From Fig. 1 and the estimates regarding dust masses and resulting extinctions made in Sect. 3.1.1, it becomes clear that G11 is located close to the centre of a large molecular cloud. The K-band extinction, estimated in Sect. 3.1.1 under the assumption of G11 being located halfway into the observed cloud, amounts to 3.8mag. This matches fairly well with the mean extinction of 3.5mag measured by the comparison from free-free and Br $\ensuremath {\gamma }$ radiation in Sect. 3.2. The peak value of 4.3mag derived by that comparison cannot be compared directly to the dust-based estimate because it appears on much smaller spatial scales than the resolution of the 1.3mm map can reproduce.

We find the density of the ionized gas of 4.3 $\ensuremath{~ 10^{3}}$ $\ensuremath{{\rm cm}^{-3}}$ derived in Sect. 3.2 to be slightly lower than the general hydrogen density of 3.6 $\ensuremath{~ 10^{4}}$ $\ensuremath{{\rm cm}^{-3}}$ based on the dust mass. For the outer regions, the millimetre-continuum measurements yield a lower value of 5.8 $\ensuremath{~ 10^{3}}$ $\ensuremath{{\rm cm}^{-3}}$ . This means that the electron density measured from the radio continuum at 2cm is of the same order as the measured halo density of the G11 molecular cloud, but falls short of the measured core density.

The coarse spatial resolution of the 1.3mm map would not reveal the presence of a potential cavity blown free of dust by the central O star. The density inside this cavity might be significantly lower than in the rest of the cloud core. Indeed, when judging from Figs. 6 and 7, the ionized region seems to be located at the rim of a region of much higher extinction immediately to the south (no fainter background sources south of G11 in Fig. 6). When judging the position from Fig. 1, the UCH II is located on the southern rim of the cloud core. This means that a small central cavity might indeed exist.

$\begin{figure} \par\includegraphics[angle=90,width=8cm]{figure8.ps} \end{figure}$

Figure 8: K' image of G341. The logarithmic gray scale ranges from 0.14mJy/ $\Box$ $^{\prime \prime }$ to 33.8mJy/ $\Box$ $^{\prime \prime }$ . The image was subject to a maximum entropy filtering algorithm (see text). The position of the hydroxyl maser from Caswell (1998) is marked by a cross. The contours are from the H-band image, ranging from 0.007mJy/ $\Box$ $^{\prime \prime }$ to 4.7mJy/ $\Box$ $^{\prime \prime }$ in steps of 1.58 (0.5 mag)

$\begin{figure} \par\includegraphics[angle=90,width=8cm]{figure9.ps} \end{figure}$

Figure 9: H-band image of G341. The logarithmic gray scale ranges from 0.23mJy/ $\Box$ $^{\prime \prime }$ to 52.6mJy/ $\Box$ $^{\prime \prime }$ . The image was subject to a maximum entropy filtering algorithm (see text). The contours are from the N-band map. Contour levels are 84, 118, 151, 185, 218, and 252mJy/ $\Box$ $^{\prime \prime }$ . Again, the cross marks the position of the hydroxyl maser from Caswell (1998)

4.1.2 The central engine

Figure 11 shows the spectral energy distribution of G11. The data are taken from the IRAS point source catalogue (including the LRS spectrum), the MSX catalogue (see Sect. 3.4) and from our observations. Apart from the steeply rising shape of the SED up to 100 $\ensuremath {\mu }$ m, which is typical for UCH IIs, we notice a strong emission feature around 7 $\ensuremath {\mu }$ m. This feature is probably due to polycyclic aromatic hydrocarbons (PAHs). When we integrate the flux using a simple trapezoidal integration, we end up with a total luminosity of 7.2 $\ensuremath{~ 10^{4}}$ $\ensuremath{{L}_{\odot}}$ . This corresponds to a difference of 12.14mag from the bolometric absolute magnitude of a G2V star. According to Straizys (1995), the resulting bolometric absolute magnitude of -7.53mag points to a star of spectral type O7ZAMS. We note that although the error of the rough integration may be considerable, we expect it to be no larger than about 30%. This estimate comes from a comparison with an artificial SED for G11 that peaks at 200 $\ensuremath {\mu }$ m at 400Jy and then stays constant out to 1.3mm.

Since both the SED and the ionization point to an O7ZAMS star, it appears that the determination of the spectral type from the colour magnitude diagram is contaminated by some additional flux in K and H. Since the de-reddening procedure described in Sect. 3.6.1 places the central star exactly on the ZAMS within the photometric errors of about 0.1mag, it seems unlikely that large errors arise from this step. Another error source might be the slope of NIR reddening vector, but the existing reddening laws differ little in this near-infrared region. Contamination of the Kmagnitude by Br $\ensuremath {\gamma }$ -flux might be a reasonable cause, but our measurements show that the Br $\ensuremath {\gamma }$ -flux contributes less than 0.1mag to the total K-magnitude, whereas the difference in brightness between an O7 and an O5ZAMS star is about 0.7mag. This brightness difference means that we measure about twice the flux in the NIR than expected from a star of spectral type O7.

Two possible explanations remain: first, the surrounding dust might scatter light into the line of sight that would not have reached the observer without this dust. The resulting anisotropic radiation distribution might resemble a source brighter in the NIR than is actually present. Future polarization measurements might clarify whether this is the case or not. Second, an unresolved binary system or even a very compact cluster might be contained in what we identify as the ionizing source. The most massive star allowed to contribute to the total luminosity of the system, besides an O7 star, would be of spectral type B0. An unresolved binary consisting of an O7 and a B0 star would have a combined K-band magnitude of 13.4mag, close to the measured 13.3mag and within the 0.1mag error limit. The colour index of the system would not be affected by the binary. The total luminosity of the system would be 30% above the result of the trapezoidal integration, only just within the limit of the above error estimate. A somewhat fainter star of spectral type B would be a better fit here. Several other stars of spectral type B have been identified inside G11 (see Table 5), so this might be asuggestive of an embedded cluster. However, our detection limit does not allow us to probe for more deeply embedded or lower mass members.

4.1.3 Star formation efficiency

The total mass of the G11 region is 1 $\ensuremath{~ 10^{4}}$ $\ensuremath{{M}_{\odot}}$ . As the central heating object appears to be of spectral type O7, its mass should be about 25 $\ensuremath{{M}_{\odot}}$ . Two additional stars have been identified inside G11, both of spectral type around B1 and thus contributing a total mass of about 20 $\ensuremath{{M}_{\odot}}$ . Assuming that these stars contain the most significant part of the total (young) stellar mass inside G11, we derive an absolute lower limit for the star formation efficiency of about 0.6% for G11. Together with the total luminosity of 7.2 $\ensuremath{~ 10^{4}}$ $\ensuremath{{L}_{\odot}}$ , we derive an L/M ratio of about 7 for the complete region while that of the central source is 2500 $\ensuremath{{L}_{\odot}}$ / $\ensuremath{{M}_{\odot}}$ . The L/M value is often used as a measure for the star formation efficiency. The values obtained here compare well with the results discussed by Henning et al. (2000b) for other regions of massive star formation.

4.2 The nature of G341

4.2.1 Fragmented cloud cores

The millimetre continuum map (Fig. 2) shows the image of a filamentary and fragmented cloud. The region of G341 contains two clearly distinct cores. These cores are separated by a projected distance of about 1.6 pc. This distance points to two independently collapsing cores of an extended cloud rather than to a binary system of two massive young stars. The object G341 studied in this paper by AO and other observations is close to the northwestern core as can be seen from Figs. 8 and 4. The southeastern core is the known IRAS source 16487-4423, probably another region of massive star formation.

In the following discussion, we will concentrate on the northwestern core, which is also the location of the OH maser (Caswell 1998). The projected distances between the individual infrared sources visible in our AO image in Fig. 8 range from 2 $^{\prime \prime }$ to 4 $^{\prime \prime }$ (7000-14 000AU). These distances are comparable to those between the massive stars in the central area of the Orion nebula near the Trapezium region. A similar range of distances was found for the "pseudo Trapezium systems'' by Abt & Corbally (2000).

4.2.2 The central sources

From Fig. 10 we learn that most of the sources inside the region covered by our AO image are probably of spectral types O and B. The brightest source in the K'-band image (see Fig. 8) close to the OH maser position of Caswell (1998) is not plotted in the colour-magnitude diagram. Its H-K colour is 5.5mag. This might indicate that we are seeing extended emission from heated dust or scattered light. Scattering certainly does occur, as can be seen from the polarization map in Fig. 5.

The rather steep spectral index $a=\frac{{\rm d}(\log(\lambda F_{\lambda}))}{{\rm d}\log(\lambda)} \approx 2.3$ between 2.2 and 10.6 $\mu$ m would qualify this object as Lada class I, following Wilking et al. (1989). Dereddening can obviously not be applied to determine the spectral type of the embedded star. Using our four data points at 1.6, 2.2, 10.6, and 1300.0 $\ensuremath {\mu }$ m for a simple trapezoidal integration, we derive a total luminosity of 195 $\ensuremath{{L}_{\odot}}$ . Due to the lack of data points between 10.6 $\ensuremath {\mu }$ m and the millimetre measurement, this should be seen as an absolute lower limit because we miss the peak emission at mid-infrared/far-infrared wavelengths. We note that the spectral properties of the central source are very similar to those of the luminous young stellar object MIR1 found by Feldt et al. (1998) slightly north of the cometary UCH II G45.45+0.06.

Evidence for an embedded massive star comes from the association of the source with an OH maser(Caswell 1998). OH masers in star-forming regions are generally associated with (compact) UCH IIs. It is mostly accepted that the OH masers are located in the compressed shell between the ionization and shock front (see, e.g., Elitzur 1987). The H II/OH masers are probably excited by a combination of collisions and far-infrared continuum radiation. Luminous sources with a low geometrical dilution factor are required to explain the high brightness temperatures of the OH masers (Pavlakis & Kylafis 1996).

Another hint towards a region of high-mass star formation is the mass of the cloud core of 400 $\ensuremath{{M}_{\odot}}$ . This mass is comparable to the mass of G11's cloud core (620 $\ensuremath{{M}_{\odot}}$ ) and the core to the southeast, which is known to be associated with a luminous IRAS source and has a mass of 370 $\ensuremath{{M}_{\odot}}$ .

4.2.3 No visible ionization

Starting from our hypothesis that there is indeed a young high-mass star embedded in G341, we need to find a reason why no detectable ionization occurs in its immediate surroundings.

Due to the probable combination of reddening and excess emission it is virtually impossible to determine the spectral type of the embedded star from photometry. If we rather arbitrarily assume a spectral type of O8, we end up with a factor of 40 fewer UV photons than for the central source of G11. Such a star would emit 10^48.35 Lyman continuum photons per second (Panagia 1973). If we assume that this is also the number of recombinations in the nebula and that the ratio of recombinations to Br $\ensuremath {\gamma }$ photons is about $N_{\rm c} / N_{\rm Br\ensuremath{\gamma} } \sim 70$ (Hummer & Storey 1987), we can derive the expected Br $\ensuremath {\gamma }$ flux in Wm^-2 from

$\begin{displaymath}I_{\rm Br\ensuremath{\gamma} }=1.1~10^{-61}\left(\frac{N_{\rm... ...\mbox{s}^{-1}}\right) \left(\frac{D}{\mbox{kpc}}\right)^{-2} . \end{displaymath}$

(4)

The above estimate holds for the assumption that the free-free emission from the ionized gas is optically thin. However, this may not be the case. The strong, extended K'-band emission indicates very dusty surroundings very close to the central source. UV photons might be absorbed before they can cause ionization of gas atoms. In such a scenario, the central star might be of earlier spectral type than derived above and still no detectable ionization would be expected. In Sect. 3.3, we derived a lower limit of 2.8 $\ensuremath{~ 10^{5}}$ $\ensuremath{{\rm cm}^{-3}}$ for the hydrogen density around the embedded source from the N-band flux. Computing the UV optical depth using Ryter (1996) and Mathis (1990), we find that such a density would lead to an optical depth of 1 for energetic UV photons after a line of sight of about 100AU in length. Should this density be present very close to the star, all of the ionizing radiation might be absorbed within a few hundred AU. Another possible cause for the lack of visible ionization might be the infall of material to be ionized (see also below).

Another complication is the second brightest star in Fig. 8 at position (+5 $^{\prime \prime }$ , -1.5 $^{\prime \prime }$ ). This star appears in the colour-magnitude diagram about 0.7mag short of the ZAMS after de-reddening, as can be seen from Fig. 10. The rough determination of the extinction in Sect. 3.1.1, the uncertainty of the embedding depth plus additional circumstellar contributions can easily account for this shortfall. Hence, this second brightest source is almost certainly a star of spectral type O. Yet it doesn't produce a visible HII region. The source might be located at a much smaller distance and thus be not as bright in absolute magnitudes as it appears in Fig. 10, but this is unlikely because of the extinction which fits the one measured towards G341 very well. On the other hand, this source might be a fully evolved O star which has no detectable HII of its own any more. The projected distance towards G341 is about 22200AU. The linear size of G341 in Fig. 8 is about 11000AU. Thus the solid angle under which G341 is seen from the O star is about 0.16sr. Assuming the O star to be of spectral type O7, this means that 10^46.72Lyman continuum photons per second should hit the dust shell of G341. Accordingly, gas atoms at the surface of this dust shell should be ionized. Again, we can convert this number into an expected Br $\ensuremath {\gamma }$ flux using Eq. (4), and we end up with an expected flux density of $F_{\rm Br\ensuremath{\gamma} } \sim 18$ mJy. This is still above our detection limit but only by a factor of about three. Intermediate absorption by dust or a slight error in the assumption of a spectral type can explain why we would not detect Br $\ensuremath {\gamma }$ radiation from this type of ionization.

4.2.4 A UCHII precursor?

So far, we have been unable to determine exactly the type of the central source embedded in G341. However, we have argued that it might be a massive star and if it is, it must be so heavily enshrouded by dust that all of its ionizing radiation is absorbed before it can produce a detectable HII region or the HII region is suppressed by the accretion of matter (see, e.g., Yorke 1986; Henning 1990; Testi et al. 1997). The OH maser and the total mass of the cloud core are arguments in favour of a massive star-forming region. For a discussion of similar objects using radio continuum and OH maser data, we refer to Forster & Caswell (1999).

The strong excess emission in K' shows that large amounts of hot dust exist very close to the star. Its morphology would be governed both by local density variations and the temperature gradient. A very crude comparison to the model of Testi et al. (1997) shows that we should expect a colour index of K-H of about 5 mag on the central peak close to the star and about 1.5mag in outlying regions. We derived a colour index of 5.5mag for the central peak in K and measure a gradually decreasing colour index towards north/northeast which reaches 2.5mag at the northern end of the extended emission. Differences in foreground extinction and the density profiles can easily account for the differences. We should note however, that in contradiction to the model of Testi et al., in G341, intensity variations do not happen on scales of 1 $\ensuremath{~ 10^{-4}}$ pc but on scales of 1 $\ensuremath{~ 10^{-2}}$ pc.

4 Discussion

4.1 The nature of G11