Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.arcetri.astro.it/~lt/preprints/nh3_ctc/vla.ps.gz Äàòà èçìåíåíèÿ: Tue Sep 11 16:56:44 2007 Äàòà èíäåêñèðîâàíèÿ: Sat Dec 22 07:15:14 2007 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: massive stars

A&A manuscript no.
(will be inserted by hand later)
Your thesaurus codes are:
08(02.13.3; 08.06.2; 09.13.2; 13.19.3)
ASTRONOMY
AND
ASTROPHYSICS
25.2.1997
The molecular environment of H 2
O masers:
VLA ammonia observations
C. Codella 1 , L. Testi 2 , and R. Cesaroni 3
1 Max­Planck­Institut f¨ur Radioastronomie, Auf dem H¨ugel 69, D­53121 Bonn, Germany
2 Dipartimento di Astronomia e Scienza dello Spazio, Universit`a di Firenze, Largo E. Fermi 5, I­50125 Firenze, Italy
3 Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, I­50125 Firenze, Italy
Received date; accepted date
Abstract. We present the results of single dish and in­
terferometric observations of ammonia towards 5 sources
selected from a sample of H 2 O and OH masers associ­
ated with star forming regions. The Medicina telescope
was used to observe the NH 3 (1,1), (2,2), and (3,3) inver­
sion transitions. High resolution maps in the NH 3 (2,2) and
(3,3) lines and in the 1.3 cm continuum were then ob­
tained with the Very Large Array. The main result of this
research is to confirm the belief that H 2 O masers form
in hot dense molecular cores which are sites of massive
star formation. We also find evidence for the H 2 O maser
phase to be prior to the appearance of an ultracompact
Hii region around the embedded high mass star(s).
Key words: masers -- stars: formation -- interstellar
medium: molecules -- radio lines: interstellar
1. Introduction
Water masers at 22.2 GHz have been traditionally asso­
ciated with the process of high mass star formation. In­
deed, they are often observed close to typical signposts
of massive stars, namely ultracompact (UC) Hii regions,
far­infrared (FIR) point sources, OH masers, molecular
outflows. However, such an association does not imply a
physical relation between these phenomena and, to some
extent, could be due to a bias in the method used to se­
lect the targets to search for H 2 O maser emission. In fact,
the first surveys in the 22 GHz H 2 O line were conducted
towards well known Hii regions (e.g. Genzel & Downes
1977) and gave ambiguous results: in particular, although
the detection rate was quite high, unlike OH masers H 2 O
masers did not seem to be distributed over the continuum
emission. A typical example of this situation is W3(OH)
Send offprint requests to: R. Cesaroni
(see e.g. Fig. 15 of Cohen 1989), where the OH masers sur­
round the UC Hii region, whereas the H 2 O maser spots
are located ¸ 7 00 (¸ 0:07 pc) to the east of it. Analogous
conclusions were reached by Forster & Caswell (1989),
who studied the spatial association between H 2 O and OH
masers and, to a lesser extent, Hii regions. Also, in a re­
cent H 2 O maser survey towards UC Hii regions, Hofner
& Churchwell (1996) conclude that the median projected
distance between masers and continuum peak is ¸ 0:1 pc,
too large to justify a physical relationship between the
two.
With the advent of the IRAS era, it became easy to
identify selected samples of sources from the IRAS Point
Source Catalog (PSC, IRAS 1985), mostly on the basis
of their FIR colours. Several surveys in the H 2 O maser
line were thus performed towards candidates with suitable
FIR colours, resulting in high detection rates (Wouter­
loot & Walmsley 1986; Braz and Epchtein 1987; Scalise
et al. 1989; Palla et al. 1991; Henning et al. 1992; Palla
et al. 1993). Although the association of H 2 O masers with
cold IRAS point sources turned out to be stronger than
that with Hii regions, nevertheless the spatial resolution of
IRAS was too poor (a few minutes of arc) to ensure that
a positional (and hence physical) relationship did exist.
The ``statistical'' association suggested by the high detec­
tion rates and the striking correlation between the FIR
luminosity of the IRAS source and the luminosity in the
H 2 O lines were commonly interpreted with the idea that
more energetic environments give rise to more powerful
maser emission (see e.g. Palagi et al. 1993).
What is the environment where H 2 O masers form? The
first detection of a line tracing the molecular environment
of H 2 O masers is to be ascribed to Turner & Welch (1984),
who mapped the HCN(1--0) transition around W3(H 2 O)
with high spatial resolution. Such detection was later con­
firmed in other lines by Wink et al. (1994). Later on,
Codella & Felli (1995) and Codella et al. (1996) have inves­
tigated the nature of several star forming regions through

2 C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations
single dish observations and found that water masers prob­
ably occur at the earliest evolutionary stages of high mass
stars, much before the development of an ionised region
detectable in the radio continuum. By using high spa­
tial resolution, other authors (Cesaroni et al. 1994) have
found that H 2 O masers are positionally coincident with
hot dense cores, most of which do not present free­free con­
tinuum emission and are believed to be the site of birth of
one or more massive stars. Similarly, Tofani et al. (1995)
have performed VLA observations of the water maser line
at 22.2 GHz and of the 22.2 GHz and 8.4 GHz continuum
emission of 22 molecular outflows and found no contin­
uum emission on scales ! 5 00 in 13 objects, and only weak
emission from ionised envelopes in 9 sources. In a near in­
frared (NIR) survey toward H 2 O maser sources Testi et al.
(1994b; 1995) found that most of them (? 85%) are asso­
ciated with 2:2 ¯m sources with strong NIR excess, usu­
ally not associated with radio continuum. A few objects
observed in detail with a high resolution multiwavelength
approach (Felli et al. 1997; Hunter et al. 1995; Palla et
al. 1995; Persi et al. 1996), showed that the sources that
are likely to be powering the H 2 O masers are probably the
youngest in the region and are never associated with op­
tically thin radio continuum. These findings suggest that
H 2 O maser emission is strictly related to the very first
stages of the evolution of a massive star, when no Hii re­
gion has yet developed. However, this conclusion is based
on a sample of a few objects and hence needs further high
resolution observations to become more significant on a
statistical ground.
The purpose of this paper is to confirm the previous in­
terpretation, namely to demonstrate that H 2 O maser spots
are positionally coincident with molecular cores which do
not show continuum emission from ionised gas, namely
that they switch on close to a massive (proto)star prior
to the appearance of an UC HII region. For this purpose
we needed a tracer of hot dense gas and an instrument
with spatial resolution comparable to the typical size of
the cores (a few seconds of arc): the (2,2) and (3,3) in­
version transitions of ammonia and the Very Large Array
interferometer in its C configuration satisfy these require­
ments. We have thus performed a small survey towards
five sources selected from the list of Forster & Caswell
(1989), observing at the same time the ammonia lines
and the 1.3 cm continuum. The criteria used to select the
sources observed are given in Sect. 2, while the observa­
tions are described in Sect. 3. In Sect. 4 we discuss the
results for the continuum and the line, dedicating Sect. 5
to the most complex and interesting object observed. Fi­
nally, in Sect. 6 the conclusions are drawn.
2. Selection criteria
An initial sample of 15 targets was selected on the basis
of the following criteria: (i) H 2 O maser positions known
from the literature (Forster & Caswell 1989) with an accu­
racy of Ÿ 0: 00 5 and (ii) association with a NIR source with
strong NIR excess (Testi et al. 1994b; 1995). Note that
in this paper we are treating only masers in star form­
ing regions: no maser in late type stars is considered. As
previously discussed, we wish to investigate the compact
molecular emission around the H 2 O masers and for this
purpose we have used the VLA interferometer in its C con­
figuration, which ensures resolution comparable to that of
the maser and NIR data, as well as sensitivity to struc­
tures ! ¸ 25 00 in size. However, in order to get an estimate
of the line strength and properly center the bandwidth of
the VLA observations for each source, we have also per­
formed single dish observations by means of the Medicina
antenna. NH 3 (1,1) emission was detected in 13 sources out
fo 15 (Testi et al. 1994a), confirming the expectation that
molecular emission is virtually present in all star forma­
tion type H 2 O masers.
Given the time restriction for our VLA observations,
we had to limit our sample to 5 sources, chosen among the
strongest NH 3 emitters, according to the Medicina spec­
tra. Table 1 lists names, coordinates, local standard of rest
(LSR) velocities, and the calculated kinematic distances,
d, computed using the VLSR observed at Medicina and the
rotation curve of Brand (1986), which assumes for the Sun
an orbital velocity of 220 km s \Gamma1 and a galactocentric dis­
tance of 8.5 kpc. The coordinates correspond to the H 2 O
maser reference feature in Forster & Caswell (1989).
Table 1. List of observed sources
Source ff ffi VLSR d
(1950) (1950) (km s \Gamma1 ) (kpc)
G12.68\Gamma0.18 18 10 59.2 --18 02 43 55.6 11.5
G16.59\Gamma0.05 18 18 18.1 --14 33 15 60.0 4.7
G23.01\Gamma0.41 18 31 56.1 --09 03 04 77.0 10.7
G24.78+0.08 18 33 30.4 --07 14 43 110.0 7.7
G28.87+0.07 18 41 08.3 --03 38 35 101.0 7.4
3. Observations and data reduction
3.1. Single dish Medicina observations
The Medicina 32­m VLBI antenna 1 was used in single
dish mode to observe the NH 3 (1,1), (2,2), and (3,3) in­
version transitions. The observations were made during
several runs in 1993 and 1994. At the frequency of the
NH 3 (1,1) (23694.496 MHz), (2,2) (23722.634 MHz) and
(3,3) (23870.130 MHz) transitions the half power beam
width (HPBW) is 1: 0 9. The zenith system temperature
ranged from 120 K to 160 K depending on weather con­
ditions. The antenna efficiency was 38% with a maximum
1 The Medicina 32­m radiotelescope is operated by the Isti­
tuto di Radioastronomia CNR, Bologna, Italy

C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations 3
gain of 0.11 K Jy \Gamma1 . The intensity scale of the spectra was
calibrated on the continuum source DR21, resulting in an
uncertainty of 20%. The spectra were corrected for tele­
scope gain changes with elevation. The pointing accuracy
was ¸ 20 00 . The observations were made using position
switching with 3--5 minutes integration on source. Several
scans for each source were obtained for a total integration
time between 35 and 45 minutes. This gave a r.m.s. noise
of about 0.1 K in main beam brightness temperature. The
spectra were obtained with a 1024­channel autocorrelation
spectrometer. A 12.5 MHz bandwidth was used: the corre­
sponding spectral resolution at 23.7 GHz is 0.155 km s \Gamma1
and the total velocity coverage is 160 km s \Gamma1 . The con­
version factor from main beam brightness temperature to
flux is 5.6 Jy K \Gamma1 .
3.2. Interferometric VLA observations
We used the Very Large Array interferometer 2 to measure
the (2,2) and (3,3) inversion transitions of ammonia and
the 1.3 cm continuum emission towards the sources listed
in Table 1. The observations were carried out in December
1994 using the C configuration of the array. The (2,2)
line and the continuum were observed simultaneously by
centering the first IF at the rest frequency of NH 3 (2,2)
and the second IF at 23737.810 MHz. For the (3,3) line
only one IF was used, centred at the rest frequency of
NH 3 (3,3). The bandwidth was chosen for both lines large
enough to cover not only the \DeltaF = 0 main lines, but also
the \DeltaF = 1 satellites. On­line hanning smoothing was
applied in all cases. Further details of the observations
are given in Table 2.
The AIPS package developed at NRAO was used for
calibrating the data and producing maps. The UV data
of the continuum were averaged in such a way to pro­
duce a single channel from the original seven. All data,
line or continuum, were Fourier transformed using nat­
ural weighting and then cleaned by standard methods
(task MX of AIPS). Only towards G24.78 both line and
continuum emission were detected; in this case contin­
uum subtraction was performed by different methods and
precisely: (i) line­free channels were averaged and then
subtracted in the UV­plane by task UVLIN; (ii) a cube
with continuum plus line was produced, a continuum map
was made by averaging line­free channels, and finally sub­
tracted from the cube; (iii) a continuum map obtained
from the second IF was subtracted from the cube with
continuum plus line. The results were all consistent: we
thus chose method (iii) which gives a better signal­to­
noise due to the larger bandwidth of the continuum ob­
servations. In the case of G24.78 the continuum emission
was strong enough to allow self­calibration. We thus used
2 The National Radio Astronomy Observatory is operated
by Associated Universities, Inc., under cooperative agreement
with the National Science Foundation
task ASCAL to determine phase corrections to the con­
tinuum data and then transferred such corrections to the
NH 3 (2,2) line data, observed simultaneously with the con­
tinuum. For G24.78 both natural and uniform weighting
were used to produce the maps. For all sources the natural
weighted maps were convolved with a gaussian to produce
continuum or line channel maps with resolution 1: 00 5. Note
that the half power width (HPW) of the synthesised beam
reported in Table 2 is an average value which refers to the
maps obtained with natural weighting.
Fluxes per beam can be converted into beam bright­
ness temperatures using the relation
TSB(K) = 2:16 F(mJy beam \Gamma1 )
\Theta min ( 00 ) \Delta \Theta max ( 00 ) (1)
where \Theta min and \Theta max indicate respectively the minor and
major HPWs of the beam.
4. Results and discussion
4.1. Continuum maps
Fig. 1. Contour plot of the radio continuum emission in
G12.68--0.18. The filled triangles mark the position of the H2O
and the open squares that of the OH maser spots detected
by Forster & Caswell (1989). The ellipse in the top right cor­
ner shows the HPBW. The contours are drawn at 7, 10, and
13 mJy beam \Gamma1
Water masers may reasonably be associated with early
type stars: evidence for this can be found, for example,
in Palagi et al. (1993), who show (see their Fig. 10b) that
more than 90% of the H 2 O masers in star forming re­
gions have luminosities greater than ¸ 10 4 L fi , namely
that of a B0.5 star (see Panagia 1973). Such early type
stars are expected to develop UC Hii regions. What we
wish to verify is that H 2 O masers show up in a very early

4 C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations
Table 2. Instrumental parameters for the VLA observations
Parameter NH3(2,2) Continuum NH3(3,3)
Number of sources observed 5 5 4
Total observing time 8 hr
Number of antennas 27
IF band center 23722.634 MHz 23737.810 MHz 23870.130 MHz
Number of channels 31 7 31
Channel bandwidth (a) 195.313 kHz 3125 kHz 390.625 kHz
Velocity resolution (a) 2.468 km s \Gamma1 39.467 km s \Gamma1 4.906 km s \Gamma1
Primary beam HPW 2: 0 3
Synthesized beam HPW 1: 00 2 \Theta 1 00
Largest structure visible ¸ 25 00
Primary flux density calib. 3C 286 (2.42 Jy)
Phase and bandpass calib. 1741\Gamma038 (4.2 Jy)
1908\Gamma201 (2.5 Jy)
(a) on­line Hanning smoothing applied
Fig. 2. Same as Fig. 1, for G24.78+0.08. The two detected
components (A and B) are labelled. The contours are drawn
from 4 to 60 mJy beam \Gamma1 each 7 mJy beam \Gamma1 . The HPBW is
shown in the bottom left corner
stage of the evolution of an early type star. It is thus
important to study the continuum emission possibly as­
sociated with the H 2 O maser spots: if the associated Hii
region is pointlike or absent at all, then this indicates the
embedded star(s) to be very young. Our results confirm
this scenario. In fact, continuum emission is detected only
towards G12.68 and G24.78: in the former, the Hii re­
gion looks extended and not associated with the H 2 O or
OH maser spots (see Fig. 1); in the latter instead (see
Fig. 2), the emission comes partly from a pointlike Hii
region (component A) coincident with the H 2 O and OH
masers, and partly from a resolved Hii region (component
B), offset by ¸ 5 00 from the H 2 O and OH masers.
Table 4. Upper limits for the continuum emission at maser
position
Source Fpeak TSB
(mJy beam \Gamma1 ) (K)
G12.68\Gamma0.18 ! 1:2 ! 2:2
G16.59\Gamma0.05 ! 0:36 ! 0:64
G23.01\Gamma0.41 ! 0:37 ! 0:66
G28.87+0.07 ! 0:34 ! 0:61
In Table 3 we report the main parameters of the con­
tinuum emission in the two sources detected, i.e. the posi­
tion of the peak, the flux measured at this position, F peak ,
the corresponding synthesised beam brightness tempera­
ture, TSB , the observed angular diameter at half power of
the emitting region, \Theta HP , the deconvolved angular diam­
eter, \Theta S , and the integrated flux density over the whole
emitting region, S š . For the non­detected sources we give
in Table 4 the upper limits corresponding to 3oe RMS val­
ues, for the peak flux and corresponding brightness tem­
perature.
In Table 5 the main physical quantities of the related
Hii regions are given, namely: the physical dimension of
the clumps, D; the excitation parameter, U , the elec­
tron density, n e , and the Lyman continuum luminosity,
N Ly , computed using the formulae of Schraml & Metzger
(1969), under the assumption of optically thin emission at
1.3 cm; and the luminosity, L ? , and spectral type of the
star, derived from U using the tables of Panagia (1973).
We point out that the values for G12.68 and G24.78B are

C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations 5
Table 3. Parameters of the continuum emission
Name ff ffi Fpeak TSB \Theta HP \Theta S Sš
(1950.0) (1950.0) (mJy beam \Gamma1 ) (K) ( 00 ) ( 00 ) (mJy)
G12.68\Gamma0.18 (a) 18:10:59.8 --18:02:29 15 81 17 17 990
G24.78+0.08 A 18:33:30.40 --07:14:42.7 76 150 1:0 ! 0:3 77
G24.78+0.08 B 18:33:30.5 --07:14:47 15 30 3:0 2:7 70
(a) this source is overresolved in our observations, and hence fluxes and sizes represent only lower limits.
Table 5. Derived physical quantities for the Hii regions
Name d D U ne NLy L? Spectral
(kpc) (pc) (pc cm \Gamma2 ) (cm \Gamma3 ) (10 47 s \Gamma1 ) (10 4 L fi ) Type
G12.68\Gamma0.18 11.5 0.96 76 2:0 10 3 150 33.0 O6
G24.78+0.08 A 7.7 ! 0:01 24 ? 2:9 10 5 5.2 3.3 O9.5
G24.78+0.08 B 7.7 0.10 24 1:0 10 4 4.7 3.2 O9.5
given only for the sake of completeness, since these Hii
regions are not associated with the maser environment.
In the following, we discuss the constraints set by
the previous results on the Hii region possibly associated
with the maser clusters. We treat the detected and non­
detected sources separately, stressing that for ``detected''
we mean sources with continuum emission positionally
coincident with the masers: according to such definition
G12.68 belongs to the non­detected sample.
4.1.1. Continuum detected sources: G24.78+0.08A
From Table 5 one sees that G24.78A is very compact and
dense, as expected for a very young Hii region, whereas
G24.78B looks more extended. This is consistent with
the VLA observations at 6 cm of Becker et al. (1994):
in a 4 00 beam they detect a 32.2 mJy point source at
ff = 18 h 33 m 30: s 53 and ffi = \Gamma7 ffi 14 0 46: 00 4. Such position
is almost coincident with that of G24.78B, which proves
that most of the emission at 6 cm originates from compo­
nent B. However, since the angular resolution is compara­
ble to the separation between G24.78 A and B, it is hard
to estimate the contribution of component A to the mea­
sured 6 cm flux: if one assumes G24.78A to be an order
of magnitude fainter than B, i.e. ¸ 4 mJy, then the spec­
tral index between 6 cm and 1.3 cm turns out to be ¸ 1:9,
which proves the 1.3 cm free­free continuum emission of
G24.78A to be optically thick. In turn, this indicates that
the G24.78A Hii region is very small and compact. Anal­
ogously, one can demonstrate that G24.78B is instead op­
tically thin at 1.3 cm, with a turn­off wavelength between
the thick and thin regimes of ¸ 4 cm.
We conclude that the values of Table 5 (derived under
the optically thin approximation) are only lower limits in
the case of G24.78A; in particular, the spectral type of the
ionising star is likely to be much earlier than O9.5. This is
in agreement with the luminosity of the IRAS point source
associated with G24.78 (IRAS 18355--0713A; see Palagi et
al. 1993), which amounts to 2:8 10 5 L fi and corresponds
to an O6 star.
4.1.2. Continuum non­detected sources
For all sources of our sample but G24.78, no continuum
emission has been detected at the position of the masers.
How can one use upper limits to set sensible constraints on
the parameters of an undetected Hii region? If as discussed
above (see Sect. 4.1) H 2 O masers are associated with em­
bedded early type stars, then the lack of free­free contin­
uum emission can be explained in three manners: (i) the
embedded star(s) is so young that no Hii region could
yet develop; (ii) the Hii region is too small (i.e. young) to
be detected; (iii) the Hii region is too extended and faint
to be detected. Both (i) and (ii) are consistent with the
scenario that we wish to prove; case (iii) instead would in­
dicate that the Hii region is old: we want to demonstrate
that this is not likely for our sources.
For a spherical, homogeneous, isothermal Hii region,
with given electron temperature, the peak brightness tem­
perature at a given frequency can be expressed as a func­
tion only of its Str¨omgren radius, R S , and of N Ly . Once the
distance to the source and the HPBW of the telescope are
known, it is possible to compute the brightness tempera­
ture which would be measured with the synthesised beam,
TSB (calc), as a function of R S and N Ly , and compare it
with the upper limits in Table 4. In Fig. 3 we plot curves
of constant TSB corresponding to these upper limits. Only
points of the R S --N Ly plane falling below such curves (i.e.
satisfying the condition TSB (calc) Ÿ TSB (measured)) are
acceptable. We note that each curve can be divided in
three parts: the first, at low R S , is parallel to the y­axis
and corresponds to the case of an optically thick and un­

6 C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations
resolved Hii region; the second, at intermediate values of
R S , is parallel to the x­axis and indicates that the Hii re­
gion is still unresolved but optically thin; and the third,
at high R S , satisfies the approximate relation N Ly / R S
2 ,
typical of a resolved optically thin Hii region.
Fig. 3. Curves corresponding to constant values of peak bright­
ness temperature in the synthesised beam, TSB , of a spherical,
homogeneous, isothermal Hii region. We have assumed an elec­
tron temperature of 9000 K and a gaussian beam with HPW
of 1: 00 5. RS and NLy are respectively the Str¨omgren radius of
the Hii region and the Lyman continuum of the ionising star;
note that TSB is an increasing function of NLy . On the right
hand side axis the values of NLy corresponding to the spectral
types as of Panagia (1973) are indicated. The contours corre­
spond to 3oe upper limits in units of TSB , and precisely: 2.2 K
for G12.68, 0.64 K for G16.59, 0.66 K for G23.01, and 0.61 K
for G28.87
When looking at Fig. 3, one must take into account
that we are searching for UC Hii regions, namely for
R S ! 0:1 pc. The figure clearly shows that in all cases
if any UC Hii region exists at the position of the masers,
then could not be detected if either too small (! 0:001 pc)
and optically thick, or too faint (N Ly ! 10 47 sec \Gamma1 ) and
optically thin. In order to discriminate between these
two possibilities, one can estimate the bolometric lu­
minosity, which can then be converted into the corre­
sponding N Ly by means of the tables of Panagia (1973).
Beside G24.78 (see Sect. 4.1.1), two more objects of
our sample are associated with a source of the IRAS
PSC (Palagi et al. 1993): G16.59 (IRAS 18182--1433) and
G28.87 (IRAS 18411--0338). Incidentally, we note that the
IRAS colours of these objects satisfy the criteria set by
Wood & Churchwell (1989b) for identifying IRAS sources
associated with UC Hii regions: this strengthens the idea
that we are dealing with embedded early type stars. The
bolometric luminosity, L IRAS , estimated by integrating
the four IRAS fluxes, are 2:3 10 4 L fi for G16.59, and
1:9 10 5 L fi for G28.87. These imply N Ly ' 2 10 47 sec \Gamma1
for G16.59 and N Ly ' 10 49 sec \Gamma1 for G28.87. By as­
suming these two values for N Ly in Fig. 3, one derives
R S ! 3 10 \Gamma4 pc for G16.59 and R S ! 5 10 \Gamma4 pc for
G28.87.
A word of caution must be spent on using the IRAS
fluxes to compute N Ly : in fact, it is well known (see e.g.
Wood & Churchwell 1989a and Churchwell et al. 1990)
that IRAS luminosities are usually higher by a factor ¸ 5
than the luminosities derived from the number of ionising
photons required to explain the radio continuum fluxes
of known UC Hii regions. This means that the effective
value of N Ly may be smaller by factors of 10 and 100 re­
spectively for G28.87 and G16.59: such a reduction is not
large enough to invalidate our method for setting an upper
limit on the Hii region size.
Unfortunately, the same method cannot be applied to
G12.68 and G23.01, since they do not have an IRAS coun­
terpart. However, it seems plausible that also in these
cases the non­detection of compact continuum emission
might indicate that the Hii region is still too compact or
not even formed. In fact, VLA observations of G12.68 at
2 cm and 3.6 cm in the B­array configuration (Kurtz, priv.
comm.) did not detect any continuum emission at the po­
sition of the masers, thus confirming our hypothesis.
We conclude that very likely no or a not­yet­developed
UC HII region is associated with the environment of the
H 2 O masers.
4.2. Line spectra
With the Medicina antenna, NH 3 (1,1) and (2,2) line emis­
sion has been detected towards all objects of our sam­
ple. Moreover, NH 3 (3,3) line emission has been searched
for, and detected, towards all the sources but G28.87, for
which the (2,2) line turns out to be very faint.
Using the VLA, we have searched for NH 3 (2,2) and
(3,3) towards all five sources of Table 1, but G12.68, for
which only the (2,2) line was observed with a shorter in­
tegration time than the other sources, due to time re­
strictions. In Fig. 4 the Medicina and VLA spectra are
compared. The Medicina NH 3 (1,1) spectra are shown at
full resolution (0:155 km s \Gamma1 ), whereas the NH 3 (2,2) and
(3,3) spectra have been smoothed to the VLA spectral
resolution (2:468 km s \Gamma1 and 4:906 km s \Gamma1 , respectively)
to allow a direct comparison. The VLA spectra have been
obtained by integrating over the whole area where NH 3
emission is detected, with the sole exception of G12.68
and G23.01: in these sources no emission is visible in the
VLA channel maps of the NH 3 (2,2) line and hence we
have integrated the NH 3 (2,2) emission over a large area
(¸ 60 arcsec 2 ) around the H 2 O maser spots. The spectra
obtained in this manner are also shown in Fig. 4: a faint
feature is seen at the expected VLSR , so that we consider
this a positive detection.
From the comparison between the Medicina and the
VLA spectra we can immediately deduce that: (i) the hy­
perfine satellites (indicated by the vertical marks in Fig. 4)

C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations 7
Fig. 4. Comparison between spectra observed at Medicina and at the VLA. The Medicina NH3(1,1) spectra are shown at
full spectral resolution, whereas the NH3(2,2) and (3,3) spectra have been smoothed to the VLA resolution. The VLA spectra
have been obtained by averaging the emission over a suitable area (see text). The vertical marks indicate the positions of the
hyperfine satellites
do not show up in the NH 3 (2,2) and (3,3) spectra of Medic­
ina, whereas they are detected in the VLA spectra of
G23.01, G24.78 and G28.87; and (ii) the main line inten­
sity ratios between the Medicina and the VLA spectra are
?1 for G12.68, G23.01, and G24.78, and ¸1 for G16.59
and G28.87 (see Table 6). These findings suggest that a
compact optically thick core is present in most sources,
surrounded in some cases by an optically thin envelope,
which in G23.01 and G24.78 contributes substantially to
the NH 3 emission seen by the single dish telescope.
Tables 7 and 8 contain the observed ammonia spectra
parameters: the name of the source, the line transition, the
main beam brightness temperature (TMB ), the LSR veloc­
ity (V LSR ) and the full width at half maximum (\DeltaV 1
2
) of
each hyperfine component. The last column gives the to­
tal optical depth over all components of a transition (Ü tot ;
see Ungerechts et al. 1986), calculated for those spectra
where the satellites have been detected. The parameters
of Tables 7 and 8 have been obtained using the TAUFIT
program which takes into account the hyperfine structure
of the lines, assuming gaussian profile for each component

8 C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations
Table 6. Main line intensity ratios between the Medicina and
VLA data
Source NH3(2,2) NH3(3,3)
G12.68\Gamma0.18 ?4 -- (a)
G16.59\Gamma0.05 1.4\Sigma0.7 !1.6
G23.01\Gamma0.41 AE1 6.2\Sigma2.3
G24.78+0.08 4.9\Sigma1.2 5.3\Sigma1.2
G28.87+0.07 !2.0 -- (b)
(a) not observed with the VLA
(b) not observed with Medicina
and relative intensities consistent with the local thermo­
dynamic equilibrium (LTE) approximation. The data re­
garding the VLA observations have been obtained from
the spectra at the ammonia peak position (see Sect. 4.3).
It is worth noting that in Table 8 also the fit parameters
relative to a second molecular component of G24.78 (here­
after indicated as G24.78+0.08M) are presented: this cor­
responds to the ammonia emission associated with the wa­
ter maser spots in the G24.78 region at ff ' 18 h 33 m 31 s
and ffi ' \Gamma7 ffi 14 0 40 00 (see Sect. 5). Since no NH 3 peak
is seen at this position, we have used the NH 3 spectra
obtained by averaging the NH 3 (2,2) and (3,3) emission
within a radius of 1: 00 35 centred on the H 2 O maser posi­
tion: this has been arbitrarily chosen equal to the angular
radius of the ammonia core seen towards G24.78A.
Table 7. Results of fits to the spectra observed at Medicina
Source line TMB VLSR \DeltaV 1
2
Ü tot
(K) (km s \Gamma1 ) (km s \Gamma1 )
G12.68 (1,1) 1.0\Sigma0.1 55.18\Sigma0.07 3.1\Sigma0.2 4.6\Sigma0.8
(2,2) 0.52\Sigma0.08 55.3\Sigma0.1 3.6\Sigma0.4 -- (a)
(3,3) 0.4\Sigma0.1 55.5\Sigma0.1 4.1\Sigma0.6 -- (a)
G16.59 (1,1) 0.9\Sigma0.1 59.51\Sigma0.07 2.4\Sigma0.2 1.8\Sigma0.8
(2,2) 0.27\Sigma0.07 59.6\Sigma0.2 2.5\Sigma0.5 -- (a)
(3,3) 0.08\Sigma0.08 59.9\Sigma0.4 2\Sigma1 -- (a)
G23.01 (1,1) 1.0\Sigma0.1 76.58\Sigma0.08 3.9\Sigma0.1 4.4\Sigma0.6
(2,2) 0.74\Sigma0.08 77.07\Sigma0.08 3.4\Sigma0.3 -- (a)
(3,3) 0.3\Sigma0.1 76.5\Sigma0.3 6\Sigma2 -- (a)
G24.78 (1,1) 1.1\Sigma0.1 109.56\Sigma0.05 2.7\Sigma0.1 3.9\Sigma0.5
(2,2) 0.76\Sigma0.06 110.14\Sigma0.06 3.3\Sigma0.2 -- (a)
(3,3) 0.6\Sigma0.1 109.9\Sigma0.1 4.2\Sigma0.6 -- (a)
G28.87 (1,1) 0.31\Sigma0.05 100.61\Sigma0.09 3.3\Sigma0.2 3.0\Sigma0.8
(2,2) 0.04\Sigma0.04 102.6\Sigma0.6 4.4\Sigma0.9 -- (a)
(a) satellites not detected (optically thin main line assumed)
From these results we conclude that we detect NH 3
emission with the VLA towards all of the sources observed,
although in one case (G12.68) the NH 3 (2,2) line is de­
tected only after averaging over a region a few seconds
of arc in size around the position of the masers. In the
next section we shall investigate the distribution of the
ammonia emission in detail.
Table 8. Results of fits to the spectra taken with the VLA.
In all cases but G24.78 M the spectrum at the peak position in
the map was used
Source line TSB VLSR \DeltaV 1 2
Ü tot
(K) (km s \Gamma1 ) (km s \Gamma1 )
G12.68 (2,2) !4
G16.59 (2,2) 41\Sigma5 59.2\Sigma0.3 4.4\Sigma0.6 -- (a)
(3,3) 48\Sigma5 59.8\Sigma0.3 5.4\Sigma0.8 -- (a)
G23.01 (2,2) !3
(3,3) 41\Sigma7 79.0\Sigma0.4 7.0\Sigma0.6 7.7\Sigma0.9
G24.78 (2,2) 37\Sigma6 110.3\Sigma0.3 5.9\Sigma0.4 8.9\Sigma0.8
(3,3) 44\Sigma6 109.4\Sigma0.3 5.5\Sigma0.4 14.9\Sigma0.9
G24.78 M (b) (2,2) 24\Sigma4 110.7\Sigma0.3 5.7\Sigma0.5 8\Sigma2
(3,3) 25\Sigma3 111.2\Sigma0.5 7\Sigma2 -- (a)
G28.87 (2,2) 21\Sigma5 103.2\Sigma0.1 2.3\Sigma0.3 13.3\Sigma0.8
(3,3) 25\Sigma4 103\Sigma1 3\Sigma1 11.6\Sigma0.9
(a) satellites not detected (optically thin main line assumed)
(b) spectra obtained by averaging the NH3 (2,2) and (3,3) emission over
a region with radius 1: 00 35 centred on the H2O masers at
ff ' 18 h 33 m 31s and ffi ' \Gamma7 ffi 14 0 40 00
4.3. Line maps
The contour plots of the main line integrated intensity
for all the detected sources are shown in Fig. 5. The
positions of all the maser spots detected by Forster &
Caswell (1989) are indicated in the contour maps by filled
triangles (H 2 O masers) and open squares (OH masers).
Also shown are the positions of the 2:2 ¯m sources with
strong NIR excess detected by Testi et al. (1994b; 1995).
The association between the H 2 O masers and the NIR
source has been stressed and discussed at length by Testi
et al. (1994b; 1995); what Fig. 5 shows is that an even
closer association does exist between masers and ammo­
nia emission, indicating that maser emission is connected
with the presence of compact molecular gas. A special
case is represented by G24.78, for which one of the two
groups of H 2 O spots lies relatively far (¸ 10 00 ) from the
NH 3 emission peak, but still inside the ammonia clump.
The occurrence of absorption and emission makes G24.78
quite an interesting case: this will be analysed in detail in
Sect. 5. Only towards G12.68 was no NH 3 core detected,
but one must take into account that: (i) G12.68 is the most
distant source in our sample; (ii) only the NH 3 (2,2) line
has been observed; and (iii) the integration time is ¸ 40%
of that used for the other sources. However, since the in­
tegrated spectrum discussed in the previous section shows
a faint NH 3 (2,2) line, it is very likely that also in the case
of G12.68 an ammonia clump would show up in a better
S/N map, at the maser position.
For each ammonia transition, Figs. 6 and 7 show the
maps of G24.78 and G28.87 obtained by averaging the
channels in the main line (upper panels) and those in the
four satellites (lower panels). One can see that the emis­
sion due to the satellites is detected from the region where

C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations 9
Fig. 5. a. Contour plots of the integrated main line intensity for G16.59--0.06 and G23.01--041 for the lines detected. The filled
triangles and open squares mark respectively the positions of the H2O and OH masers from Forster & Caswell (1989). The cross
represents the position of the 2:2 ¯m source with strong NIR excess detected by Testi et al. (1994b; 1995); the size of the cross
indicates the 1oe positional uncertainty. The thick ellipse in the lower right corner of each image represents the HPBW. Contour
levels are: --8, 8 to 32 by 4 mJy beam \Gamma1 for G16.59--0.05 NH3(2,2) and (3,3); and --8, 8 to 40 by 4 mJy beam \Gamma1 for G23.01--0.41
NH3(3,3)
the main line is strongest, thus confirming the existence
of an optically thick core coincident with the H 2 O masers.
In Fig. 6 also the 1.3 cm continuum map is displayed
(grey scale), showing the coincidence between the ab­
sorption in the NH 3 lines and the continuum of the
G24.78A UC Hii region. In Fig. 8 we plot the spectra of
the NH 3 (2,2) and (3,3) lines corresponding to the absorp­
tion peak: the high ratio between satellites and main line
clearly indicates that the UC Hii region is surrounded by
optically thick material.
4.4. Derived parameters of the ammonia cores
From Boltzmann plots for the Medicina data, assuming
that NH 3 is populated according to LTE, one can esti­
mate the rotational temperature and total column den­
sity performing a linear fit to the data (see e.g. Cesaroni
et al. 1992). The results are presented in Table 9. One
can see that the single dish measurements are sensitive to
cool molecular gas with temperature of 10--20 K, which is
very likely distributed over more extended regions than
those seen in the VLA observations, as already noted in
Sect. 4.2. This applies in particular to the NH 3 (1,1) tran­
sition. In fact, the ratio between the NH 3 (1,1) and (2,2)
column densities indicates a lower rotational temperature

10 C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations
Fig. 5. b. Same as Figure 5 but for G24.78+0.08 and G28.87+0.07. Contour levels are: --12, --8, 12, 18 to 46 by 4 mJy=beam
for G24.78+0.08 NH3(2,2); --7, 13, 25 to 67 by 6 mJy beam \Gamma1 for G24.78+0.08 NH3 (3,3); and --8, 8, 10 to 18 by 4 mJy beam \Gamma1
for G28.87+0.07 NH3(2,2) and (3,3)
(T rot ) than the ratio between NH 3 (2,2) and (3,3): this sug­
gests that higher excitation transitions sample hotter re­
gions, in agreement with the findings of other authors (see
e.g. Cesaroni et al. 1992).
The previous conclusion is confirmed by the VLA data,
as shown in Table 10 where we summarise the derived
physical parameters of the ammonia clumps: observed
(\Theta HP ) and deconvolved (\Theta S ) angular diameter, physical
diameter (D), kinetic temperature (T k ), ammonia col­
umn density (NNH3 ), ammonia mass (MNH3 ), virial mass
(M vir ), H 2 volume density (nH2 ), and ammonia abundance
with respect to H 2 (XNH3 ).
The observed diameter of the molecular clump corre­
sponds to the full width at half power (FWHP) of the
maps in the main line. The true diameters have been esti­
mated by deconvolution of a gaussian beam from a gaus­
sian source. For G16.59 the average of the diameters mea­
Table 9. Results of linear fits to Boltzmann plots for the
Medicina data
Source T rot N tot
(a)
(K) (10 14 cm \Gamma2 )
G12.68\Gamma0.18 17\Sigma2 14.6\Sigma0.9
G16.59\Gamma0.05 17\Sigma4 1.88\Sigma0.09
G23.01\Gamma0.41 13\Sigma1 25\Sigma 2
G24.78+0.08 18\Sigma1 10.8\Sigma0.5
G28.87+0.07 7\Sigma2 41\Sigma16
(a) beam averaged
sured in the (2,2) and (3,3) maps has been used, whereas
for the other sources we give the diameter in the inversion

C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations 11
Fig. 6. Contour maps of integrated NH3(2,2) and (3,3) main line (upper panels) and satellite (lower panels) emission towards
G24.78+0.08. The grey scale represents the 1.3 cm continuum. Contour levels correspond to --6, 8 to 50 by 7 mJy beam \Gamma1 for
the NH3(2,2) main line and satellites, and to --7, 7 to 67 by 6 mJy beam \Gamma1 for the NH3(3,3) main line and satellites. White
contours represent negative levels
Fig. 8. Spectra of the ammonia lines at the position of the
absorption deep (i.e. that of the G24.78 A UC Hii region). The
dashed lines indicate the zero level. The vertical ticks mark the
positions of the main line and hyperfine satellites
transition for which the signal to noise ratio is higher:
NH 3 (3,3) for G23.01 and G28.87, and (2,2) for G24.78.
In order to derive the NH 3 column density at the peak
position (NNH3 ) in each molecular core, we have used the
method described, e.g., by Ungerechts et al. (1986) and
the NH 3 partition function. From NNH3 and the core di­
ameter, D, it is possible to calculate the total number of
NH 3 molecules in the core (N ), assuming the core to be
spherical and homogeneous:
N = ú
6 NNH3 D 2 (2)
The ammonia mass (MNH3 ) is then easily estimated
from N . Note that, since G23.01 has not been detected in
the (2,2) line, we have derived the total column density for
this object from the (3,3) column density adopting LTE
at temperature T k .
Assuming that the cores are virialised, one can esti­
mate their total masses from (MacLaren et al. 1988):
M vir (M fi ) = 0:509 d(kpc) \Theta S (arcsec)
h
\DeltaV 1
2
(km s \Gamma1 )
i 2
(3)
where \DeltaV 1
2
is the ammonia linewidth derived from the
VLA observations (Table 8). The NH 3 abundance is then

12 C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations
Fig. 7. Maps of the integrated NH3(2,2) and (3,3) main line (upper panels) and satellite (lower panels) emission towards
G28.87+0.07. Contour levels correspond to --8, 8 to 18 by 3 mJy beam \Gamma1 for the NH3(2,2) main line and satellites and to --8, 8
to 20 by 4 mJy beam \Gamma1 for the NH3(3,3) main line and satellites
Table 10. Parameters of the ammonia cores derived from the VLA data
Source \Theta HP \Theta S D Tk NNH 3 MNH 3 M vir nH 2 XNH 3
( 00 ) ( 00 ) (pc) (K) (10 16 cm \Gamma2 ) (10 \Gamma5 M fi ) (M fi ) (10 6 cm \Gamma3 ) (10 \Gamma8 )
G16.59\Gamma0.05 5.0 4.9 0.112 54 1.1 1 225 6.2 0.5
G23.01\Gamma0.41 3.5 3.3 0.171 58 34 71 886 6.9 9.4
G24.78+0.08 2.7 2.4 0.090 87 42 24 330 17.5 8.6
G24.78+0.08 M (a) 2.7 (b) 2.4 0.090 29 62 36 305 16.2 13.9
G28.87+0.07 2.1 1.8 0.065 37 10 3 34 4.8 10.4
(a) parameters derived from a mean spectrum over \Theta HP (see text)
(b) assumed equal to \Theta HP of G24.78+0.08
given by XNH3 = (2=17) MNH3 =M vir , while the hydrogen
density (nH2 ) has been calculated from the virial mass and
the clump diameter.
The kinetic temperature (T k ) reported in Table 10 is
the VLA peak TSB of the NH 3 line corrected for the
beam filling factor, i.e. divided by \Theta S
2 =\Theta HP
2 . This is in
good agreement with the rotational temperature derived
from the ratio between the NH 3 (2,2) and (3,3) lines, as
seen with the VLA. Note that, for all sources where both
NH 3 (2,2) and (3,3) have been detected, the peak values of
TSB in these two lines are very similar (with a maximum
difference of 15%): this indicates that the two transitions
are optically thick and confirms that their brightness tem­
perature is indeed a good estimate of the line excitation
temperature, or, equivalently, of T k .
From Table 10, one sees that the ammonia column den­
sities are ¸ 10 16 --10 17 cm \Gamma2 , and the NH 3 abundances are
¸ 10 \Gamma8 --10 \Gamma7 . Such values are about an order of magni­
tude smaller than those derived by Cesaroni et al. (1994)
for a sample of hot cores close to UC Hii regions. However,

C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations 13
these authors mapped NH 3 (4,4), a higher excitation tran­
sition that very likely arises from denser hotter regions
than the (2,2) and (3,3) lines, which can partly explain
the discrepancy between their and our results. This seems
confirmed also by the slightly lower T k of our objects as
compared to those of Cesaroni et al. (1994), who find T k
ranging from 50 to 165 K.
The fundamental conclusion we wish to stress is that
our VLA observations confirm the existence of compact
( ! ¸ 0.1 pc) cores positionally coincident with the H 2 O and
OH masers: such cores turn out to be hot (40--90 K) and
massive (30--900 M fi ), thus suggesting that high mass star
formation is going on inside them.
5. The case of G24.78+0.08
Among the objects of our sample, G24.78+0.08 turns out
to be the most complex. In particular, we note that (see
Fig. 6):
-- the ammonia clump extends over a region of ¸ 0:35 pc,
larger than in the other sources, although an optically
thick core is seen towards the OH and H 2 O masers;
-- a pointlike UC Hii region (G24.78A) is very likely em­
bedded in such core, as suggested by the NH 3 absorp­
tion seen towards it (Fig. 8);
-- the ammonia emitting region contains two centers of
star formation, as indicated by the existence of two
groups of H 2 O maser spots, one (hereafter H 2 O--A) co­
incident with the NH 3 absorption (i.e. with the UC Hii
region G24.78A) and close to the OH masers, the other
(hereafter H 2 O--M) located ¸ 9 00 (0.34 pc) to NE from
the former, at the position called G24.78 M.
This scenario is reminiscent of the morphology of
W3(OH), where OH maser emission is seen close to the
UC Hii region, whereas H 2 O masers lie in a molecular core
located ¸ 7 00 from it. This is commonly interpreted with
the existence of two massive stars in different evolution­
ary phases, one still very young embedded in the molec­
ular core, the other, older, at the center of the UC Hii
region. Such an interpretation can probably apply also to
G24.78, with G24.78A playing the role of W3(OH) and
G24.78M that of W3(H 2 O). A few important differences
are there though: first of all, although G24.78M lies in
the NH 3 clump, no NH 3 optically thick core is seen by us
at its position; moreover, unlike W3(OH), H 2 O masers are
seen also towards the UC Hii region G24.78A; finally, such
UC Hii region is smaller than W3(OH) (!0.01 pc as op­
posed to 0.02 pc). The difference in size can be explained
in terms of an earlier evolutionary status of the former
UC Hii region with respect to the latter: this would also
be consistent with the idea that H 2 O masers (unlike OH
masers) form prior to the appearance of an Hii region and
disappear when this evolves.
Another noticeable feature of G24.78 is that in each of
the three groups (OH, H 2 O--A, H 2 O--M) the maser spots
are aligned along a straight line. By using the positional
and velocity information (Forster, priv. comm.), we have
studied the velocity gradient of the masers. As a first step,
a least square fit to the spots in each maser group has
been done. Interestingly, the resulting three directions are
almost parallel, being characterised by the following angles
(from north to west): 44 ffi (OH), 66 ffi (H 2 O--A), and 46 ffi
(H 2 O--M).
This seems to indicate a preferred direction in the
molecular clump where the masers are embedded. Such
direction might be related to the magnetic field, as pre­
dicted by the models of Shu et al. (1993) and McKee et al.
(1993) for the magnetically supercritical collapse. If this
is the case, then two scenarios are possible:
1. According to Elitzur et al. (1989) H 2 O masers form
behind shocks: in particular, in the assumption of a
YSO emitting an isotropic wind through a medium
threaded by a magnetic field, the model predicts that
the brightest masers are those located in a ring around
the magnetic equator (Heiles et al. 1993). In this case,
the masers would trace a direction perpendicular to
the magnetic field.
2. If H 2 O masers are strictly related to bipolar flows as
suggested, e.g., by Felli et al. (1992), then the direction
described by the maser spots should be parallel to the
outflow axis and hence to the direction of the magnetic
field. In fact, the NH 3 clump in G24.78 looks slightly
flattened perpendicularly to the masers: this might in­
dicate that the maser distributions are parallel to the
direction along which the collapse has occurred, i.e.
the magnetic field direction.
In an attempt to further investigate this issue, in
Fig. 9, we plot the LSR velocities of the maser spots
against the projected distance along the direction given
by the linear fit. The dashed line stands for the VLSR
of the NH 3 (2,2) emission at the peak position, which is
110 km s \Gamma1 (Table 10). It is worth noting that this veloc­
ity does not differ from those of the ammonia absorption
and emission seen respectively against G24.78A and to­
wards the H 2 O--M group. Figure 9 shows that the VLSR
of the maser spots changes steadily along each group, but
around the centre it shows a spike. This is particularly
evident in the H 2 O--A group. Such effect suggests the ex­
istence of a systemic velocity field (like accretion, expan­
sion, or rotation) close to the central object powering the
masers. However, it is difficult to distinguish among dif­
ferent models, at least in the case of H 2 O--M. As for the
OH and H 2 O--A masers, one sees that almost all spots
are blue­shifted with respect to the NH 3 gas. This can be
explained in a model where the UC Hii region G24.78A
has formed inside the NH 3 core just at the SE border of
it, on the opposite side of the core with respect to the
observer: in this scenario, most of the molecular gas lies
among the UC Hii region and the observer (giving rise to
NH 3 absorption) and to the NW of G24.78A (originat­

14 C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations
ing the NH 3 emission). Therefore, if the maser spots are
bullets shot away from the central stellar object against
the surrounding dense molecular gas, then only those shot
towards the observer can be seen, which hence have blue­
shifted VLSR , as observed.
Although the previous facts seems to favour the in­
terpretation of masers as originating in outflows, we con­
clude that at present, both hypotheses above (1. and 2.)
are highly speculative and in no way we can rule out any
of them. Better signal­to­noise ammonia maps with higher
spectral resolution are in order, to obtain detailed infor­
mation about the physics of the molecular gas. Also, VLA
maps of the H 2 O masers done with higher angular reso­
lution than that of Forster & Caswell (1989), would be
worth to studying the proper motions of the maser spots
and hence the intrinsic velocity field of the gas close to
young massive stellar objects.
Fig. 9. Plot of LSR velocity versus projected distance for the
maser spots of the three maser groups of G24.78+0.08: OH
(empty squares), H2O--A (empty triangles) and H2O--M (filled
triangles). The projected distance is given in seconds of arc
(1 00 =0.037 pc) measured from a ``center of mass'' weighted ac­
cording to the maser intensities. The dashed line indicates the
LSR velocity of ammonia (¸ 110 km s \Gamma1 )
6. Conclusions
We have used the VLA for mapping the NH 3 (2,2) and
(3,3) and the 1.3 cm continuum emission towards 5 sources
selected from the Forster & Caswell (1989) sample of
H 2 O and OH masers. We detect ammonia emission in all
sources, but in one case the emission could not be mapped
mostly because of the short integration time used. As for
the 4 sources where the spatial distribution could be prop­
erly studied, we can draw the following conclusions:
1. NH 3 cores are seen toward the positions of the H 2 O
and OH masers;
2. only in one case we detect an UC Hii region coincident
with the NH 3 core: such UC Hii region is likely to be
very young, still at the very beginning of its develop­
ment;
3. the ammonia cores are small (¸ 0:1), hot (30--90 K),
and massive (¸ 300 M fi ), i.e. similar to the hot cores
seen by other authors (e.g. Cesaroni et al. 1994) in
regions of massive star formation;
4. in one case (G24.78+0.08) there seem to be indication
of the fact that H 2 O masers appear in a stage of stellar
evolution prior to that of OH masers and disappear
when an UC Hii region develops (see also e.g. Codella
& Felli 1995 and Codella et al. 1996).
The most relevant result of our search is that H 2 O
masers originate in hot massive cores, which are likely
sites of very young massive (proto)stars prior to the de­
velopment of an UC Hii region.
Acknowledgements. We thank M. Felli for critically reading the
manuscript, R. Foster for providing the VLSR and positions of
the masers spots, S. Kurtz for making us available the results of
the continuum observations of G12.68--0.18, and L. Moscadelli
for useful discussions. This research has made use of the Sim­
bad database, operated at CDS, Strasbourg, France.
References
Becker R.H., White R.L., Helfand D.J., Zoonematkermani S.
1987, ApJS, 91, 347
Brand J. 1986, Ph.D. Thesis, Leiden University
Braz M.A., Epchtein N. 1987, A&A, 176, 245
Cesaroni R., Churchwell E., Hofner P., Walmsley C.M., Kurtz
S. 1994, A&A, 288, 903
Cesaroni R., Walmsley C.M., Churchwell E. 1992, A&A, 256,
618
Churchwell E., Walmsley C.M., Cesaroni R. 1990, A&AS, 83,
119
Codella C., Felli M. 1995, A&A, 302, 521
Codella C., Felli M., Natale V. 1996, A&A, 311, 971
Cohen R.J. 1989, Rep. Prog. Phys., 52, 881
Elitzur M., Hollenbach D.J., McKee C. 1989, ApJ, 346, 983
Felli M., Palagi F., Tofani G. 1992, A&A, 255, 293
Felli M., Testi L., Valdettaro R., Wang J.­J. 1997, A&A, in
press
Forster J.R., Caswell J.L. 1989, A&A, 213, 339
Genzel R., Downes D. 1995, A&AS, 30, 145
Heiles C., Goodman A.A., McKee C.F., Zweibel E.G. 1993, in
``Protostars and Plantes III'', eds. E.H. Levy & J. Lunine
(Tucson: Univ. Arizona Press), p. 320
Henning Th., Cesaroni R., Walmsley C.M., Pfau W. 1992,
A&AS, 93, 525
Hofner P., Churchwell E. 1996, A&AS, 120, 283
Hunter T.R., Testi L., Taylor G.B., et al. 1995, A&A, 302, 249

C. Codella et al.: The molecular environment of H2O masers:VLA ammonia observations 15
IRAS Point Source Catalogue 1985, Joint IRAS Science Work­
ing Group (U.S. GPO, Washington)
Kurtz S., Churchwell E., Wood D.O.S. 1994, ApJS, 91, 659
McKee C.F., Zweibel E.G., Goodman A.A., Heiles C. 1993, in
``Protostars and Plantes III'', Eds. E.H. Levy & J. Lunine
(Tucson: Univ. Arizona Press), p. 327
MacLaren I., Richardson K.M., Wolfendale A.W. 1988, ApJ,
333, 821
Palagi F., Cesaroni R., Comoretto G., Felli M., Natale V. 1993,
A&AS, 101, 153
Palla F., Brand J., Cesaroni R., Comoretto G., Felli M. 1991,
A&A, 246, 249
Palla F., Cesaroni R., Brand J., et al. 1993, A&A, 280, 599
Palla F., Testi L., Hunter T.R., et al. 1995, A&A, 293, 521
Panagia N. 1973, AJ, 78, 929
Persi P., Roth M., Tapia M., et al. 1996, A&A, 307, 591
Scalise E. Jr., Rodr'iguez L.F., Mendoza­Torres E. 1989, A&A,
221, 105
Schraml J., Mezger P.G. 1969, 156, 269
Shu F.H., Najita J., Galli D., Ostriker E., Lizano S. 1993, in
``Protostars and Plantes III'', Eds. E.H. Levy & J. Lunine
(Tucson: Univ. Arizona Press), p. 3
Testi L., Codella C., Cesaroni R. 1994a, in ``IV workshop inter­
disciplinare sull'uso delle antenne di Medicina e Noto'', F.
Palagi and R. Cesaroni eds., p. 173, Osservatorio Astrofisico
di Arcetri
Testi L., Felli M., Persi P., Roth M. 1994b, A&A, 288, 634
Testi L., Felli M., Persi P. 1995, Mem. S. A. It., vol. 66, p. 677
Tofani G., Felli M., Taylor G.B., Hunter T.R., 1995, A&AS,
112, 299
Turner J.L., Welch W.J. 1984, ApJ, 287, L81
Ungerechts H., Walmsley C.M., Winnewisser G., 1986, A&A,
157, 207
Wink J.E., Duvert G., Guilloteau S., et al. 1994, A&A, 281,
505
Wood D.O.S., Churchwell E. 1989a, ApJS, 69, 831
Wood D.O.S., Churchwell E. 1989b, ApJ, 340, 265
Wouterloot J.G.A., Walmsley C.M. 1986, A&A, 168, 237
This article was processed by the author using Springer­Verlag
L A T E X A&A style file L­AA version 3.