HOT GAS LINES IN T TAURI STARS

The COS and STIS DAO data were all taken in time-tagged mode.

• 1.  
  COS observations. The FUV (1150–1790 Å) spectra were recorded using multiple exposures with the G130M grating (1291, 1327 Å settings) and the G160M grating (1577, 1600, 1623 Å settings). These provide a velocity resolution of Δv ∼ 17 km s⁻¹ (R ∼ 17,000) with seven pixels per resolution element (Osterman et al. 2011; Green et al. 2012).The maximum exposure time was accumulated in the region of the Si iv resonance lines and the exposures ranged from 3 ks (for most stars with two-orbit visits) to a maximum of 8 ks for a few stars with four-orbit visits. Use of multiple grating positions ensured full wavelength coverage and reduced the effects of fixed pattern noise.All observations were taken using the primary science aperture (PSA), which is a 25 diameter circular aperture. The acquisition observations used the ACQ/SEARCH algorithm followed by ACQ/IMAGE (Dixon & the COS/STIS Team 2011). The absolute wavelength scale accuracy is ∼15 km s⁻¹ (1σ), where the error is dominated by pointing errors. We obtained one dimensional, co-added spectra using the COS calibration pipeline (CALCOS) with alignment and co-addition obtained using the IDL routines described by Danforth et al. (2010).
• 2.  
  STIS observations. Those targets that are too bright to be observed by COS were observed using the STIS E140M echelle grating, which has a spectral resolution of Δv ∼ 7 km s⁻¹ (R ∼ 45,000), over the region 1150–1700 Å. All the observations were obtained using the 02 × 02 "photometric" aperture, during two HST orbits.The STIS FUV spectra were calibrated and the echelle orders co-added to provide a single spectrum using the IDL software package developed for the StarCAT project (Ayres 2010). This reduction procedure provides a wavelength-scale accuracy of 1σ = 3 km s⁻¹. Non-DAO STIS data were taken directly from StarCAT.

For a given pointing, errors in the positioning of the target within the aperture result in an offset in the wavelength scale. As indicated above, these are supposed to be 15 km s⁻¹ for COS and 3 km s⁻¹ for STIS, according to the instrument observing manuals (Dixon & the COS/STIS Team 2011; Ely & the COS/STIS Team 2011). In addition, the geometric correction necessary to account for the curvature of the COS FUV detector may make longer wavelength features appear redder than they really are. The offset depends on the exact position of the target on the detector. In most cases this intra-spectrum wavelength uncertainty is <10 km s⁻¹, but in a few extreme cases it will introduce a ∼15 km s⁻¹ shift from the reddest to the bluest wavelengths in a single COS grating mode (see Figure 4 from Linsky et al. 2012).

To determine how important the errors due to pointing and calibration are in the COS data, we focus on the H₂ lines. We use the line measurements from K. France (2013, private communication; see also France et al. 2012) for P(2)(0–5) at 1398.95 Å, R(11)(2–8) at 1555.89 Å, R(6)(1–8) at 1556.87 Å, and P(5)(4–11) at 1613.72 Å. France et al. (2012) show that in the case of DK Tau, ET Cha (RECX 15), HN Tau, IP Tau, RU Lup, RW Aur, and V1079 Tau (LkCa 15), some H₂ lines show a redshifted peak and a blueshifted low-level emission, which makes them asymmetric. Ignoring these stars, the four H₂ lines are centered at the stellar rest velocity, with a standard deviations of 7.1, 6.5, 6.1, and 12.1 km s⁻¹, respectively. The large scatter in the P(5)(4–11) line measurement is partly the result of low signal-to-noise in this region of the spectrum. Unlike the results reported by Linsky et al. (2012), we do not observe a systematic increase in the line center as a function of wavelength, neither star by star nor in the average of all stars for a given line. This is likely the result of the different acquisition procedure followed here. For the STIS data, the average standard deviation in all the H₂ wavelengths is 4.4 km s⁻¹.

This shows that the errors in the COS wavelength scale are smaller than the 15 km s⁻¹ reported in the manuals and that no systematic shifts with wavelength are present. For the rest of the paper we will assume that the pointing errors in the COS data result in a velocity uncertainty of ±7 km s⁻¹ (the average of the first three H₂ lines considered) while the STIS errors are 5 km s⁻¹. We do not correct the spectra for the H₂ velocities, as it is not clear how to implement this correction in the case of stars with asymmetric H₂ lines.

If optically thin, both C iv lines form should have the same shape. Differences between the components can help us discover the presence of extra sources of absorption or emission. To exploit this redundancy, in Figure 16 we plot both members of the C iv doublet, with the 1550 Å line scaled to match the 1548 Å one. The scaling is done by matching the line peaks or the line wings from ∼0 to ∼150 km s⁻¹. We expect this scaling factor to be 2 for optically thin or effectively thin emission. However, as we discuss in Section 5.2.4, the opacity characteristics of the broad and narrow line components are different, and the overall scaling factor may not be the best predictor of opacity.

Figure 16 shows that the line wings tend to follow each other closely, at least until about +200 to +300 km s⁻¹, when the 1548 Å line start to bump into the 1550 Å one. The 1548 Å line is usually contaminated by the H₂ line R(3)1–8 (1547.3 Å), at −167 km s⁻¹ with respect to its rest velocity (e.g., DK Tau). Figure 16 also reveals examples of extra emission at ∼−100 km s⁻¹ which do not appear in the 1550 Å line (AK Sco, DE Tau, DK Tau, DR Tau, HN Tau A, RU Lup, and UX Tau A). We tentatively identify emission from Fe ii (1547.807, −73.6 km s⁻¹ from the rest velocity of the 1548 C iv line), C ii (1547.465; −139.8 km s⁻¹), and Si i (1547.452 Å, −142.3 km s⁻¹; 1547.731 Å, −88.3 km s⁻¹), for these stars, as responsible for the emission to the blue of the C iv 1548 Å line.

To measure the C iv line flux listed in Table 8 we subtracted the continuum and interpolated over the H₂ R(3)1–8 line and the Si i, C ii, Fe ii-complex, when present. The resulting spectrum is then integrated from −400 km s⁻¹ to 900 km s⁻¹ of the 1548 C iv line. We also detect the CO A-X (0–0) absorption band at 1544.4 Å (−730 km s⁻¹ from the 1548 Å line; see France et al. 2011; McJunkin et al. 2013) in a significant fraction of the sample. The wing of the CO absorption may extend to the blue edge of the 1548 C iv line. However, its impact in the overall C iv flux is negligible.

Note that the red wings of each C iv line for DX Cha, RW Aur A, DF Tau, and RU Lup do not follow each other, and AK Sco and CS Cha have extra emission features near the 1550 Å.

For DX Cha we will argue in Section 8 that the peculiar shape of the C iv lines can be explained by the existence of a hot wind.

The strange appearance of C iv lines of RW Aur A is due to a bipolar outflow (see Figure 5). France et al. (2012) show that the H₂ lines from RW Aur A are redshifted by ∼100 km s⁻¹ in at least two progressions ([v', J'] = [1, 4] and [1, 7]), as they originate in the receding part of the outflow. In the observations we present here, the C iv doublet lines are blueshifted by ∼−100 km s⁻¹, as can be seen from the position of the 1550 C iv line. This points to an origin in the approaching part of the outflow. The net result of these two outflows is that the blueshifted 1548 C iv line is buried under the redshifted H₂ emission. RW Aur A is the only unequivocal example of this coincidence in the current data set. For HN Tau A the H₂ R(3) 1–8 line is also redshifted, although the redshift (+30 km s⁻¹) is within the 2σ of the error introduced by the pointing uncertainty (see also France et al. 2012). Therefore, HN Tau A may be another object for which we are observing two sides of the outflow, although the velocities of the blueshifted and redshifted sides do not match as well as in RW Aur.

For DF Tau, the apparent red wing in the 1550 Å line is an artifact of the line scaling, because for this star the ratio between the two C iv lines is almost 3, indicating either extra absorption or emission in one of the components. Unlike the case in RW Aur A, redshifted H₂ emission is not responsible for the extra emission, as the R(3) 1–8 line is observed at −167 km s⁻¹. Other observations show that the ratio between the two lines remains high among different epochs (Section 5.4). The origin of this extra emission in the 1548 Å line or absorption in the 1550 Å one remains unexplained in this work.

For RU Lup, the extra "bump" 200 km s⁻¹ to the red of the 1548 Å C iv line is also present in other epoch observations of the system and remains unexplained in this work. AK Sco and CS Cha also present anomalous profiles, but of a different kind. AK Sco shows a broad emission −250 km s⁻¹ from the 1550 Å line, not present in the 1548 Å line. CS Cha also shows and extra emission in the 1550 Å line at ∼−100 km s⁻¹ and ∼+200 km s⁻¹.

5.2.1. The Velocity of the Peak Emission as a Function of Luminosity and Accretion Rate

The velocity of peak emission for each line is defined as the mean velocity of the top 5% of the flux between −100 and +100 km s⁻¹ (for the 1548 C iv line) or between +400 and +600 km s⁻¹ (for the 1550 Å line). This is calculated from spectra that have been smoothed by a five-point median. The velocity of peak listed in Table 5 is the average of both lines.

The C iv line is centered or redshifted for 21 out of 29 CTTSs, uncorrelated with line luminosity or accretion rate (Figure 6, top). The most significant exceptions are: DK Tau which shows two emission peaks, one blueshifted and the other redshifted, present in both lines; HN Tau A, for which both C iv lines have an emission peak at ∼−80 km s⁻¹ from their rest velocity; ²⁰ RW Aur A, for which the peak of the 1550 Å line is at −86 km s⁻¹. In the case of AK Sco there is extra emission in the 1548 Å line that may be due to Si i but there is also extra emission ∼−250 km s⁻¹ from the 1550 Å line. It is unclear therefore, whether the C iv lines are centered or blueshifted, although both the Si iv and the N v lines (which follow the C iv lines) are well centered. Those four stars are among the higher accretors in the sample and have jets seen in the high-velocity components of forbidden lines (Hartigan et al. 1995). Four other CTTSs have slightly blueshifted peaks, but with velocities > − 4 km s⁻¹.

The average velocity at maximum flux for the C iv CTTSs lines is  km s⁻¹ (Table 9). In this paper the uncertainties refer to uncertainties in the calculated mean values, not including the four stars with strong blueshifts (DK Tau, HN Tau A, and RW Aur A for which outflowing material is contributing to the profile, nor AK Sco, for which the two C iv lines are different from each other). We have argued that the velocity uncertainties for this COS data set are ∼7 km s⁻¹. A one-sample Kolmogorov–Smirnov (KS) test confirms that the probability of the observed distribution being normal centered at zero, with 1σ = 7 km s⁻¹ is negligibly small. The conclusion is the same if we take 1σ = 15 km s⁻¹, the nominal wavelength error for COS. In other words, the overall redshift of the CTTSs C iv lines is significant.

Table 9. Average Line Kinematic Parameters

Parameter	WTTSs C iv	CTTSs C iv	WTTSs He ii	CTTSs He ii
	(km s−1)	(km s−1)	(km s−1)	(km s−1)
Non-parametric Averages (Section 5.2)
VMax	10.6 ± 4	18.3 ± 4 a	4.1 ± 5	7.1 ± 3 a
FWHM	92 ± 10	210 ± 18	85 ± 11	96 ± 9
Skewness	−0.005 ± 0.01	0.018 ± 0.01	−0.007 ± 0.01	0.0126 ± 0.004
Parametric Averages (Section 5.3)
VNC	1.6 ± 2	26.6 ± 6 a	3.9 ± 3	6.0 ± 3 a
VBC	17.2 ± 5	39 ± 10 a	−2.2 ± 1	53 ± 20 a
FWHMNC	91 ± 20	118 ± 10	83 ± 10	85 ± 10
FWHMBC	223 ± 40	334 ± 30	178 ± 20	335 ± 30
Velocity Differences b
− VMax	7.7 ± 6 a	0	14.2 ± 6 a	11.0 ± 4
− VMax	0	−7.7 ± 6 a	6.5 ± 6	3.4 ± 6 a
− VNC	24.8 ± 7 a	0	22.7 ± 7 a	19.7 ± 6
− VNC	0	−24.8 ± 7 a	−2.2 ± 5	−4.4 ± 3 a
− VBC	21 ± 20 a	0	39.2 ± 10 a	−4 ± 10 a
− VNC	38 ± 10 a	26 ± 17	39 ± 10 a	42 ± 10
− VNC	4.4 ± 3 a	−19.7 ± 6	2.1 ± 4 a	0
− VNC	2.2 ± 5	−39.2 ± 10 a	0	−2.1 ± 4 a

Notes. All calculations include the H₂ velocity correction (Section 2.1). ^aDoes not include HN Tau, RW Aur, AK Sco. ^bVelocity differences measured in the same spectrum (for example − ) are not subject to errors due to pointing.

Download table as:  ASCIITypeset image

The WTTSs are also redshifted, with  km s⁻¹. A two-sided KS test comparing the CTTSs and WTTSs V_Max distribution gives p-value = 0.6, which indicates that they are consistent with the null hypothesis. In other words, although the mean velocity of the WTTSs is ∼2σ less that the mean CTTS velocity, the observations are consistent with the two quantities having the same distribution. Linsky et al. (2012) have shown that for rotation periods such as those observed in T Tauri stars, the Si iv and C iv lines in low-mass dwarfs present redshifts of ∼7 km s⁻¹, the result of gas flows produced by magnetic heating. Given the scatter in their sample, our redshifts in WTTSs are consistent with theirs.

5.2.2. The Line Width as a Function of Luminosity and Accretion Rate

5.2.3. The Skewness as a Function of Luminosity and Accretion Rate

5.2.4. The Line Scaling as a Function of Luminosity and Accretion Rate

5.2.5. The Role of the Inclination in the Measured Luminosity and Accretion Rate

5.2.6. Conclusions from the Non-parametric Analysis

Different regions of the T Tauri system may be contributing to the C iv lines. In the context of the magnetospheric accretion paradigm, the pre- and post-shock regions should be the dominant sources of the observed line emission. Shocks in an outflow, hot winds, and the stellar atmosphere may also contribute to the emission. To examine the line kinematics of the different regions we decompose each C iv line in one or two Gaussian functions. We are not asserting that the mechanism giving origin to the lines produces Gaussian shapes, although turbulent flows will do so (Günther & Schmitt 2008). Representative decompositions are shown in Figure 2.

The primary goal of the Gaussian decomposition is to obtain widths and centroids for the main line components. According to the magnetospheric accretion model, post-shock emission lines should have small velocity centroid distribution about the stellar rest frame velocity, and if no turbulence is present, the lines should be narrow. Pre-shock emission lines, which likely originate in a larger volume upstream from the accretion flow (Calvet & Gullbring 1998), should have larger velocity centroid distribution in the stellar rest frame and broader lines.

Table 6 presents the results of fitting one or two Gaussians to each of the C iv lines, after subtracting the continuum and interpolating over the H₂ line R(3)1–8, and the Si i, C ii, Fe ii-complex, when present. For each target, we assume that the C iv lines are always separated by 500.96 km s⁻¹, and that they have the same shape. This results in a four-parameter fit when fitting one Gaussian to each line: height for the 1548 Å line; height for the 1550 Å line; width σs; centroid velocity for the 1548 Å line. When fitting two Gaussians to each line, we have an eight-parameter fit: For the 1548 Å line, the heights of the broad (A_BC) and narrow (A_NC) components; analogous parameters for the 1550 Å line; σ for the BC; σ for the NC; centroid velocities for the broad (V_BC) and narrow (V_NC) components of the 1548 Å line.

Table 6 also indicates whether the data were taken with COS (possible systematic wavelength error assumed to be 7 km s⁻¹) or STIS (possible systematic wavelength error of 3 km s⁻¹). We also list the parameters derived for the multi-epoch observations of BP Tau, DF Tau, DR Tau, RU Lup, and T Tau N that we will consider in Section 5.4.

Wood et al. (1997) analyzed the C iv lines of 12 stars with spectral types F5 to M0. They found that the observed line profiles could be better fit with both a narrow and a broad Gaussian component, than with a single Gaussian component. For the type of stars included in their sample (dwarfs, giants, spectroscopic binaries, and the Sun), the C iv line is a transition-region line, as it is in the WTTSs. Because of this, we fitted both NCs and BCs to all the WTTSs, except V410 Tau. For V410 Tau, each C iv line can be well fitted with only one NC, although this may just be the result of the low S/N in the spectrum.

When comparing the two C iv lines in Section 5.1, we mentioned that AK Sco, CS Cha, DX Cha, DF Tau, RU Lup, and RW Aur A were objects for which the line doublet members had different shapes and/or extra unidentified emission in one of the doublet members. We do not perform the Gaussian decomposition for RW Aur A or DX Cha. For AK Sco, we list the parameters derived from the Gaussian fits (Table 6), but we do not use these results when exploring correlations. For the rest, we interpolate the profiles over the apparent the extra emission.

For four objects (CY Tau, DM Tau, ET Cha, and UX Tau A) we decompose the C iv lines in only one broad Gaussian component. For the rest of the CTTSs, the decomposition requires both a narrow and a broad Gaussian components, and Figure 7 shows the distributions of velocity centroids and FWHM. Average velocity and FWHM values are given in Table 9. Typical FWHMs of CTTS NCs range from 50 to 240 km s⁻¹, with an average of 130 km s⁻¹, while BCs widths range from 140 to a 470 km s⁻¹, with an average of 350 km s⁻¹. The velocity centroids range from −100 km s⁻¹ to 200 km s⁻¹. The BC velocity is larger than the NC velocity in 70% of the CTTS sample, and the distribution of BC velocities tend to be more positive ( km s⁻¹) than that of the NC velocities ( km s⁻¹), giving some of the profiles the characteristic "skewed to the red" shape.

5.3.1. The Optical Depth as a Function of Accretion

5.3.2. The Kinematic Predictions of the Magnetospheric Accretion Model

Line flux variability in CTTSs occurs on all time scales, from minutes to years. It is therefore relevant to ask how does the line shape change in time and what is the impact of these changes on the general statements we have made.

High-resolution multi-epoch observations of the C iv lines are available for a subset of our objects, plotted in Figure 10. The spectra were obtained from Ardila et al. (2002; GHRS observations: BP Tau, DF Tau, DR Tau, RU Lup, T Tau, and RW Aur), Herczeg et al. (2005; STIS: RU Lup), Herczeg et al. (2006; STIS: DF Tau), and the MAST archive (HST-GO 8206; STIS: DR Tau). The epoch of the observations is indicated in the figure. Note that the apertures of the three spectrographs are different. The GHRS observations were performed with the large science aperture (2'' × 2'' before 1994, 174 × 174 after 1994), the PSA for COS is a circle 25 in diameter, and the STIS observations where obtained with the apertures 02 × 006 (for T Tau) or 02 × 02 (for the other objects).

For DR Tau the DAO observations show double-peaked C iv emission as well as strong emission in the blue wing of the 1548 Å line. We have identified the additional blue wing emission as a mixture of H₂, Si i, C ii, and Fe ii. The parameters that we have listed for DR Tau in Table 6 provide a good fit to the overall profile, but not to the double-peaked C iv emission. In the GHRS observation (red, from 1995) the extra Si i, C ii, and Fe ii emissions are not present, leaving only a low S/N H₂ line. The low-velocity peak of the C iv line observed in the DAO data is not present in the GHRS data. In addition to this 1995 observation, Ardila et al. (2002) describe a 1993 observation (not shown in Figure 10), which shows blueshifted emission at −250 km s⁻¹ present in both line components. The STIS observations from HST program GO 8206 (PI: Calvet) show a strong H₂ line. The centroids of the C iv lines are either redshifted by ∼200 km s⁻¹ or the centers (within ±150 km s⁻¹ of the rest velocity) are being absorbed. If redshifted, this is the largest redshift observed in the sample, although a comparable shift is seen in V_BC for CS Cha (Figure 14(c)). We may be observing extended C iv emission that is not seen in the narrow-slit STIS observations (Schneider et al. 2013), C iv absorption from a turbulent saturated wind or the disappearing of the accretion spot behind the stellar limb as the star rotates.

T Tau shows a change in the H₂ emission and a small decrease in the strength on the NC. Walter et al. (2003) and Saucedo et al. (2003) show that the H₂ emission around T Tau is extended over angular scales comparable to the GHRS aperture, and the smaller STIS flux is due to the smaller aperture.

For DF Tau, the ratio between the two line members remain anomalously high, ∼3 in all epochs. For BP Tau, a decrease in the flux is accompanied by a decrease in both components, although the NC decreases more strongly. Changes over time in the velocity centroids are not significant.

For RW Aur A the difference between GHRS and STIS DAO observations is also dramatic. As we have shown, the DAO observations can be explained by assuming that we are observing two sides of a bipolar outflow. The GHRS flux is larger, and dominated by three peaks, the bluemost of which is likely the R(3)1–8 H₂ line. The other two do not match each other in velocity and so they cannot both be C iv. Errico et al. (2000) have suggested that the early GHRS observations may be affected by Fe ii absorption. We do not find evidence for Fe ii absorption within the C iv lines for any other star, nor for the DAO observations of RW Aur, and so we discard this possibility. Based on the spectroscopic and photometric variability, Gahm et al. (1999) have suggested that a brown dwarf secondary is present in the system, although its role in the dynamics of the primary is uncertain. The COS aperture is larger, suggesting that the changes are due to true variability.

Dramatic changes are also seen in RU Lup, as noted by Herczeg et al. (2006) and France et al. (2012). The C iv line, which is almost absent in the GHRS observations is 4× stronger in the STIS observations. The DAO observations presented here are similar to the latter, although the strength of the extra Si i, C ii, and Fe ii emissions is also variable. The excess emission in the blue wing of the 1548 Å line described in Section 5.1 is also present in the STIS observations.

For all of the multi-epoch CTTSs observations except RW Aur A, Figure 6 (fourth row) shows the path that the line ratio (the optical depth indicator) follows as a function of line luminosity. The changes in the value of the line ratio are not correlated with the changes in the C iv line luminosity, as we observed before.

Most CTTSs C iv line profiles can be decomposed into narrow ( km s⁻¹) and broad ( km s⁻¹) components, with the BC redshifted with respect to the NC in 70% of the CTTSs sample. The fractional contribution to the flux in the NC increases (from ∼20% to 40% on average) and may become more optically thick with increasing accretion rate. Strong narrow components will be present in objects of high accretion rate, but high accretion rate by itself does not guarantee that the C iv lines will have strong NCs.

The component velocities in about 12 out of 23 CTTSs are roughly consistent with predictions of the magnetospheric accretion model, in the sense that V_BC > 0, V_NC > 0 and V_BC ≳ 4V_NC. For most of the 12 the NC velocity seems too large or the BC velocity too small, compared to predictions, although this may be the result of errors in the wavelength calibration. For 11 of the CTTSs the kinematic characteristics of the C iv line cannot be explained by the magnetospheric accretion model: these objects (AK Sco, DF Tau, DK Tau, DR Tau, GM Aur, HN Tau A, IP Tau, RU Lup, SU Aur, T Tau N, and V1190 Sco) have BC velocities smaller than their NC velocities, or one of the velocity components is negative. An examination of the systematic pointing errors which produce offsets in the wavelength scale lead us to conclude that these do not impact the conclusions significantly.

Multi-epoch observations reveal significant changes in morphology in all the lines, from one epoch to the next. The line velocity centroids remain relatively constant in low accretion rate objects such as DF Tau and BP Tau, but the overall line appearance change significantly for high accretion rate CTTSs, like DR Tau. In the case of RU Lup, the observed variability of the line may be consistent with the accretion spot in C iv coming in and out of view.

We find that, except for HN Tau A and RW Aur A, the CTTSs He ii lines are well-centered or slightly redshifted: the average V_Max ignoring these two stars is 7 ± 3 km s⁻¹ (Figure 11). As with C iv, we conclude that the error on the COS wavelength scale for He ii should be smaller than the nominal value. This is because only 7 out of 23 objects (again ignoring HN Tau A and RW Aur A) have V_Max < 0, a situation expected to occur with negligible probability if the wavelength errors are normally distributed around zero km s⁻¹. The difference in redshift between CTTSs and WTTSs is not significant.

The C iv lines are redder than the He ii lines. For the CTTSs  km s⁻¹. For WTTSs, the difference is 7 ± 6 km s⁻¹ (Table 9).

The FWHM for He ii CTTSs range from ∼50 to 400 km s⁻¹, with an average of 96 ± 9 km s⁻¹, which is significantly narrower than in C iv. On the other hand, the FWHM of He ii lines of CTTSs and WTTSs are consistent with being drawn from the same sample, according to the KS test.

The average skewness for the CTTS He ii sample is 0.01 ± 0.01, whereas the skewness of the WTTS sample is −0.01 ± 0.01: the difference between the two populations is not significant. The He ii lines are significantly more symmetric (as measured by the absolute skewness) than the C iv lines.

As with C iv, neither the velocity shift, FWHM, or skewness are correlated with line luminosity, accretion rate or inclination.

Most He ii lines require a narrow and a broad Gaussian component to fit the core and the wings of the line, respectively. The average of the ratio of the He ii NC line luminosity to the total He ii line luminosity is 0.6 and uncorrelated with accretion rate, and it is the same in CTTSs and WTTSs. This large contribution of the NC gives the lines their sharp, peaked appearance. As shown in Figure 7 the BC of the He ii lines span a much broader range than those of C iv, and their distributions are significantly different.

Overall, this Gaussian decomposition confirms the conclusions from the non-parametric analysis. The average value of V_NC for He ii is the same for the CTTS and WTTS ( − = 2 ± 3 km s⁻¹), the C iv CTTS line is redshifted with respect to the He ii line ( − = 20 ± 6 km s⁻¹), and for WTTSs the velocities of He ii and C iv are the same ( − = −2 ± 5 km s⁻¹).

Based both on the non-parametric analysis and the Gaussian decomposition we conclude that the He ii line is comparable in terms of redshift and FWHM in WTTSs and in CTTSs. The line is blueshifted with respect to C iv CTTSs but has the same velocity shift as a C iv WTTS line, and as the NC of the CTTSs in C iv. We discuss this further in Section 8.

The two Si iv lines are separated by 1938 km s⁻¹ (see Figures 14 and 15 in the Appendix). The wavelength of the Si iv doublet members coincides with the bright H₂ lines R(0) 0–5 (1393.7 Å, −9 km s⁻¹ from the 1394 Å line), R(1) 0–5 (43 km s⁻¹ from the 1394 Å line), and P(3) 0–5 (1402.6 Å, −26 km s⁻¹ from the 1403 Å line). The strong narrow line between the two Si iv lines is H₂ P(2) 0–5 (1399.0 Å, 1117 km s⁻¹ away from the 1394 Å Si iv line). Additional H₂ lines (R(2) 0–5, 1395.3 Å; P(1) 0–5, 1396.3 Å; R(11) 2–5, 1399.3 Å) are observed between the doublet members of DF Tau and TW Hya. For this paper, the H₂ lines are considered contaminants in the spectra and we will ignore them. A detailed study of their characteristics has appeared in France et al. (2012). The panels in figures also show the O iv line at 1401.16 Å, sometimes observed as a narrow emission line blueward of the 1403 Å Si iv line (see, for example, DE Tau). We have also indicated the position of the CO 5–0 bandhead (France et al. 2011). In the case of WTTSs, no H₂ lines are observed in the Si iv region and narrow well-centered Si iv lines characterize the emission.

To the extent that the Si iv lines can be seen under the H₂ emission, their shape is similar to that of the C iv lines, as was suggested in Ardila et al. (2002) (see, for example, IP Tau in Figure 14(c)). A notable exception is the WTTS EG Cha, for which the Si iv lines have broad wings that extend beyond ±500 km s⁻¹, not present in C iv. Because the observations were taken in TIMETAG mode, we have time-resolved spectra of the Si iv region that shows sharp increase in the count rate (∼10 × in 300 s) followed by a slow return to quiescence over 1000 s. These observations indicate that the WTTS EG Cha was caught during a stellar flare. The broad line wings are observed only during the flare. A detailed study of this flare event is in preparation.

The two N v lines are separated by 964 km s⁻¹. Figures 14 and 15 also indicate the positions of the H₂ lines R(11) 2–2 (1237.54 Å), P(8) 1–2 (1237.88 Å), and P(11) 1–5 (1240.87 Å). In addition to the H₂ lines, we note the presence of N i absorptions (Herczeg et al. 2005) at 1243.18 Å and 1243.31 Å (1055 km s⁻¹ and 1085 km s⁻¹ from the 1239 Å line).

The two N i lines are clearly seen in the 1243 N v member (the red line) of EP Cha (Figure 14(d)). N i absorptions are observed in most CTTSs (exceptions: AA Tau – N i in emission, DN Tau, EG Cha; Uncertain: AK Sco, CV Cha, and CY Tau). Some objects (e.g., RU Lup) show a clear wind signature in N i, with a wide blueshifted absorption which in this object absorbs most of the N v line.

The 1239 Å line of N v (the blue line, not affected by extra absorption) and the 1550 Å line of C iv (the red line, not affected by H₂) have comparable wing extensions, velocity centroids, and overall shapes. As for C iv and Si iv, the doublet lines of N v should be in a 2:1 ratio if effectively thin although the presence of the N i line makes estimating the ratio impossible. For CTTSs there are other unidentified absorption sources in vicinity of the 1243 Å N v line. These can be seen most easily in the spectra of DE Tau(−32 km s⁻¹), TW Hya (−87 km s⁻¹, −32 km s⁻¹), GM Aur, and V4046 Sgr (50 km s⁻¹) with respect to the rest velocity of the 1243 N v line. It is unclear if these are N i absorption features in a highly structured gas flow or absorptions from a different species.

For the WTTSs both N v lines are copies of each other, and copies of the WTTS Si iv, C iv, and He ii lines. Therefore, the extra absorption in the CTTSs are due to the "classical" T Tauri Star phenomena: they may represent absorption in a velocity-structured wind, or in a disk atmosphere. In particular, the N i feature is likely due to absorption in the CTTSs outflow or disk. We note here the similar excitation N i absorption lines are observed at 1492.62 Å and 1494.67 Å for all stars that show N i absorption in the N v region. A detailed study of the wind signatures in the sample will appear in a future paper.

Figure 12 shows the relationship between the C iv luminosity and the Si iv and N v luminosities.

To measure the flux in the Si iv lines (Table 8), we have integrated each of them between −400 and 400 km s⁻¹, interpolating over the contaminating H₂ lines. The Si iv lines are much broader than the H₂ ones, and at these resolutions can be separated from them, at least in the cases in which the Si iv lines are actually observed. Assuming the H₂ lines are optically thin, the emission in R(0) 0–5 should be 2× weaker than P(2) 0–5 and R(1) 0–5 should be 1.4× weaker than P(3) 0–5, as explained in Ardila et al. (2002). In 11 objects no Si iv line is seen under the H₂ emission. For these stars, we give the 3σ upper limit to the flux, over the same velocity range we used to measure the C iv line in the same star.

Table 8 also lists the flux in the N v lines. Note that the flux in the 1243 Å line is the observed flux, and it has not been corrected for N i absorption or by any of the other absorptions mentioned above.

Figure 12 shows that the C iv and Si iv luminosities of CTTSs are correlated (, ignoring the non-detections and the WTTSs). To understand the nature of this correlation, we perform the following calculation. We assume that each line has a contribution from two regions, a pre- and a post-shock, and that the ratio of the contributions between the two regions is the same in Si iv and in C iv. In other words, . This is likely appropriate if all the emission regions contributing to the line are optically thin. If this is the case, the total luminosity in each line can be shown to be proportional to the post-shock luminosity, and we have

further assuming that the post-shock gas is an optically thin plasma in collisional equilibrium and the emission measure (EM) is the same for both lines over the emitting region. N_x is the fraction of the species in that ionization state, at that temperature, Ab_x is the number abundance of the element and C_x is the collisional excitation rate. We assume a Si/C number abundance ratio of 0.13 for the present day Sun (Grevesse et al. 2007). To calculate the collision rate we use the analytical approximation (Burgess & Tully 1992; Dere et al. 1997):

where ϒ is the thermally averaged collision strength. We use Chianti V. 7.0 (Dere et al. 1997; Landi et al. 2012) to calculate this quantity and assume that the Si iv and the C iv emission come from a plasma at log(T) = 4.9 (K) and log(T) = 5.0 (K), respectively.

The relationship = 0.111 is indicated with a solid line in Figure 12 (top). A model that does not assume the same EM for Si iv and C iv will move the solid line up or down (modestly: a 30% larger emission measure in Si iv than in C iv moves the solid line up by 0.1 dex), but will not change the slope.

To perform the same comparison between C iv and N v requires a correction from the observed values in the latter, because we only fully observe the 1239 Å doublet member, as the 1243 Å doublet member is generally absorbed by N i. Therefore, we assume that the ratio of the flux in the 1239 Å to the 1243 Å N v lines is the same as in the C iv lines and calculate the total flux that would be observed in both N v doublet lines in the absence of N i. This is N v flux plotted against C iv in Figure 12 (bottom). For N v the observed relationship is , ignoring the WTTSs. The expected relationship between the luminosities of both lines is , using N/C abundance of 0.25, and maximum ionization temperature of log(T) = 5.3 (K).

In the case of Si iv the observed linear fit (dashed line in Figure 12) is not consistent with our simple model (solid line, Figure 12). For the current assumptions, V4046 Sgr and TW Hya show a flux deficit in Si iv (or an excess in C iv), while objects such as CV Cha, CY Tau, DX Cha, RU Lup, and RW Aur present a flux excess in Si iv (or a deficit in C iv). In average, the WTTSs are above the line predicted by the model. For N v versus C iv this simple model succeeds in explaining the observed linear correlation.

Based on the absence of Si iii lines as well as the weak Si ii lines in TW Hya, Herczeg et al. (2002) proposed that for this star the silicon has been locked in grains in the disk and does not participate in the accretion. Consistent with this idea, Kastner et al. (2002) and Stelzer & Schmitt (2004) found depletion of Fe and O in TW Hya. For V4046 Sgr, Günther et al. (2006) found a high Ne/O ratio, similar to that which was found in TW Hya and suggesting that a similar process may be at work.

More generally, Drake et al. (2005) argues that the abundance of refractory species reflects the evolutionary status of the circumstellar disk. We see no evidence of this in our sample. The distribution of accretion rates for stars above and below the model line for Si iv/C iv is not significantly different. Furthermore, we find that 80% ± 10% of the non-TD CTTSs are above the Si/C model line, compared to 60% ± 20% of the TD stars. This difference is not significant either. It is, however, noteworthy that our analysis suggests that most CTTSs are Si iv-rich.

Four out of the five WTTSs for which we measure the Si iv luminosity lie above the model line. This is reminiscent of the first ionization potential (FIP) effect seen in the Sun (e.g., Mohan et al. 2000) in which elements in the upper solar atmosphere (the transition region and the corona) show anomalous abundances when compared to the lower atmosphere. Upper atmosphere elements with low FIPs (FIP < 10 eV, like silicon) show abundance excesses by 4×, on average, while high FIP elements (FIP > 10 eV, like carbon and nitrogen) show the same abundances between the photosphere and the corona. Possible explanations for the effect include the gravitational settling of neutrals in the chromospheric plateau (Vauclair & Meyer 1985) and/or diffusion of neutrals driven by electromagnetic forces (Hénoux 1998; Laming 2004). The fact that the luminosity in N v for WTTSs follow the expected relationship with C iv, is consistent with the Si iv excess flux in WTTSs being due to the FIP effect (Wood & Linsky 2011).

However, abundance studies of active, non-accreting, low-mass stars based on X-ray observations, find a mass-dependent inverse FIP-effect abundance pattern in Fe/Ne (see Güdel 2007; Testa 2010 for reviews). If this applies to WTTSs, we would expect to see a silicon deficit in them, as compared to carbon. If we were to move the model (solid) line from the top panel of Figure 12 upward, in order to make the WTTSs silicon-poor, most CTTSs would become silicon-poor, suggesting that the disk grain evolution observed in TW Hya starts at younger ages. This is speculative and requires further study. We note here that Telleschi et al. (2007) find a mass-dependent inverse FIP-effect abundance pattern also in CTTSs.

These statements are only intended to identify candidates for further study. What we can tell from our observations is that there is considerable scatter of CTTSs to both sides of the linear relation between and . EM analyses for each star are necessary before significant conclusions can be drawn from this sample regarding abundances. These are possible with the DAO data set, but they are beyond the scope of this paper.

If the lines originate primarily in an accretion shock, one can make at least four predictions: (1) the C iv and He ii line profiles should be redshifted, with the latter having smaller velocities, (2) the line emission should be well localized on the stellar surface, (3) if increases in the accretion rate are due to density increases, the post-shock gas emission will be quenched, due to burying of the post-shock column, and (4) the velocity of the narrow and broad components of the C iv lines should be related as V_BC ≳ 4 V_NC.

Indeed, we observe C iv and He ii redshifted, with the former redshifted more than the latter. This, however, is not a very constraining prediction. Linsky et al. (2012) have shown that C iv lines in dwarfs are redshifted by an amount correlated with their rotation period and Ayres et al. (1983) have shown that the C iv lines in late-type giants are redshifted with respect to the He ii lines by variable amounts of up to ∼20 km s⁻¹. So, for non-CTTSs gas flows in the upper stellar atmosphere may result in line shifts comparable to those we observe in our sample.

The predictions regarding the localization of the emitting region in the stellar surface are difficult to test with this data set, because, as we have argued in Section 5.2.5, the sample is not large enough for the inclination to serve as a discriminator of the line origin. On the other hand, there is a large body of evidence (see the references in the introduction) suggesting that the accretion continuum and the (optical) line emission are well localized on the stellar surface. No unequivocal rotational modulation of the lines studied here has been reported in the literature, but this may be because the available data set does not provide enough rotational coverage. The change in line flux observed among the different RU Lup and DR Tau epochs (Figure 10) may result from rotational modulation.

If we assume that the NC of C iv is due to post-shock emission, we would naively expect its importance to diminish for high accretion rates as the post-shock is buried in the photosphere (Drake 2005). However, we observe the opposite, with the NC flux increasing with respect to the BC flux as the accretion rate increases. One possible explanation is that increased accretion produces larger accretion areas on the star rather than much higher densities in the accretion column. This has been shown for BP Tau by Ardila & Basri (2000). Larger areas provide more escape paths for the photons emitted from a buried column. Area coverages as small as 0.1% of the stellar surface of a 2 R_☉ star have a radius of ∼10⁵ km, which is larger than the deepest likely burials (Sacco et al. 2010).

A complementary insight comes from X-ray observations. Models that assume a uniform density accretion column also predict quenching of the X-ray flux, due to absorption of X-rays in the stellar layers, for accretion rates as small as a few times 10⁻¹⁰  M_☉ yr⁻¹ (Sacco et al. 2010). However, models by Romanova et al. (2004) indicate that the accretion column is likely non-uniform in density and the denser core may be surrounded by a slower-moving, lower-density region. Sacco et al. (2010) argue that most of the observed X-ray post-shock flux should come from this low-density region. In addition, post-shock columns are believed to be unstable to density perturbations, and should collapse on timescales of minutes (Sacco et al. 2008). The lack of observed periodicities in the X-ray fluxes indicate that multiple incoherent columns should be present. Orlando et al. (2010) also conclude that the presence of multiple columns with different densities is necessary to explain the low accretion rates derived from X-ray observations.

In addition, models of low-resolution CTTSs spectra from the near-UV to the infrared indicate that the observed accretion continuum is consistent with the presence of multiple accretion spots (Ingleby et al. 2013).

In summary, we observe that the C iv flux does not decrease with accretion, which suggests that if burying is occurring, it does not affect the observed flux substantially. This may be because the aspect ratio of the accretion spots is such that the post-shock radiation can escape without interacting with the photosphere, and/or that the flux we observe is emitted from unburied low-density edges of the accretion column, and/or that multiple columns with different densities and buried by different amounts are the source of the emission.

8.1.1. He ii Emission in the Pre-shock Gas

8.1.2. Kinematic Predictions of the Accretion Shock Model

8.1.3. A Contribution from the Stellar Transition Region?

We have argued that if the post-shock radiation ionizes material far away from the accretion spot, we may end up with BCs having small or even negative velocities. On the other hand, the objects for which we observe a negative velocity in the NC of C iv, or in the overall profile, present considerable challenges. These are AK Sco, DK Tau, HN Tau A, T Tau N, and RW Aur A. The He ii line matches the C iv line for RW Aur A and HN Tau A (Figure 17) but is centered at velocities closer to zero than C iv for the other stars.

If we observe a CTTS for which the accretion stream is moving toward us, the velocity components would be negative. This would be the case, for example, if we observe the stream below the disk through the inner truncation hole of the accretion disk. The only case in which this is a possibility is T Tau N as this is the only target for which V_NC < 0, V_BC < 0, and ∣V_NC∣ < ∣V_BC∣. This requires an inclination close to face-on, which the system has, and perhaps a large difference between the stellar rotation axis and the magnetic field axis.

Outflows present a more likely explanation for the blueshifted profiles. Outflow phenomena are common in CTTSs and high speed shocks between the jet material and the ISM may result in C iv emission. The C iv blueshifted emission in DG Tau (Ardila et al. 2002) is clearly related to the beautiful outflow imaged with HST/NICMOS (Padgett et al. 1999) and the C iv emission is likely produced by shocks in the jet (Schneider et al. 2013). In order to generate high temperature (>10⁵ K) gas via an outflow shock, velocities larger than 100 km s⁻¹ in the strong shock limit are necessary (Günther & Schmitt 2008), resulting in post-shock (observed) C iv velocities >25 km s⁻¹. The five objects we are considering have absolute NC velocities or velocities at maximum flux larger than this value. Therefore, at least energetically, it is possible for the emission to be produced by a shock in the outflow. Both HN Tau A and RW Aur A are known to have outflows (Hirth et al. 1994; Hartigan et al. 1995), and velocities in the approaching and receding jets that are comparable to the ones we observe in C iv and H₂ (Melnikov et al. 2009; Coffey et al. 2012).

For objects such as AK Sco and DK Tau the accretion shock and the outflow regions may both be contributing to the emission. For HN Tau A and RW Aur, if we accept that the observed profiles originate primarily in an outflow, the lack of accretion shock emission becomes puzzling. These are high-accretion rate objects, and perhaps in these conditions the low-density region in the periphery of the accretion column is not present, and so the shock is truly buried. Or maybe we are observing the objects in a rotational phase such that the accretion spot is away from us. The GHRS observations of RW Aur (red trace, Figure 10) show the presence of additional emission components to the red of the nominal C iv lines, which may be due to the accretion spot.

Overall, the relationship between H₂ asymmetries and hot line shapes remains to be fully explored. The analysis of DAO data by France et al. (2012) shows that some H₂ emission lines in DK Tau, ET Cha (RECX 15), HN Tau, IP Tau, RU Lup, RW Aur, and V1079 Tau (LkCa 15) are asymmetric, presenting redshifted peaks and low-level emission to the blue of the profiles. For HN Tau A and RW Aur, the peak of the H₂ line R(6) (1–8), at 1556.87 Å is shifted +19 km s⁻¹ and +88 km s⁻¹ respectively, from the stellar rest frame, larger than would be expected from pointing errors alone and suggests that outflows that are fast enough in C iv are accompanied by H₂ flows away from the observer. For RW Aur A, we observe that the (redshifted) H₂ emission covers the (blueshifted) C iv emission (Figure 5).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

If the gas heating occurs very close to the star, or if the wind is launched hot, one would expect to observe P-Cygni-like profiles in the hot gas lines. Dupree et al. (2005) argued that the asymmetric shape of the O vi profiles of TW Hya is the result of a hot wind in the star. Because the two C iv lines are close to each other, blueshifted wind absorption in the 1550 Å line will decrease emission in the red wing of the 1548 Å line. Johns-Krull & Herczeg (2007) compared both C iv lines and concluded that their similarity, as well as the absence of absorption below the local continuum (as seen in many neutral and singly ionized lines), suggest that a hot wind is not present in the case of TW Hya.

Within our sample, DX Cha is the only object that may have a high temperature wind, as it shows a deficit in the red wing of the 1548 Å line for C iv, compared to the red wing of the 1550 Å line and a very sharp blue cutoff in the C iv and the Si iv lines. A blueshifted absorption is seen in the 1548 Å C iv line, suggesting outflow speeds of up to 400 km s⁻¹. A blueshifted absorption is also seen in the 1403 Å line of Si iv (the red doublet member). However, note that there is an O iv emission line at −340 km s⁻¹ of the line, and its presence may give the illusion of a wind absorption. The narrow H₂ lines observed in the Si iv profiles suggest that the putative wind is collimated or inhomogeneous, as the H₂ lines are not absorbed by the wind. The difference between both members of N v is not due to a wind but to absorption of the 1243 Å line by circumstellar absorbers.

DX Cha also shows absorption features in the region near He ii, although it is not clear that they are related to the outflows. Lower temperature outflows, like those observed in other CTTSs (Herczeg et al. 2005) are also observed in the DAO spectra of DX Cha: the Si ii λ1526.71 and 1533.43 (not shown here) present very clear P-Cygni profile with wind absorption up to 600 km s⁻¹ from the star. Further examination of the outflows will appear in a future paper.

In this analysis, we have considered DX Cha as one more member of the overall sample, in order to contrast the characteristics of Herbig Ae stars with those of CTTSs. It has the largest mass (∼2.2 M_☉, Böhm et al. 2004), and earliest spectral type (A7.5) in the sample.

DX Cha is a spectroscopic system with a K3 secondary and an average separation of ∼0.15 AU between components (Böhm et al. 2004). Even at these small separations, small circumstellar disks may be present, in addition to the circumbinary disk (de Val-Borro et al. 2011). To add to the complexity of the system, observations by Tatulli et al. (2007) are consistent with the presence of a wind launched in the 0.5 AU region of the disk. Testa et al. (2008) show that two different temperature plasmas are responsible for the X-ray emission. The overall flux is emitted from a relatively high-density region and dominated by the primary star. They argue that the hot component is created in the companion's corona, while the low-temperature component originates in an accretion shock.

In the observations presented here, DX Cha does appear peculiar when compared to the CTTSs. The system has the largest Si iv luminosity of the sample, and the second largest C iv luminosity. The 1550 Å C iv line has the second largest FWHM of the sample, and as we have noted, the C iv lines are not alike in shape, suggesting extra emission or absorption in one of the members. The Si iv lines are unlike any other Si iv or C iv lines in shape, in that they show a very sharp, blue cutoff. As indicated above, DX Cha is also the only clear candidate in the sample for the presence of a hot wind.

For DX Cha we observe a strong continuum in the He ii region and no clear emission line. If in Figure 14 we identify the depression at +60 km s⁻¹ in the He ii panel of DX Cha as He ii in absorption, this would be the largest redshift of any He ii line in our sample, and larger than the 11.5 km s⁻¹ observed for the peak of C iv. The absence of the He ii line in emission is put in context by Calvet et al. (2004), who present low-resolution UV spectra of accreting objects with masses comparable to DX Cha, all of which show He ii in emission. For all of them /L_☉ ∼ 10⁻⁵ to 10^−4.5, comparable to the luminosities of other stars observed here. This makes the absence of an He ii line a mystery.

Are these characteristics the continuation of the standard accretion rate phenomena to larger masses, the consequence of the close companion and/or an outflow, or new phenomena related to the weak magnetic fields associated with Herbig Ae objects? A larger sample of high-resolution UV spectra of higher mass objects is required to answer these questions.

• 1.  
  The most common (50%) C iv line morphology is that of a strong, narrow emission component together with a weaker, redshifted, BC (see, for example, BP Tau; Figure 1). In general, the C iv CTTSs lines are broad, with significant emission within ±400 km s⁻¹ of the lines. When compared to the WTTSs lines, the C iv lines in CTTSs are skewed to the red, broader (FWHM  ∼  200 km s⁻¹ for CTTSs, FWHM ∼100 km s⁻¹ for WTTSs), and more redshifted (20 km s⁻¹ for CTTSs, 10 km s⁻¹ for WTTSs); see Table 9. The 1548 Å member of the C iv doublet is sometimes contaminated by the H₂ R(3)1–8 line and by Si i, C ii, and Fe ii emission lines.
• 2.  
  Overall, the C iv, Si iv, and N v lines in CTTSs all have similar shapes. The 1243 Å N v line is strongly absorbed by circumstellar N i and both lines of the Si iv doublet are strongly affected by H₂ emission (Figures 14 and 15). We do no detect H₂ emission within ±400 km s⁻¹ of the hot gas lines for WTTSs.
• 3.  
  In general, the He ii CTTSs lines are symmetric and narrow, with FWHM  ∼  100 km s⁻¹. The FWHM and redshifts are comparable to the same values in WTTSs. The He ii lines also have the same FWHM as the NC of C iv in CTTSs. They are less redshifted than the CTTSs C iv lines, by ∼10 km s⁻¹, but have the same redshift as the WTTSs. A comparison of the Gaussian parameters for C iv and He ii is shown in Figure 7.

• 1.  
  We confirm that the C iv line luminosities are correlated with accretion luminosities and accretion rates, although the exact correlation depends on the set of extinctions that is adopted (Figure 4). C iv and He ii luminosities are also correlated with each other, and the C iv/He ii luminosity ratio is a factor of three to four times larger in CTTSs than in WTTS. This confirms that the ratio depends crucially on accretion phenomena and it is not an intrinsic property of the stellar atmosphere.
• 2.  
  We do not find any significant correlation between inclination and C iv V_Max, FWHM, or skewness, or between inclination and any of the Gaussian parameters for the C iv decomposition (Figure 6). We conclude that the data set is not large enough for inclination to be a strong descriptor in the sample. The variability observed in the C iv lines of RU Lup and DR Tau (Figure 10) may be consistent with the C iv accretion spot coming in and out of view.
• 3.  
  We do not find correlations among V_Max, FWHM, or skewness, with each other or with accretion rate or line luminosity. On the same token, we do not find correlations between the width of the Gaussian components and accretion rate or line luminosity.
• 4.  
  The ratio between the flux in the 1548 Å C iv line to the flux in the 1550 Å C iv line is a measure of the opacity of the emitting region. For CTTSs, 70% (19/29) of the stars have blue-to-red C iv line ratios consistent with optically thin or effectively thin emitting regions. DF Tau is anomalous, with a blue-to-red C iv line ratio close to 3, indicating either emission in the 1548 Å line or absorption on the 1550 Å one.
• 5.  
  We find that the contribution fraction of the NC to the C iv line flux in CTTSs increases with accretion rate, from ∼20% up to ∼80%, for the range of accretion rates considered here (Section 5.2.4, Figure 8). This suggests that for some stars the region responsible for the BC becomes optically thick with accretion rate. As a response to the accretion process, the C iv lines develop a BC first.
• 6.  
  We show resolved multi-epoch observations in C iv for six CTTSs: BP Tau, DF Tau, DR Tau, RU Lup, RW Aur A, and T Tau N. The kind of variability observed is different for each star, but we confirm that the NC of BP Tau, DF Tau, and RU Lup increase with an increase in the line flux (Section 5.4, Figure 10).

• 1.  
  The Si iv and N v line luminosities are correlated with the C iv line luminosities (Figure 12). The relationship between Si iv and C iv shows large scatter with respect to a linear relationship. If we model the emitting region as an optically thin plasma in coronal ionization equilibrium, we conclude that TW Hya and V4046 Sgr show evidence of silicon depletion with respect to carbon, as has already been noted in previous papers (Herczeg et al. 2002; Kastner et al. 2002; Stelzer & Schmitt 2004; Günther et al. 2006). Other stars (AA Tau, DF Tau, GM Aur, and V1190 Sco) may also be silicon-poor, while CV Cha, DX Cha, RU Lup, and RW Aur may be silicon-rich. The relationship between N v and C iv shows significantly less scatter.
• 2.  
  For WTTSs, we observe silicon to be generally more abundant than expected within our simple model. This is reminiscent of the FIP effect. However, an inverse FIP has been observed in the Fe/Ne abundance ratios of active, low-mass, non-accreting stars, which should behave similarly to WTTSs (Güdel 2007). It is unclear whether the discrepancy is due to the simplicity of the assumed model or if the Si/C ratios in WTTSs behave differently than the Fe/Ne ratios in active stars. This points to the need for a more sophisticated model of the relationship between Si iv, C iv, and N v both in WTTSs and CTTSs.

• 1.  
  Our sample covers a range of ages from ∼2 Myr to ∼10 Myr, and a wide range in disk evolutionary state. We do no detect changes in any of the measured quantities (line luminosities, FWHM, velocity centroids, etc.) with age. We do not find systematic differences in any quantity considered here between the CTTS subsample with TDs and the whole CTTS sample.
• 2.  
  We do not find any significant difference in the accretion rates of the Si iv-rich objects compared to those Si iv-poor, nor in their disk characteristics.

• 1.  
  We find no evidence for a decrease in any of the Gaussian components of C iv that could be interpreted as burying of the accretion column in the stellar photosphere due to the ram pressure of the accretion flow (Drake 2005; Sacco et al. 2010). In particular, we do not observe a decrease in the strength of the NC as a function of accretion, but exactly the opposite. In the context of the magnetospheric accretion model, we interpret this as an argument in favor of a line origin in either (i) an accretion spot with an aspect ratio such that the post-shock photons can escape the columns (f  >  0.001), and/or (ii) the low-density, slow-moving periphery of an inhomogeneous accretion column (Romanova et al. 2004), and/or (iii) multiple, uncorrelated accretion columns of different densities (Sacco et al. 2010; Orlando et al. 2010; Ingleby et al. 2013), which appear as the result of increasing accretion rate.
• 2.  
  The accretion shock model predicts that the velocity of the post-shock emission should be 4 times smaller than the velocity of the pre-shock gas emission. If we identify the broad Gaussian component with the pre-shock emission and the NC with the post-shock emission, we find that V_BC ≳ 4 V_NC is the case in only 2 out of 22 CTTSs. Accounting for possible pointing errors, 10 more objects could be within the accretion shock predictions (see Figure 9). For six systems the Gaussian decomposition is impossible (DX Cha, RW Aur) or requires only one Gaussian component (CV Cha, ET Cha, DM Tau, and UX Tau A).
• 3.  
  When compared to the predicted infall velocities, the measured V_BC values of these objects imply large (∼90°) inclinations of the flow with respect to the line of sight. Because this is unlikely in general, we suggest that the accretion flow responsible for the emission may not be at the free-fall velocity (for example, if the emission comes from the slow-region at the edge of the column). Another alternative is that we are observing emission from multiple columns with different lines of sight, for which the ratio of pre- to post-shock emission is not the same. Finally, radiation from the post-shock gas may be ionizing regions far from it, which, because of the shape of the magnetic field channeling the flow, may have line-of-sight velocities unrelated to the post-shock ones. Our models of the pre-shock gas show that pre-shock sizes can range from a few tens of kilometers to ∼1 R_☉ for typical parameters. These concepts may also explain cases for which the NC velocity is positive, but the BC velocity is small or negative (DF Tau, DR Tau, GM Aur, IP Tau, RU Lup, SU Aur, and V1190 Sco).
• 4.  
  The accretion shock model predicts that, because the formation temperature of He ii is lower than that of C iv, the post-shock line emission from the former should have a lower velocity than the latter. We confirm this, by identifying the NC emission with the post-shock gas emission. However, we note that gas flows in the stellar transition region may result in similar velocity offsets between C iv and He ii.
• 5.  
  Observationally, the amount of flux in the BC of the He ii line is small compared to the C iv line. We model the pre-shock column and conclude that for typical parameters, it can produce line ratios between C iv and He ii within the observed range. In other words, the weakness of the BC component in He ii is consistent with its origin in the pre-shock gas.
• 6.  
  Overall, we favor the origin of the emission lines in an accretion shock, but we cannot rule out that the true transition region is contributing part of the NC flux.

We find three different types of profiles that show evidence of outflows (Section 8.2).

• 1.  
  For HN Tau A and RW Aur A, most of the C iv and He ii line fluxes are blueshifted and the peaks of the H₂ lines are redshifted (France et al. 2012). The C iv and He ii emission in this case should be produced by shocks within outflow jets. For these stars we may be observing the opposite sides of an outflow simultaneously (Section 5.1, Figure 5).
• 2.  
  For the CTTSs AK Sco, DK Tau, T Tau N, we find that the NC velocity is negative. These represent less extreme examples of objects such as HN Tau A and RW Aur, although the observed velocities are high enough that internal shocks in the outflow can result in hot gas emission. For these stars we do not observe any redshifted NC emission that would correspond to the post-shock.
• 3.  
  Within our sample, DX Cha is the only candidate for a high temperature wind, as it shows a deficit in the red wing of the 1548 Å line for C iv, compared to the red wing of the 1550 Å line, and blue absorption in 1548 Å C iv line (Figure 13). Lower temperature outflows in Si ii are also observed in DX Cha.

Certain well-known objects are tagged as peculiar in this work. In addition to DX Cha, HN Tau, and RW Aur, mentioned before, the most peculiar are the following

• 1.  
  AK Sco. Even when accounting for extra emission lines, the C iv lines are different form each other, perhaps as a result of it being a spectroscopic binary. It may be Si-rich.
• 2.  
  DF Tau. The 1548-to-1550 ratio is too large. The BC velocity is small compared to expectations of the accretion model.
• 3.  
  DK Tau. The BC velocity is negative, the C iv profile is double-peaked, may be Si-rich.
• 4.  
  DR Tau. For C iv, the BC velocity is small compared to expectations, the thermalization depth for the NC is small, but the BC flux ratio is anomalously high. The C iv lines are double-peaked. Multi-epoch observations show either absorption within 150 km s⁻¹ of the line center, or evidence for rotational modulation in the profile. It may be Si-rich.
• 5.  
  RU Lup. The red wings of the C iv lines are different from each other, the BC velocity, the thermalization depth for the NC is small, and it may be Si-rich. In addition, some of the H₂ lines present a low level blueshifted emission. As in DR Tau, we may be observing rotational modulation of the C iv profile in multi-epoch observations.

High-spectral resolution UV observations provide a crucial piece of the puzzle posed by young stellar evolution. They sample a much hotter plasma than optical observations and complement and enhance X-ray data. However, the data considered here has two important limitations. One is the lack of time-domain information, which makes it difficult to understand what is the average behavior of a given target. The other is the small range of accretion rates well-covered by objects. Future observational work should focus on resolving these limitations. Even with these limitations, it is clear that with this work we have only scratched the surface of this magnificent data set.