HIGH-RESOLUTION ULTRAVIOLET RADIATION FIELDS OF CLASSICAL T TAURI STARS^*

By bringing together the reconstructed Lyα emission profiles, the "H₂ cleaned" hot gas emission lines, and measurements of the uncontaminated FUV continuum, we can explore the relations between these different components and empirically constrain the origin of the strong Lyα and FUV continuum emission in these sources. The prevailing picture for the formation of the UV continuum excess is that it arises from ionized gas (T_cont ∼ 1.5–3 × 10⁴ K) in an optically thin pre-shock region at the base of the accretion column (Calvet & Gullbring 1998; Costa et al. 2000). In this picture, the FUV continuum may be an extension of the Balmer continuum used for accretion rate determination in the NUV (see, e.g., Ingleby et al. 2013); and the connection between the NUV and FUV continua has been explored previously for DF Tau (France et al. 2011b). In the following subsection, we show that for a subsample of six of our targets with nearly simultaneous FUV and NUV observations, the FUV and NUV continua are characterized by a significantly different slope. This demonstrates that the connection proposed by France et al. (2011b) was most likely an artifact of comparing non-simultaneous observations, although this cannot be conclusively demonstrated for all of our targets in our sample without contemporaneous FUV and NUV data.

We find that while some of our sources display a monotonically decreasing FUV continuum towards shorter wavelengths (BP Tau, DF Tau, RU Lup; see also Herczeg et al. 2005), many of the FUV continua are flat across the FUV band (GM Aur, RECX-15, UX Tau, V4046 Sgr). This suggests that multiple physical processes likely contribute in the complex emitting region near the stellar photosphere and the inner accretion disk. Our aim here is to provide the data for use in detailed models of disk chemistry and evolution, and therefore we do not attempt to present a comprehensive model for the continuum emission region.

The hot gas emissions from C iv and N v have been long considered to be related to accretion processes (Johns-Krull et al. 2000). A detailed spectral line analysis of a larger sample of COS and STIS observations has shown that these lines are likely produced by hot gas in accretion spots near the stellar photosphere, the edges of accretion columns, or multiple accretion columns of varying densities (Ardila et al. 2013). The He ii lines however appear to be dominated by a combination of emission from the magnetically active stellar atmosphere and the pre-shock region. Lyα is presumably generated at several places in the near-star environment; however, we are not aware of any previous work that explores the observational connection between Lyα luminosity and accretion processes (see Muzerolle et al. 2001 and Kurosawa et al. 2006 for a detailed description of Balmer lines).

In order to empirically constrain the relationship between Lyα and the FUV continuum and accretion processes in the following subsections, we assume that the integrated C iv flux is a proxy for the mass accretion rate (Johns-Krull et al. 2000; Ardila et al. 2013). It is critical to have this quasi-simultaneous accretion diagnostic because accretion rates and UV fluxes from CTTSs are known to vary by factors of ∼2 on timescales of months to years (Valenti et al. 2000). Therefore, accretion rates in the literature will not necessarily be well correlated with the UV emission at the time of our observations. In Figure 5, top left, we demonstrate the tight correlation between the accretion-dominated C iv and N v flux for our stars (see also the larger sample from Ardila et al. 2013). For all of the correlations shown in Figure 5, we present the Spearman rank correlation coefficient (ρ) and the likelihood that this population is drawn from a random sample (i.e., not correlated; n). In general, |ρ| > 0.5 and n < 0.05 indicates that a real correlation exists between the two quantities. For instance, [ρ, n] = [0.81, 1.3 × 10⁻⁴] for the plot of N v and C iv, indicating a strong correlation, as expected.

Zoom In Zoom Out Reset image size

4.1.1. FUV Continuum

The FUV continuum flux (in units of erg cm⁻² s⁻¹), evaluated at 1 AU from the central star, is computed by integrating the polynomial fit to the binned intra-emission-line HST spectra (Appendix A.3) from 912 to 1650 Å (Figures 2 and 11–13). The FUV continuum is shown to be tightly correlated with the C iv emission (Figure 5, top right, [ρ, n] = [0.93, 1.5 × 10⁻⁷]). We interpret this as strong evidence that the CTTS FUV continuum is generated by accretion processes, either directly or powered by accretion luminosity. However, it may originate in a spatially separate region from the NUV Balmer continuum, as mentioned above.

In Figure 6, we compare data for six of our targets (DM Tau, DR Tau, GM Aur, HN Tau, RECX-11, and RECX-15) with FUV and NUV observations separated by less than a few hours. Linear fits to the FUV (excluding the molecular quasi-continuum) and NUV regimes are shown overplotted as solid orange and dashed pink lines, respectively. The observed FUV continuum is shown to be brighter than predicted from a simple extrapolation of the NUV continuum, and has a significantly shallower slope (see also Herczeg et al. 2004, 2005). We note that the S/Ns of individual binned data points are relatively high, demonstrating that the excess FUV continuum flux is statistically significant in all cases. The average S/N per binned continuum point ranges from 2.5 to 14.8 from 1140 to 1340 Å and 0.9–17.0 from 1660 to 1740 Å (Table 4). While it seems clear that the FUV continuum is related to accretion processes, these differences argue that the FUV and NUV continua may be formed in spatially distinct regions near the interaction of the accretion streams and the stellar photosphere. Recent models of the NUV Balmer excess have incorporated multiple accretion components with a range of energy fluxes and densities (e.g., Ingleby et al. 2013), and future work may be able to connect the FUV and NUV continua by incorporating these different components.

Zoom In Zoom Out Reset image size

Our analysis finds that the FUV continuum extends to λ < 1216 Å in all cases, ruling out the possibility that this flux originates from H i two-photon (2s² S_1/2–1s² S_1/2) emission. Furthermore, H i two-photon emission is suppressed in high-density regions (n_H > 1.5 × 10⁴ cm⁻³), consistent with the scenario in which this emission is the high-frequency extension of the Balmer continuum generated in the relatively dense pre-shock region at the base of the accretion column (Calvet & Gullbring 1998). A future space-borne instrument capable of simultaneous FUV and NUV spectroscopy or photometry could provide more insight on the potential link between the FUV continuum and the NUV Balmer continuum in CTTSs.

4.1.2. Lyα Line Emission

The intrinsic Lyα flux (in units of erg cm⁻² s⁻¹), evaluated at 1 AU from the central star, is computed by integrating the reconstructed Lyα line profile from 1211 to 1221 Å. We note that we take the intrinsic reconstructed line profile as opposed to the line profile seen by the H₂ molecules at the disk surface after absorption by atomic hydrogen in the outflow (see Figure 8). A discussion of the Lyα profiles including the neutral outflow absorption component is presented by Schindhelm et al. (2012b), and the uncertainty on the intrinsic Lyα flux used here is ∼10%–30%, depending on the detectability and S/N of the H₂ fluorescence lines used for the reconstruction (France et al. 2012b).

The middle panels of Figure 5 show the correlation between Lyα and the FUV continuum ([ρ, n] = [0.71, 1.9 × 10⁻³]) and the C iv emission ([ρ, n] = [0.77, 4.8 × 10⁻⁴]). The Lyα flux is observed to correlate with both quantities, but with less significance than the correlations between C iv and N v, and C iv and the FUV continuum. This is possibly confirmed by another representation on the bottom panels of Figure 5, where the ratio of Lyα/FUV continuum is plotted against the published mass accretion rates (left, [ρ, n] = [−0.31, 0.24]) and the C iv fluxes (right, [ρ, n] = [−0.50, 0.047]). The plots are qualitatively similar; Figure 5 (bottom left) does not show a significant anti-correlation between Lyα/FUV continuum and the mass-accretion rate, Lyα/FUV continuum and C iv show a weak anti-correlation suggesting that the FUV continuum is more tightly correlated with the accretion than the Lyα emission.

The production of Lyα is still likely dominated by accretion processes. Figure 7 shows the observed Lyα emission lines from two representative CTTSs from this work (GM Aur and BP Tau) and two WTTSs (TWA 7 and LkCa19). In all cases, the CTTS Lyα lines are much stronger and broader than WTTS Lyα profiles for a similar age and stellar mass. All CTTSs display broad Lyα wings to v_Lyα > 500 km s⁻¹, and many show Lyα emission wings in excess of ±1000 km s⁻¹. There may be several physical reasons that the correlation between Lyα and C iv is weaker than between the FUV continuum and C iv, including uncertainties in the Lyα reconstruction process, particularly when there are multiple sources of Lyα photons in the systems (stellar atmosphere, outflows and/or disk surface) and resonant scattering of Lyα photons in the gas-rich CTTS circumstellar environment.

Zoom In Zoom Out Reset image size

In order to understand the UV-driven photochemistry of the protoplanetary environment, one needs constraints on both the absolute flux and the shape of the illuminating radiation field. The absolute flux of the radiation environment determines the photoexcitation and photodissociation rates for disk molecules and the SED determines the primary scattering and absorption mechanisms for diffusing FUV photons through the disk (Bergin et al. 2003; Bethell & Bergin 2011; Fogel et al. 2011). In Table 2, we give the total, reddening-corrected flux from our targets (evaluated at 1 AU to enable inter-comparison) integrated over the 912–1650 Å bandpass. In the second column, this quantity is given in units of the average interstellar radiation field from Habing (1968). The average incident flux from the accreting central star is ∼10⁷ times the contribution from scattered OB starlight at 1 AU from the central star, although the shapes of the CTTS radiation field and the interstellar field are very different.

By spectrally resolving the various components of the FUV field and including the reconstructed Lyα emission, we can provide an inventory of the FUV luminosity sources in ∼1–10 Myr CTTSs. Table 2 shows the relative contribution of each component to the total FUV irradiance from each target. As argued by previous authors (Bergin et al. 2003; Herczeg et al. 2004; Schindhelm et al. 2012b), Lyα dominates the FUV radiation output from these sources, with an average fractional luminosity of 88.1% ± 7.3%. ⁹ The FUV continuum is the second largest energy source, with 8.4% ± 5.2%. The contribution from the C iv λλ1548, 1550 Å doublet is 2.1% ± 1.6%, and the remaining hot gas emission lines make up <5% in all cases, although the exact contribution is rather uncertain because we assumed the TW Hya C iii and O vi fluxes, scaled by the relative C iv fluxes, for the remaining targets. Lower ionization species likely contribute <1% to the total field strength (see Appendix A.2).

The primary transitions leading to dissociation of H₂ and CO, through the H₂ B–X and C–X band systems (radiatively dissociative) and the CO E–X band system (predissociative), are located at λ ⩽ 1110 Å. The λ ⩽ 1110 Å radiation field also plays an important role in the production of photo-electrons that heat the gas in the disk surface layers (Pedersen & Gómez de Castro 2011). The 912–1110 Å bandpass contains between 0.1% and 1.8% of the total FUV flux, with an average (excluding DR Tau) of 0.5 ± 0.4%. On average, the 912–1110 Å CTTS radiation field has a flux of ∼150 erg cm⁻² s⁻¹ at 1 AU from the central star.

In this subsection, we quantify the impact of low-resolution data on the determination of the total FUV flux from CTTSs. Previous large FUV spectroscopic samples of CTTSs were acquired at low-spectral resolution (R ∼ 80–1000; Johns-Krull et al. 2000; Valenti et al. 2000; Ingleby et al. 2009, 2011b; Yang et al. 2012). Most of these observations did not have access to Lyα directly due to instrumental bandpass limitations and/or contamination by geocoronal emission. Furthermore, low-resolution prevents the necessary measurements of individual H₂ or CO fluorescent emission lines necessary for reconstructing the Lyα emission profile. Consequently, most quoted measurements of the FUV radiation field strength are integrated over ≈ 1230–1700 Å (Ingleby et al. 2011b; Yang et al. 2012), excluding the dominant FUV flux source, H i Lyα. Low-resolution data are also insufficient for separating the accretion-powered FUV continuum from the wealth of narrow molecular emission lines and the molecular continuum (the "1600 Å Bump"). In order to quantify the differences between high- and low-resolution CTTS FUV radiation fields, data for 10 objects with previous low-resolution FUV observations from HST were retrieved from the Yang et al. (2012) atlas of low-resolution T Tauri star spectra hosted at MAST.

The average flux ratio between the observed low-resolution 1235–1700 Å flux from the Yang et al. (2012) atlas (F(1235–1700)_LR) and the 1235–1700 Å fluxes from this work (F(1235–1700)_fit), F(1235–1700)_LR/F(1235–1700)_fit, is 1.41 ± 0.52 (Table 3). The average being greater than unity most likely reflects the inclusion of Lyα-pumped H₂ and CO fluorescence in low-resolution data and the scatter is most likely the time-variability that is characteristic of these sources (recall these data were acquired over a multi-year baseline; and see, e.g., Giovannelli et al. 1995; Gómez de Castro & Fernández 1996). The two sources which display the largest FUV variability are DM Tau and SU Aur, showing a factor of ∼3 variability between the two epochs. If we screen sources farther than 3σ from the mean of the distribution, the average drops to F(1235–1700)_LR/F(1235–1700)_fit = 1.06 ± 0.34.

Table 3. Comparison to Low-resolution CTTS Spectra

Target	HST Mode a	b	c
AATAU	ACS	1.18	0.07
BPTAU	STIS	0.70	0.21
DETAU	ACS	1.68	0.27
DMTAU	STIS	2.73	0.31
DRTAU	ACS	0.77	0.55
GMAUR	ACS	0.99	0.14
GMAUR	STIS	0.96	0.14
HNTAU	ACS	1.04	0.17
LKCA15	STIS	1.43	0.19
SUAUR	STIS	3.24	0.60
V4046SGR	ACS	0.76	0.05
Average		1.41 ± 0.52	0.25 ± 0.17

Notes. ^a HST low-resolution mode used for comparison. Low-resolution data taken from Yang et al. (2012). ACS refers to ACS/SBC PR130L mode, STIS refers to STIS G140L. ^bRatio of the observed low-resolution spectrum from 1235 to 1700 Å to the extracted, intrinsic CTTS spectrum from 1235 to 1700 Å. ^cRatio of the observed low-resolution spectrum from 1235 to 1700 Å to the extracted, intrinsic CTTS spectrum from 1150 to 1700 Å, including the reconstructed Lyα emission line.

Download table as:  ASCIITypeset image

Comparing the ratio of the observed low-resolution 1235–1700 Å flux from the Yang et al. (2012) atlas and our 1150–1700 Å fluxes that include the reconstructed Lyα profiles (F(1150–1700)_fit), we find F(1235–1700)_LR/F(1150–1700)_fit =0.25 ± 0.17. Applying the 3σ screen, F(1235–1700)_LR/F(1150–1700)_fit = 0.17 ± 0.08. The general result is that even with the additional contribution from H₂ emission, and the uncertainty associated with the time-variability of the sources, low-resolution FUV spectra that do not include a reconstructed Lyα emission line underestimate the total FUV flux by a factor of ≈ 6. While a factor of ≈ 6 is the average flux underestimate, Table 3 shows that in the extreme cases (e.g., AA Tau and V4046 Sgr), the low-resolution 1235–1700 Å data underestimate the true intrinsic stellar+accretion FUV radiation field strength by factors of ⩾15.

Our conclusion from this comparison with low-resolution data in the literature is that ACS/SBC PR130L and STIS G140L-based flux estimates underestimate the true FUV radiation field strength from CTTSs by approximately an order of magnitude. Our findings quantitatively support the assertions of previous authors (Bergin et al. 2003; Herczeg et al. 2004; Schindhelm et al. 2012b) that Lyα must be included to make an accurate of estimate of both the strength and the shape of the FUV radiation field in protoplanetary environments around low-mass stars. It should also be emphasized that the largest uncertainty on the inferred local FUV radiation field strength is likely the uncertainty on the line-of-sight reddening, both interstellar and circumstellar. Typical dispersions on the derived values of A_V in the literature are 0.3–1.0 mag, the relative contributions of interstellar and circumstellar grains are unclear (see, e.g., McJunkin et al. 2014 and references therein), and the scattering and extinction properties of grains in regions like Taurus may be quite different than those found in the diffuse interstellar medium (Calvet et al. 2004). With empirically derived UV radiation fields now available, the uncertainty on the extinction is likely the limiting factor on our knowledge of the absolute flux of the energetic radiation environments around young stars.

The H i Lyα emission line dominates the energy output of CTTSs in the FUV bandpass, contributing ∼70%–90% to the total 912–1700 Å flux (Herczeg et al. 2004; Schindhelm et al. 2012b). This line, produced in the protostellar atmosphere, accretion shocks, and extended outflows (Walter et al. 2003) can be hundreds of km s⁻¹ wide (Schindhelm et al. 2012b), with luminosities ranging from roughly 0.25%–40% L_☉. Due to resonant scattering of neutral hydrogen in the interstellar and circumstellar media, the intrinsic Lyα radiation field cannot be directly measured for any star other than the Sun. In order to produce an accurate estimate of the local Lyα flux incident on the disk surface, the emission lines must be reconstructed using additional observational constraints. Wood et al. (2002) and Wood & Karovska (2004) pioneered a technique of Lyα emission profile reconstruction using the fluorescent H₂ lines as a tracer of the Lyα radiation field incident on the disk surface. This method, adapted to CTTS disks by Herczeg et al. (2004), uses the measured H₂ emission line strengths for a given [v', J'] → [v'', J''] progression to determine the total absorbed Lyα flux at a given pumping wavelength. The grid of Lyα fluxes is then fit with an intrinsic line shape and intervening absorber to infer the local Lyα profile and reproduce the observed molecular fluorescence spectra. We have recently extended this technique to model the Lyα radiation fields and FUV CO emission from 7 stars (Schindhelm et al. 2012a).

With the advent of larger medium-resolution CTTS samples observed with the HST-COS, the fluorescent H₂-based Lyα reconstruction can be applied to larger target samples. In the present work, we adopt the reconstructed Lyα radiation fields from Schindhelm et al. (2012b) as the input to our intrinsic CTTS FUV spectra. They use the fluxes from 12 fluorescent H₂ progressions (France et al. 2012b) to infer the intrinsic Lyα radiation fields, assuming a single component Gaussian Lyα emission line, a blueshifted H i outflow component, and uniform covering fraction for the absorbing H₂. In addition to the 14 targets that were presented in the Schindhelm et al. sample, we have added new Lyα line reconstructions of DF Tau and DR Tau. DR Tau could not easily be fit by our single-component model, with enhanced fluorescent emission pumped near the Lyα line core. We interpret this as a second source of Lyα photons, most likely produced in an outflow, as has been observed in T Tau (Walter et al. 2003; Saucedo et al. 2003). The Lyα spectra are presented in Figure 8. The intrinsic profiles (normalized to the flux at 1 AU) are shown in black and the outflow-absorbed profiles as observed at the disk surface are shown in red.

The accretion shocks (including pre- and post-shock regions and the heated stellar photosphere near the base of the accretion columns) and stellar chromosphere/transition regions will produce other strong emission lines formed at temperatures from T_form ∼ 10⁴–3 × 10⁵ K. The higher ionization lines are above the typical temperatures of the intervening interstellar material, and therefore are not subject to strong attenuation before reaching Earth. For hot gas lines formed near the stellar surface, we use the observed emission spectra, subtracted for a local continuum level, and then combined with the FUV continuum fit described below. The emission lines included in our FUV radiation field are C iii λ977, O vi λλ1032,1038, H i Lyα λ1216 (described above), N v λλ1239, 1243, C iv λλ1548, 1550, and He ii λ1640 Å. We do not include emission lines of silicon because these lines are weak and occasionally depleted in CTTS accretion spectra (Herczeg et al. 2002; Ardila et al. 2013). Inclusion of the C iii and O vi lines is important because hot gas lines will contribute a significant fraction of the stellar+shock flux at λ < 1100 Å where most of the H₂ and CO absorption bands reside. However, only TW Hya, DF Tau, RU Lupi, and V4046 Sgr have moderate S/N spectra in the FUSE archive, owing to the faintness of these sources at the shortest UV wavelengths. In order to present a uniform analysis, we take the emission profiles of the highest quality sub-1100 Å data (TW Hya) as representative of all of the CTTS spectra. The TW Hya C iii and O vi emission line profiles are then scaled to the C iv flux level for each source. This may introduce substantial errors to the sub-1100 Å data, but in the absence of a self-consistent model for the entire FUV bandpass, this simple prescription was adopted.

Emission from N v and C iv are blended with several fluorescent H₂ emission lines; in order to present the intrinsic hot gas lines without contamination from the surrounding circumstellar material, we have developed a technique to fit and remove these lines. For N v, we load the 1237.1–1243.9 Å spectra into the multi-Gaussian fitting routine described in France et al. (2012b) and fit all of the observed features with Gaussian emission lines convolved with the appropriate wavelength dependent HST+COS line-spread function (LSF; Kriss 2011). ¹⁰ Emission lines with FWHM < 100 km s⁻¹ and within ±40 km s⁻¹ of the rest wavelength of a known H₂ emission line pumped by Lyα [B–X (2–2)R(11) λ1237.54 Å, B–X (1–2)P(8) λ1237.87 Å, and C–X (1–5)R(9) λ1240.87 Å] were then subtracted from the data. For C iv, the 1546.1–1553.9 Å spectra were fitted and H₂ lines [B–X (1–8)R(3) λ1547.34 Å and B–X (3–7)P(17) λ1551.76 Å] subtracted. The "H₂ cleaning" process is displayed graphically in Figure 9. The final N v, C iv, and He ii profiles used in the FUV radiation fields, scaled to the flux at 1 AU, are displayed in Figures 10(a)–(c), respectively.

We compared the "H₂ cleaned" N v λ1238 and C iv λ1548 + λ1550 raw line fluxes (without correction for reddening or scaling to 1 AU) to the fluxes for these lines presented by Ardila et al. (2013). We find that our C iv fluxes agree to better than 10% and the N v fluxes derived in this work agree to within 10%–20% of those measured by Ardila et al. The minor residual differences are most likely attributable to the H₂ removal procedures employed; fitting and removing the contaminating H₂ lines (this work) versus interpolating over wavelengths contaminated by H₂ (Ardila et al.). The differences are slightly larger for N v, where the relative contribution of H₂ fluorescence is larger.

The inclusion of lower ionization state lines is complicated by abundance variations and absorption from outflows and the interstellar medium (see, e.g., Johns-Krull & Herczeg 2007). The strongest magnetospheric lower ionization emission lines in the spectra of WTTSs are Si iii λ1206 and C ii λλ1334, 1335 Å. As noted above, some CTTSs show no silicon emission in their accretion spectra (Si ii, Si iii, Si iv; see also France et al. 2010), possibly due to depletion of refractory elements into grains during the evolution of the disk in the first 10 Myr. However, other CTTSs show broad Si iii emission lines (e.g., BP Tau), likely formed near the stellar/magnetospheric accretion region. C ii lines are usually blended with strong H₂ features and strong ISM+CSM absorption lines, which makes the intrinsic C ii line flux extremely challenging to recover. In order to assess the relative importance of these low-ions to the total FUV radiation field strength, we fit the Si iii profiles (when present) and integrated over the 1334–1337 Å region covered by C ii because the actual C ii emission lines could not be clearly identified in all cases. These low-ion line fluxes were then compared to the flux in the C iv doublet. We find only two stars where F(C ii)/F(C iv) or F(Si iii)/F(C iv) approach unity: HN Tau and RECX-15. A quick inspection of Table 2 shows that these stars are anomalously weak C iv emitters, therefore the low ions are not particularly strong in these cases, but instead the higher ions are particularly weak. Excluding these two sources, we find 〈F(C ii)/F(C iv)〉 = 0.12 ± 0.07 and 〈F(Si iii)/F(C iv)〉 = 0.05 ± 0.04. Therefore, these emission lines are not included in the compiled radiation fields as they contribute of order 0.2% to the total FUV irradiance from the central stars, although we caution that this number is highly uncertain in the case of C ii; it is possible that this emission line may contribute as much as 1%–2% of the total FUV radiation field.

The NUV continua from CTTS are consistent with shock-generated Balmer continuum emission, and can be modeled to determine the mass accretion rate onto the star when the interstellar reddening is known (e.g., Calvet & Gullbring 1998; Ingleby et al. 2013). In TW Hya, the FUV continuum has been shown to be in excess of what is predicted by simply scaling the Balmer continuum to shorter wavelengths (Herczeg et al. 2004); contamination by atomic and molecular emission lines and low flux levels make the true FUV continuum essentially impossible to measure without moderate spectral resolution and low instrumental backgrounds for all but the brightest CTTSs. Previous studies of the FUV continuum have suggested two primary origins for this emission; a hot accretion component (Costa et al. 2000; Herczeg et al. 2004; France et al. 2011b) and an H₂ molecular dissociation quasi-continuum generated by collisions with non-thermal electrons (Bergin et al. 2004; Herczeg et al. 2004; Ingleby et al. 2009). France et al. (2011a, 2011b) presented the first detailed observations of the FUV continuum, concluding that while a combination of accretion continuum, electron-impact H₂ emission, and CO fluorescence (excited by Lyα and C iv photons) can reproduce some features of the spectra, there are significant discrepancies between the predicted and observed spectral features from this process.

We have created new FUV continuum spectra for the 16 targets considered in this study. Owing to the various H₂ and CO pumping transitions populated in different targets, each CTTS molecular emission spectrum is essentially unique. In order to find true continuum regions, we searched the spectra by hand and identified 210 spectral points where a 0.75 Å (approximately 10 spectral resolution elements) emission line-free window can be identified between 1138 and 1791 Å. We refer to these as "binned FUV continuum spectra." Each individual binned FUV continuum point is taken as the mean of the flux in this ±0.375 Å spectral window free of molecular and atomic line emission. The error on each individual binned FUV continuum point is taken to be the standard deviation about the mean flux in the ±0.375 Å spectral window. A comparison of the observed spectra and the binned FUV continua spectra is shown in detail for the 1245–1275 Å spectral region in Figure 11. The binned continuum spectra are then corrected for interstellar reddening, assuming the optical extinctions given in Table 1 and R_V = 3.1.

In order to estimate the true underlying continuum, we fitted the spectra in three broad regions free from residual features in the binned FUV continuum spectra (over the intervals Δλ = 1160–1190 Å, 1245–1330 Å, and 1663–1685 Å) with a second order polynomial. In Figure 12, we show the short-wavelength portion of the binned FUV continuum spectrum with the polynomial fit overplotted as the red dashed line. In Table 4, we present the average flux and the average S/N per binned continuum point for each target in the sample, in both short-wavelength (1140–1340 Å) and long-wavelength (1660–1700 Å) bands free of molecular continuum emission. While the average S/N per binned continuum point is >2.5 for all stars in the short-wavelength portion of the HST bandpass used in this work (1140–1340 Å), we caution that the data quality of the continuum for the fainter targets (DE Tau, DR Tau, LkCa15, SU Aur, TW Hya, and UX Tau A) makes the short-wavelength continuum fitting less reliable. In these cases, simply taking the average FUV continuum level may be more appropriate than the polynomial fit to the FUV continuum.

We have previously described the use of a first order linear fit (France et al. 2011b), but we employ a curvature term here to account for uncertainty in the reddening correction. While most of our spectra maintain appreciable S/N to λ ∼ 1760 Å, we truncate the fits at 1685 Å to accommodate the long wavelength cut off of the STIS E140M data used for TW Hya. The binned continua are shown on a linear scale, with the polynomial continuum fits overplotted, in Figure 2 and the extrapolation of our continuum fit to λ < 1140 Å is shown compared with archival FUSE observations in Figure 3.

The complete binned FUV continuum spectra are shown scaled to the flux at 1 AU in Figure 13. There are several residual features in this continuum spectrum: (1) emission excess between 1200 and 1230 Å that may be a very broad (±4000 km s⁻¹) Lyα component, (2) CO A–X absorption bands between 1400–1520 Å (e.g., AA Tau and DE Tau; McJunkin et al. 2013), (3) absorption from low-ionization atomic species most likely arising in outflows (e.g., DR Tau and RU Lupi; Herczeg et al. 2005), and (4) an excess emission feature spanning 1520–1660 Å in some targets (e.g., AA Tau, DM Tau, GM Aur, RECX-15, V4046 Sgr). This latter feature is the "1600 Å Bump" and we will present a detailed analysis of this quasi-continuous emission feature in a future work. Figure 12 shows the binned spectra at the short-wavelength end of the HST bandpass; broad Lyα emission lines can be seen in the binned continuum spectra. The center of the Lyα emission line is contaminated by geocoronal emission in most cases and has been removed, resulting in the zero-flux points seen in from 1210 to 1220 Å in Figure 12.