapp (1994) in K', who dis­
covered more than 300 sources around S235 B, down to
a magnitude of about ¸ 17:0. In the same region we de­
tect 144 sources in our K image, down to a magnitude of
¸ 15:0. The large difference in the number of sources de­
tected is primarily due to the different sensitivities of the
two data sets. Our observations are unable to reveal the
low luminosity tail of the K band luminosity function, but
are in good agreement for the high luminosity part of the
distribution, after correcting for the difference between K'
and K.

M. Felli et al. 7
Fig. 7. The (J­H,H­K) colour­colour diagram of the total (43)
sources with detection in all the three bands in the S235 A­B
complex region. The exciting stars of S235 A and B are marked
as well as the K­band sources (M1,M2 and M3) near the H2O
maser. The dotted line is used to separate the sources with
IR excess from reddened MS stars allowing for errors on the
colours.
In Fig. 7 the (J--H,H--K) colour­colour diagram is plot­
ted for the sources detected by us in the cluster area
defined by Hodapp. The solid line, labelled MS, marks
the position of unreddened main sequence stars. The two
dashed lines show the position of the reddening belt for
main sequence stars. A dotted line (J \Gamma H) = 1:75(H \Gamma
K) \Gamma 0:35 is drawn parallel to the reddening law. This line
takes into account the error in the colours, and is used to
separate sources with infrared excess (to the right of the
line and represented as crosses in Figure 7) from reddened
MS stars.
Out of the 91 sources detected in all the three bands, 20
(22%) show near infrared excess and are probably young
stars, the other are reddened MS stars. An important
point that can be settled with our multi­band observa­
tions, is that most of the sources with infrared excess are
located near the maser position between S235 A and B.
This implies that the H 2 O maser (and its powering source,
see Sect. 3.7) is at the center of the embedded cluster,
and not S235 B, as supposed by Hodapp (1994). More­
over, since the stars with IR excess are believed to be
the youngest, this also suggests a gradient in the star for­
mation process, in which the stars near the centre of the
cluster and around the maser are the least evolved, while
S235 A? and B? represent more evolved stages.
A few sources which will be discussed in more detail
are labelled in Fig. 7, with observed parameters given in
Table 3.
3.4. S235 A
S235 A is classical HII region with flux density of 270 mJy
at 5 GHz, size 20 00 (0:17 pc) and electron density 1:3 \Theta
10 3 cm \Gamma3 (Israel & Felli 1978). The central star is a B0.5
type star with luminosity of 1:1 \Theta 10 4 L fi (Thompson et
al. 1983).
The centre of the radio continuum emission (which is
representative of the ionized gas distribution over the en­
tire HII region) and the stellar near IR source approxi­
mately coincide, as expected for a constant density spher­
ical HII region ionized by a central star. However, the Hff
and diffuse K band emission are offset from the position
of S235 A?. In Figure 8 we have overlayed the central part
of the K map (full contours) on the Hff photo of Krassner
et al. (1982). It can be clearly seen that S235 A? coin­
cides approximately with the centre of the radio contin­
uum emission (whose outer boundaries are outlined by the
dashed line in Figure 8), but lies to the west of the diffuse
Halpha nebulosity. This offset suggests that the star is still
surrounded by a thick dust envelope on all sides except to
the east, where ionized gas is less obscured and become
visible in Hff and in diffuse K band emission (most proba­
bly reflected star­light). About 5 00 south of S235 A? there
is another Hff nebulosity (or perhaps a fainter extension
of the upper one). Also in this case there is a K band
point source at its edge (this time the eastern edge of the
Hff emission), which has possibly a similar explanation.
However, in this case no radio continuum peak is found at
the same position, which is located towards the southern
boundary of the radio HII region. Presumably, the spec­
tral type of this star must be later and the luminosity
smaller than that of S235 A?.
S235 A? is the second brightest NIR source in the clus­
ter. We find diffuse Brfl emission around the source (al­
ready detected by Tokunaga & Thompson 1979), and hot
molecular hydrogen emission in the belt­like photon dom­
inated region (PDR) to the south of it (Figure 9). The H 2
is distributed around the ionized gas, preferentially at the
interface between the HII region and the C 34 S cloud core.
This confirms that S235 A? is interacting with the molec­
ular core that hosts the maser and its powering source.
Evans et al. (1981) found far IR emission peaking close
to S235 A, and this was confirmed by the detection of
IRAS05375+3540 at approximately the same position of
S235 A? (within 10 00 ). The primary source for the far in­
frared emission seems to be the cool dust around the com­
pact HII region S235 A, which in turn is energized by
S235 A?. Possible contributions to the FIR emission from
S235 B? and from the exciting source of the H 2 O maser
must be weaker than that of S235 A.
As already noted in Sect. 2.3, S235 A is on the edge
of the C 34 S clump. One might speculate that star for­
mation starts first from the outer edge of the molecular
cloud and that star formation inside the molecular peak
(as witnessed by the maser and the near IR source, see sec­

8 M. Felli et al.
Table 3. Positions and magnitudes of some near­infrared point sources.
# R.A. (1950.0) DEC. (1950.0) mJ e a
J mH e a
H mK e a
K Name
1 5 : 37 : 30:0 35 : 36 : 59 13:4 0:1 11:7 0:1 10:8 0:1 S235 C?
2 5 : 37 : 30:8 35 : 40 : 00 10:9 0:1 9:8 0:1 8:8 0:1 S235 B?
3 5 : 37 : 31:2 35 : 40 : 50 12:1 0:1 10:8 0:1 10:1 0:1 S235 A?
4 5 : 37 : 32:0 35 : 40 : 13 ? 17:3 -- 15:4 0:1 13:4 0:1 M1
5 5 : 37 : 32:3 35 : 40 : 16 16:8 0:2 15:0 0:1 13:6 0:1 M2
6 5 : 37 : 32:5 35 : 40 : 12 ? 17:3 -- 15:8 0:2 14:4 0:1 M3
a ) the errors quoted are 1oe uncertainties, note that the photometric calibration accuracy has been included, and, being ¸ 10%,
is the most important contribution to the errors quoted
30 s
31 s
32 s
5 37 33
h m s
+35 40'00"
o
20"
40"
+35 41'00"
o
20"
Fig. 8. Overlay of the K band emission (full contours) on the
Hff photo (gray­scale) of Krassner et al. (1982). The matching
between the two images is obtained with the three stars present
in both. The cross is the H2O maser position. The dashed line
is the lowest contour of the 6 cm radio continuum emission
from Israel & Felli (1978). The full triangle marks the position
of the radio continuum peak. The three crosses mark the posi­
tions of M1, M2 and M3. M2 and M3 are too faint to be well
represented in the contour map.
tion 3.7) is induced by the expansion of the S235 A HII
region. This sort of sequential (and possibly induced) star
formation from the edges to the core of a molecular cloud
has been observed also in other regions (see e.g. the case
of the S155/Cepheus B interface, Testi et al. 1995).
According to Tofani et al. (1995), S235 A? is the least
probable candidate for the excitation of the maser. In fact,
it is at a large distance from it, 0:29 pc, and the small
solid angle subtended from S235 A? would require a very
large stellar luminosity. Also the H 2 O maser is far beyond
28 s
29 s
30 s
31 s
32
33 s
34 s
5 37 35
h m s
+35 40'00"
o
20"
40"
+35 41'00"
o
20"
28 s
29 s
30 s
31 s
32
33 s
34 s
5 37 35
h m s
+35 40'00"
o
20"
40"
+35 41'00"
o
20"
Fig. 9. Full contours:
H2 emission; dashed contours: Brfl emission. Contour values
are: 20; 33; 45; 58; 70 \Theta 10 \Gamma15 erg s \Gamma1 cm \Gamma2 arcsec \Gamma2 for the H2 ;
20; 53; 85; 118; 150 \Theta 10 \Gamma15 erg s \Gamma1 cm \Gamma2 arcsec \Gamma2 for the Brfl.
The positions of the K band point sources S235 A?, S235 B?,
and M1 are marked with filled circles; the position of the H2O
maser is marked with a plus sign (note that the symbol is much
larger than the position accuracy).
the molecular/HII region interface and beyond the hot
molecular hydrogen emission.
3.5. S235 B
From our observations, the near infrared spectrum of
S235 B? is consistent with a reddened stellar photosphere
with a moderate NIR excess (see Figure 7). The visual ex­
tinction appears to be in the range 8 \Xi 12 mag dependent
upon the spectral type and assuming a standard extinc­
tion law. This estimate is in agreement with that of Evans
& Blair (1981), who derived an extinction of ¸ 9 mag
from the 13 CO column density. The presence of the 8.6
and 11.3 ¯m emission features is also suggestive of hot

M. Felli et al. 9
dust (Krassner et al. 1982). Alternative interpretations of
S235 B have been discussed (Krassner et al. 1982, NY and
Tofani et al. 1995), and could be: i) that it is an optically
thick UCHII region self absorbed in the radio continuum,
ii) or, more probably, that S235 B is an ionized expanding
envelope around a young star, which is optically thick in
the radio continuum and moderately thin in the Brackett
lines (Panagia & Felli 1975, Simon et al. 1981). Ionized en­
velopes of this type have also been found associated with
low luminosity YSOs.
The overlay of the K image on the Hff photo in Fig­
ure 8 shows that the optical nebulosity has a sharp edge
to the west and a diffuse tail on the opposite side. S235 B?
is located at the position of the sharp edge, and again sug­
gests a configuration similar to that of S235 A, in which
the star is surrounded by a thick envelope on all sides ex­
cept a small opening from which ionized gas can freely
escape.
The Brfl emission from this source is essentially un­
resolved in our images (i.e. ! 3 00 , much smaller than the
10 00 size of the Hff nebulosity). The integrated line emis­
sion is F(Brfl) = (2:0 \Sigma 0:4) \Theta 10 \Gamma12 erg s \Gamma1 cm \Gamma2 , which
is in excellent agreement with the value given by Krassner
et al. (1982). The peak of the Brfl emission is coincident
with the K­band point source.
In the region from which the Hff radiation is observed
no diffuse (down to 5 mJy at 5 GHz with a 10 00 beam,
Israel & Felli 1978) nor compact radio emission (down to
0.3 mJy at 8.4 GHz with a 0:1 00 beam, Tofani et al. 1995)
has been detected. This excludes the possibility that the
Hff emission is produced by an optically thin HII region.
If, instead, S235 B were an ionized expanding envelope,
we would expect a radio continuum flux given by:
S 8:5GHz = 0:63 \Theta 10 12
`
F(Brfl)
erg s \Gamma1 cm \Gamma2
'
mJy
(from Simon et al., 1983). Thus, even assuming AV = 0,
a radio continuum flux of 1:3 mJy at 8:5 GHz is expected.
Since such radio emission is not detected, the most plausi­
ble explanation is that the expanding envelope is only par­
tially ionized. The part of the envelope closer to the star
(from which the optically thin Brfl line radiation is pro­
duced) is ionized, the outer parts of the envelope, which
would emit most of the radio continuum radiation, are
neutral. Due to the high extinction toward the source, the
ionized envelope is not observed directly in Hff: the dif­
fuse emission that we see in this line is probably scattered
radiation, which escapes from a hole in the dust cocoon
around S235 B? toward the southeast and is reflected in
our direction.
The weak FIR emission from this heavily obscured
source implies that it is much less luminous than S235 A?.
Considering the distance from the H 2 O maser, it would
be difficult for this source to be the energy supply for the
maser. In fact, Evans et al. (1981) claim that the FIR lu­
minosity of S235 B? is less than 10 times that of S235 A?,
even though their measurements have been obtained with
very large beams and could be contaminated by extended
emission as well as by the contribution from a large num­
ber of faint sources.
Moreover, considering also its location at the edge of
the C 34 S cloud, we suggest that this source is in a rather
evolved stage, and it is blowing away the parent molecular
environment.
3.6. S235 C
S235 C is an optically thin HII region coinciding with an
optical nebula. A faint star (probably the B0.5 star ex­
citing the radio emission) is detected at the center of the
nebula in the Palomar plates (Israel & Felli 1978). Our
observations are completely consistent with this picture:
we observe (Figure 1) a point source inside a diffuse neb­
ulosity. Unfortunately our Brfl image does not extend so
far to the south from the maser position. In Table 3 the
NIR photometry of the source is reported, assuming a B0.5
spectral type, the NIR colours are consistent with a visual
extinction of the order of 16 magnitudes.
3.7. The H 2 O maser
In Figure 2 a cross is plotted on the K band image at the
maser position. There are three weak K band sources near
the maser positions (all within 10 00 ). The sources are re­
ported in Table 3 and in Fig. 7 and are barely visible in
the K­band grey­scales presented in Fig. 2 and also visi­
ble in the K' image of Hodapp (1994). They are labelled
M1, M2, and M3, in order of increasing distance from the
maser position (nearly coincident with the C 34 S peak).
The separation between the maser and the closest NIR
source is 5:0 00 , corresponding to 44 mpc.
While two of these sources (M2 and M3) show a mod­
erate NIR excess in the colours, like many of the cluster
stars, source M1, which is detected only in the K and H
images, shows colours very different from all the sources
in the field. This is evident from Figure 7 when compared
with the sources detected in all the three NIR bands, but
it is also true if we compare M1 with the other sources
detected in H and K only, none of which show such large
value of the H--K colour index. Its characteristic colours
resembles those of the NIR sources revealed in a large
survey of NIR sources associated with H 2 O masers (Testi
et al. 1994; 1996), characterized by values of the H--K
colour of the order or greater than 2 and distances be­
tween the NIR source and the maser spots of the order of
30 \Sigma 20 mpc.
With the large AV implied by the molecular observa­
tions the K band emission may be better explained by a
hot dust envelope or disk close to the stellar source rather
than by intrinsic stellar emission. In fact, if most of the
stellar flux is absorbed by dust at high temperature (of
the order of 10 3 K) near to the star and converted into

10 M. Felli et al.
near IR emission, the K band flux can be increased by
several orders of magnitude with respect to the ``naked''
stellar emission.
The high density conditions existing in a cocoon (of
the order of – 10 6 cm \Gamma3 ) may explain the lack of radio
continuum emission because of self absorption effects.
That M1 represents a YSO in a very early phase, re­
sponsible for the maser excitation, is also supported by
its coincidence with the molecular peak observed in C 34 S
and the high peak in brightness temperature observed in
13 CO. The first implies very high densities and the second
may suggest an increase of temperature in the inner core
of the cloud.
The large velocity differences of the maser components
with respect to the molecular gas point to highly energetic
winds/jets from the central source. Why then we do not
see similar high velocities in the molecular gas? Perhaps in
these early stages the acceleration occurs only very near
to the central YSO and involves only a very small fraction
of the molecular cloud gas, with the creation of high veloc­
ity and higly collimated jets with intrinsecally very little
mass. In other words the transfer of momentum from the
collapsing cloud to the surrounding molecular gas has just
begun. The interaction of these high velocity jets with the
surrounding molecular cloud may give rise to the masers.
The high maser variability suggests that these jets are far
from being a steady phenomenon (a wind) but are formed
by recurrent episodic ejections of small duration.
4. Discussion
4.1. Morphology of the S235 A­B complex
The new picture of this region that comes out of our data
is that of a progressive star formation activity from the
outside to the inner regions of a molecular cloud. In fact,
the peak of the molecular emission coincide with a cluster
of deeply embedded stellar objects. One of the objects de­
tected (M1, in Table3), is believed to be one of the earliest
in the cluster, with the (proto­)star itself embedded in its
parental dust cloud. The star may not be directly visible,
and the observed K­band source may be the emission of
the hot dust envelope around the YSO. At the edge of the
core is located S235 B?, probably an intermediate mass
object, which is blowing away the gas and dust material
still around it, as witnessed by the high value of the ex­
tinction toward the source, and, at the same time, the Hff
nebulosity on the side of the star. Just outside the molec­
ular core is S235 A?, which is a more evolved massive star
with a well developed HII region arount it.
The important aspect that comes out of the present
work is that the H 2 O maser and its associated exciting
star represent the youngest sources to be found in the clus­
ter (see also Testi et al. 1994; Hunter et al. 1995; Palla et
al. 1995; Persi et al. 1996). While originally H 2 O masers
had been found close to HII regions, and associated ge­
netically to them, it is now becoming clear that this as­
sociation is fortuitous, or more clearly, derived from the
fact that other stars in the same cluster may have al­
ready formed their own HII region, but are not neces­
sarely connected to the masers. Instead, the maser seems
to be present only in the earliest phases, when no thermal
continuum radio emission is detectable. The study of the
regions around the H 2 O masers is thus of fundamental im­
portance in the understanding of the earliest (proto­)stars
and their interaction with the ambient medium. Due to the
complexity of the morphology of the regions in which these
objects are forming, high resolution observations are es­
sential to separate the youngest objects from more evolved
HII regions and to avoid the confusion introduced by the
presence of nearby more evolved stars. In fact in all the
regions that we have investigated in detail (this one as
well as the ones discussed in: Hunter et al. 1995; Palla
et al. 1995; Persi et al. 1996), the previous low resolution
(and low sensitivity) observations had not been able to
reveal the true connection between H 2 O masers and the
earliest stellar evolutionary phases.
4.2. Star forming efficiency in the cloud core
From our molecular observations is it also possible to de­
rive an estimate of the star forming efficiency in the molec­
ular cloud. If we assume a spherical uniform cloud with a
density of n(H 2 ) ¸ 6 \Theta 10 5 cm \Gamma3 , we obtain a total mass
M ¸ 450 M fi , while the virial mass is M v ¸ 150 M fi . This
two values are in good agreement, if we take into account
the uncertanties in the density and distance determina­
tions.
The total mass of the stars within the cloud core is
more difficult to estimate, since we do not know the spec­
tral types of each source. From our K­band image and
from the K' image of Hodapp (1994) we can estimate that
the number of stars within the half power level in the
C 34 S (3--2) map is of the order of 12 \Xi 20. Assuming a
mean mass of 1 M fi per star, we find that the mass al­
ready converted into stars is ¸ 5 \Xi 15% of the molecular
mass.
5. Conclusions
We presented near infrared broad band and narrow band
images of the S235 A­B star forming region. Their anal­
ysis, together with new molecular data brings to the fol­
lowing new picture of the complex:
1. The presence of an embedded stellar cluster in the
S235 A­B region is confirmed. The center of the clus­
ter appears to be located between the two nebulosi­
ties, where the molecular core and the youngest cluster
members are located.
2. Brfl emission occurs mainly in the vincinity of S235 A
and B, and hot molecular hydrogen is distributed
around S235 A, especially in a belt­like region of the

M. Felli et al. 11
south it, at the interface between the HII region
and the molecular core. The morphology of the emis­
sion confirms that S235 A? and B? are (moderately)
evolved YSOs, that are emerging from the parental
cloud.
3. The driving source of the H 2 O maser has been identi­
fied on the near infrared images. On the basis of the
infrared results and the C 34 S, 13 CO molecular lines
observations, the newly discovered NIR object M1 is
suggested to be the powering source of the H 2 O maser.
This object, which does not coincide with any de­
tected radio continuum source, appears to be one of
the youngest in the region and to be located in the
innermost part of the cloud core, at the center of a
stellar cluster. These characteristics suggest that the
source is probably the first manifestation of a massive
YSO, a precursor of the UCHII phase.
4. A large fraction of the mass of the cloud core (¸ 10%)
has already been converted into young stars.
Acknowledgements. One of the authors (WJJ) thanks the Os­
servatorio Astrofisico di Arcetri for offering a fellowship to him,
and also acknowledges support from the National Natural Sci­
ence Foundation of China.
We thank R. Cesaroni and C. M. Walmsley for letting us
use the molecular data before publication and their LVG code
for the C 34 S data.
References
Anglada, G., Estarella, R., Pastor, J., Rodriguez, L.F.,
Haschick, A.D., 1996, ApJ, in press
Bertoldi, F., & McKee, C.F., 1992, ApJ, 395, 140
Blair, G.N., Davis, J.H., Dickinson, D.F., 1978, ApJ, 226, 435
Cesaroni, R., Churchwell, E., Hofner, P., Walmsley, C.M.,
Kurtz, S., 1994, A&A, 288, 903
Cesaroni, R., Walmsley, C.M., K¨ompe, C., Churchwell, E.,
1991, A&A, 252, 278
Cesaroni, R., Felli, M., Walmsley, C.M., 1996, in preparation
Codella, C., Felli, M., Natale, V., Palagi, F., Palla, F., 1994,
A&A, 291, 261
Codella, C., Felli, M., 1995, A&A, 302, 521
Elitzur, M., Hollenbach, D.J., McKee, C.F., 1989, ApJ, 346,
983
Evans, N.J.II, Beichman, C., Gatley, I., Harvey, P., Nadeau,
D., Sellgren, K., 1981, ApJ, 246, 409
Evans, N.J.II, Blair, G.N., 1981, ApJ, 246, 394
Felli, M., Palagi, F., Tofani, G., 1992, A&A, 255, 293
Felli, M., Testi L., Persi P., 1996, in preparation
Gennari S., & Vanzi L., 1994, Technical Report 8/94, Arcetri
Astrophysical Observatory
Henkel, C., Haschick, A.D., G¨usten R., 1986, A&A, 165, 197
Ho, P.T.P., Martin, R.N., Barrett, A.H., 1981, ApJ, 246, 761
Hodapp, K.­W., 1994, ApJS, 94, 615
Hunt, L.K., Lisi, F., Testi, L., Baffa, C., Borelli, S., Maiolino,
R., Moriondo, G., Stanga, R.M., 1996a, A&AS, 115, 181
Hunt, L.K., Migliorini, S., Testi, L., Stanga, R., Casali, M.M.,
Aspin, C., Hawardeen, T.G., Baffa, C., Lisi, F., 1996b, AJ,
submitted
Hunt, L.K., Testi, L., Borelli, S., Maiolino, R., Moriondo, G.,
1994, Technical Report 4/94, Arcetri Astrophysical Obser­
vatory
Hunter, T.R., Testi, L., Taylor, G.B., Tofani, G., Felli, M.,
Phillips, T.G., 1995, A&A, 302, 249
Israel, F.P., & Felli, M., 1978, A&A, 63, 325
Krassner, J., Pipher, J.L., Sharpless, S., Herter, T., 1982, A&A,
109, 223
Kutner, M.L., Ulich, B.L., 1981, ApJ, 250, 341
Lafon, G., Deharveng, L., Baudry, A., de La No¨e, J., 1983,
A&A, 124, 1
Lisi, F., Baffa, C., Hunt, L.K., 1993, in Infrared Detectors and
Instrumentation, SPIE Vol. 1946, p.594
Lo, K.Y., Burke, B.F., Hashick, A.D., 1975, ApJ, 202, 81
Nakano, M., & Yoshida, S., 1986, PASJ, 38, 531 (NY)
Palla, F., Cesaroni, R., Brand, J., Caselli, P., Comoretto, G.,
Felli, M., 1993, A&A, 280, 509
Palla, F., Testi, L., Hunter, T.R., Taylor, G.B., Prusti, T., Felli,
M., Natta, A., Stanga, R.M., 1995, A&A, 293, 521
Panagia, N., Felli, M., 1975, A&A, 39, 1
Persi, P., Palagi, F., Felli, M., 1994, A&A, 291, 594
Persi, P., Roth, M., Tapia, M., Marenzi, A.R., Felli, M., Testi,
L., Ferrari--Toniolo, M., 1996, A&A, 307, 591
Plume, R., Jaffe, D.T., Evans, N.J.II, 1992, ApJS, 78, 505
Rodriguez, L.F., Moran, J.M., Ho, P.T.P., Gottlieb, E.W.,
1980, ApJ, 235, 845
Sandell, G., Hoglund, B., Kislyakov, A.G., 1983, A&A, 118,
306
Simon, M., Felli, M., Cassar, L., Fischer, J., Massi, M., 1983,
ApJ, 266, 623
Simon, M., Righini­Cohen, G., Felli, M., Fischer, J., 1981, ApJ,
245, 552
Snell, R.L., Dickman, R.L., Huang, Y.­L., 1990, ApJ, 352, 139
Stutzki, J., Jackson, J.M., Olberg, M., Barrett, A.H., Win­
newisser, G., 1984, A&A, 139, 258, 1984
Testi, L., 1993, Technical Report 10/93, Arcetri Astrophysical
Observatory
Testi, L., Felli, M., Persi, P., Roth, M., 1994, A&A, 288, 634
Testi, L., Felli, M., Persi, P., 1996, in preparation
Testi, L., Olmi, L., Hunt, L.K., Tofani, G., Felli, M., Goldsmith,
P., 1995, A&A, 303, 881
Thompson, R.I., Thronson, H.A., Campbell, B., 1983, ApJ,
266, 614
Tofani, G., Felli, M., Taylor, G.B. Hunter, T.R. 1995, A&AS,
112, 299
Tokunaga, A.T., & Thompson, R.I., 1979, ApJ, 233, 127
Wilking, B.A., Mundy, L.G., Blackwell, J.H., Howe, J.E., 1989,
ApJ, 345, 257
Wootten, A., Sargent, A., Knapp, G., Huggins, P.J., 1983, ApJ,
269, 147
This article was processed by the author using Springer­Verlag
L A T E X A&A style file L­AA version 3.