Astron. Astrophys. 359, 113-130 (2000)

5. X-ray detectability of the embedded T Tauri star population

We use the information given both by the ISOCAM survey and by our HRI deep exposure to study the X-ray detected TTS population of the Oph dense cores. We restrict the following studies to the HRI /ISOCAM overlapping area. ⁹ This area comprises 98 Class II sources (classified from ground-based and ISO observations), including 52 new ISOCAM Class II sources, ¹⁰ and 35 Class III sources (characterized as YSO from X-ray or radio observations, and classified as Class III sources from ground-based and ISO observations), including 21 new Class III sources (HRI or PSPC X-ray sources without IR excess observed by ISOCAM ). We will call these sources the "TTS sample".

Our HRI observation detected a large number of sources, yet these constitute only 30 of the "TTS sample". In this section, we examine the reasons why the other TTS were not detected, and in particular whether the undetected TTS form a separate population of genuinely X-ray weak objects.

5.1. X-ray vs. stellar luminosities

First, to know more about the X-ray properties of the members of the whole TTS sample (with upper limits if they are not detected with the HRI ), we examine whether a correlation exists between the X-ray luminosity and the stellar luminosity (both corrected from extinction), and if so, whether it is the same as the one found in  Oph by CMFA. We chose for each Core F X-ray source its lowest X-ray luminosity (including HRI upper limits) to minimize the effects of X-ray variability. For TTS undetected by the HRI , we estimate count rate upper limits () ¹¹

To establish the existence of a linear correlation between and , we performed three statistical tests using ASURV: Cox's proportional hazard model, the generalized Kendall test, and Spearman's test. The probability of the null hypothesis (i.e. that this correlation is not present) is for each of the three tests. Thus, a strong linear correlation between and is indeed present. We found the linear regression coefficients by using the Estimation Maximization (EM) algorithm under Gaussian assumptions and the Buckley-James method. The Buckley-James method gave results similar to those of the EM algorithm, but with a larger uncertainty on the slope. As the Buckley-James method is semi-nonparametric, this suggests that the residuals of the linear correlation may be non-Gaussian. We thus conservatively keep the slope uncertainty given by the Buckley-James method. The - correlation is then given by (see Fig. 5): log(/erg s^-1) = (1.00.2) log() + 30.1. We note that the censoring fraction is so high that the correlation line misses most of the data points and depends entirely on the location of the few lowest detections. The correlation dispersion may be due to X-ray variability, and also to TTS spectral type and age differences: Neuhäuser et al. (1995) points out that the ratio increases with decreasing effective temperature, and shows a variation of with age.

Fig. 5. Intrinsic X-ray and stellar luminosity correlation for the "T Tauri star sample". The squares represent the T Tauri stars detected with the HRI in Core A observation, and the T Tauri stars detected with the HRI in Core F observations for which the lowest X-ray luminosity of the 3 observations is not an upper limit. Downward triangles correspond to the T Tauri stars detected with the HRI in Core F observations for which the lowest X-ray luminosity of the 3 observations is an upper limit. Arrows are the hundred upper limits for the HRI undetected T Tauri stars. The highest upper limits correspond to high extinction sources (see Fig. 7). The solid line shows the correlation found using ASURV : log(/erg s-1) = (1.00.2) log() + 30.1.

This correlation spans three orders of magnitude in and two in . The slope of this correlation, a, is equal to 1.0, and the TTS X-ray luminosity is then approximatively given by the simple proportionality: . We thus confirm that the characteristic for TTS in  Oph is 10^-4, with a large dispersion up to a level . There is no evidence for the "saturation" effect seen at this level in late type main sequence stars, and attributed to the complete filling of the stellar surface by active regions (Fleming et al. 1989).

A similar correlation between and was found for Class II and Class III sources in the previous  Oph study of CMFA (the method used to estimate was different, but very similar numerically; see 3.2), but also in other star-forming regions: Chamaeleon (Feigelson et al. 1993; Lawson et al. 1996), Taurus-Auriga (Neuhäuser et al. 1995), and IC 348 (Preibisch et al. 1996). However, the slopes may not be identical: while Feigelson et al. (1993), Preibisch et al. (1996), and CMFA find the same slope as above. On the other hand, Lawson et al. (1996) found , on a better characterized, enlarged X-ray source sample in Chamaeleon, stessing the importance of having a sample as complete as possible. Nevertheless, the fact that we find the same slope as CMFA with an enlarged sample of X-ray sources in the same cloud is certainly a good internal consistency check between the PSPC and the HRI .

We find that the majority of the X-ray luminosity upper limits are above the - correlation. Only 5 of the X-ray undetected TTS are below the correlation, mixed with X-ray detected TTS. This is consistent with the idea that all TTS in  Oph may be X-ray emitters with . Therefore, the TTS undetected by the HRI do not make up a separate population, but must have X-ray properties comparable to that of the detected population, verifying the same correlation.

5.2. The X-ray undetected T Tauri star population

Using the previous correlation between the stellar and X-ray luminosities, we can now estimate for each member of the TTS sample the intrinsic X-ray luminosity in the ROSAT energy band, ¹² and compare it with the HRI detection threshold to understand the X-ray detectability of the TTS sample with the HRI . However, the comparison is not straightforward, since the HRI detection threshold depends on both instrumental effects and extinction along the line of sight.

Fig. 6 shows the instrumental effects: the HRI count rate threshold () increases away from the pointing axis. We interpret this dependence as the consequence of the point spread function degradation and reduced sensitivity off-axis of the ROSAT mirrors (David et al. 1997). With the X-ray spectrum assumptions described in x4.1, we have determined using EXSAS the conversion factor, f, between the HRI counts and the apparent X-ray luminosity (i.e., in the absence of extinction) in the ROSAT energy band (0.1-2.4 keV), . We find: erg cts^-1 s for pc. The minimum X-ray luminosity for a HRI detection, , ranges from erg s^-1 on-axis (angle=0´) to erg s^-1 off-axis (angle=19.2´) (see Fig. 6).

Fig. 6. HRI count rate thresholds (3.25 ) vs. off-axis distance (bottom scale ). Left (resp. right) scale: HRI count rate (resp. apparent X-ray luminosity) threshold (3.25 ). Plus signs (asterisks) represent Core A (F) T Tauri stars of the "T Tauri star sample" undetected with the HRI .

In case the X-ray sources suffer some extinction equivalent to magnitudes, the values of on-axis and off-axis must be corrected to obtain the corresponding intrinsic minimum X-ray luminosities as a function of : if a source is heavily extincted, this minimum may be up to two orders of magnitude higher or more than in the absence of extinction (see for example the high values of the upper limits of Fig. 5).

Fig. 7 plots the X-ray luminosities of the TTS sample as a function of (or ). These X-ray luminosities were estimated from the stellar luminosities using the correlation discussed in the previous section. The points are compared with the HRI threshold curves , computed both on-axis and off-axis. The HRI detected TTS (crossed dots) are found to be rather bright ( erg s^-1) and weakly extincted (). The undetected TTS have estimated X-ray luminosities below the computed HRI detection threshold. In particular, we understand why the new ISOCAM Class II sources (Bontemps et al. 2000), characterized both by low stellar luminosities ( ) (and thus presumably low predicted X-ray luminosities, erg s^-1), and relatively high extinctions (), could not have been detected with our HRI observation. ¹³

Fig. 7. Detectability of the "T Tauri star sample" with the HRI vs. extinction. Right scale: stellar luminosity determined from near-IR photometry; left scale: X-ray intrinsic luminosity predicted from the - correlation (see Fig. 5); top scale: extinction expressed in ; bottom scale: extinction expressed in . "New" ISOCAM Class II sources are IR sources with IR excess discovered by ISOCAM . "New" Class III sources are IR sources characterized as YSO candidates from X-ray or radio observations, and for which ISOCAM observed no IR excess. Crosses indicate T Tauri stars detected with the HRI . Solid curves show the detection threshold for the ROSAT   HRI (on- and off-axis), assuming a Raymond-Smith spectrum with =1 keV and an exposure time of 77 182 s. Several T Tauri stars were detected despite being below the nominal HRI detection threshold: these sources have likely been detected in a flaring state (see text for details). Dashed-dotted curves show the XMM-EPIC (medium filter) detection threshold for the same spectrum and exposure time; we assumed a noise of cts s-1 for the EPIC-pn , and deduced the detection threshold for the all EPIC instrument applying a 0.75 factor (see the XMM Users' Handbook; Dahlem et al. 1999). The dotted curves show the XMM-EPIC (medium filter) detection threshold for an high plasma temperature of 5 keV (flare). The X-ray detection of the "T Tauri star sample" is limited by the HRI sensitivity. XMM-EPIC thanks to its increased sensitivity and its energy range (0.2-12 keV), is less sensitive to extinction, and should be able to detect numerous new Class III sources, as well as the low-luminosity Class II sources discovered by ISOCAM

Now ISOCAM cannot per se recognize Class III YSO among its sources without IR excess, but X-rays can. However, a reliable census of Class III sources in  Oph is de facto limited by the sensitivity of X-ray observations: Fig. 7 shows that the number of detected Class III sources decreases for low and high ; roughly speaking Class III sources are mainly detected above erg s^-1 (or equivalently =0.35 ), and below .

This strongly suggests that unknown Class III sources may exist. We can estimate their number at least in regions at the periphery of cloud cores, by using the fact that the WTTS/CTTS ratio (or equivalently the Class III/Class II source ratio) is , and also that the HRI is equally sensitive to Class III and Class II sources (see x4). In the HRI/ISOCAM overlapping area, this ratio is 19/22 ; on a comparable area Martín et al. (1998) also found a WTTS/CTTS ratio . Since the "TTS sample" comprises 88 Class II sources and 35 Class III sources above , we predict that Class III sources remain to be discovered in X-rays in the HRI/ISOCAM overlapping area above erg s^-1. These sources are not seen now either (i) because they are too faint in X-rays () -or equivalently from the existence of an proportionality, too faint in stellar luminosity ()- or (ii) too absorbed (), or a combination of both. XMM-Newton will be an ideal tool to reveal such a large number of unknown Class III sources in the future (see 7), but, as shown in the next section, we can already figure out their nature to a large extent.

© European Southern Observatory (ESO) 2000

Online publication: June 30, 2000
helpdesk.link@springer.de