4. Comments on individual regions
4.1. The relation between the W3 H II regions and the dense molecular W3 core
Fig. 6 shows an overlay of the radio continuum emission at 6 cm with the emission of the rotational transition J=2-1 in C18 O measured with the IRAM 30-m telescope (Tieftrunk et al. 1995). The C18 O emission traces the dense molecular gas in the W3 core. The peak positions of C18 O and C34 S clumps given by Tieftrunk et al. (1995) are marked as filled dots in Fig. 6 (cf. Roberts et al. 1996); the positions of the m peaks SMS 1, 2 and 3 (Ladd et al. 1993) as crosses. There is a distinct anti-correlation between the compact H II regions and the dense molecular cores. The larger sources such as W3 A, D and H have apparently ionized and cleared away most of the remnant molecular gas from which they formed. These appear to cluster around the molecular cloud cores. Ultracompact H II regions such as W3 C and F still seem to be partially embedded within the parental molecular cloud, their bright rims pointed toward the very dense molecular gas within these clouds. Only the hypercompact continuum sources W3 M and Ca are directly associated with denser molecular gas and submillimeter sources.
4.2. The hypercompact continuum sources (d AU)
The cluster of hypercompact continuum sources W3 Ma-g , with diameters of AU (cf. Table 3), is situated between W3 A and B. A first analysis of this region was made by Claussen et al. (1994), who found seven hypercompact continuum sources. This cluster is associated with the infrared double IRS 5 at , (Megeath et al. 1995). We find that the cluster is embedded in a very dense ( cm-3) and hot ( = 100 K) molecular clump (Tieftrunk et al. 1995) and is associated with the submillimeter source SMS 1 (Ladd et al. 1993). In Fig. 7a we show a uniform weighted image of the continuum emission of the W3 core at 4.9 GHz. The angular resolution is , combining data obtained from the D,C,B and A array configurations. Fig. 7b shows a resolution image of the 14.9 GHz radio continuum emission of W3 M. We find eight hypercompact continuum sources, also detected in our maps at 4.9 GHz and 22.5 GHz. The dotted contour in Fig. 7b indicates the size and orientation of the compact infrared double IRS 5 measured at 5µm, the cross shows the uncertainty in position (Megeath et al. 1995). The axis of the bipolar outflow observed in CO J=2-1 has the same position angle, when projected on the sky plane, the red wing emission extending toward the SW and the blue wing peak emission almost centered on the IRS 5 position (cf. Fig. 3, Mitchell et al. 1991).
Roberts et al. (1993) observed the Zeeman splitting in the 1420 MHz line of H I toward W3 with a resolution of using the VLA. Toward IRS 5 a highly collimated and very strong ( G) line-of-sight magnetic field has been detected with a gradient of G per arcsec, which self-reverses within of IRS 5. They present an hourglass model where the collapse of the surrounding molecular cloud resulted in a pinched magnetic field, self-reversed at the center. The direction of the bipolar outflow and the geometry of the S-shaped cluster of hypercompact continuum sources may be suggestive of a strong coupling with this magnetic field.
Willner (1982) showed that the near-IR to far-IR spectrum of W3 IRS5 is close to that of a blackbody, indicating that an ionizing star would be embedded in an optically thick dust shell. Oldham et al. (1994) and Campbell et al. (1995) present radiative-transfer dust-cloud models to fit the mm and submm data obtained toward IRS 4 and IRS 5. Campbell et al. (1995) base their model on the assumption that the dust cloud is powered by a group of seven B0.5-B0 ZAMS spectral type stars (Clausen et al. 1994), embedded in a dust free cavity of pc diameter. For the 8 hypercompact continuum sources, W3 Ma-g, we determine radii of pc with a maximum separation of pc between W3 Ma and g. If each hypercompact continuum source toward IRS 5 is indeed being ionized by a single stellar object, we would find these to be of ZAMS spectral type B3-B1. From recent near-infrared observations of the cluster toward IRS 5, however, Megeath et al. (1995) found a large deficit of low mass stars relative to the proposed number of high mass stars ionizing W3 M. They believe this is due to a misidentification of the radio continuum sources toward IRS 5 with H II regions ionized by B0 stars. Indeed, from the spectral indices between 6 cm and 1.3 cm, we find that it is not clear that every hypercompact continuum source is associated with an ionizing star. For the most intense hypercompact continuum sources W3 Md1/d2 and Ca we determined spectral indices between 2 cm and 1.3 cm, and negative indices, , between 6 cm and 2 cm. For W3 g and f we also find a flux density minimum for the 2 cm emission. The sources W3 a,b,c show a flat spectral index between 6 cm and 2 cm and negativ spectral indices, , between 2 cm and 1.3 cm. Thus, we believe that the cluster of hypercompact continuum sources toward W3 M may be far from being a simple association of ZAMS type B stars. In fact, these hypercompact continuum sources are somewhat reminiscent of the variable thermal and nonthermal cm continuum sources with similar diameters of pc and mean flux densities of a few mJy, monitored toward the "Orion Radio Zoo" by Felli et al. (1993). They discuss the possible nature of these hypercompact radio continuum emission sources near the Trapezium cluster in great detail.
In Fig. 7c we show the 1.3 cm radio continuum emission of the ultracompact H II region W3 C and its hypercompact radio continuum companion W3 Ca with an angular resolution of . W3 Ca at = , = (1950.0), is less than NE of W3 C. It has a diameter of AU. Images published as early as 1976 (Harris & Wynn-Williams 1976) actually showed this hypercompact radio continuum region, but it had not been resolved into a seperate object. Our sensitive high resolution images at 1.3 cm, 2 cm and 6 cm propose that W3 Ca, not W3 C, is the radio continuum region directly associated with the pointlike infrared source IRS 4. The m peak of IRS 4 has been measured at = , = (1950.0) by Ladd et al. (1993). Toward these sources there is dense molecular gas with cm-3 (Roberts et al. 1996, Tieftrunk et al. 1995). The position of this dense clump is identical with the m FIR continuum source SMS 2 (Ladd et al. 1993). The luminosity provided by a B2-B1 ZAMS star ionizing W3 Ca and a B0-O9.5 ionizing W3 C is in good agreement with the measured FIR luminosity of (Ladd et al. 1993). As with W3 Md1/d2 the spectral index of W3 Ca is negativ between 6 cm and 2 cm and positiv between 2 cm and 1.3 cm.
4.3. The ultracompact H II regions (d AU)
4.3.1. W3 C
Fig. 7c shows the ultracompact drop-like H II region W3 C with an angular diameter of less than , a deep central minimum and an irregular brightened edge. From our data we find that W3 C has a very high peak electron density of cm-3 and a low mass of of ionized hydrogen. W3 C lies less than south of the m peak SMS 2 at the edge of the dense molecular clump associated with this submillimeter source (Tieftrunk et al. 1995). SMS 2 has 20% of the luminosity of SMS 1 (Ladd et al. 1993), and the molecular clump is a factor 5 less dense. We find that W3 C is not associated with the pointlike source IRS 4, but rather W3 Ca, the hypercompact region NE of W3 C. The morphology of W3 C and Ca indicates two separate ionizing sources. Ladd et al. (1993) find a m peak at = .5 , = (1950.0) which is west of the m peak IRS 4. The discrepancy between the m and m peak is identical to the spatial separation of W3 C and Ca, as seen in our high resolution data. The reason for the separation in the infrared is unclear, but the energy distributions of pointlike and extended components show the growing influence of dust emission with aperture size for IR measurements. Extended components, such as W3 C, completely dominate over pointlike components, such as W3 Ca, at longer wavelengths. The aperture used to measure the m flux may not resolve a source ionizing W3 Ca, which is from W3 C. If IRS 4/W3 Ca is more deeply embedded in dense gas it may be expected to exhibit an IR maximum shortward of the maximum of W3 C. Thus, W3 C may be more extended than the probably younger hypercompact continuum source, because it is surrounded by less dense neutral gas.
The velocity gradient we find from our H66 recombination line data (Fig. 3) is indicative of an expanding shell around W3 C. The edges have velocities of km s-1 blueshifted with respect to the center with minimum emission. Fig. 7c shows that the southern bulge is strongly edge-brightened SW and NE of its center, a small "funnel" connecting the two lobes. Broad linewidths in the C18 O data of Tieftrunk et al. (1995) and red and blue pedestal features in CO and 13 CO are observed and may indicate a bipolar outflow (Hasegawa et al. 1994). If the double-peaked morphology of the continuum emission is related to a symmetric density inhomogenuity, this, in turn, may be related to the bipolar outflow (cf. W3 B). South of W3 C there is very dense molecular gas associated with the submillimeter source SMS 3 observed at m (Ladd et al. 1993). The southern edge of W3 C may be ionization bound (the free path length of the ionizing photons is shorter in dense gas - enough neutral gas is present for all ionizing photons to produce an ionization). The diffuse ionization tail expanding to the north, away from the dense center toward lower density regions of ambient material, could be density bound (the free path length of the ionizing photons is longer in diffuse gas - some ionizing photons will escape beyond the ionization). Here the ionization front may expand into the ambient medium as described by the champagne-flow model (Bodenheimer et al. 1979). Our H66 recombination line data shows that the southern peak of W3 C has a slightly blue-shifted maximum in respect to the northern part. This champagne flow may be the cause for the turbulent motion of the gas north of W3 C, indicated by the observed reversal in the gradients of the neutral (Tieftrunk et al. 1995) and ionized (Roelfsema & Goss 1991) gas.
4.3.2. W3 E
In Fig. 8 we show the 6 cm continuum emission of W3 B, E, F, G, M and the 2 cm continuum emission of W3 J and K. Megeath et al. (1995) detected a spectral type O7-B0 ZAMS with a shell of diffuse near-IR emission embedded in W3 E. We find an ionized gas mass for W3 E of and the Lyman continuum photon flux indicates a B0.5-B0 ZAMS.
4.3.3. W3 F
The ultracompact H II region W3 F shows a bright rim when observed with high angular resolution. This source is shown in Fig. 8. It is coincident with the infrared source IRS 7 at = .1, = (1950.0) (Wynn-Williams et al. 1972). From near-IR images, Megeath et al. (1995) detect a star in the center of W3 F. From our data we find that a B0-O9.5 spectral type ZAMS is ionizing W3 F. Megeath et al. (1995) find this star to be anomalously luminous at m, possibly because of strong excess emission from circumstellar dust and proposed that W3 F must be very deeply embedded. Our C18 O images when combined with multi-wavelength infrared images (Hackwell et al. 1978) show that W3 F lies southwest of the dense molecular dust cloud centered on IRS 5. We find high peak electron density at very low H II mass and a high electron gas temperature, K.
The tail region of W3 F is facing away from the dense molecular gas in the NE. From our H66 recombination line data we find that the ionized gas in this region has a velocity of -38 km s-1. The northern arc has a peak velocity of km s-1 and is blue-shifted in respect to the neutral gas. Thus, the velocity gradient from head to tail of km s-1 (cf. Fig. 4), when seen in relation to the velocity gradient of the ambient gas, may support the model of a bow shock created by a star (van Buren & Mac Low 1992) moving toward the dense molecular core NE of W3 F. The bright edge to the NE may then indicate an ionization front created by the ram pressure of the star within W3 F. However, if the molecular clump to the northeast is also surrounding W3 F as suggested by Megeath et al. (1995), the exciting star may carve a photoionized arc into this denser neutral gas. The center of the clump would be on the convex side of the W3 F arc, where continuous mass injection from the clump could maintain a recombination-front bound ultra-compact H II region (Redman et al. 1996). The diffuse ionization front expanding SW away from the dense gas could be density bound. At K, the thermal pressure of the H II region may ( ) produce a champagne-flow toward the lower density gas in the SW.
4.3.4. W3 G
W3 G lies between the two dense molecular condensations of the W3 core (cf. Fig. 6). No counterpart in the infrared or optical has been associated with W3 G. From our data, we deduce the ionizing source would have to be of spectral type B0.5-B0 ZAMS. It is unclear why such an ionizing source is not detected, unless it is located behind the bulk of dense molecular gas as seen in Fig. 6. The cm continuum emission of W3 G peaks in the SW where the neutral gas is densest.
Images at low and high resolution from our high sensitivity 1.3 cm, 2 cm or 6 cm data do not show the existence of the source W3 L south of IRS 7 given by Colley (1980).
4.4. The compact H II regions
4.4.1. W3 A
W3 A is a shell-like and asymmetrically edge-brightened H II region. Fig. 9 shows an image of the radio continuum emission at 2 cm of the compact radio source W3 A. W3 F and M can be seen to the SW in this image. From our data we find no cm continuum emission toward the double point source IRS 11, which is associated with a dense ( cm-3) clump with a warm dust component (Hayward et al. 1989). Measurements of the visual extinction (Hayward et al. 1989) and dust toward W3 A (Hackwell et al. 1978) suggest that the column density of neutral gas must be low. Fig. 6 shows that the main part of W3 A is no longer embedded in dense molecular gas (cf. Roberts et al. 1996). The remnant shell of W3 A is ionized by a group of young stellar objects: the embedded O5-O6 spectral type star IRS 2 and the O5-B1 star IRS 2a (Wynn-Williams et al. 1972, Beetz et al. 1974, 1976, Harris & Wynn-Williams 1976), which are located in the center of W3 A. IRS 2b lies in a centrally condensed cloud of warm dust seen in mid-IR images in the NW of W3 A (Hayward et al. 1989). It may have just entered its dust clearing phase and may be causing the edge-brightening of ionized gas seen in the NW of our W3 A image. IRS 2c lies east of IRS 2 and is less luminous. It has the same extinction and has been associated with the same H II gas (Hayward et al. 1989).
Dickel et al. (1983) described W3 A as an emerging blister on the edge of a warm dust cloud, representing an early stage in the "champagne flow" model of Bodenheimer et al. (1979). It is more likely, however, that W3 A is actually in a late stage of its expansion. Compared to the ultracompact H II regions the bulk emission of W3 A appears spherical and homogeneous. Spectral indices increase toward regions of strong radio continuum emission. The turbulent small structures in the expanding shell, that is trunks and wiggles, are of an angular size comparable to the ultra- and hypercompact radio continuum regions. These ionized eddies seen at high resolution (cf. Fig. 9) are probably the remains of the expansion of the ionization front into very inhomogeneous ambient neutral gas over time. The broad ionization front seems to curve around compact dense molecular clumps (Tieftrunk et al. 1995) embedded in the ambient gas. To the south our data shows extended low emission and here photons may escape into the lower density ambient medium. Lower resolution CO images (Dickel et al. 1980, Thronson et al. 1985) and far-IR observations (Werner et al. 1980) show very little emission SE of W3 A, so this density gradient must be present. From the 's of the H76 and the H110 lines, Roelfsema & Goss (1991) concluded that the H II gas in W3 A is expanding away from its central stars with a velocity of km s-1. Our H66 recombination line data of W3 A has a low signal-to-noise ratio, but we find km s-1 toward several positions in the center of W3 A, whereas toward the edges 's peak at -43km s-1 to -45km s-1. These velocity profiles are consistent with the model of an expanding shell.
4.4.2. W3 B
The diffuse infrared emission of the more confined asymmetrically bright shell-like region W3 B had been labeled IRS 3 (Hackwell et al. 1978). New infrared images (Megeath et al. 1995) have resolved a very compact source IRS 3a at m located at = , = (1950.0) (cf. Fig. 8). If this star is ionizing W3 B, we find a Lyman continuum flux, which corresponds to an O7-O6.5 spectral type ZAMS star. From Fig. 2, two distinct maxima in the cm continuum, separated by , can be seen in the south of W3 B. In Fig. 2, these two peaks can also be distinguished from the different velocities seen in the H66 recombination line data. The cause of this double-peaked emission, similar to the bipolar structure observed toward W3 C, is unclear. Within the positional uncertainties of the IR measurements, the location of the cavity between the two peaks is coincident with the emission peak of IRS 3a.
W3 B shows strong limb-brightening to the NE and in the SW and S, where the compressed ridge of the dense molecular gas can be found. In Fig. 6 we see that W3 B is located between the two dense molecular condensations. Our images show an extended filamentary feature in the NW, with a high electron temperature of K. This limb is oriented toward molecular gas of low density. From infrared data, Megeath et al. (1995) find very low extinction toward this NW region. The dust found toward W3 B is depleted by a factor of and well mixed with the ionized gas according to Hackwell et al. (1978). Thus, W3 B is most likely an emerging blister only partly embedded in the molecular cloud. Roelfsema & Goss (1991) find a north-south velocity gradient and suggest a rotation of W3 B around an axis perpendicular to this gradient at km s-1. From our first-moment analysis, we interpret this velocity gradient as the result of ionized gas in the tail region moving at a different velocity from the gas expanding in the head. Our H66 recombination line data shows that the southern bulge of W3 B is expanding at a velocity of km s-1 with a central velocity of -43km s-1, the found for the ambient molecular gas (Tieftrunk et al. 1995). The gas in the tail region is moving at -35 km s-1. If W3 B is an emerging blister, more deeply embedded in the S and SE where IRS 3a ionizes denser gas, then the blueshifted limb (cf. Fig. 2) in the NW may be expanding more rapidly toward the north following the density gradient of the neutral ambient gas. Qualitatively, W3 B exhibits the characteristics which would place it, on an evolutionary timescale, between the young W3 C and old W3 A region. W3 B is most likely ionization bound along its southern edge, where the edge of the dense molecular gas has been detected.
4.4.3. W3 D
W3 D is a very weak and extended H II region with a brighter edge facing W3 C, where there is denser molecular gas (cf. Fig. 6&7). W3 D, C and Ca presumably all formed in the same dense volume of molecular gas. W3 D is associated with the infrared source IRS 10 (1977). From the measured flux density, we find that a stellar type O8.5 ZAMS may be ionizing the dense neutral gas. The velocity gradient from km s-1 in the NW to km s-1 in the SE matches the velocities of the neutral gas (Tieftrunk et al. 1995). We believe that W3 D may still be ionization bound in the SE, but the ionized gas is rapidly expanding toward us in the NW according to the density gradient seen in the neutral ambient gas. In the SE and S, the direction of the velocity gradient of the neutral Tieftrunk et al. (1995) and the ionized gas (Roelfsema & Goss 1991) is the reverse of that found on larger scale. From our continuum data, we propose that the reversal of this velocity field may be caused by the champagne flow of W3 C rather than effects related to "magnetic braking" proposed earlier (Tieftrunk et al. 1995).
4.5. The diffuse H II regions
4.5.1. W3 H
The diffuse H II region W3 H is located north of the dense molecular core of W3 (cf. Fig. 6&7). W3 H is the most northernly source of the W3 core. Column densities of the molecular gas are low toward this expanded H II region. The probable exciting star of W3 H, has been identified as a O6-B2 spectral type with a visual extinction of (Beetz et al. 1976, Ogura & Ishida 1976, Cohen & Lewis 1978, Schultz et al. 1978, Axon & Ellis 1976). This is consistent with our observations. From our H II continuum and molecular images Tieftrunk et al. (1995) and the C18 O BIMA maps of Roberts et al. (1996), covering a larger field of view, we can see that W3 H and the exciting star are not associated with any dense molecular gas.
4.5.2. W3 J & K
W3 J and K are very diffuse and expanded H II regions over an arcmin south of the dense molecular gas associated with the H II sources forming the W3 core (cf. Fig. 1&8). The lack of emission from denser molecular material or dust continuum indicates that these regions are older than the core regions and the ionizing sources have already dispersed their natal cloud. Although edge-brightening is still observed, as for W3 J to the SW, these features are extremely low brightness, compared to core sources. The ionizing sources may be photoevaporating neutral gas remnant in filaments of low density. Colley (1980) proposed that W3 J and K are blister type H II regions in a very evolved evolutionary state. We find that stars of spectral type O8.5-O8 and O7-O6.5 ZAMS could produce the observed Lyman continuum photon fluxes toward W3 J and K, respectively. Positions of an O9-B3 spectral type star with a visual extinction of toward W3 J and an O5-B2 star with toward W3 K (Beetz et al. 1976, Schultz et al. 1978, Axon & Ellis 1976, Cohen & Lewis 1978) are given in Fig. 8.
© European Southern Observatory (ESO) 1997
Online publication: July 3, 1998