9 Discussion

   
9.1 NIR lines vs. accretion/wind flows

The presence of accretion/wind flows in TTS is signaled by redshifted/blueshifted absorption features respectively dipping below the continuum in their line profiles. Whenever Pa ${\rm\beta }$ and Br ${\rm\gamma }$ display absorption features they are redshifted and dip below the continuum (CW Tau's blueshifted absorption at Pa ${\rm\beta }$ is the exception). These features must arise from infalling gas. The velocities corresponding to those features are in good agreement with those expected for free-falling material from a few stellar radii out. In the context of the magnetospheric accretion scenario, these velocities can be combined with an observational determination of the stellar radius in order to obtain lower limits for the stellar mass. Bonnell et al. (1998) used velocities provided by Pa ${\rm\beta }$ and Br ${\rm\gamma }$ data presented here to obtain such mass estimates.

Modeling carried out in the context of magnetospheric accretion (e.g. Hartmann et al. 1994) shows that the presence of a redshifted absorption results from seeing the infalling gas against the hot region where the accretion shock occurs. The geometry of the system and the contrast between the temperature of the infalling material and that of the shock region are pivotal in determining whether an absorption appears in the line profile and, if so, the strength of the absorption feature. In fact, they seem to be much more important than the value of the accretion rate itself. It is thus not surprising that plotting accretion rates from Hartigan et al. (1995) and from Gullbring et al. (1998) against the EW of the absorption feature in the IPC Pa ${\rm\beta }$ line profiles shows no significant correlation between the two quantities. Also, no obvious differences in accretion rates are found between stars displaying IPC or type I profiles. As suggested from the modeling, the lack/presence of redshifted absorption in some line profiles does not seem to have an obvious relation with stars accreting matter at different rates or, at least, the accretion rate is not the most important factor determining the actual shape of the line profiles.

To investigate the role of the system's geometry in determining the shape of the line profile we used inclinations available in the literature (Bouvier et al. 1995) for the stars in our sample. From the stars with inclinations in Bouvier et al. (1995), fifteen have Pa ${\rm\beta }$ IPC or type I profiles and nine have Br ${\rm\gamma }$ IPC or type I profiles. Figure 7 shows the distribution of inclinations for those stars in bins of $30^{\circ}$.

Figure 7: Distribution of inclinations for IPC and type I profiles. Top panel - Pa ${\rm\beta }$; Bottom panel - Br ${\rm\gamma }$. Inclinations from Bouvier et al. (1995)

Figure 8: Accretion Rate vs. emission EW in Pa ${\rm\beta }$ IPC lines. Top panel - accretion rates from Hartigan et al. (1995); Bottom panel - accretion rates from Gullbring et al. (1998)

These distributions do not show any clear separation in inclination angle for stars with IPC and with type I line profiles. However, the Pa ${\rm\beta }$ IPC line profiles tend to occur in systems with ibetween $30^{\circ}$ and $60^{\circ}$ but rarely above $60^{\circ}$ or below $30^{\circ}$. This is a result obtained from a relatively small number of stars and should be regarded with extreme caution. It should be noted that the magnetospheric accretion model predicts redshifted absorption for systems observed at high inclinations. If systems at intermediate inclinations are really more prone to display IPC structure in their line profiles then the flow geometry in the magnetospheric accretion models envisioned thus far is only a rough approximation of the way in which matter really accretes onto the star. Another indication that this is the case is the fact that one sees so many type I profiles, which is not predicted by current magnetospheric accretion models.

Since the presence of a redshifted absorption feature is also sensitive to the temperature of the accretion shock, one could, in principle, try to correlate the appearance of IPC structure in a given star with the temperature of its accretion shock. However, the latter is not observationally constrained well enough for individual objects.

What about the origin of line emission? According to Muzerolle et al. (1998a), if hydrogen line emission originates in infalling matter in the context of the magnetospheric accretion model, one should expect to see a correlation between emission line strength and accretion rate.

Plotting accretion/wind rates versus the EW of type I line profiles show no apparent relation between the rate at which mass falls onto or flows away from the stars and the strength of the Pa ${\rm\beta }$or Br ${\rm\gamma }$ (where the lines strenghts are normalised to the continuum). Pearson's correlation coefficients are 0.44 and 0.41 respectively for $\dot{M}_{\rm acc}$ vs. EW and $\dot{M}_{\rm wind}$ vs. EW. The number of data points in those plots (not shown) is 16 for $\dot{M}_{\rm acc}$ vs. EW and 14 for $\dot{M}_{\rm wind}$ vs. EW.

On the other hand, plotting accretion rates versus the EW of the emission component in the Pa ${\rm\beta }$ IPC lines (Fig. 8) shows that there seems to be a trend in the sense that stars with larger accretion rates tend to have larger EW in the emission component of the IPC profiles.

Figure 9: Wind Mass Loss Rate vs. emission EW in Pa ${\rm\beta }$ IPC lines. Mass loss rates from Hartigan et al. (1995)

For this relationship Spearman's $\rho$ rank corrrelation is 0.72 with a significance of 0.008. A larger number of points in Fig. 8 would be necessary to claim a definite correlation between the quantities in question.

Plotting the wind mass loss rate, as determined by Hartigan et al. (1995), versus the EW of the emission component of the IPC Pa ${\rm\beta }$ line profiles hints that no correlation exists between the two quantities (see Fig. 9). This is somewhat surprising given the positive correlation between mass accretion rate and wind mass loss rate observed by Hartigan et al. (1995).

While the amount of emission in IPC profiles does seem to behave according to predictions from existing magnetospheric accretion models, the amount of emission in type I profiles does not seem to follow that trend. One possible explanation for the lack of correlation is that the main determinant of the strength of line emission in these NIR lines is not the amount of accreting matter. Alternatively, radiative transfer and/or geometrical effects may smooth out clear relationships between those two quantities.

Calvet & Hartmann (1992) were partially motivated by the fact that the Balmer line profiles observed for many T Tauri stars are generally symmetric and "centrally peaked'', i.e. type I profiles, which turned out to be very difficult to explain by wind models. That work and its development (Hartmann et al. 1994) show that infall models in a magnetospheric accretion scenario can produce the desired Balmer line profiles with no obvious signature for accretion. Hartmann et al. (1994) shows that profiles of hydrogen lines arising in magnetospheric accretion flows can be of type I but displaying a distinctive asymmetry due both to geometrical and radiative transfer effects. This asymmetry is in the sense that line profiles have Af's larger than unity and that line peaks occur at slightly blueshifted velocities. Edwards et al. (1994) use this asymmetry in Balmer lines as a signature for infall.

As discussed in Sect. 6.2 above, Pa ${\rm\beta }$ and Br ${\rm\gamma }$ line profiles have their line peaks ocurring at slightly blueshifted velocities and their Af's are larger than unity (recall Figs. 4 and 5). Applying the same criteria used by Edwards et al. (1994) one can argue that Pa ${\rm\beta }$ and Br ${\rm\gamma }$ type I profiles result from infall in a magnetospheric accretion scenario. A different geometry and/or different physical conditions in the accretion flow relative to those considered by Hartmann, Calvet, Muzerolle and co-workers seem to be necessary though. The observation of line profile variability implying variable accretion (e.g. Johns & Basri 1995b) and the presence of localized accretion spots (e.g. Unruh et al. 1998) indeed confirm that axially symmetric accretion models, such as those considered by Hartmann and co-workers, are not truly applicable in T Tauri stars. The virtue of these models is that they do explain, in a qualitative fashion, the general shape of the observed Pa ${\rm\beta }$ and Br ${\rm\gamma }$ line profiles. When quantitative comparisons are done, the model Pa ${\rm\beta }$ and Br ${\rm\gamma }$ lines are too narrow, by about $100\ {\rm km\ s}^{-1}$ in both FWHM and HWZI and also far too intense (by factors of a few in peak intensity). An alternative to infall models for the origin of line emission are wind models. As thoroughly discussed by Calvet et al. (1992) and by Calvet & Hartmann (1992), the latter models seem to have great difficulties in explaining type I profiles. With the appropriate choice of parameters, "stochastic'' wind models (Grinin & Mitskevich 1991 and Mitskevich et al. 1993) are able to produce type I profiles for lines of "intermediate optical depth''. These were calculated with the intention of comparing the results with observations of the infrared CaII triplet. Detailed calculations, specific for the Pa ${\rm\beta }$ and Br ${\rm\gamma }$ lines, are needed in order to establish whether a clumpy structure for a wind can produce type I profiles. As referred to at the beginning of Sect. 3, a number of narrow photospheric absorption lines fall on top of the Pa ${\rm\beta }$ emission line (see FE99). These photospheric lines are easily seen in many of the Pa ${\rm\beta }$ lines observed. As they are photospheric in origin, the fact they are seen imply that either Pa ${\rm\beta }$ is optically thin or, alternatively, that the filling factor of the line emitting region is only a small fraction of the stellar disk, allowing for a direct view of the stellar photosphere.

   
9.2 Pa $\,{\rm\beta }$ vs. Br $\,{\rm\gamma }$

How different is the information conveyed by Pa ${\rm\beta }$ with respect to that given by Br ${\rm\gamma }$?

Almost half of the stars in the studied sample with line emission either at Pa ${\rm\beta }$ or Br ${\rm\gamma }$ have both lines displaying similar characteristics. The most conspicuous differences are stars that have IPC profiles at Pa ${\rm\beta }$ but type I at Br ${\rm\gamma }$ or that display emission at Pa ${\rm\beta }$ but no emission at all at Br ${\rm\gamma }$. The former can result either from the physical conditions where the lines are formed or from the non-simultaneity of the Pa ${\rm\beta }$ and Br ${\rm\gamma }$ line profiles (they were obtained one day apart). Br ${\rm\gamma }$ lines that have type I profiles when the corresponding Pa ${\rm\beta }$ are IPC tend to have Af's larger than the average for their class, with more emission blueward of the line rest velocity. This tends to point to the former explanation, with Br ${\rm\gamma }$ optically thinner than Pa ${\rm\beta }$ and hence less prone to display redshifted absorptions. If this is the case, it can provide constraints on the physical conditions of the infalling gas.

Lack of detected Br ${\rm\gamma }$ emission while Pa ${\rm\beta }$ is in emission shows that for some stars the physical conditions in the hydrogen gas that surrounds them are such that population of level 7 is not too significant while a reasonable amount of atoms have level 5 populated.

A significant percentage of the Br ${\rm\gamma }$ IPC lines tend to peak more towards the blue than the Pa ${\rm\beta }$ IPC lines (Sect. 6.2 above). The blueward shift is of about $60\ {\rm km\ s}^{-1}$. It is worthwhile noting that amongst the Br ${\rm\gamma }$ type I line profiles a significant number also have the velocity of the line peak shifted to the blue relative to $v_{\rm peak}$ in the Pa ${\rm\beta }$ lines (see Fig. 4). An explanation for this shift would be that, due to their different optical depths, Pa ${\rm\beta }$ and Br ${\rm\gamma }$ sample regions where gas moves at different speeds, resulting in line profiles peaking at different velocities. Surprisingly the magnetospheric accretion line profile modeling carried out by Hartmann, Calvet, Muzerolle et al. does not show any relative shift in the peak velocity between lines of very different optical depths, such as H ${\rm\alpha}$, H$\beta$, H$\gamma$, and Br ${\rm\gamma }$.

9.3 Multiple systems

Multiplicity amongst the population of the T Tauri stars is well known (Ghez et al. 1993; Leinert et al. 1993; Richichi et al. 1994; Mathieu 1994; Simon et al. 1995). The separation between components in many of those systems is such that the NIR lines observed might be influenced by the presence of the companion stars. From the 50 stars in the sample presented in this work, the following 16 are, according to the references above, binary or multiple systems with separation between components smaller than the slit width used for the observations reported here. They are: DD Tau, DF Tau, DQ Tau, FS Tau, FX Tau, GG Tau, GH Tau, HP Tau, LkCa7, RW Aur, T Tau, UY Aur, V773 Tau, VY Tau, XZ Tau and ZZ Tau. Of these 16 binary/multiple stars, 9 have Pa ${\rm\beta }$ in emission, from which 6 are of type I, and 3 are IPC profiles; from the 6 displaying Br ${\rm\gamma }$ in emission, 5 are of type I and 1 is an IPC profile. Amongst the T Tauri stars that are part of a multiple system type I line profiles are the most common, followed by the IPC profiles. This is in tune with the results obtained for the whole sample. Excluding the binary/multiple systems from the classification scheme does not change the results significantly.

   
9.4 Line profile variability

A few TTS for which spectra are presented in this work were observed more than once and some considerations regarding variations in the observed NIR line profiles are due. The stars for which the NIR lines were observed more than once are shown in Table 8 where the dates of the observations are also indicated. Observations in consecutive nights of the same line for the same star were only carried out during the December 1995 run, only at Pa ${\rm\beta }$ line and only for two nights. The night to night variation observed is not very dramatic, in the sense that, for these stars, the type of line profile did not change, despite changes in equivalent width and/or in the position of absorption features.

For some of the stars, more drastic changes are observed to occur in spectra taken roughly one year apart. DR Tau changes from type I into type II R and GK Tau changes from type I into an IPC line profile. The Br ${\rm\gamma }$ lines of DR Tau and GG Tau and the Pa ${\rm\beta }$ line of GI Tau change significantly in terms of equivalent width (due to line profile variations) but preserve the type of profile. The type of variations that can occur in NIR lines are diverse and must reflect the dynamical activity and/or changes in the physical characteristics of the emitting region(s) (the change of the Pa ${\rm\beta }$ profile of GK Tau from type I to IPC surely indicates either that an accretion region came into view or that accretion activity in the star changed between the first and second observing runs). For a given star, monitoring these lines over at least two rotation periods (typically a week to a week and a half) would tell us how the infalling matter behaves, since any variability displayed, especially in a redshifted absorption component, should be associated with an accretion flow. As an example, if a magnetospheric accretion scenario is correct and the rotation axis of the star is tilted relative to the magnetic axis one expects to find a correlation between the line profiles and the phase of the rotation period of the star (Johns & Basri 1995b). A better understanding of how and where these lines are formed in TTS certainly requires a variability study of these lines.