7 Discussion

 

   
7.1 Theoretical models for multiple star formation

It is now widely believed that fragmentation during protostellar collapse is the major process for forming multiple stars in low-density SFRs as discussed here (e.g. Clarke et al. 2000). Our results generally are in line with this assumption:

• We have some additional support from our data for coeval formation of the components (Sect. 4.4). Capture processes of independently formed single stars (that would produce a widespread distribution of relative ages) should not play a dominant role in the formation of the binaries discussed here;
• There are only a few companions with masses $M~\le~0.1~\,~M_{\odot}$ which is a typical mass of a T Tauri stars' disk (Beckwith et al. 1990). Furthermore these low mass companions do not preferentially occur at separations $d\le 100~{\rm AU}$ that are comparable to typical disk radii. This is difficult to reconcile with the idea that a large number of companions is formed from disk instabilities;
• Fragmentation during protostellar collapse does not lead to a preference of distinct mass ratios or to a dependence of mass ratio from other binary parameters like the components' separation (e.g. Ghez et al. 2000). We have indeed not seen any such correlations (Sect. 6.3).

Regarding the last item one has however to consider that there is a large time span between the end of numerical simulations of fragmentation during protostellar collapse and the state of dynamically stable T Tauri multiple systems as observed by us. Particularly, at the end of the simulations presented by Burkert & Bodenheimer (1996) and Burkert et al. (1997) only $\approx$10% of the parent cloud's mass has condensed into fragments. Therefore it is a highly important question how subsequent accretion processes influence the properties of young multiple systems. Since it is not possible today to cover the entire evolution of a molecular cloud core into a binary system with one simulation, theory has to take a different approach.

For this purpose Bate (1997, 1998) and Bate & Bonnell (1997) have simulated the behaviour of a "binary'' formed out of two point masses that are situated in a cavity within a surrounding gas sphere. The mass ratio of the binary and the angular momentum of the infalling material are variable initial conditions. The result of these simulations is that in the course of the accretion process the system's mass ratio increases and approaches unity if the total cloud mass is accreted. Mass ratios close to unity should be more probable in close systems than for wide pairs.

This is not in agreement with our result (Sect. 6.3) that there is no significant preference of mass ratios close to unity and that the mass ratio is probably independent of the components' separation. So it seems that the initial conditions used by Bate & Bonnell are somehow unrealistic and that the "final'' stellar masses are more dependent on the fragment masses than on the following accretion processes. One explanation for this could be that in the course of protobinary evolution most of the initial mass is condensed into fragments, before accretion becomes important. Another idea is that there is some process that halts accretion before a large amount of the remaining cloud mass is accreted onto the fragments. In any case, our knowledge about this issue is preliminary and additional theoretical and observational effort is necessary to decide what physical processes determine stellar masses.

   
7.2 Implications for binary statistics

Multiplicity surveys by Leinert et al. (1993) and Ghez et al. (1993) led to the surprising result that there is a significant overabundance of binaries in the Taurus-Auriga SFR compared to main sequence stars in the solar neighbourhood. This result was further proved by the follow-up studies done by Simon et al. (1995) and Köhler & Leinert (1998). Although this paper is not directly concerned with binary statistics, we can draw some conclusions that further support the idea that this binary excess is real and not a result of observational biases.

If one compares the binary frequency among young and evolved stars one has to take into account that due to evolutionary effects companions can be relatively bright in their PMS phase, but invisible on the main sequence stage. This is particularly the case for substellar companions. One has further to consider that the multiplicity surveys were done at infrared wavelengths in SFRs, but in the optical range for main sequence stars. So there might be a bias that supports the detection of very red "infrared companions'' (IRCs, see Sect. 3.2) in the vicinity of PMS stars. Another problem is that the surveys in Taurus-Auriga and the solar neighbourhood could be not directly comparable if they were sensitive to a different range of mass ratios and thus stellar masses.

If the presence of substellar companions or IRCs in the vicinity of T Tauri stars in Taurus-Auriga were a common phenomenon, this could at least partially explain the observed binary excess in this SFR. Our results presented in Sects. 3.2 and 6.1 show that this is not the case: Köhler & Leinert (1998) have found that after applying a statistical correction for chance projected background stars there are $48.9\pm 5.3$ companions per 100 primaries (including single stars) in Taurus-Auriga. We have denoted only 5 out of 40 companions (for which we have given masses in Table 2) as candidates for substellar objects based on masses derived from the D'Antona & Mazzitelli (1998) PMS evolutionary tracks (Sect. 6.1). With respect to the Baraffe et al. (1998) model this number is even lower. If we take the mentioned 5 out of 40 companions as an estimate for the real number of brown dwarf companions in Taurus-Auriga and subtract this from the companion frequency given by Köhler & Leinert (1998) this value diminishes to $42.8\pm 6.0$. This is still far above the value of $25.3\pm 3.9$ that was given by Duquennoy & Mayor (1991) for G-dwarfs in the solar neighbourhood. Furthermore we have found that only 3 out of 51 companions in Taurus-Auriga are detectable in the H-band and at longer wavelengths, but were missed at $1.25\,\mu{\rm m}$, so IRCs are probably not a frequent phenomenon. The binary frequency does not have to be corrected for IRCs, because their successors in the main sequence phase will be "normal'' stellar companions (see Koresko et al. 1997 for estimates of IRCs' masses).

Duquennoy & Mayor (1991) have claimed that their sample is complete for mass ratios $M_2/M_1\ge 0.1$. It has already been mentioned by Köhler & Leinert (1998, Sect. 5.2) that the completeness limit of the binary surveys in Taurus-Auriga is in any case not lower, the actual value dependent on the mass-luminosity relation used. We can further prove this result, because there are only 2 out of 51 systems with mass ratios less than 0.1 with respect to the D'Antona & Mazzitelli (1998) PMS model and 1 out of 50 considering the Baraffe et al. (1998) tracks (Table 3). We conclude that the observed binary excess in Taurus-Auriga compared to nearby main sequence stars is neither the result of a higher sensitivity in mass ratio nor a consequence of a large frequency of substellar or infrared companions. The strange overabundance of binaries in Taurus-Auriga remains a fact also after this more detailed analysis of the systems found by Leinert et al. (1993).