5 Discussion

Attention has been focused in the past years on the duplicity of solar-type- and low-mass stars in nearby associations, due to the development of speckle interferometry and adaptive optics. The interesting result is the large frequency of wide binaries, resolved in the observations. It should be noticed that the rate of spectroscopic binaries is not yet determined, and a fortiori it is not known if any of these "visual'' components is itself a spectroscopic binary, producing triple systems.

5.1 Binary frequency in various environnements

5.1.1 "Rich'' clusters

Duplicity of massive stars in very youg clusters and associations also contains important information for the understanding of star- and cluster formation, but, probably due to technical difficulties, - the spectra are sometimes difficult to understand -, much less efforts have been put in this mass domain. In this context, NGC 6231 is an important cluster because it is outside its parent cloud and contains 14 O-type stars. To compare the characteristics of NGC 6231 with those of other clusters, we have searched the open-cluster database (Mermilliod 1995) for available information on spectroscopic binaries in other O-type clusters.

The binary frequencies of 8 galactic open clusters with at least 6 O-type (primary) stars are collected in Table 4, which lists the cluster names, the number of O-type stars (counting only the primaries), the number of spectroscopic binaries and the binary frequency. This table shows that the overall binary frequency among O-type stars varies from cluster to cluster from values as high as 80% in IC 1805 and NGC 6231 to values as low as 14%, in Tr 14. In all cases the duplicity is well documented and based on extensive radial-velocity surveys, even if few orbits have been completely determined.

 

 

Table 5: Open clusters with one or two O-type stars in binary or multiple systems

Cluster	No	HD	Sp.T.	Binarity	Period	Visual	$N_{\rm star}$	$N_{\rm O}$	Multiplicity
NGC 2362	23	57061	O9 II	SBE	1 $\hbox{$.\!\!^{\rm d}$ }$3	Aa 0 $\hbox{$.\!\!^{\prime\prime}$ }$15	5	2	Quintuple
				SB1O	154 $\hbox{$.\!\!^{\rm d}$ }$9	AB 8 $\hbox{$.\!\!^{\prime\prime}$ }$2
NGC 7380	2	215385	O6 V + O7 V	SB2OE	2 $\hbox{$.\!\!^{\rm d}$ }$1		2	2	SB2
NGC 6193	1	150136	O5 V + O6:	SB2O	2 $\hbox{$.\!\!^{\rm d}$ }$7	AB 1 $\hbox{$.\!\!^{\prime\prime}$ }$6	5	3	Quintuple
	2	150135	O7 V	SB2?		AC 9 $\hbox{$.\!\!^{\prime\prime}$ }$6
NGC 1502	2	25639	O9.5 V	SB2OE	3 $\hbox{$.\!\!^{\rm d}$ }$1	Ba 0 $\hbox{$.\!\!^{\prime\prime}$ }$07	6	1	Multiple
	1	25638	B0 II	SB1O	2 $\hbox{$.\!\!^{\rm d}$ }$8	AB 17 $\hbox{$.\!\!^{\prime\prime}$ }$9, Aa 6 $\hbox{$.\!\!^{\prime\prime}$ }$0
NGC 6604	1	167791	O8 If +				3	3	Triple, A
			O5-8 V + O5-8 V	SB2E	3 $\hbox{$.\!\!^{\rm d}$ }$3				B: SB2
NGC 6383	1	159176	O7 V + O7 V	SB2OE	3 $\hbox{$.\!\!^{\rm d}$ }$4	Aa 0 $\hbox{$.\!\!^{\prime\prime}$ }$27	5	2	Quintuple
						AB 5 $\hbox{$.\!\!^{\prime\prime}$ }$4, Ba 0 $\hbox{$.\!\!^{\prime\prime}$ }$7
Tr 37	466	206267	O6	SB2O	3 $\hbox{$.\!\!^{\rm d}$ }$7	Aa 0 $\hbox{$.\!\!^{\prime\prime}$ }$09, AB 1 $\hbox{$.\!\!^{\prime\prime}$ }$6	4	1	Quadruple
Tr 24	515	152623	O7 V	SB1O	3 $\hbox{$.\!\!^{\rm d}$ }$9	Aa 0 $\hbox{$.\!\!^{\prime\prime}$ }$23	3	1	Triple
NGC 6871	1	190918	WN5 + O9.5 I	SB2O	112 $\hbox{$.\!\!^{\rm d}$ }$4	AB 6 $\hbox{$.\!\!^{\prime\prime}$ }$6	3	2	Triple
NGC 2264	131	47839	O7 V((f)) + O9.5 V	SB1O	23.6y	AB 2 $\hbox{$.\!\!^{\prime\prime}$ }$91	3	2	Triple
NGC 1976	1891	37022	O6	Cte		Cc 0 $\hbox{$.\!\!^{\prime\prime}$ }$037	>8	1	Trapezium
	1865	37020	B0	SB1E	65 $\hbox{$.\!\!^{\rm d}$ }$4	Aa 0 $\hbox{$.\!\!^{\prime\prime}$ }$22
NGC 6823	1	344782	O7 V((f))				>5	2	Trapezium
	3		O7 V((f))
IC 1848	1	17505	O6.5 V(f)	SB2?		AG 23 $\hbox{$.\!\!^{\prime\prime}$ }$7	4	2	Multiple
	2	17520	O8 V			Gg 0 $\hbox{$.\!\!^{\prime\prime}$ }$3
IC 1590	248	5005	O6.5 V((f)) + O8 Vn			AB 1 $\hbox{$.\!\!^{\prime\prime}$ }$5, AC 3 $\hbox{$.\!\!^{\prime\prime}$ }$9	4	3	Quadruple
	250		O9 Vn			AD 8 $\hbox{$.\!\!^{\prime\prime}$ }$9

Among the seven O3-O9 stars in Tr 14, only one has been definitively proved to be a binary (García et al. 1998), with a period of 5 $\hbox{$.\!\!^{\rm d}$ }$03 (Levato et al. 2000). Tr 14 has the lowest number of binaries and is the most dense clusters among those listed in Table 4.

The orbital periods, mainly determined in NGC 6231 (HCB and present study) and Tr 16 (Levato et al. 1991), are preferently shorter than 10 days.

5.1.2 "Poor'' clusters

The very young open cluster NGC 6193 has also quite a large binary frequency (Arnal et al. 1988), but it contains only two O-type stars and most binaries are early-B stars. It has been included in Table 5 which presents a list of 14 very young open clusters which contain only one or two O-type stars. These clusters are considered separately from the "richer'' ones because the population characteristics strongly differ between the two kinds of clusters.

In these "poor'' clusters, all O-type stars, $\Theta^1$ Ori C being the only exception, are spectroscopic binaries, often SB2, and even eclipsing. The orbital periods of 8 among 11 O-type binaries are around 3 $\pm$ 1 days and this accumulation is quite surprising. In addition, the hard binaries have other companions and the multiplicity is usually larger than 3. The overall appearance of these poor clusters is therefore quite different from that of the "rich'' clusters. The systems which contains an inner short-period spectroscopic binary are in fact highly hierarchised, as deduced from the estimations of the periods for the close "visual'' components computed by Mason et al. (1998). However, a few ones, like Orion and NGC 6823, form trapezium systems.

Whether the companions are optical or visual is a question which is difficult to answer with available data. Indeed, no separate photometry or radial velocity exist which could prove that the companions fall on the right place in the colour-magnitude diagram. However, companions with separations less than a few arcsec are certainly physical, while in the case of separations of the order of 15 to 20 arcsec, the question of the gravitational link may be raised, especially in a cluster. In Table 5, each O-type spectroscopic binary has at least one close companion and is therefore a triple or a quadruple system. Even if companions with larger separations are not physically associated but are simply cluster members, nearly all systems have a multiplicity larger than 2. It should be mentioned that some young and dense open clusters, like NGC 1502, 2264 and 6871, have been improperly included in the ADS catalogue with 10 or even 20 components. A plot of these stars just reproduces the cluster map.

No information on periods are unfortunately available for the O-type stars in NGC 6823 and in IC 1590 and 1848. These clusters are listed at the bottom of Table 5 because the O-type stars are part of multiple- or trapezium systems. Further open clusters have not been included in Table 5 because of the lack of information on the binarity and multiplicity of their O-type stars (e.g. Tr 27, St 16, NGC 6514, NGC 6618).

A list of 37 galactic open clusters with 1 to 7 O-type stars has been published by Mermilliod & García (2000). It would be important to observe the O-type stars and determine their binary and multiplicity status to extend the present statistics. However, only 4 additional clusters have more than 5 stars, and the sample of clusters presented in Table 4 cannot be increased significantly.

5.2 Cluster structure

We have at present two evidences that there may be a causal link between cluster density and binary frequency. On the one side, wide binaries seem to be more numerous in loose aggregates (i.e. in Taurus-Auriga) than in more dense clusters (i.e. in Orion) as demonstrated by Duchêne (1999) and Mathieu et al. (1999). On the other side, Penny et al. (1993) have proposed that the higher than normal density in the central part of Tr 14 is due to the lack of binaries. One can therefore wonder whether the cluster density is governed by the binary population or if the binary population depends on the cluster density. It is probably valid for wide binaries, but we are here dealing with short-period spectroscopic binaries and the condition of survival are not the same. A dense central region would have a tendency to destroy binaries and to become more dense. If few binaries are initially formed, the process of cluster formation could directly form a dense cluster. But, in light of the results on duplicity reported in the present paper, we can wonder if single massive stars are ever formed in cluster environment.

In rich clusters with many O-type stars, the massive O stars have characteristics similar to those of the other main sequence stars: they appear to be single or double, but the rate of multiple system seems to be low, they are found over most cluster area, as in IC 1805. On the contrary, in clusters which contain only few massive stars, but may contain many low-mass stars, like the Orion Trapezium cluster, the O-type stars are members of multiple or trapezium systems which are very often located close to the cluster center. The observed multiplicity is generally higher than 3.

The question is: do dense clusters destroy their binaries or is there any coupling producing different binary populations in different environnements: few binaries in very dense clusters (Tr 14), many binaries in rich, less dense, clusters (NGC 6231, Tr 16, Cr 228), more multiple systems in less rich clusters (Orion cluster, NGC 2264) and wide binaries in poor aggregates (Taurus-Auriga)? This result is to be also considered in light of the correlation found by Abt & Sanders (1973) between the binary frequency and the ratio of the mean rotational velocities between the cluster and field stars.

It has been proposed that the binary fraction decreases with the time (Ghez et al. 1993; Patience et al. 1998), or that the physical conditions of the cloud, from which the cluster was born, are the responsables of the incidence of binary systems. Durisen & Sterik (1994) proposed that the binary fraction is established during the formation process, without many later disruptions. They pointed out that a natural prediction of both cloud- and disk-fragmentation models is that the binary fraction is higher in colder star-formation regions. Moreover, they proposed that the cloud temperature could also influence the orbital-period distribution.

The observational evidences we have collected show that we have a clear difference in structure and multiplicity characteristics between "rich'' and "poor'' clusters. But we agree that, due to the available data, we have considered only the two extreme cases. Observations of the clusters listed by Mermilliod & García (2000) would be very important to complete the overall picture.

Consequently the overall structure and, probably, the evolutionary history of these clusters are different.