In this section, we try to derive the nature of G45 from the measured properties. From the colour information, we determine the spectral types of the point sources and argue that G45 is a cluster of young, massive stars. To be able to do this, we first have to measure the extinction towards these sources. This information is extracted from a comparison between our Br data as well as the 6 cm and 3.6 cm VLA data by WC89 and W96.
Going further, we speculate on the cause of the ionization and the possible scenarios of the star formation sequence in the cluster. In the end, we will draw a parallel to the Orion Trapezium and KL/BN regions by comparing parts of G45 with these regions. Although the Orion region is not in the UCH II phase any more and the emission measure is two orders of magnitude lower there (Felli et al. 1993), we can still compare the dense cluster of stars at G45's southern edge with the Trapezium cluster and the larger part of it with the KL region. Perhaps MIR1 may even be an analogue to the BN object.
To derive the extinction towards the embedded sources in G45, we use the 6 cm map to determine the emission measure and thus the expected Br emission. By comparing this prediction to the actually measured Br flux, we get the extinction towards G45 at the line's wavelength. This extinction is assumed to be constant throughout the band. Several difficulties arise during this process: The 6 cm VLA map was obtained in the A configuration of the VLA. This means, the largest visible structures in the map can be of order across (WC89). From Fig. 4, as well as from the 3.6 cm map from W96 we learn however, that weak emission extends over a much larger area. Thus, we have to account for possible losses of flux in the 6 cm map. This is done by predicting the 6 cm flux from the 3.6 cm flux assuming a spectral index of 0.1. By comparing this to the real 6 cm map, smoothed to the same resolution, we get a map of fraction of missing flux, which is shown as contours in Fig. 7b. At the location of G45, the fraction of flux that is missing generally is of the order of 50%, i.e. we see only half of the flux we should in the 6 cm map. This means that by predicting the Br flux from this map alone leads to a serious underestimation and thus the comparison to the measured flux also underestimates the extinction. On the other hand, we can not use the 3.6 cm map to predict the Br emission, because the resolution of this map is too low and we would thus average the extinction over too large an area. As the Br image shows that most of the emission comes from a very small area, the measured extinction would again be largely underestimated by using the 3.6 cm map alone. To get out of this dilemma, we tried to correct the 6 cm map for the missing flux by using the now known fraction of missing flux derived above. The resulting corrected 6 cm map is then used to predict the Br flux and thereby determine the extinction in a way outlined in Watson et al. (1998): First, we convolve the corrected 6 cm image to match the resolution of our Br image. Then we use the familiar expressions to determine the brightness temperature at each point of the corrected 6 cm map (all units are cgs):
(e.g. Rohlfs 1990).
In this formula, is the solid angle of the synthesized beam, and is the flux density in mJy, both after the convolution to the new resolution! The quantity c denotes the speed of light, k is the Boltzmann constant and the frequency at which the data were taken. We combine this with
where is the optical depth, the electron temperature, the observed frequency, EM the emission measure, and a a correction factor of order 1 (Mezger & Henderson 1967). The combination gives the emission measure as
Once we have determined the emission measure, we can calculate the measured Br flux after Osterbrock (1989) to be:
The factor of 0.9 represents the assumption that only 90% of the nebula consists of Hydrogen, the rest of Helium. For , which comprises the transition coefficients and the level populations via a temperature dependency, Hummer & Storey (1987) give
The solid angle is now the pixel size of the Br image.
The emission measure computed from the corrected radio flux has its maximum at 21.7 pccm-6 with a mean value of 3.0 pccm-6, assuming the Br beam of 2.8 sr to be uniformly filled with 104 K gas.
The result of the whole procedure can be seen in Fig. 7. Fig. 7a shows the measured Br image with the continuum emission already subtracted. Some remnants of bright stars can be seen which are due to imperfect alignment of the line- and the continuum image during subtraction. Fig. 7b gives the prediction as computed from the 6 cm image. The ratio of the predicted and the measured Br flux yields the extinction towards the nebula in Br. We assume the resulting value to be constant over the -band and consequently use it as extinction. Also shown in this part of the figure are the contours denoting the missing 6 cm-flux, which show that at the location of G45 the missing flux fraction is about 50% with little variation across the region of interest. Fig. 7c shows the resulting extinction map superimposed as contours on the deconvolved -image. The extinction was calculated only in areas where both the flux in the 6 cm map and in the Br image was above 1.5 times the corresponding sky sigma level. In Fig. 7c, this means that outside the contours we do not have any information on the extinction, not that there is none! The resulting mean value of the extinction is 2.5 mag with a peak value of 3.6 mag. From the map in Fig. 7c it becomes clear that the extinction towards most of the detected point sources has a value of about 2.5 mag. The extinction values at the positions of the point sources are used to de-redden their colours and thus to determine their spectral types in the next section.
4.2. Point sources
In 1989, Wood & Churchwell wrote in their fundamental paper on UCH II regions: "Multiple ionizing stars are clearly producing the UCH II region complexes we have seen, and it is possible that close binaries or clusters of young stars are responsible for ionizing individual UCH II regions as well... If the Trapezium were at a distance of 5kpc, its angular diameter would be only , comparable to the size of many UCH II regions." Now it turns out that this was indeed a prophetic remark. The resolution of our AO images reveals for the first time that G45 contains more than a dozen single compact sources.
In the deconvolved image shown in Fig. 3, we labelled 15 point sources inside the VLA map, which is superimposed as contours in the figure, with letters from a to o. Table 2 gives the results of the photometric measurements of these sources.
To confirm the presumed stellar nature of these point sources, we
looked at their locations in a colour-magnitude diagram. Such a
diagram is shown in Fig. 8. For comparison, a zero-age main sequence
(ZAMS) as it would appear at a distance of 6.6 kpc is also shown. The
grey shaded area denotes the track for reddening by interstellar dust
grains after Rieke & Lebofsky (1985). It is clearly visible that
the reddening track allows the sources to be ZAMS stars of spectral
types O to B. To perform a further check on the nature of these
unresolved sources, we used the measured
-extinction values from Fig. 7c to de-redden the
detected point sources. Combining these extinction values with the
reddening vector from Rieke & Lebofsky (1985) results in the
locations marked with primed letters in the diagram. Some remarks have
to be made concerning this result:
Second, the error bars are relatively large. The main error contributions come from the uncertainty of the predicted and the measured Br flux. We assume a flux error of 25% in the corrected 6 cm flux and thus almost the same in the predicted Br flux. Together with the high noise in the Br image from the continuum subtraction and calibration uncertainties, we get an extinction uncertainty of 0.44 mag. In total, we note that de-reddening the sources confirms their nature as high-mass ZAMS stars with the restriction that additional sources of extinction may exist that are not measured by the Br method and that uncertainties are fairly large.
4.2.2. Other point sources
Two even more interesting sources are revealed by taking a look at the MIR images in Fig. 5: All show the strong point source MIR1 at position (,+10:005). No trace of this source is found in the H image. Its spectral energy distribution is shown as the lower line in Fig. 6. We believe that this source is one of the youngest within the G45 complex; This issue is discussed in Sect. 4.5.
Another, weaker point source (MIR2 ) is visible at position (+3:008,+4:008). From the superimposed radio contours in Fig. 5 it becomes clear that the 10 and 12 µm position of this source is coincident with a radio peak, while it appears shifted by 0:005 to the west at 3.5 µm. Additionally, there are 4 NIR point sources grouped along the northeastern side of this formation at a distance of approximately which corresponds to 6600 AU. Such positional shifts between wavelengths can be caused by surrounding dust or even a disk (see, e.g. Feldt et al. 1998 on Cha IRN, Stecklum et al. 1997 on GGD 27 or Close et al. 1997 on HL Tau). This object clearly not coincides with any NIR source.
Fig. 1 shows still more distinct sources, the two most interesting ones being the two mentioned before with the extended features attached to them at the positions (,+10:008) and (,+9:008) north of the chain-like arc. The extensions might be due to outflow phenomena, but for a save identification, high-resolution narrow-band images (H2 2-1 (S1)) which trace the shocked molecular gas are needed. A first criterion in this direction would be the non-detection in our narrow band NIR images, which is the case at least for the object (,+9:008). At the other location, the resolution of the Br and its accompanying continuum image does not permit to decide where the extended emission in these images originates.
As described in the preceding sections, most of the described point sources seem to be of stellar nature and thus to be young, massive stars. Their young age is generally indicated by the strong reddening or sometimes by attached extended features. Additional point sources can be seen in the VLA map shown in Fig. 3. The radio peaks are usually not coincident with NIR or MIR sources. Testi et al. (1997) have proposed an evolutionary sequence to explain that very young massive stars may not be visible at radio wavelengths, because their radio continuum is too much self-absorbed. Molinari et al. (1998) give the high mass infall rates at early stages of the massive star's life as a quenching mechanism for the radio emission.
However, the radio peaks are generally located close to NIR sources, which might indicate that they trace dense clumps of material which is ionized by the surrounding (young, massive) stars. They might even be identified as proplyd-like objects inside G45. Such disks around low-mass stars that are being ionized by nearby massive stars (see, e.g. McCullough et al. 1995 for an overview on the phenomenon) often exhibit some features of UCH II s themselves (see, e.g. Stecklum et al. 1998 on G5.97-1.17). While the detection of G5.97-1.17 would not have been possible at the distance of G45 within the WC89 sample due to sensitivity constraints, a brighter proplyd might well be detected. Given the shorter projected distances inside G45 of only 1300 AU between source a and its neighbouring radio peak (G5.97-1.17 is at 4900 AU projected distance from Her 36), the larger distance towards the observer can be compensated and we can expect comparable beam brightness temperatures for the radio peak in G45 and for G5.97+1.17. Of course, this requires similar properties of the two systems, like density and size of the evaporating structure. Actually, the brightness temperature of the peak north of source a is 2300 K and that of the source south of l, m, and n is 3500 K, while that of G5.97+1.17 is only 1500 K. Assuming the required similar properties of the evaporating structures, source a has to be of slightly earlier spectral type than Her 36 (Stecklum et al. 1998 follow Woolf 1961 in their assumption that Her 36 is of spectral type O7) to the additional Lyman continuum flux for the higher measured beam brightness temperature. We do not have any information about the extinction towards a, but its non-detection in H renders such a spectral type improbable. Sources l, m, n, and o might easily provide enough photon flux to ionize a region of high density gas and thus be responsible for the radio peak 0:008 southwest of them. Note that this object also appears as a thermal infrared source in the MIR images. In conclusion, we can not determine whether the radio peaks are ionized by the surrounding stars or internally or both, but all three possibilities clearly exist.
From the colour-magnitude diagram, we only have identified massive stars. We can estimate the range of the ZAMS we can cover with our method from the detection limit of mag and mag. Assuming a mean extinction of 2.5 mag in and thus 4.0 mag in H for our assumed reddening values, we would be able to detect and identify the entire range of O and B stars down to around B8. A glance at Fig. 8 confirms that our latest identified object b is of spectral type around B1. The fact that we do not see intermediate mass stars is striking, but does not exclude the presence of lower mass are present inside G45. Assuming coevality, such stars would of course be in an earlier stage of their development and thus probably not appear on the ZAMS. When we estimate the luminosity of pre-main sequence (PMS) stars from 0.4 to 2.5 solar masses at the age of 1Myr from D'Antona & Mazzitelli (1994) using the polynomial fit method from Meyer (1996), we expect a maximum apparent magnitude of for a PMS star of 2.5 . This estimate takes into account only the interstellar extinction and reddening towards G45 from Neckel & Klare (1980). If such a star is located inside the VLA map region with its mean extinction of 2.5 mag in , we are not able to detect it.
However, we do have some candidate objects that possibly represent stars in an earlier evolutionary stage than the ZAMS. The first one is object MIR1 , which is by far the most luminous MIR source. From Fig. 6 we learn that it contributes 10% of G45's infrared flux alone. It does not have a radio counterpart on the 6 cm map, but it is clearly visible in the new VLA map taken at 1.3 cm by Hofner et al. (1998). Also the two sources with the attached extended features might belong to the category of pre-ZAMS stars. As these sources at present are only detected in our high-resolution K-band image, their final identification will be the task of future studies of this region.
In general, none of the NIR sources show up in the mid-infrared images. From the sensitivities and resolutions of these images given in Table 1, this is what one would expect: Our brightest detected source, k, can be expected to be of =11.8 mag and =12.0 mag. For this estimate, we used the intrinsic colour indexes for an O5 star from Straizys (1995) and the reddening vector from Rieke & Lebofsky (1985). Although the source should show up in the L-band image, the large PSF-FWHM of this image prohibits the identification of single sources in this crowded field. Fig. 5 shows that MIR emission is highest close to the NIR point sources, whether this comes from the unresolved point sources themselves or from concentrations of warm dust cannot be decided from the data.
Table 1. Summary of observations
Overall, G45 seems to consist of the bright NIR sources close to a radio and MIR source at the southern edge, the chain of identified O- and B-stars along the radio ionization front and a couple of sources in an earlier phase of their evolution to the north. The linear distances fit that of massive star formation sites in our vicinity. From our biased sampling of the mass range it seems not unlikely that the initial mass function of this cluster extends towards lower mass stars which we are unable to identify at present.
4.3. The nebula
In this section, we will discuss the large scale morphological properties of the ionization nebula of G45. The large-scale structure of G45 is best visible in Fig. 4. Here, the Br emission comes from a shell-like structure with an additional arm stretching to the north. The ultracompact H II region G45 itself forms only the small, central part of this structure. This result is absolutely consistent with the findings of Garay et al. (1993) and W96, whose VLA image exhibits the same structure. As G45 is only a part of the shell-like structure, a closer look at its environment becomes necessary.
From the spherical shape and the large size of the ionized area, one might be tempted to conclude that G45 was not the primary cause for its ionization. Several possible sources of the large scale ionization provided by an external source can be thought of:
Now that we have rejected the possibilities that a non-associated source is responsible for the ionization and thus for the formation of G45, we have to take a closer look at the internal energy budget. Later on we will draw a sketch of the formation of G45 which does not require any external influence. Even the large shell provides a radio flux of only 3.7 Jy at 3.6 cm according to W96. Using Kurtz, Churchwell, and Wood's (1994) estimate of the necessary rate of Lyman UV photons to provide the ionization for this flux, we get a rate of s-1. According to model calculations by Panagia (1973), this rate can be provided by a single O6 star. Even at our adopted distance of 6.6 kpc and the corrected radio flux of 7.4 Jy, the required flux of s-1 ionizing photons per second can be provided by a single star of spectral type O5.5 ZAMS. In the absence of errors other than those of our photometry, our determination of spectral types from the colour indices plotted in Fig. 8 yields that source k should easily be able to provide the ionizing Lyman flux alone. Consequently, the energy budget of the shell structure does not require any ionizing source other than the stars we already see.
4.4. The role of dust in UCH II s
All these estimates were made in the absence of significant absorption of UV photons by dust. However, the MIR maps do show that at 10 µm the extended emission follows very well the ionization as seen in the radio map (see Fig. 5). Especially the 12 µm image, which does not contain the Ne II line proves that this is due to warm dust as noted above. Of course this is what one would expect, as the UV photons that ionize the gas also tend to heat up the dust grains - if a sufficient number of such grains exist. At 3.5 µm, the correlation between radio and MIR emission is less obvious. The differences in the distribution of the emission between the three MIR maps might well be explained by temperature effects. Additionally, these images prove that the distribution of hot dusts also appears clumpy and irregular with a concentration along the ionization front.
To estimate the amount of UV absorption, we first determine the dust mass from our measured 12 µm flux by using
Eq. (8) uses the mean emission measure pc cm-6and assumes G45's shape as a half sphere with radius AU to determine the average gas density. The factor represents the conversion to the mean radius of the half sphere. This results in an electron density of 1.2 cm-3, which is assumed to be equal to the gas density. Here, we use Jy from our 12 µm image inside an aperture of diameter, pc, THz as source dependent constants. is interpolated from Ossenkopf & Henning (1994) to be roughly m2kg-1 for 11.7 µm and the determined gas density of cm3. k is the Boltzmann constant, c the speed of light and h the Planck constant. For a temperature of 100 K we derive a dust mass of 0.02 . Two remarks have to be made concerning this result: Firstly, the above formula assumes that the emission is optically thin - which should be the case at 12 µm if any UV photons get through the dust to provide ionization - which is obviously the case. However, the IRAS spectra show that this wavelength is still inside a broad absorption feature (see below), thus the derived mass is more a lower limit than an actual determination. Secondly, the result strongly depends on the dust temperature, therefore even this lower limit is a very crude estimate. Using this estimate together with
where is again taken from Ossenkopf & Henning (1994), this time for 2.15 µm, is the total mass of dust divided by the volume of the assumed half-sphere, and L the length of the line of sight through the absorbing matter (i.e. in this case equal to R), we determine a K optical depth of or mag. This converts into a UV extinction of mag after Mathis (1990). This result is checked against Ryter's (1996)
where the last factor represents the conversion from to after Mathis (1990). Using the gas density from above as Hydrogen density and converting it to a column density by multiplying with the radius of our half-sphere, we also determine mag. We note, that the latter method assumes a gas-to-dust-ratio of 100, while our mass estimates from above using the dust temperature of 100 K delivers a gas mass of 2.3 (also assuming a pure Hydrogen gas) and thus a gas-to-dust-ratio 111. This good agreement may serve as further justification for the crude estimate we introduced in the guess of the temperature.
Of course a UV extinction of would mean that almost no UV photons are available for ionization, but as we have detected at least 15 stars inside G45 and may assume that they are all (not only the identified eight) able to deliver sufficient ionizing flux, each of them only has to supply ionization in of the volume. Assuming spherical geometries, this results in a reduction of the radius of the spheres ionized by individual components and thus the extinction suffered by photons while travelling through these spheres by to . After Aannestad (1989), this means that still more than 99% of all UV photons are absorbed before they can provide sufficient ionization. Thus, even our 15 stars (if they all are hot stars like the identified eight) cannot provide sufficient UV flux to ionize the region. We note, that 15 O5 stars would provide a total luminosity of 313 . This is much more than the 144 measured by WC89 from IRAS data which translate to 66 at our distance of 6.6 kpc. Thus, some of the stars have to be of later spectral types, diminishing the UV photon flux even further. To get out of this dilemma, which underlines the importance of dust inside this kind of regions, we have to speculate about the distribution of dust inside the H II region. Although our MIR images cannot resolve it, the clumpy appearance of the MIR emission is visible in Fig. 5. Varying optical depths inside G45 could thus provide enough free paths for the ionizing photons to escape and ionize the surrounding gas.
We note that the result derived above also means, that our measured Br extinction largely is due to foreground dust. The additional internal extinction of AK 0.4 mag towards the centre of the cluster shifts our stars in Fig. 8 even more towards the ZAMS. However, we do not know, which stars are embedded how deeply, therefore we do not correct for this additional extinction in Fig. 8. The interstellar extinction towards G45 should only amount to mag according to Neckel & Klare (1980) and Mathis (1990) while we have measured 2.8 mag towards the ionized region. from the comparison to AK 0.4 mag inside G45, we can derive that the major fraction of (cold, i.e. invisible in the MIR) dust is situated between us and the source. A similar scenario is assumed for the Orion Trapezium Cluster, where the O stars are thought to have have cleared a bubble in the surrounding dust (Felli et al. 1993).
Unfortunately, the three MIR images are of significantly different quality in terms of resolution and sensitivity. This renders quantitative statements on the dust temperature or optical depths of possible silicate features very difficult and thus we will not try to make one. The IRAS LRS spectrum of the source does show a pronounced absorption feature between 8 and 12.5 µm, but its centre wavelength is close to 11 µm, untypical for a (pure) silicate feature. As this feature comprises almost our entire N band filter, we decided it would be of little use to compare the 11.7 µm image and the N-band image to derive an optical depth. Also the wavelength base is very short between 10.5 µm and 11.7 µm, especially as 11.7 µm still is within the absorption feature. Additionally, it has to be marked, that the large IRAS beam probably measured the total of G45.45+0.06. Thus, without further information on the SED in this part of the spectrum, we will refrain from statements on temperature and optical depth.
One of the original goals of our investigations was to identify and classify the ionizing source of G45. However, during the previous sections it has become clear that this is a very difficult task. On the one hand, it is not immediately obvious if G45 is internally ionized at all, on the other hand we did not find the strong point source that coincides in position at all the observed wavelengths like e.g. Watson et al. (1998) claim to have identified in G29.96-0.02.
In the previous section we ruled out the possibilities of external ionization for reasons of the energy budget, which does not require any ionizing flux other than that from the stars inside G45 (except for the dust problem), and because no traces of possible ionizing sources like supernovae or clusters of OB stars can be found in reasonable vicinity to G45.
We will now try to draw a complete while still coarsely simplified picture of "G45 - The Orion nebula's younger brother". For an extensive review on the Orion regions see e.g. Genzel & Stutzki 1989 and Felli et al. 1993.
The stars labelled and n are situated close to a radio point source as well as to a clump visible in the mid-infrared. Although the flux from and n can not be measured due to their small separation in the un-deconvolved image, they belong to the brightest sources in the field. Therefore and for reasons explained below, we believe that star formation in G45 had its origin here. Massive stellar winds and the expanding ionization front swept through the ambient matter and formed the structure we see today.
The morphology of the radio shell can be explained by means of density gradients in the molecular material. The expansion towards the southeast happened faster and did not create visible star formation sites. In contrast, the ionization front seems to have hit dense clumps of material to the north, creating the cometary structure of G45 and triggering sequential star formation in dense and/or already collapsing molecular cloud cores. The result of this procedure is now visible as the chain of ZAMS-stars grouped along the ionization front. Unfortunately, there are no high-resolution maps at wavelengths that trace the distribution of cold dust to confirm our hypothesis. The 1.3 mm map of Mooney et al. (1995) shows an almost spherically symmetric distribution around the position of G45 at a resolution of . This resolution is of course too low do detect any details on scales that are visible in our NIR images and the VLA maps.
North of the ionization front as visible in Fig. 3, we detect several sources that show characteristics of stars in an evolutionary phase earlier than the ZAMS. The most obvious of these is source MIR1 which exhibits very strong MIR emission and provides approximately 10% of the infrared luminosity of G45. The spectral index of this object with between 2.2 and 12 µm would qualify this object for Lada class I following Wilking et al. 1989. At least two more objects with extended features are visible north of the front in Fig. 1. All of these objects need to undergo further examinations to determine their nature.
The apparent distances between the single point sources range from 0:004 to about . At the adopted distance of 6.6 kpc, this implies projected physical distances of around 2600 AU to 13000 AU. This distance range fits very well that of the central part of the Orion nebula around the Trapezium region, another hint that G45 may be a young twin of that region in the making.
The surroundings of G45 which include the UCH II s G45.48+0.13 and G45.47+0.05 appear clumpy and fragmented on a larger scale (W96). The latter authors also speculate on triggered formation of UCH II s which had its origin at G45, because it is the most evolved UCH II in the region.
Drawing the parallel to the Orion region, we would identify the stars o, l, m, and n as our Trapezium cluster where it all started. The necklace along the ionization front with its extended emission of NIR and MIR light would then be an analogue of the KL region, with several young massive stars embedded in the local concentrations of hot dust. While in Orion not all embedded point sources have been identified as self-luminous YSOs (Genzel & Stutzki 1989), all our sources that could be identified turned out to be young massive stars.
These most probably have been formed as a consequence of the ionization and shock front passing through these clumps. North of that front, source MIR1 might even be an analogue of the BN object, although its nature is much more uncertain up to now than that of BN. On the large scale, the younger UCH II s at distances around 57 pc might then be counterparts for the other OB associations in the Orion region which are located at comparable distances from the Trapezium and KL/BN. In Orion, the other OB associations are believed to be older than the Trapezium cluster while G45 itself is believed to be the oldest and most evolved UCH II in the Region (W96 on G45 and Genzel & Stutzki 1989 on Orion). W96 speculate on G45 having triggered star formation in the other UCH II s nearby - a scenario which has been introduced by Elmegreen & Lada (1977).
This parallel is another strong evidence that the formation of massive stars usually happens in clusters and leads to induced star formation in its vicinity. With the new techniques of high resolution imaging, it is now possible to enlarge the sample of these massive star formation sites towards more distant objects. Our example of G45 proved that this enlargement includes the detection of objects in an earlier phase of their evolution than the Orion region. More examples of the benefits of these high-resolution imaging techniques applied to massive star formation sites will be added in the near future.
© European Southern Observatory (ESO) 1998
Online publication: October 22, 1998