A&A 390, 945-960 (2002)
DOI: 10.1051/0004-6361:20020799
J. Knödlseder1 - M. Cerviño2 - J.-M. Le Duigou1 - G. Meynet3 - D. Schaerer4 - P. von Ballmoos1
1 - Centre d'Étude Spatiale des Rayonnements, CNRS/UPS, BP 4346,
31028 Toulouse Cedex 4, France
2 -
LAEFF (INTA) Apdo. 50727, Madrid 28080, Spain
3 -
Observatoire de Genève, 1290 Sauverny, Switzerland
4 -
Observatoire Midi-Pyrénées, 14, avenue Édouard Belin,
31400 Toulouse, France
Received 22 March 2002 / Accepted 22 May 2002
Abstract
Gamma-ray and microwave observations of the Cygnus region reveal an
intense signal of 1.809 Me V line emission, attributed to radioactive
decay of 26 Al, that is closely correlated with 53 GHz free-free emission,
originating from the ionised interstellar medium.
We modelled both emissions using a multi-wavelength evolutionary
synthesis code for massive star associations that we applied to the
known massive star populations in Cygnus.
For all OB associations and young open clusters in the field, we
determined the population age, distance, and richness as well
as the uncertainties in all these quantities from published photometric
and spectroscopic data.
We propagate the population uncertainties in model uncertainties by means
of a Bayesian method.
The young globular cluster Cyg OB2 turns out to be the dominant 26 Al
nucleosynthesis and ionisation source in Cygnus.
Our model reproduces the ionising luminosity of the Cygnus region very well,
yet it underestimates 26 Al production by about a factor of 2.
We attribute this underestimation to shortcomings of current
nucleosynthesis models, and suggest the inclusion of stellar rotation
as possible mechanism to enhance 26 Al production.
We also modelled 60Fe nucleosynthesis in the Cygnus region, yet the
small number of recent supernova events suggests only little 60Fe
production.
Consequently, a detection of the 1.137 Me V and 1.332 Me V decay lines
of 60Fe from Cygnus by the upcoming INTEGRAL observatory is not
expected.
Key words: nuclear reactions, nucleosynthesis, abundances - open clusters and associations: general
OB associations and young open clusters constitute the most prolific nucleosynthesis sites in our Galaxy. The combined activity of stellar winds and core-collapse supernovae ejects significant amounts of freshly synthesised nuclei into the interstellar medium. Radioactive isotopes, such as 26 Al or 60Fe, that have been co-produced in such events, may eventually be observed by gamma-ray instruments through their characteristic decay-line signatures. Indeed, galactic 1.809 Me V gamma-ray line emission attributed to the radioactive decay of 26 Al has been observed by numerous gamma-ray telescopes (see Prantzos & Diehl 1996 for a review). In particular, the COMPTEL telescope provided the first image of the Galaxy in the light of this isotope, showing an asymmetric ridge of diffuse emission along the galactic plane with a prominent localised emission enhancement in the Cygnus region (Diehl et al. 1995). The Cygnus emission has been interpreted as the result of 26 Al ejection in Wolf-Rayet winds and during core collapse supernova explosions from a nearby (1-2 kpc) massive star population which probably is part of the local spiral arm structure (del Rio et al. 1996).
Gamma-ray line emission is not the only tracer of this activity.
Knödlseder et al. (1999a) demonstrated that the galactic 1.809 Me V emission is
closely correlated to galactic free-free emission as observed in the
microwave domain.
The free-free emission mainly results from the ionisation of the
interstellar medium by the UV flux of O stars, hence it traces the
massive star population.
The observed correlation is indeed one of the strongest arguments in
favour of prolific 26 Al production by massive stars.
In general, a wealth of distinct massive star populations of different
ages, sizes, or metallicities contribute to the emission along a line
of sight through the Galaxy, and the observed correlation allows only
conclusions about the average properties of the contributing
populations.
In that way, the observations suggest that the equivalent O7V star 26 Al
yield, defined as the average amount of 26 Al ejected per number of
O7V star (measured by their equivalent ionising production), has a
galaxywide constant value of
(Knödlseder 1999).
It is surprising, however, that the correlation between 1.809 Me V
and microwave free-free emission also holds for the Cygnus region.
Both gamma-ray and microwave data show a localised emission enhancement
towards Cygnus, similar in size and relative intensity, resulting in
an equivalent O7V star 26 Al yield of
(see Sect. 4).
Within the uncertainties this yield is identical to the galactic value.
In contrast to the Galaxy, however, only few massive star associations
contribute to the observed emission in Cygnus, and it is not expected
that they have the same properties as the Galaxy as a whole.
In particular, the galactic metallicity gradient leads to an average
galactic abundance that is supersolar, and indeed only a supersolar abundance is
able to reconcile theoretical 26 Al yields with the observed galactic 26 Al
mass (Knödlseder 1999).
In contrast, massive star populations in Cygnus show slightly subsolar
abundances (e.g. Daflon et al. 2001) and since
26 Al yields are believed to depend on metallicity (e.g. Prantzos &
Diehl 1996) the nucleosynthetic properties of the
Cygnus region should deviate from those of the average Galaxy.
To understand the observations, we present in this paper a bottom-up model of the Cygnus region where we aim to explain the gamma-ray and microwave data from the underlying stellar populations. For this purpose we developed a multi-wavelength evolutionary synthesis model that we presented in Paper I of this series (Cerviño et al. 2000). We put considerable effort into the characterisation of the massive star populations in Cygnus with particular emphasis on the involved uncertainties (distance and age uncertainty; coeval or continuous star formation). We incorporate these uncertainties into our model by means of a Bayesian method and determine confidence intervals for all quantities to assess the predictive power of our approach. Despite the resulting uncertainties, we will demonstrate that the gamma-ray observations provide important clues on nucleosynthesis physics in massive stars. In particular we will demonstrate the shortcomings of current theoretical nucleosynthesis models in explaining 26 Al production and discuss possible modifications that may improve the models.
The evolutionary synthesis model we employed in this work is described in
detail in the first paper of this series by Cerviño et al. (2000).
In summary, the evolution of each individual star in a stellar population
is followed using Geneva evolutionary tracks with enhanced
mass-loss rates (Meynet et al. 1994).
Stellar Lyman continuum (Lyc) luminosities are predicted using the CoStar
atmosphere models of Schaerer & de Koter (1997), supplemented by the
Schmutz et al. (1992) atmospheres for the Wolf-Rayet phase.
At the end of stellar evolution, stars initially more massive than
are exploded as Type Ib supernovae, while stars
of initial mass within
and
are assumed to
explode as Type II SNe.
Nucleosynthesis yields have been taken from Meynet et al. (1997) for the
pre-supernova evolution, from Woosley & Weaver (1995) for Type II
and from Woosley et al. (1995) for Type Ib supernova explosions.
Note that Type II SN yields have only been published for stars without
mass loss and Type Ib yields have only been calculated for pure Helium
stars.
In order to obtain consistent nucleosynthesis yields for Type II
supernovae we followed the suggestion of Maeder (1992) and linked
the explosive nucleosynthesis models of Woosley & Weaver (1995) to the Geneva
tracks via the carbon-oxygen core mass at the beginning of carbon burning.
For Type Ib SN we used the helium-core mass at the beginning of helium core
burning for the link.
In order to predict the present gamma-ray line emission from a massive
star association we have to estimate the production of radioactive
isotopes in the past.
We do this by determining the actual number of stars
within a given initial mass interval
that
is not affected by incompleteness at the lower end and evolutionary
effects at the upper end.
We then extrapolate this population back to the past using a Salpeter
initial mass function (IMF) of slope
that is normalised
to the number of stars we observe today.
In practice, we generate an initial stellar population by randomly
sampling the IMF until the number of stars falling in the mass interval
amounts to
.
We then use the evolutionary synthesis code to follow the time evolution
of the 26 Al and 60Fe yields, the Lyc luminosity, and the distribution
of spectral types within the population.
The stellar systems we study in this work only comprise a few hundred to a few thousand stars, making the high-mass end of the population generally sparsely sampled. However, massive stars provide a considerable fraction of the association nucleosynthesis and ionising power, and the early evolution of the system will depend rather sensitively on the actual choice of stellar masses. In particular, different Monte Carlo samples of the same initial stellar population may lead to quite different evolutions of the association observables, leading to considerable uncertainties in our model predictions. We decided to include these uncertainties in our analysis by means of a Bayesian method. Instead of investigating the time evolution of a quantity X for a specific population, we base our analysis on the probability density function (PDF) p(X|t) which quantifies the probability for X to be equal to a value x at the time t. We approximate this function for a given association by repeating the evolutionary synthesis calculations for independently sampled stellar populations, from which we obtain the frequency of observing the value x at time t. In this way, the uncertainty arising from the extrapolation to the past, which manifests in an uncertainty about the initial massive star population, is reflected in the width of the PDF. Examples of PDFs obtained for massive star associations in the Cygnus regions are given in Cerviño et al. (2000).
To predict the today value of the quantity X, we marginalise over the
(uncertain) age t of the association using
The starting point of our study of massive star populations in Cygnus
is the WEBDA database of open clusters of Mermilliod (1998).
From this database we established a list of open clusters within the
limits
and
,
a region which
largely encloses the 1.809 Me V and microwave free-free emission
features.
We preselected young clusters from this list by requiring either an
earliest spectral type of B3V or earlier (corresponding to a star of
8
initial mass), or a cluster age of less than
yr.
Older clusters do not house any potential core collapse progenitor
anymore, and their nucleosynthesis activity is therefore negligible.
The lack of stars more massive than
10
also reduces their
ionising flux to an imperceptible level.
Of course, not for all known open clusters an age estimate is
available, but those clusters are then often rather distant and/or
do not contain luminous members.
Consequently, they do not contribute significantly to the emissions
that we aim to study in this work.
For the OB associations in the Cygnus region we gathered the relevant information from the literature. Our main resources were the compilations of Humphreys (1978) and Garmany & Stencel (1992), complemented with data from the SIMBAD database and some recent investigations (see the appendix for a detailed description of our database).
The association parameters relevant to our study are the age ,
the distance d, the richness, expressed as the number of
stars
in a specific initial mass interval
,
and the slope
of the IMF
(in the following we deliberately choose the term association when we
talk both about OB associations and young open clusters).
Literature values for these parameters are only available for some of
the associations, and have generally been derived by different methods.
Also the choice of the stellar evolutionary tracks and luminosity
calibrations impacts the results, and may lead to differences among
the derived parameters.
Additionally, the uncertainty of the association parameters has rarely
been determined, and if so, also not in a consistent manner.
For these reasons we decided to re-determine the relevant association
parameters and their uncertainties from the basic stellar data in our
database.
In the following we describe our method and present the resulting
parameters for all young stellar associations in Cygnus.
The results of the analysis are summarised in Tables 1
and 2.
Details for each association are given in the appendix.
As a first step we determine the reddening for each association from
spectroscopic and photometric information.
For each star with spectral classification we used the spectral type to
intrinsic colour calibration of FitzGerald (1970) complemented by information
from Schmidt-Kaler (1982) for the earliest spectral types to determine
their colour excess
E(B-V)=(B-V)-(B-V)0 and
E(U-B)=(U-B)-(U-B)0.
From these we calculate the slope
of the reddening curve
for each association by averaging the slopes of all member stars with
known spectral type.
Outliers were iteratively removed from the average until all
remaining stars lie within 2 rms of the average.
The slopes typically vary from 0.67 to 0.84 with a few exceptions
with poorly defined values due to the proximity of the association
(Cyg OB7, Lac OB1, Roslund 5).
We find a trend towards steeper slopes (typically around
)
for the more reddened associations, in particular in the area
of Cyg OB1, OB2 and OB9, i.e. the central part of the Cygnus X region.
This confirms findings by other authors
(e.g. Massey & Thompson 1991) of an anomalous reddening
in this area.
For the couple of open clusters for which no spectral data are
available, the canonical value of
has been adopted.
In view of the dispersion in the value of
in the Cygnus area this
choice is certainly somewhat arbitrary, but the impact of the exact
value on the resulting cluster parameters is rather small, leading to
a negligible additional uncertainty in our study.
As a second step the distance of each association is estimated.
For associations with spectroscopic information we employed the
method of spectroscopic parallax.
For each MK classified star we compute
,
where
is the visible extinction.
The association distance is then calculated by averaging DM
where again 2 rms outliers were removed iteratively.
The absolute visual magnitudes
have been extracted from
calibrations of Vacca et al. (1996) for stars earlier than B1,
complemented by data from Humphreys & McElroy (1984) and Schmidt-Kaler (1982)
for later spectral types.
For the ratio
of absolute-to-selective absorption we apply the
canonical value of 3.1, although we recognise that deviations from
this value are eventually observed in star forming regions
(Mathis 1990).
For Cyg OB2, for example, Massey & Thompson (1991) found
,
and
indeed a variable extinction analysis of our data suggests a similar
value.
This deviation results in a distance modulus error of 0.2 mag for
this association, inferior to the statistical and systematical uncertainty
of 0.5 mag.
Cyg OB2 is certainly an extreme case due to the large reddening of
the association, and we believe that variations in
should not
dramatically alter our results.
Anyways, we have little choice since generally the available data do
not allow to perform a reliable variable extinction analysis for the
associations, and we fear that a poorly determined value of
introduces a larger error than a rather solid mean value of 3.1
that may not precisely apply to all cases.
For some of the open clusters for which not enough spectral type
information was available, we estimated the distance from the hot
cluster stars using the reddening-free parameter
,
where
is the mean cluster reddening
slope as given in Table 1.
For Q<-0.4, corresponding to spectral types earlier than B5 or so,
there exist unique relations between intrinsic colour (B-V)0, absolute
visual magnitude
,
and Q for each luminosity class.
We determined these relations by fitting polynomial functions to
calibration tables that we compiled from FitzGerald (1970),
Schmidt-Kaler (1982), Humphreys & McElroy (1984), and Vacca et al. (1996).
To estimate the luminosity class we apply an iterative approach.
For each star with Q<-0.4 we assume as initial estimate a luminosity
class of V (this is certainly a reasonable assumption for young unevolved
clusters).
Based on this estimate we use the calibration relations to determine
(B-V)0 and
from Q.
From these quantities we derive E(B-V),
,
and DM,
leading to a first cluster distance estimate by averaging over DM.
Cluster stars which we misclassified in luminosity class will clearly
show up as outliers in DM with respect to the average.
Hence, for stars with DM more than 1 rms above the average we reduce
the luminosity class (in the sense V
I) while for stars with
DM more than 1 rms below the average we increase the luminosity class
(in the sense I
V) for the next iteration.
We then repeat the entire procedure for the updated luminosity classes.
For all clusters in our sample, this scheme led to convergence in
DM after 2-3 iterations.
We tested the validity of our procedure by applying the method to clusters for
which DM is known from spectroscopic parallaxes and found satisfactory
agreement with the photometric results.
For Cyg OB2 and many of the clusters, field star contamination is a serious problem in our database. Field stars may be recognised by particularly low or high E(B-V) values with respect to bona fide association members (e.g. Massey et al. 1995). For contaminated associations, we therefore excluded those stars whose inferred colour excess is outside the range spanned by member stars with MK classification. Most clusters show a clear upper main sequence which allows for an unambiguous definition of stellar reddenings of cluster members. For the cases where we applied such a selection, the E(B-V) range of cluster members is specified in Col. 5 of Table 1.
To perform such a selection, E(B-V) has been determined for each star
from either spectroscopic or photometric data.
In the latter case, the iterative procedure described above has been
used for hot stars with Q<-0.4.
For ,
the relation between (B-V)0 and Q becomes
ambiguous, corresponding to the well known "knee'' in the colour-colour
diagram.
Nevertheless, from the intersection of the reddening lines with the
intrinsic colour-colour tracks we can determine all possible E(B-V)
values (3 at maximum) that are consistent with the observed colours, and
we exclude all stars that have no solution that is compatible with the
specified colour excess interval.
To remove remaining field or background stars we employed for most
clusters another selection criterion based on the distance
moduli of the stars.
From bona fide members with MK classification we define a distance moduli
interval
and exclude all stars for which
,
where
is
the distance modulus of the association.
Distance moduli for stars with MK classification or hot (Q<-0.4) stars
were determined from spectroscopic and photometric parallaxes,
respectively.
For the remaining stars, distance moduli were again estimated from
the intersection of the reddening lines with the intrinsic colour-colour
tracks.
Our next step consists of constructing H-R diagrams (HRDs) for all
associations.
For this purpose we transform the spectroscopic and photometric
information for all member stars to
and
.
If a star has a MK classification, we used the spectral type to determine
both the effective temperature and bolometric correction B.C. using the
calibration of Vacca et al. (1996), complemented by data from Humphreys & McElroy (1984) and
Schmidt-Kaler (1982).
From the absolute visual magnitude and the bolometric correction, the bolometric
luminosity is then derived using
For stars without MK classification, we again distinguish between hot (Q<-0.4)
and cold ()
stars.
For hot stars we determined
and B.C. from Q, while for
cold stars we derived both parameters from the intrinsic colour (B-V)0.
Using the mean association distance DM and extinction
we
assign a bolometric luminosity to each star using
We estimate the intrinsic colour
(B-V)0 = (B-V) - E(B-V) for each star by
assuming a reddening identical to the mean reddening of the association.
Our analysis is based on polynomial functions that we fitted to the calibration
tables of intrinsic stellar parameters.
In general, the relations depend of the luminosity class which we
estimate for each star from the absolute visual magnitude
(unless it has already been estimated by the iterative
procedure described in Sect. 3.2).
Here we make use of the rapid variation of
with luminosity
class, and we search the calibration tables for the luminosity class
that comes closets to the pair
and (B-V)0 of stellar
parameters.
Although this method is certainly not very accurate, the impact of a
misclassification in luminosity class is only moderate.
Typically,
varies about
0.02-0.1 dex and B.C. about
0.3-0.7 mag between subsequent luminosity classes.
![]() |
Figure 1:
H-R diagram of Cyg OB2. The filled circles are stars for which
spectral classification is available, the open circles are stars
for which we have only photometry. Solid lines are the stellar
tracks of Meynet et al. (1994), dotted lines are the corresponding
isochrones for intervals of 2 Myr, starting from ![]() |
Open with DEXTER |
From the H-R diagrams we estimated the age of the associations.
For this purpose we compared the location of the association stars in
the HRDs to the theoretical evolutionary tracks of Meynet et al. (1994) for
solar metallicity (Z=0.02).
The same tracks are used for the evolutionary synthesis calculations,
hence in this respect our models are consistent with the population
input data.
As an example we show the HRD for Cyg OB2 in Fig. 1 which
may be compared to Fig. 16 of Massey & Thompson (1991) who made a similar analysis.
From the upper part of the HRD (masses
)
we infer the ages
of the massive stars from the Meynet et al. (1994) isochrones that we
superimposed on the diagrams.
For Cyg OB2, for example, this method leads us to an age estimate of
1-4 Myr, comparable to the finding of Herrero et al. (1999).
The results for all associations are summarised in Table 2. Instead of determining a best fitting age, we always tried to specify an age range that reflects the dispersion of the upper main sequence. The dispersion comes partly from the systematic limitation of our method (knowledge of stellar calibrations, neglection of stellar rotation and binarity, measurement or classification errors), but could also reflect a period of continuous star formation. In particular, some associations like Cep OB1, Cyg OB8 or Roslund 4 show a considerable scatter in the upper HRD which could even reflect several epochs of massive star formation. For Cep OB1 we decided to split the population into two subgroups of different age since the data are difficult to reconcile with a single star formation event. Yet, a large apparent age spread could also signal that the association presents not a physically associated group but rather a chance projection of massive stars in the sky. For this reason we excluded Cyg OB4 from our study since Hipparcos data did not confirm the reality of this stellar group (de Zeeuw et al. 1999). Other associations such as Cyg OB8 or Vul OB1 are also highly doubtful, yet it will turn out that they provide only a negligible contribution to the overall luminosities in the Cygnus region, hence we do not introduce an important uncertainty by including them into our study.
Name | ![]() |
mass range (![]() |
Cyg OB1 |
![]() |
15- 85 |
Cyg OB2 |
![]() |
15-120 |
Berkeley 86 |
![]() |
7-40 |
Berkeley 87 |
![]() |
7-60 |
IC 4996 |
![]() |
5-40 |
NGC 6823 |
![]() |
7-60 |
NGC 6871 |
![]() |
7-40 |
NGC 6910 |
![]() |
5-40 |
NGC 6913 |
![]() |
7-60 |
NGC 7235 |
![]() |
7-40 |
NGC 7380 |
![]() |
7-40 |
Trumpler 37 |
![]() |
7-60 |
The comparison of the H-R diagram with the evolutionary tracks allows
an estimation of the initial richness of the association.
We determined the richness by counting the number of stars
over
a sufficiently large initial mass interval
.
We tried to choose the lower mass limit
sufficiently high to
avoid any bias from population incompleteness due to the limiting
magnitudes of the association surveys.
Conversely, the upper mass limits
were chosen sufficiently
low so that the evolutionary turn-off has no impact on the richness
determination.
The results are summarised in Cols. 8 and 9 of Table 2.
The only exception to this procedure is Cyg OB2 for which we rely on
the recent mass determination of Knödlseder (2000), based on
an analysis of 2MASS near infrared data.
In fact, the data that are available in the visible waveband for Cyg
OB2 are heavily affected by interstellar absorption, and their use would
considerably underestimate the total mass of this association.
In contrast, the 2MASS data do not allow a precise spectral
type determination and the quoted number of 120 O stars rather
reflects the fact that there were initially 120 stars more massive than
in Cyg OB2, although some may already have
evolved, having either turned in early B-type or even Wolf-Rayet stars.
The quoted number of 120 O stars should therefore present an
upper limit.
For some associations our stellar database contains enough objects
to allow an estimation of the initial mass spectrum.
We fitted these spectra by power-law initial mass functions and
summarise their slopes
in Table 3.
We also quote the mass intervals that have been used for the fitting
which again were chosen to minimise any bias due to evolved stars,
population incompleteness, and field star contamination.
Yet we recognise that the results are still affected by systematic
uncertainties, as demonstrated by the steep slope of
for NGC 7235 which shows a heavy field star
contamination (Massey et al. 1995).
Formally, we derive a weighted mean IMF slope of
which is close to the canonical Salpeter value of -1.35 and
compatible with the value of
that has been
determined by Massey et al. (1995) for 12 galactic OB associations and open
clusters.
To illustrate the systematic uncertainties in IMF slope
determinations, we compare our result for Cyg OB2 (
)
to
the value of
determined by Massey & Thompson (1991) from the same
stellar data, and the recent analysis of Knödlseder (2000) of
2MASS near infrared data that suggests
.
In view of these uncertainties, we believe that the assumption of a
uniform canonical Salpeter law for our evolutionary synthesis modelling
of all associations is a reasonable approximation that is compatible
with the observational data.
![]() |
Figure 2:
Left: contour map of 1.809 Me V gamma-ray line emission in
the Cygnus region (Plüschke 2001) superimposed
onto a greyscale image of microwave 53 GHz free-free emission
(Bennett et al. 1992).
The prominent emission feature extending from
![]() ![]() ![]() |
Open with DEXTER |
The discovery of an extended X-ray ring-like source surrounding the Cygnus X region by Cash et al. (1980) has been interpreted as a superbubble blown by the massive star winds and supernova explosions in the Cyg OB2 association. Anomalous stellar proper motions have been reported for stars in associations located near the edge of the bubble (Comerón et al. 1993), and have been interpreted as the sign for triggered star formation following gravitational instability of the expanding shell (Comerón & Torra 1994). However, the energetic requirements needed to explain the expansion motion are difficult to reconcile with the mechanical power provided by Cyg OB2 (Comerón et al. 1998). In addition, the superbubble is equally well, if not even better, explained as a superposition of multiple objects at different distances that aline along the local spiral arm that is seen tangentially in this direction (Bochkarev & Sitnik 1985; Uyaniker et al. 2001).
Also our cluster analysis seems difficult to reconcile with the scenario of an
expanding bubble that triggered formation of the associations
surrounding Cyg OB2.
Firstly, the central association Cyg OB2 is one of the youngest in the
area (1-4 Myr) while the surrounding associations Cyg OB1, OB3,
OB7, and OB9 are slightly older (2-6 Myr).
In the scenario where Cyg OB2 triggered the formation of the other
associations, however, the situation should be vice versa.
Secondly, only Cyg OB1 and OB3 are situated at distances that are
compatible with that of Cyg OB2 while Cyg OB7 and OB9 are likely
foreground objects.
In addition, Cyg OB2 is not situated at the centre of the X-ray superbubble
and the other OB associations correlate only partly with the structure
(Uyaniker et al. 2001), making a physical relation
of these objects highly questionable.
Thirdly, the expansion age determined by Comerón et al. (1998) from
stellar proper motions fits surprisingly well our age estimate for Cyg
OB2, supporting the idea that the expanding stars are dynamically
ejected runaway stars from Cyg OB2 (Comerón et al. 1998).
Comerón et al.'s list of candidate members of the Cygnus expanding structure
contains four O-type stars, which compared to the total of 120 O-type
stars in Cyg OB2 (Knödlseder 2000)
would result in a runaway star fraction of
for this association.
Compared to other young massive star associations, such as the
Trapezium cluster in Orion or the
Orionis star-forming region
(Hoogerwerf et al. 2001), such a proportion seems
typical, and we believe that at least part, if not all, of the anomalous
stellar proper motions in Cygnus are indeed explained by runaway stars.
Using simple approximations we can already make a first order estimate
of the ionising flux that we expect from the stellar populations in
the Cygnus region.
The CoStar models of Schaerer & de Koter (1997) predict a Lyman continuum luminosity of
ph s-1 for a star of spectral type O7V, and assuming
that all 204 O stars that we find in the associations of the Cygnus region are
indeed O7V stars, we derive a total luminosity of
ph s-1.
Assuming further that all stars lie at a typical distance of 1.8 kpc, and
using the relation
Before we present the results of our evolutionary synthesis modelling
we want to discuss the gamma-ray and microwave observations that
triggered our investigation.
Maps of 1.809 Me V gamma-ray line emission and 53 GHz free-free emission
of the Cygnus regions are shown in superposition in Fig. 2.
The spatial correlation between both emission features is striking.
However, we want to caution the reader of overinterpretations of the
apparent structures in the map.
The gamma-ray feature has a total significance of only ,
and the detailed morphology of the image is heavily biased by instrumental
noise (Knödlseder et al. 1999b).
Formally, the COMPTEL instrument had an angular resolution of
(FWHM), yet the small amplitude of the signal probably
does not allow inferences at angular scales below
.
This is comparable to the angular resolution of
(FWHM) for
the COBE DMR experiment which was used to derive the free-free
emission map.
Observations of galactic radio-continuum emission from Cygnus at smaller
angular scales indicate that most of the emission originates from a region
between
and
with emission peaks around
(e.g. Wendker 1970).
This region coincides with the OB associations Cyg OB1, Cyg OB2, Cyg OB8,
and Cyg OB9 and a large number of the youngest open clusters in our
database.
However, the small-scale radio maps that are available were generally
obtained at lower frequencies where synchrotron emission contributes
significantly to the signal.
Although the Cygnus emission is primarily optically thin thermal
radiation, a superimposed smooth non-thermal component makes the
determination of absolute free-free emission levels difficult
(Wendker et al. 1991).
Despite the poor angular resolution, we therefore prefer using the 53 GHz
free-free emission map of Bennett et al. (1992) for the estimation of the ionising
flux in Cygnus.
At this frequency the contribution from synchrotron and thermal dust emission
is small within the galactic plane, and can easily be removed by
modelling their spatial distribution at adjacent frequencies.
The image shown in Fig. 2 results from such a separation
performed on DMR data (Bennett et al. 1992).
Smaller angular scales are probably even not desirable for a global
flux comparison, although they bear valuable information about the
interplay between the massive stars with the interstellar medium.
Firstly, at angular scales that are smaller than the associations,
the flux distribution may be influenced by individual massive stars,
which we do not attempt to model in our study.
Secondly, the correlation between free-free emission and the massive
star distribution may anyway break-down at scales comparable to the
association dimensions due to the lack of gas in their ploughed interiors.
For example, Huchtmeier & Wendker (1977) did not find any H II region inside
Cyg OB2 despite the large number of ionising stars that is present.
They argue that the association is devoid of hydrogen and the ionising
flux is converted into free-free emission by the surrounding gas,
leading to a diffuse extended emission component.
Thirdly, the long lifetime of 26 Al (
Myr) allows the
isotope to travel substantial distances before the emission of the
gamma-ray decay photons, and the detailed 1.809 Me V emission
structure should depend considerably on the interaction and
deceleration of the ejecta in the surrounding interstellar medium.
Morphology studies that address the question of ejecta propagation in
the Cygnus region will become possible with the upcoming
INTEGRAL observatory
(e.g. Knödlseder & Vedrenne 2001), but the
existing COMPTEL data limits our analysis to only a global investigation.
We determined the 1.809 Me V and 53 GHz free-free fluxes from the
Cygnus region by integrating the respective skymaps over the area of
interest.
A diffuse galactic ridge emission that underlies the features in the
Cygnus region has been subtracted from the flux estimates in order to
extract the emission that is correlated to the Cygnus associations.
The resulting flux estimates are only weakly sensitive on the precise
location of the integration boundaries.
To safely cover the investigated region we selected a longitude interval
from
to
for the flux determination and
determined the galactic ridge background from two adjacent intervals
at
and
.
We selected a rather large latitude window of
to assure
that we recover also the tails of the emission (in particular the
COMPTEL skymap shows extended high latitude wings due to the low
significance of the reconstructed emission).
The resulting 1.809 Me V gamma-ray line flux amounts to
ph cm-2 s-1 while the 53 GHz
free-free emission flux equals
Jy.
The quoted errors reflect both the statistical and the systematical
uncertainties in the flux determination, in particular those
introduced by the selection of the integration boundaries.
The flux measurements can be converted into the equivalent O7V star
26 Al yield using
Quantity | Measured value |
S1809 |
![]() |
S1137 | - |
S53 |
![]() |
![]() |
![]() ![]() |
![]() |
![]() ![]() |
We used the method described in Cerviño et al. (2000) and Sect. 2 to estimate the nucleosynthesis yields and Lyc luminosity for each association. The results are compiled in Table 5. For each association 100 statistically independent evolutionary synthesis models have been calculated to approximate the PDFs for all quantities of interest. Based on our age determinations, we marginalised the age uncertainty using Eq. (1) with bounded Gaussian-shaped prior probability distributions. The mean and standard deviation of the Gaussians were defined by the mean and the half-width of the age boundaries quoted in Col. 2 of Table 2; the prior density was set to zero for ages outside the age boundaries. The 26 Al and 60Fe yields as well as the Lyc luminosities that result from this procedure are given in Cols. 2-4 of Table 5.
Alternatively, we also tried box-shaped prior distributions that were bounded by our age estimates, but the results were only slightly different. In fact, Eq. (1) performs a weighted average of the PDFs where the age prior p(t) defines the weights that are attributed to each age. From our association analysis we derive a possible range of ages, yet the data are not sufficiently accurate to derive details about the age distribution. Therefore, any prior distribution that reflects reasonably well the fact that the association age is comprised between two limits should provide a valid prior for our analysis. The only difference between the Gaussian and the box-shaped prior is that the first slightly favours models with an age close to the mean age between the boundaries, while the second gives equal weight to all ages that are compatible with the assigned limits. Since the observed age spread is at least partly due to measurement uncertainties, and since we defined our age boundaries as two extreme limits that enclose the age distribution of the most luminous stars, we feel that a bounded Gaussian-shaped prior reflects most properly the information that we have at hand about the association ages.
We also want to emphasise that our age marginalisation is equivalent to the assumption of a continuous star formation over an interval defined by the age boundaries. In this interpretation, the form of the age prior specifies the time evolution of the star formation rate. Since this rate is generally unknown, a rather flat prior distribution, like those we discussed above, is certainly appropriate. The interpretation of the apparent age distribution in the H-R diagram is usually a difficult enterprise due to the blending between measurement uncertainties, spectral misclassification, neglection of binarity and rotation, non-member pollution, and a possible non-coeval star formation. We cannot distinguish between all these uncertainties either, yet our formalism includes all of them into the model results.
The nucleosynthesis yields and ionising luminosity were converted into
gamma-ray line and microwave free-free fluxes by marginalising the
distance uncertainty for each association (Eq. (2)).
For the 1.809 Me V gamma-ray line flux due to the radioactive decay
of 26 Al we used the relation
![]() |
(6) |
![]() |
(7) |
Our evolutionary synthesis model also predicts the spectral type distribution, and we quote in Cols. 8-9 of Table 5 the predicted number of O-type and Wolf-Rayet stars. These predictions can be directly compared to the respective columns of Table 2, and we globally find a good agreement although the median values often fall below the observed values. The exception to the rule is Cyg OB2 for which the models predict between 51 and 116 O stars while the observations suggest 120. However, as discussed in Sect. 3.6, the 120 O stars quoted in Table 2 are rather an upper limit since the 2MASS analysis did not account for possible evolutionary effects. Yet, Cyg OB2 houses evolved stars and even shows indications for non-coeval and ongoing star formation (e.g. Torres-Dodgen et al. 1991; Massey & Thompson 1991; Parthasarathy et al. 1992; Pigulski & Kolaczkowski 1998; Herrero et al. 1999).
![]() |
Figure 3: Evolution of the number of O stars, Wolf-Rayet stars, supernova remnants, and 1.809 Me V gamma-ray line flux as function of age for Cyg OB2. We assumed a lifetime of 105 yr for each SNR. |
Open with DEXTER |
The effect of the age uncertainty on the stellar population in Cyg
OB2 is illustrated in Fig. 3.
For a coeval star formation event, the number of O stars drops rapidly
from the initial value of 120 to zero during the first 5 Myr of the
association evolution.
Clearly, there is a considerable number of O-type stars in Cyg OB2
(e.g. Massey et al. 1991), hence the associations
should be younger than 5 Myr.
Wolf-Rayet stars appear around 2 Myr after the starburst while the first
supernovae explode about 4 Myr after the birth of the association.
The fact that there are Wolf-Rayet stars associated with Cyg OB2
suggests an age Myr, the presence of 2 WC-type stars even
indicates an age above 3 Myr.
On the other hand no supernova remnant has been detected within the
boundaries of Cyg OB2 (Wendker et al. 1991), hence
the association should be younger than
4 Myr.
The 1.809 Me V gamma-ray line flux raises from zero to a maximum level
of
ph cm-2 s-1 within
3 Myr (for an assumed distance
modulus of DM=11.0 mag), and Cyg OB2 could be situated just at the maximum
of the 1.809 Me V gamma-ray light curve, which is about a factor of two
above the median value quoted in Table 5.
Nevertheless, the present reasoning only holds for a strictly coeval
star formation event, and the evidence for ongoing star formation in
Cyg OB2 holds against narrowing down the age boundaries for the
association.
Clearly, more observational data is required to improve our knowledge
about the age distribution in Cyg OB2 before we safely can improve
upon the nucleosynthesis yield estimation for this association.
![]() |
Figure 4:
Predicted 1.809 Me V (filled dots and solid error bars) and
1.137 Me V gamma-ray line fluxes (open dots and dashed error
bars) as function of predicted 53 GHz free-free flux.
The arrow indicates the displacement due to a decrease of 2 mag
in the distance modulus or an increase of the richness by a
factor of 6.3. The two solid lines indicate equivalent 26 Al
O7V star yields of
![]() |
Open with DEXTER |
This example illustrates that the apparent age spread in the HRDs presents an important source of uncertainty for the final flux estimates, at least for the youngest associations. In general, however, the dominating source of uncertainty depends on the association richness and its age, but also on the nature of the quantity of interest. For Cyg OB2, for example, the 53 GHz flux prediction is much less affected by the age uncertainty than the 1.809 Me V flux estimate. While the 1.809 Me V flux varies by more than two magnitudes between 1-4 Myr, the number of O stars changes only by a factor of two, implying a 53 GHz flux variation by a factor of less than six. For slightly older associations, such as Trumpler 37, the situation is reversed: the 1.809 Me V flux varies only slowly while the ionising flux now drops quickly due to the disappearing of the most massive stars in supernovae events.
If the association is sufficiently rich so that sampling effects are
negligible, the distance uncertainty turns out to be the dominating
factor of the flux dispersion.
Typically, the relative distance uncertainty for an association
amounts to
which directly translates into a flux uncertainty
of
.
Sampling effects are important when only few stars contribute to the
association luminosities, i.e. either for very young or for very poor
populations.
Cyg OB7 fulfils both conditions.
For this association, a considerable fraction of the evolutionary synthesis
samples did not lead to any 26 Al production at all.
In those cases, the initial mass function has not been populated by
sufficiently massive stars to lead to a noticeable 26 Al production
during the conceivable age range of Cyg OB7 (2-5 Myr).
We indicate such a possibility by quoting infinite negative errors in
Table 5.
To investigate the correlation between gamma-ray line flux and
free-free emission in the Cygnus region we plotted the predicted
1.809 and 1.137 Me V fluxes versus the 53 GHz flux
(Fig. 4).
All three quantities should scale with distance and association richness,
and indeed we find a general trend of increasing gamma-ray intensity
with increasing free-free flux.
A few associations with low gamma-ray fluxes
(
)
deviate
from this tendency, yet they represent those poorly populated clusters
that show a tremendous flux dispersion due to poor sampling of the
high-mass IMF (see above).
To illustrate the distance and richness impact, we superimposed an
arrow in Fig. 4 that indicates the displacement
due to a decrease of 2 mag in the distance modulus or an equivalent
increase of the association richness by a factor of 6.3.
For the 1.809 Me V flux, the associations are situated along this line
of displacement, yet there occurs a clear discontinuity around
.
All associations with
are older than 10 Myr, an
evolutionary phase during which 26 Al is exclusively ejected by supernovae
events (the only exception to this rule are Berkeley 94 and 96, yet they are
so sparsely populated that no 26 Al production is expected).
In contrast, all associations with
are younger
than 10 Myr, hence they are dominated by 26 Al ejection through stellar winds
(Cerviño et al. 2000).
The discontinuity in Fig. 4 represents a change in
the equivalent O7V star 26 Al yield, and we find typical values of
and -4 for
and >0,
respectively.
The discontinuity in the correlation coefficient may eventually lead to a break-down in the tight correlation between 1.809 Me V and 53 GHz emission, which should occur however at rather low 1.809 Me V flux levels of about 10-6 ph cm-2 s-1. At this level, it is expected to find 1.809 Me V emission features that are uncorrelated to microwave free-free emission (or for which the free-free emission should not exceed an intensity of 1 Jy). Given the sensitivity limit of the spectrometer SPI on INTEGRAL, it is unlikely that this break-down will be observable in the near future.
For the 1.137 Me V flux we find only a weak correlation with free-free
emission, which is mainly limited to
,
i.e. older
associations that are dominated by supernova explosions.
For those, a typical equivalent O7V star 60Fe yield of
is found, yet the low level of the associated
free-free flux (
Jy) will inhibit the observation of this
correlation.
For
,
1.137 Me V fluxes are highly uncertain due to
the small number of supernova events that contribute to 60Fe production
(recall that no large amounts of wind-ejected 60Fe is expected from
massive stars; see Cerviño et al. 2000).
The youngest associations, such as Cyg OB2 or Cyg OB3, show no 60Fe
production at all, despite the huge amounts of ionising photons that is
produced by their large number of O-type stars.
![]() |
Figure 5: Predicted median 1.809 Me V gamma-ray line intensity distribution (left) and 53 GHz microwave free-free intensity (right). |
Open with DEXTER |
By far the most luminous association in both 1.809 Me V and 53 GHz
free-free emission is Cyg OB2, which corresponds to the isolated dot
in the upper-right corner of Fig. 4.
The median equivalent yield of Cyg OB2 amounts to
,
somewhat lower than the observed value
of
in the Cygnus region.
In other words, Cyg OB2 seems to underproduce 26 Al with respect to
ionising photons, which is mainly related to its extremely young age.
Most of the massive stars are still on the main sequence where they
produce large amounts of ionising photons, but do not yet eject
copious amounts of 26 Al in the interstellar medium.
For the entire Cygnus region, we find an equivalent O7V star
26 Al yield of
,
about a factor of two
below the observed value.
Apparently, our model underpredicts 26 Al production with respect to
the ionising luminosity by about a factor of 2, and we will suggest later
that this underprediction is most likely related to shortcomings of
current nucleosynthesis models for massive stars.
It is interesting to note that our predicted
value is very
close to the steady-state value of
,
which
results for a stellar population with a constant star formation rate
(Cerviño et al. 2000).
Since the Cygnus region is largely dominated by a single association,
Cyg OB2, we believe that this finding is merely a coincidence.
Indeed, our predicted equivalent O7V star 60Fe yield of
for the entire Cygnus region falls
more than a factor of two below the steady-state value of
(Cerviño et al. 2000),
emphasising the relative rareness of supernova events in the Cygnus region
with respect to a steady-state population.
To illustrate the spatial distribution of the 1.809 Me V and 53 GHz
emission predicted by our model, we show in Fig. 5 intensity
maps for both observables that are based on the median flux values quoted in
Table 5.
Although we have no information about the spatial distribution of the
respective emission within and around the associations, we
distributed the flux within the association boundaries using
Gaussian-shaped density profiles to account for the spatial extent of
the stellar populations.
We recall that the true distribution could be more widespread due to
the possible lack of ionised gas in the ploughed cluster interiors and
the propagation of the radioactive ejecta before decay
(cf. Sect. 4.1).
In addition, when comparing these distributions to the observed maps
in Fig. 2, the angular resolution of the telescopes of
(COMPTEL) and
(DMR) has to be considered.
Taking these factors into account, the morphological resemblance of
our model maps to the observations is quite reasonable.
We predict free-free and 1.809 Me V emission maxima at
which spatially coincides with the observed
maximum of the 53 GHz map, and is also close the maxima in the 1.809 Me V COMPTEL map.
The observations show copious emission in the longitude range
,
and also our model predicts the bulk emission
in this region.
Obviously, the most prolific sources of 1.809 Me V and 53 GHz
emission are (in the quoted order) the OB associations Cyg OB2, Cyg OB3,
Cyg OB9, and Cyg OB1 that form the core of the Cygnus X complex (see
Fig. 2).
Cyg OB2 is by far the most important source in the area for which we
predict median fluxes of
ph cm-2 s-1 and
Jy.
Comparison of the median values with the observed fluxes from the
entire region (cf. Sect. 4.1) suggests that Cyg OB2
alone could provide around
of the ionisation in Cygnus while
it may account for only
of the observed 1.809 Me V photons.
Including both observational and modelling uncertainties weakens this
discrepancy, yet the general trend remains:
within the uncertainties, Cyg OB2 could easily explain all ionisation
in Cygnus while it can account at most for
of the
observed 26 Al.
Using successive marginalisation (Eq. (3)) we estimate the
total flux from the associations and clusters in the Cygnus X area
(
;
)
to
ph cm-2 s-1,
ph cm-2 s-1,
and
Jy.
The ionising flux is in comfortable agreement with the integrated 53
GHz emission from the Cygnus region, while the 1.809 Me V flux
falls significantly below the observed value.
The predicted 1.137 Me V flux is way below the detection limit of any
envisagable gamma-ray telescope, reflecting the extreme rareness of
supernova events in this area.
Stellar-wind ejection seems to dominate the enrichment of the
interstellar medium by nucleosynthesis products in the Cygnus X
region - more than
of the 1.809 Me V flux in our model
originates from 26 Al that is ejected through this channel.
The second feature in the model map is low-level 1.809 Me V and 53 GHz
free-free emission from some open clusters in the Cepheus
region, in particular from Trumpler 37.
Trumpler 37 is a young, rich, and nearby open cluster that is embedded
into the bright H II region IC 1396.
Distance estimates based on Hipparcos data vary between DM=9.1
and 9.9 mag (de Zeeuw et al. 1999; Robichon et al.
1999) while our analysis results in a value of
mag.
Using this larger estimate we predict a median 1.809 Me V flux of
ph cm-2 s-1.
Adopting the short Hipparcos distance of de Zeeuw et al. (1999) would
result in a flux of
ph cm-2 s-1, making
Trumpler 37 a potential source of 1.809 Me V
-ray line
emission in an area that is little affected by source confusion.
COMPTEL observations (cf. Fig. 2) do not show any feature
in this area, yet the predicted flux is considerably below the
sensitivity limit of this instrument.
With its improved sensitivity, the SPI telescope aboard the upcoming
INTEGRAL observatory may be able to detect this emission, in
particular if current nucleosynthesis models indeed underestimate 26 Al
production in massive stars (see below).
Finally, the luminosities of all associations have been combined
by successive marginalisation (Eq. (3)) to predict the
total fluxes from the Cygnus region; the results are presented in the
last row of Table 5.
For the entire Cygnus region we obtain a median 53 GHz flux of 5000
Jy, close to the observed value of
Jy.
The total predicted 1.809 Me V flux, however, is significantly below the
observations:
while the model predicts a median gamma-ray line flux of
ph cm-2 s-1 and a
confidence interval
of
ph cm-2 s-1, the COMPTEL data suggest a value
of
ph cm-2 s-1.
Our model predicts only little 60Fe synthesis in Cygnus, and
consequently, the predicted median 1.137 Me V flux amounts to only
ph cm-2 s-1.
At this level, it seems unlikely that the SPI telescope will detect
the radioactive decay of the 60Fe isotope in the Cygnus region.
On average, about
of the 1.809 Me V emission in our model
originates from 26 Al ejected by stellar winds while only
come
from core-collapse supernovae.
Also, OB associations play a dominant role in the Cygnus region since
they contribute to almost
of the 1.809 Me V emission and to
about
of the free-free luminosity.
Our model predicts 99-172 O stars and 13-34 Wolf-Rayet stars in
the Cygnus region associations, about
less than we find in
our cluster database.
We think that this discrepancy is acceptable, in particular in view of
the strong variation of the O star number with association age
(Fig. 3).
Note that in any case, the missing O stars cannot explain the apparent
underestimation of the 1.809 Me V luminosity, since lowering the age
estimates of some associations (in particular Cyg OB2) to increase the
number of O stars tends also to lower the 26 Al production.
On the other hand, an important increase in the number of O stars will
also increase proportionally the ionising flux for the Cygnus region,
leading quickly to an overproduction of Lyc photons.
One essential finding from a census of all known clusters and OB associations and our modelling effort is that we can comfortably reproduce the ionising flux from the Cygnus region, while we fall short by a factor of 2 in explaining 26 Al nucleosynthesis. There are two possibilities that may lead to such a result: either the nucleosynthesis models we employed underpredict 26 Al production by a about a factor of 2, or we overpredict the ionising flux by about a factor of 2, and in the same time, underestimate the stellar population in the Cygnus region by about the same factor. We tend to believe that the problem is more likely related to the nucleosynthesis models, and in the following we will present the elements that lead us to this conclusion.
Our prediction of the ionising flux from the Cygnus region relies on stellar atmosphere models that predict the stellar Lyman continuum luminosities as function of the stellar parameters. However, the lack of spectroscopic data in the extreme ultraviolet shortward of the hydrogen ionisation limit leaves these models basically untested in this wavelength region, and we have to consider the reliability of the employed atmosphere models in predicting Lyc fluxes.
The status of predictions of ionising fluxes has recently been
reviewed by Schaerer (1998, 1999).
From a comparison of Lyman continuum photon fluxes predicted by
various atmospheric models, Vacca et al. (1996) concluded that they are
consistent to within .
Oey & Kennicutt (1997) compared observed H II region luminosities in the Large
Magellanic Cloud to the Lyman continuum luminosities predicted from their
well-determined stellar content, using the CoStar atmosphere models of
Schaerer & de Koter (1997) that we also employed in this work.
They find that on average, H II region luminosities amount to
of the ionising luminosities predicted by the stellar models, leaving room
for about
flux overprediction by the models.
Hunter & Massey (1990) come to a similar conclusion, using the stellar Lyc
estimates of Panagia (1973) applied to small galactic H II regions.
However, accounting for the downward revision of the effective
temperature scale due to non-LTE line blanketing effects, reduces the
Lyc fluxes for a given spectral type (Martins et al.
2002) eliminating the overprediction found by the
above studies.
On the other hand, it is commonly believed that some fraction of the
ionising photons escape the H II regions and ionise the diffuse, warm,
ionised medium (WIM) of the Galaxy (e.g. Oey & Kennicutt
1998).
The WIM has been found to comprise
of the total H
luminosity in nearby star-forming galaxies (see Ferguson et al.
1996), a number that is consistent with the
of ionising photons that are on average missing in the
H II regions with respect to the model predictions.
Our study is not much affected by the fraction of ionising
photons escaping into the WIM, since our microwave flux estimate from
the Cygnus region has been obtained by integrating over a rather larger
spatial area, including both individual H II regions and the WIM
(see Sect. 4.1).
Alternatively, ionising photons may be also absorbed by dust within
the H II regions, and a recent study of Inoue et al. (2001) indicates
that
of the photons may be lost in small galactic H II
regions in this way.
Using a decomposition of DIRBE far-infrared data, Sodroski et al. (1997)
estimated an infrared excess of 1.1-1.7, corresponding to
of Lyc photon absorption within the Galaxy
(Mezger et al. 1974).
Thus instead of ionising the WIM, the apparent
flux overprediction
may be also (partly) explained by dust absorption.
Whatever the mechanism,
ionising flux overprediction by our model
should be a rather solid upper limit, which is in any case far from
the required factor of 2 that is needed to explain the 26 Al
nucleosynthesis.
In our aim to predict absolute fluxes from the OB associations and young open clusters in Cygnus, we have to be concerned about the completeness of our database. First, there might be more OB associations in Cygnus than those we listed in Table 1. Due to their extent and weak clustering, OB associations are generally difficult to identify, in particular if they are nearby. To estimate the association completeness, we compared the number of O stars in the Cygnus area from the SIMBAD database to the number of O stars in our stellar database. Many entries in the SIMBAD database originate from the O star catalogue of Garmany et al. (1982), which is complete down to a visible magnitude of V=10 mag, corresponding to a maximum DM+AV=14.3 mag for O type stars. Comparison of this limit to the distance moduli and reddenings in Table 1 shows that, except for Cyg OB2 and some of the heavily reddened or distant open clusters, the SIMBAD data should indeed be complete for the Cygnus region.
In total we find 171 MK classified O stars in the surveyed area in the
SIMBAD database while our association database contains 204 O stars.
Our excess in O stars is due to the 120 objects in the heavily reddened
Cyg OB2 association that are only partially identified in SIMBAD
(about ).
Excluding Cyg OB2 from SIMBAD and our database results in 108 O stars
in SIMBAD versus 84 O stars in our database, resulting in a
completeness of almost
.
Adding the 120 O stars of Cyg OB2 to both samples increases the O star
completeness to about
.
Thus, most known O stars are indeed located in the OB associations of
our database.
Secondly, there may be young open clusters in Cygnus that escaped
so far detection due to the heavy obscuration in this area.
Indeed, using 2MASS near-infrared data, Dutra & Bica (2001) found 21
new cluster candidates in our survey region (
),
which doubles the number of young open clusters in this area.
Also, on basis of 2MASS data, Le Duigou & Knödlseder (2002) determined the
stellar content of 15 new infrared cluster candidates in the area,
and concluded that they triple the OB population with respect to the
known open clusters in the centre of the Cygnus X region.
However, compared to the giant association Cyg OB2, they contribute
only a small fraction, and we find only an additional
of OB
stars in the newly discovered clusters candidates.
Thus, in summary, we believe that our database is fairly complete,
within about ,
and that the 1.809 Me V luminosity puzzle
can not be explained by a hidden population of massive stars.
In any case, if there would be such a hidden population, it should
also produce a considerable amount of ionising photons, which would
lead quickly to an overproduction with respect to the microwave
observations.
Throughout this work we have ignored the effects of stellar rotation, mainly because there is not sufficient stellar data available for the stellar associations we are interested in, and since there are no stellar grids available so far that allow for inclusion of rotation in evolutionary synthesis calculations. Rotation firstly affects the colour and luminosity of a star, hence it may introduce an artificial age spread in the cluster HRDs (e.g. de Geus 1990; Lamers et al. 1997). Since we accounted for age uncertainties in our approach, however, we do not believe that neglecting of rotation in the determination of the cluster parameters plays an important role.
![]() |
Figure 6:
Evolution as a function of time of the total mass of the star,
![]() ![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 7:
Continuous lines show the evolution as a function of time of the
integrated mass of 26 Al ejected by stellar winds for a non-rotating
and a rotating 60 ![]() |
Open with DEXTER |
Much more important may be the impact of rotation on stellar evolution
and in particular the nucleosynthesis of 26 Al.
To illustrate this point, we show in Fig. 6 the evolution
of the surface 26 Al abundance as a function of time for two 60
stellar
models at solar metallicity with and without rotation (Ringger 2000).
The models were computed with the same physical ingredients as used in
Meynet & Maeder (2000).
Let us note in particular that the mass-loss rates used by
Ringger (2000) are smaller by a factor of 2-3 than the mass-loss rates
used in the stellar models that have been applied to calculate the
26 Al yields in this work.
In the case of the rotating models, surface enrichments begin to occur at a much earlier stage than in the non-rotating model (see Fig. 6). Indeed rotational mixing brings freshly synthesised 26 Al at the surface well before shells, having experienced CNO processing, are uncovered by the stellar winds. This explains why the 26 Al abundance at the surface of the rotating model increases in a much smoother way than at the surface of the non-rotating model. One can see also that rotational mixing slows down the decrease in mass of the convective core and increases the Main Sequence lifetime. This results from the diffusion of hydrogen into the convective core. Some 25Mg will also migrate from the radiative envelope into the convective core where it will be transformed into 26 Al. These effects tend to increase the quantity of 26 Al ejected by the stellar winds.
Figure 7 shows the evolution of M26, the integrated mass
of 26 Al ejected by the stellar winds by the rotating and non-rotating 60
models.
This quantity is obtained by estimating the integral
![]() |
(8) |
How the results of the present population synthesis models would be affected
by rotation?
First of all, let us recall that the WR stellar models used in the present
population synthesis models implicitly account in some way for non-standard
physical mechanisms.
Indeed these models were computed with artificially enhanced mass loss rates
which enabled to account in a satisfactorily manner for the variation with
the metallicity of the number of WR to O-type stars
(Maeder & Meynet 1994).
Thus the reproduction of this observational constraints in some way calibrates
these stellar models.
If we compare the total quantity of 26 Al ejected by our rotating
60
model with the non-rotating 60
model with enhanced mass loss
rates, used in the present population synthesis model, we find a modest
26 Al yield increase of 13%.
Does this mean that rotation has to be discarded as a possible mechanism for solving the 26 Al deficit discussed in the present paper? We think it is too early to reach such a conclusion. Since the work of Ringger (2000), new improvements of the physics of rotation have been achieved especially concerning the shear diffusion coefficient and the impact of rotation on the mass loss rates (Maeder & Meynet 2000, 2001). Predictions of precise enhancement factors requires the computation of stellar models for different initial masses and rotational velocities. At the present time, no such grids of models, predicting the 1.809 Me V luminosity, have been computed. We therefore defer a more quantitatively assessment of rotation on gamma-ray line emission in Cygnus to future work.
Binarity may be another mechanism that could alter the 26 Al yields from massive stars, yet we did not account for their effects due to the involved complexity and poorly known details. In particular, only close binary systems are concerned, and there is not sufficient data available for the associations we are interested in to assess their number and mass distribution.
Two effects may impact the nucleosynthesis in close binary systems: tidal interactions in close binary systems, and mass transfer by Roche Lobe overflow. Tidal effects are expected to deform the star and therefore induce instabilities reminiscent of those induced by rotation. In particular, tidal forces and orbital motions will induce rotation even in initially non-rotating stars. To our knowledge, this effect has never been studied in nucleosynthesis calculations, despite its potentially important consequences, as for instance by homogenising the stars.
If mass transfer occurs, the removal of part of the envelope of the
donor star may favour the appearance of 26 Al on the surface, in a
similar fashion as mass-loss acts through stellar winds.
More important changes may occur for the gainer in systems with
,
as e.g. shown by the preliminary studies of
Braun & Langer (1995) and Langer et al. (1998).
The latter suggest a scenario in which mass transfer onto the
secondary leads to a rejuvenation which alters its subsequent evolution.
They predict an increase by 2-3 orders of magnitudes of the
hydrostatically produced 26 Al yield due to a reduction of the delay
between production and ejection of 26 Al, which may possibly enhance
the total 26 Al production from Type II supernovae by a factor of
about 2 (Langer, priv. communication).
Since, however, supernovae are very rare in Cygnus, it is highly
doubtful that such a mechanism could solve the 26 Al yield puzzle.
Our modelling effort of the gamma-ray line emission from OB associations and young open clusters in the Cygnus region has revealed a possible shortcoming of actual nucleosynthesis models in explaining 26 Al production. We find a 1.809 Me V flux underestimation of about a factor of 2 that is difficult to explain by other means than a modification of current nucleosynthesis models for single, non-rotating stars. From preliminary calculations, it appears difficult for rotation alone to significantly enhance 26 Al production, although this conclusion needs to be checked by more detailed calculations in the future.
Only little 60Fe production is predicted in Cygnus by our model, mainly related to the low number of recent supernova events in this region. A detection of the 1.137 Me V and 1.332 Me V lines from the radioactive decay of 60Fe by INTEGRAL would therefore present a big surprise. In the case of such a detection, 60Fe production by hydrostatic helium burning in Wolf-Rayet stars - which we have not included in our model due to the low yield predictions - should then be seriously reconsidered (Arnould et al. 1997).
In addition to 26 Al and 60Fe, massive star associations may produce further radioactive isotopes during the supernova explosion of massive stars, such as 56Co and 44Ti, yet their short lifetimes of 112 days and 87 years, respectively, make their observation impossible in absence of a very recent event. However, 56Co and 44Ti decay under positron emission (like 26 Al), and the annihilation of positrons with electrons of the interstellar medium on time scales of a few 105 years may provide a reverberation of the short-lived, extinct, radioactivities. Hence, the observation of the 511 keV positron annihilation line may provide an independent measure of the supernova activity in massive star associations (and in particular in Cygnus), although the interpretation of the observations will be complicated by the annihilation physics, the positron transport, and the positron escape fraction from the expanding supernova remnants.
Gamma-ray as well as free-free emission in the Cygnus region seems to
be dominated by a single, extremely massive association: Cyg OB2.
Indeed, Cyg OB2 is merely a prototype of a young globular cluster
than an OB association (Knödlseder 2000).
There are examples of further super star clusters in the Galaxy, such as
NGC 3603 (Moffat et al. 1994),
the Arches and Quintuplet clusters near the galactic centre
(Figer et al. 1999),
or the W49A cluster (Conti & Blum 2002).
Most of these super star clusters are partially or totally obscured in
the visible by the absorbing effects of intervening and/or local
interstellar dust, and it is unclear how many of them exist
throughout the entire Galaxy.
Making the simplifying assumption that super star clusters are
distributed uniformly throughout the Galaxy within a galactocentric
distance of 15 kpc, and taking that Cyg OB2 is apparently the only such
object within 1.5 kpc, one may expect 100 super star clusters in our
Galaxy.
Assuming that they all produce 10-2
of 26 Al, similar to
our Cyg OB2 model prediction, a total 26 Al production of 1
is
expected from these objects.
Taking into account the 26 Al yield underestimation of a factor of 2
brings this mass to 2
,
comparable to the observed galactic 26 Al
mass of 2-3
(Diehl et al. 1995;
Knödlseder 1999).
Thus, a considerable fraction of 26 Al may indeed be produced by such
super star clusters, and 1.809 Me V gamma-ray observations with
sufficient angular resolution and sensitivity may help to uncover and
to study them throughout the Milky-Way.
Our age determination of clusters in the Cygnus region sheds some doubt
on the triggered star formation scenario that has been proposed to explain
anomalous stellar proper motions in the area
(Comerón & Torra 1994).
In this scenario, the central association Cyg OB2 is supposed of
having formed a shell blown by stellar winds and supernovae, the
Cygnus superbubble, that subsequently gave birth to the surrounding
associations Cyg OB1, OB3, OB7 and OB9 due to gravitational shell instability.
However, our age estimate for Cyg OB2 is inferior to that of the
surrounding associations, making triggered star formation in this area
unlikely.
In contrast, the anomalous stellar proper motions are equally well
explained by supposing that they result from dynamically ejected
runaway stars from Cyg OB2 (Comerón et al. 1998),
in particular since the proposed expansion age of 4 Myr is
compatible with our age estimate for Cyg OB2.
Finally, we want to mention that not only the Cygnus region but also
the Galaxy as a whole suffers probably from an underproduction of 26 Al
when current, non-rotating nucleosynthesis models are considered.
By combining the same nucleosynthesis models that have been used in
this work with the atmosphere models of Schaerer & de Koter (1997), and
normalisation to the observed galactic Lyman continuum luminosity,
Knödlseder (1999) predicted a galactic 26 Al mass of 1.6 ,
significantly
below the observed value of 2-3
.
Only the inclusion of the galactic metallicity gradient, which
considerably enhances 26 Al ejection through stellar winds in the
Wolf-Rayet phase towards the galactic centre, provides a sufficient increase
of the galactic 26 Al production, bringing the model predictions in better
agreement with the observations.
Yet, metallicity is of no help in the Cygnus region, where even
slightly subsolar abundances are reported
(Daflon et al. 2001).
On the other hand, if another process, such as rotation, is needed to
explain the 26 Al production in Cygnus, it should also be at work for
the entire Galaxy, leading to a potential 26 Al overproduction if
combined with the metallicity enhancement.
In fact, if 26 Al production by massive stars would indeed follow the
dependence predicted by non-rotating stellar
models, a break-down of the tight correlation between 1.809 Me V and
galactic free-free emission would be expected due to the inverse metallicity
dependencies of both emission processes - the 1.809 Me V intensity should
be enhanced in the metal-rich inner regions of the Galaxy due to enhanced
26 Al production, while the free-free emission should be reduced due to lower
electron temperatures (see Knödlseder 1999).
The existing data show no indication for such an anti-correlation. However, it is likely that rotation reduces the metallicity dependence of Wolf-Rayet star yields. The fact that at low metallicity, there is less 25Mg, might be somewhat compensated by the fact that when the metallicity decreases, the mixing is more efficient, bringing more 25Mg from the radiative envelope into the core and more 26 Al from the convective core into the radiative envelope. The mixing efficiency increases when the metallicity decreases because, when Z decreases, the stars are more compact, the internal gradients of the angular velocity are steeper and the stars loose less angular momentum by mass loss through stellar winds (see Maeder & Meynet 2001). Secondly, when rotation is accounted for, the mass loss rate plays a less important role in the WR formation process (see Fig. 6). Thirdly, at low metallicity, stars loose less angular momentum. As a consequence they can more easily reach the break-up limit at a given stage during their evolution. Very high mass loss rates ensue even if the metallicity is low. Finally, there are some indirect indications, that the proportion of fast rotators increases when the metallicity decreases (Maeder et al. 1999).
Since rotation seems to erode the metallicity dependence of Wolf-Rayet star yields, it provides an appealing mechanism to explain the absence of a 1.809 Me V - free-free emission anti-correlation. Remains to be seen if rotation can also be a solution to the 26 Al yield puzzle - both for the Cygnus region and the Galaxy as a whole.
Acknowledgements
This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France and the WEBDA database, compiled by Jean-Claude Mermilliod, Institute of Astronomy of the University of Lausanne, Switzerland.